Design of Piping Systems


Design of Piping Systems

11 Pullman Power Products

" A Wheelabrator-Frye Company

Revised Second Edition



New York • Chichester • Brisbane • Toronto


Copyright @ 19·n. 1956 by

The M. W. Kellogg Company

All Righl$ Reserved

Reproduction or translation of any part of this work beyond that permitted by Sections 107 or 108 of the 1976 United States Copyright Act without the permission of the copyright owner is unlawful. Requests for permission or further information should be addressed to the Permissions Department, John Wiley & Sons, Inc.

Revised Second Edition

20 19 18 17 16 15 14

Nothing contained in Design of Piping System.s is to be construed B8 granting any right of manufacture, sale or use in connection with any method, apparatus Of product covered by Letters Patent, nor as insuring anyone against liability for infringement

of Letters Patent.

ISBN 0 471 46795 2

Library of Congress Catalog Card Number: 56-5573

Printed in the United States of America


A volume bearing the title Design of Piping Systems, devoted solely to the study of expansion stresses and reactions in piping systems, was privately published by The M. W. Kellogg Company early in 1941. It made available for the first time an adequately organized, comprehensive analytical method for evaluating the stresses, reactions, and deflections in an irregular piping system in space, unlimited as to the character, location, or number of concentrated loadings or restraints. It was the culmination of an intensive, widespread effort to meet the recognized need for refined analysis capable of general application to the increasing number of critical piping services required by technological progress, and to the increasingly severe problems which they posed. The timely availability of this reliable and versatile approach, now widely known as the Kellogg General Analytical Method, made it possible to provide satisfactory design for the avalanche of critical and pioneering piping requirements associated with World War II plant design, and proved to be a major step in accelerating acquaintance with accurate thermal expansion analysis and appreciation of its potentialities for more extensive application.

Since the war, technological progress and the trend to larger scale, more complex units has continued unabated, while the attendant increased pressures, temperatures, and structural complexities have resulted in larger pipe sizes, heavier wall thicknesses, and a marked increase in alloy construction. Concurrently, the wartime-fostered universal acceptance of adequate piping flexibility analysis for critical service has paved the way for more searching examination of the over-all economics of erected piping by relating potential fabrication, materials, and operating savings to increased engineering costs. Earlier concepts, which regarded piping as trivial and expendable, are fast disappearing in view of the rising costs of field corrections and loss of plant operation - and also with the recognition that piping represents an increasing percentage of initial plant expenditure.

The importance of sound piping design is now well recognized not only by designers and users, but also by authorities concerned with public safety. The Code for Pressure Piping Committee (ASA B31.1) has increased its membership and activity over the past several years and a Conference Committee has been organized, composed of the chief enforcement authorities of each State or Province that has adopted a portion or all of the Code. Significant improvements in the rules have already resulted in the revised minimum (and now mandatory) requirements for piping flexibility. With this trend, the ASA Code is now rapidly assuming the status of a mandatory Safety Code, whereas previously it had served designers and users primarily as a recommended design practice guide.

The critical shortage of engineering personnel during World War II prevented the completion of sections on other aspects of piping design that had been planned for inclusion in the original edition of Design of Piping Systems. As the shortage persisted, considerable time elapsed before resumption of work could be considered. Meanwhile, many requests for extension and suggestions for improvement were



received from readers of the text already published. Review of these and other


developments in light of extended experience led to the conclusion that a new

edition was warranted": As the work got under way, it was soon evident that broadening of the subject matter would have to be limited to treatment of the structural phase of piping design; coverage of the entire field, including fluid flow; system design and layout, valve design, piping fabrication and erection, etc., would require much more than the desired single volume.

It is the objective of this Second Edition to supplement Code rules and other readily available information with specific mechanical design approaches for entire piping systems as well as their individual components and to provide background information which will engender understanding, competent application of analytical results, and the exercise of good judgment in handling the many special situations which must be faced on critical piping. In line with this objective, the opening chapter presents a condensed treatise on the physics of materials. It is followed by a comprehensive study of the capacity of piping to carry various prescribed loadings. The utilization of materials is then considered, not only in relation to fundamental knowledge but also on the basis of conventionally accepted practices.

The present edition also includes a greatly augmented treatment of local flexibility and stress intensification, and a chapter on simplified methods of flexibility analysis contains several newly developed approaches which should prove helpful for general assessment of average piping, or in the planning stage of the design of critical piping. The Kellogg General Analytical Method, now extended to include all forms of loading, has been improved in presentation by the use of numerous sample calculations to illustrate application procedures, and by placing the derivations of the formulas in an appendix. Included in this edition are chapters on expansion joints and on pipe supports that offer, it is believed, the first broad treatment of these items with regard to critical piping. The rising significance of vibration, both structural and fluid, is recognized in the final chapter, which was also prepared especially for this edition. For ready accessibility of information, the charts and tables most frequently needed for reference have been grouped at the end of the text, and a detailed subject index has been provided.


The M.W. Kellogg Company became a subsidiary of Pullman Incorporated in 1944, and in 1975 was re-named Pullman Kellogg. In 1977, the Power Piping, Chimney and Mechanical Construction Operations of Pullman Kellogg became the Pullman Power Products division of Pullman Incorporated.


...... '

This volume is based on the broad experience, background, and mechanical engineering accomplishment of The M. W. Kellogg Company in the field of piping design. It reflects the numerous achievements and contributions of the Company to effective piping design for high temperature and pressure service. As with the First Edition, the preparation of this book has been sponsored by the Fabricated Products Division of which Waldo McC. McKee is Sales Manager. This work could be brought to realization only through the cooperation of the entire engineering staff of the Company and, in particular, of the Piping Division.

Certain individual contributions deserve specific acknowledgment. H. Wallstrom provided the major original contributions to the Kellogg General Analytical Method and its extensions (Chapter 5 and Appendix A). He was ably assisted in this work by Mrs. Catherine R. Gardiner.

Professor E. Orowan of the Massachusetts Institute of Technology, retained consultant of The M. W. Kellogg Company, is responsible for the contents of Chapter 1.

J. J. Murphy and N. A. Weil collaborated in composing Chapters 2 and 3 and assisted in the preparation of Chapters 1 and 7. Chapter 4 is the result of a cooperative effort between H. WaUstrom and N. A. Weil; L. C. Andrews is credited with the writing of Chapter 6.

Credit for the most significant contributions to Chapters 7 and 8 is due to E. F. Sheaffer. M. Yachter, assisted by S. Meerbaum, prepared Chapter 9 and Appendix B.

In addition to credits for Chapters, the following special contributions are acknowledged. J. J. Rush and M. Hartstein developed The Guided Cantilever Method of Chapter 4. L. Morrison contributed to the general phases of piping design. Valuable suggestions were supplied by M. G. Schar on Chapter 8 and by S. Chesler on Chapter 9. Credit is due to J. T. McKeon for his notable comments and assistance in reviewing and proof-reading this volume. L. Mylander is to be commended for co-ordinating portions of this work.

The task of assembling and editing the Second Edition was carried out by E. F. Sheaffer. N. A. Weil performed the review and inserted corrections for the second printing of this Edition. The entire project has been under the direction of D. B. Rossheim, who has guided the design principles and philosophies embodied in this work.

As is the case with most advances in the engineering art, the First Edition and this significantly extended Second Edition of Design of Piping Systems have greatly benefited from the research and contributions of other investigators. Their many valuable contributions are covered in the lists of references at the ends of the various chapters and in the "Historical Review of Bibliography" of Appendix A.


V icc-President, Engineering The M. W. Kellogg Company


In :M emory of


In all of his career, Mr. Rossheim's ability, dedicat.ion and friendliness were an inspiration to his associates and won f01' him everyone's affection and respect.





Chapter 1

Strength and Failure of Materials

1.1 Stable and Unstable Deformations

1.2 Plasticity 2

A. Plastic Deformation under Uniaxial Stress,

2; B. Triaxial Stress: Yield Conditions, 3;

C. Plastic Stress-strain Relationships for Tri-

axial Stress, 4.

1.3 Failure by Plastic Instability 5

A. Instability of Plastic Extension: the Ulti-

mate Tensile Strength, 5; B. Instability of the Plastic Expansion of Tubes, VesseIs, and Plates, 6; C. Ultimate Stress and Working Stress, 7.

1.4 Creep 8

A. The Andrade Analysis of the Creep Curve,

8; B. Transient Creep, 9; C. Viscous Creep,

10; D. Creep under Triaxial Stress, 11; E. The Mechanism of Creep, 11; F. Evaluation and Engineering Use of Creep Tests, 12; G. Creep Fracture, 13.

1.5 Types of Fracture; Molecular Cohesion; the

Griffith Theory 13

1.6 Ductile Fractures 15

1.7 The Brittle Fracture of Steel ("Noteh Brittle-

ness") 16

1.8 Fatigue 20

A. General Features, 20; B. The Mechanism

of Fatigue, 22; C. Influence of a Superposed Steady Stress, 23; D. Influence of a Compound State of Stress, 25; E. Influence of Notches and of Surface Flaws, 25; F. Fatigue Tests on Specimens vs, Fatigue Tests on Structural Parts, 26; G. Periodically Varying Thermal Stresses, 26; H. Thermal Fatigue, 27;

J. Damage by Overstress, 27; K. Corrosion Fatigue, 28.

Chapter 2

Design Assumptions, Stress Evaluation,

and Design Limits 30

2.1 Codes and Standards 30

2.2 Design Considerations: Loadings 32

2,3 Design Limits, Allowable Stresses, and Allowable

Stress Ranges 34


2.4 Stress Evaluation 43

a. Internal Pressure up to 3000 psi Maximum,

43; b. Internal Pressure over 3000 psi, 44;

c. External Pressures, 46; d. Expansion, 47;

c. Other Loading, 47.

2.5 Combination of Stress: Stress Intensification and

Flexibility Factors 47

2.6 Evaluation of Deflections and Reactions 48

2.7 Design Significance of Inspection and Tests 50

Chapter 3

Local Components 52

3.1 Pipe Bends: Structural Loading (Static and Cyclic) 52

3.2 Pipe Bends: Internal Pressure 60

3.3 Miter Bends 60

3.4 Bends and Miters: Summary 61

3.5 Branch Connections: Static Pressure Loading 62

3.6 Branch Connections: Repeated Loading 66

3.7 Branch Connections: Comparison with Code Re-

quirements 67

3.8 Branch Connections: Practical Considerations and

Summary 69

3.9 Corrugated Pipe 70

3.10 Bolted Flanged Connections: General Background 74

3.11 Bolted Flanged Connections: Practical Considera-

tions 77

3.12 Joints Between Dissimilar Materials 79

3.13 Other Components 81

3.14 Piping and Equipment Intereffecta 83

Chapter 4-

Simplified Method for Flexibility Analysis 90

4.1 Scope and Merits of Approximate Methods 90

4.2 Thermal Expansion 91

4.3 Preliminary Segregation of Lines with Adequate

Flexibility: Code Rules 92

4,4 Selected Chart-form Solutions 94

4.5 Approximate Solutions 97

4.6 The Simplified General Method for Square-eorner

Systems 102

4.7 Approximating the Effeot of Curved Pipe and

Other Components 107




Chapter 5

Flexibility Analysis hy the General

Analytical Method. 115

5.1 Scope and Field of Application of the General

Analytical Method 115

5.2 Calculating Aids 116

5.3 General Outline of Operations 117

5A The Solution of Simultaneous Equations 117

5.5 Single Plane Calculati~ns 119

5.6 Inclined Members and Changes in Stiffness 120

5.7 Circular Members 123

5.8 General Shape Coefficients 125

5.9 The Secondary Term 125

5.10 Effects of Direct and Shear Forces 127

5.11 Working Planes and Cyclic Permutation 127

5.12 Multiplane Pipe Lines with Two Fixed Ends 128

5.13 Hinged Joints and Partially Constrained Ends 129

5.H Skewed Members 13-1

5.15 Branched Systems 145

5.16 Intermediate Restraints 146

5.17 Calculation of Deformations at any Point 153

5.18 Symmetrical Pipe Lines 157

5.19 Inversion Procedures 157

5.20 Cold Springing 166

5.21 Weight Loading 170

5.22 Wind Loading 185

Chapter 6

Flexibility Analysis by Model Test 198

6.1 The Experimental Approach HI8

6.2 The Routinized Model Test 198

6.3 The Kellogg Model Test 200

6.4 The Kellogg Model Test Laboratory and Equip-

ment 201

6.5 Typical Model Tests 202

Chapter 7

Approaches for Reducing Expansion Effects;

Expansion Joints 210

7.1 Introduction 210

7.2 Sources of Excessive Expansion Effects 210

7.3 Approaches for Reducing Expansion Effeets 210

7.4 Packed Type Expansion Joints 212

7.5 Bellows Type Expansion Joints 214

a. Discussion, 214; b. Bellows Details, 214;

c. Support and Protection of Bellows 216;

d. Fabrication of Bellows Joints, 217; c. Establishing Purchasing Requirernenta for Bellows Joints, 219; f. Materials and Deterioration, 220; (I. Fatigue Basis for Predicting Bellows Life, 220; h. Testing and Quality Control of Bellows Joints, 222.

7.6 Expansion Joints with Built-In Constraints 223

7.7 Establishing Expansion Joint Movement De-

mands 226

Chapter 8

Supporting, Restraining, and Bracing

thc Piping 231

8.1 Terminology and Basic Functions 231

8.2 Layout Considerations to Facilitate Support 233

8.3 The Elements of the Supporting System: Their

Selection and Location 236

8.4 Fixtures 243

8.5 Pipe Attachments 248

8.6 Structures and Structural Connections 251

8.7 Erection and Maintenance of the Supporting, Re-

struining, and Bracing System 254

Chapter 9

Vibration: Pre"cntion and Control 257

9.1 Introduction 257

9.2 Fundamental Considerations in Piping Vibration 258

a. Definitions, 258; b. Types of Vibration, 258;

c. Sources of Periodic Excitation, 259; d. Vibration Prevention and Control, 259.

9.3 Structural Natural Frequency Calculations 260

a. The Spring-Mass Model, 260; b. Frequency

and Mass Effectiveness Factors for Different

End Constraints, 261; c. Variable Stiffness and Variable Mass, 263; d. Combined BendingTorsion, 264; e. Approximate Natural Frequencies of Pipe Bends with Two Members (Vibration Perpendicular to Plane of Bend),

265; f. Plates and Radial Mode in Pipe, 266.

9.4 Structural Resonance and Magnification Factors 267 9.5 Dumping of Structural Vibrations 270 a. Hydraulic Snubbers, 270; b. Elastic Foundations for Rotating Machinery, 271.

9.6 Acoustic Natural Frequency Calculations 273

a. The Organ Pipe and Resonators, 273;

b. Special Cases of Multiple Resonator Forrnulas,274; c. Piping Systems with Branches and Enlargements, 276.

9.7 Acoustic Resonance and Magnification Factors 277

9.8 Flow Pulsation Smoothing 279

a. Tuned Resonators, 279; b. Surge Tanks,

279; c. Gas Pulsation Dampener Principles,

280; d. Acoustic Expansion Tank, 281; e. Comparison of Gas Pulsation Smoothing Devices,

282; f. Hydraulic Hammer, 283; g. Magni-

tude and Direction of Forces on Piping Bends,


9.9 Illustration of Vibration Analysis of a Simple

Piping System 285

a. General Data and Estimates, 285; b. Estimates of Structural Natural Frequencies of Piping System, 285; c. Estimate of Lower Bounds of Structural Natural Frequencies, 286;

d. Effect of Elasticity of Machine Foundation,

286; e. Estimate of Hydraulic Snubber Force

and Damping Requirement for Reduction of Amplitude of Vibration, 287; f. Resonance Effect due to Wind Velocity, 287; fl. Estimate

of Acoustic Natural Frequencies, 287; h. Estimate of Acoustic Frequency of the System Corresponding to its First Harmonic (2nd Mode), 288; i. Estimates of Some Possible Resonator Frequencies, 288; j. Estimate of




Volume and Pressure Drop Requirement of Hydraulic Filters (Bottles) in the Compressor Discharge Lines, 290; k. Tuned Resonator

Geometry,290. . ...

H.lO Piping Vibration "Trouble Shooting" 291

a. Background, 291; b. Vibration Measurement, 292; c. "Trouble Shooting" Procedure,


Appendix A

History and Derivation of Piping

Flexibility Analysis 295

.\.1 History of Piping Flexibility and Stress Analysis 295 A.2 Bibliograpby on Piping Flexibility and Stress

Analysis 297

A.3 Derivation of the General Analytical Method 299

Appendix B

Derivation of Acoustic Vibration Formulas 328

B.I Multiple Resonator of nth Order 328

B.2 General Characteristic Equation for a Branched

Piping System 329

B.3 Tuned Resonator Relations 331

B.4 Simplified Surge Filter Analysis 333

Appendix C

Charts and Tables 336

C- I Properties and Weights of Pipe 336

C- 2 Thermal Expansion, Carbon and Alloy Steels 341

C- 3 Modulus of Elasticity, Carbon and Alloy Steels 342

C- 4 Chart for Criterion in Par. 620 (a) in Code for Pres-

sure Piping ASA B3Ll 343

C- 5 Length of Leg Required, Two-Member System, Both Ends Fixed, Thermal Expansion in Plane of

Members 344

C- 6 Moments and Forces, Two-Member System, Both Ends Fixed, Thermal Expansion in Plane of

Members 345

C- 7 Length of Leg Required, Two-Member System,

Both Ends Fixed, One Support Displaced in the

Direction of Adjoining Member 346

C- 8 Moments and Forces, Two-Member System, Both

Ends Fixed, One Support Displaced in the Direc-

tion of the Adjoining Member 347

C- 9 Length of Leg Requi!e'd, Two-member System,

Both Ends Fixed, One Support Displaced Normal

to Plane of Members 348

C-1O Moments and Forces, Two-Member System, Uoth

Ends Fixed, One Support Displaced Normal to

Plane of Members 349

C-11 Required Height, Symmetrical Expansion Loop 350

C-12 Moments and Forces, Symmetrical Expansion

Loop 351

C-13 Guided Cantilever Chart 352

C-14 Correction Factor I, Guided Cantilever Method 353

C-15 Design Data: Trigonometric Constants for Circu-

lar Members 354

C-16 Span V8. Stress, Horizontal Pipe Lines, Uniform

Load 356

C-17 Span VB. Natural Frequency and VB. Deflection,

Horizontal Pipe Lines, Uniform Load 357

C-18 Correction Factors for Use with Charts C-16 and

C-17 358

Appendix D

A Mutrix Method of Piping Analysis

nnd The Usc of nigital Computers

5A-l Introduction

5A-2 Derivation of the Shape Coefficient Matrix 5A-3 A Matrix Method of Piping Analysis

5A-4 An Example

5A-5 Selected Bibliography Index

359 361 362 369 372 378 379


a .

b .•........

C .

CalC"", etc .

c .

d .

e ..

i .

in .

g ..

h ..

h •.........

i ..

k .

l. .

m .

n ..

p .

q .

r ..

rio _,

Tm ••........

ro .

8 ..

Ba, 8aat 8' at etc.

t ..

U, Uo, U' 0, etc. v, VOl vloo, etc.

w .

X,II,%······ .



Definitions of Principal Symbols


Horizontal coordinate to midpoint of member in working plane.

Vertical coordinate to midpoint of member in working plane.

Distance of the working plane from the origin;

viscous damping coefficient.

Trigonometric constants. Critical damping coefficient. Diameter; inside diameter.

Unit linear thermal expansion for a temperature difference £iT; base of Napierian logarithms.

Frequency; factor. Natural frequency. Gravitational constant.

Bend characteristic (=tR/Tm1)j pitch of half corrugation of an expansion bellows; gra-

dient of pipe supports. .

Offset range of an expansion joint. Imaginary unit (= v"=l).

Flexibility factor of pipe in bending; spring


Length, span of pipe between supports. Mass.

Material constant, exponent in fatigue equation.

Pressure (load per unit area).

Plastic constraint factor; shape coefficient

known as the secondary term.


Inside radius. Mean radius. Outside radius.

Shape coefficient; steady stress component. Shape coefficients.

Time, thickness.

Shape coefficients.

Shape coefficients.

Width, unit weight load.

Unit loads in the x-, Y-, and e-directiona respectively.

Coordinate axes, coordinates of a point.



A " Area; activation energy; free end.

A.F. . . . . .. .. Attenuation factor.

B , Material constant.

a. . . . . . . . .. Cold spring factor; velocity of sound; constant.

D. . . . Diameter.

E Young's modulus of elasticity; joint efficiency.

E.. . . . . . . .. Young's modulus of elasticity at ambient


Eh. . . . . . . .. Young's modulus of elasticity at operating temperature.

F ..........

F%, F~, F •...

G .

I .

J ..

K .

L .

.M ..

M.F .

Mb ..

M'b .

M! .

M.,Mv,M •.

M' %, J,f'II, .M'.

MB · ..

N ..

0 ..

0' .

P ..

Q ..


Force component in the direction of axis indicated by subscript. Second subscript, if used, refers to the source of the force.

Shear modulus, diameter of the effective

gasket reaction on a flange.

Moment of inertia. Polar moment of inertia. Constant.


Moment .

Magnification factor.

Bending moment in the plane of the member. Bending moment transverse to the plane of

the member.

Torsional moment.

Moment component referred to origin and about axis indicated by subscript. Second subscript, if used, refers to the source of the moment.

Moment component about axis indicated by subscript. Second subscript, if used, refers to the source of the moment.

Any bending moment. Number of cycles, rpm. Origin.

Fixed end.

Point, concentrated load. Quotient, stiffness ratio, flow rate.



R .

R .

S ..

So · .

s, .

Se .. ······· .

s" .

SI ···· .

SA. .

SS .

Se .

Su ..

T ..

u ..

v ..

W : .

y .

z ..



Centerline radius of torus or curved member (pipe bend or elbow);..ratio.

U ni versal glLS constant.

Fatigue strength; stress, amplitude of alternating tensile stress component; shape coefficient; Strouhul number.

Bending stress in the plane of the member. Bending stress transverse to the plane of the member.

Allowable stress for u material at ambient temperature.

Allowable stress for a material at operating


Torsional stress.

Allowable stress range.

Resultant bending stress. Computed maximum stress range.

Ultimate tensile strength (conventional stress).

Temperature, amplitude of alternating shear

stress; period of vibration.

Velocity, energy; shape coefficient. Volume; shape coefficient.

Total uniform load.

Yield stress in uniaxial tension; resultant expansion.

Section modulus.


cc •......•..

{J ..

')' .

Ii .

e .•.........

f· . , .

flf ':::2, ~3 ••••.

r .

1] : .

O ••.••......

A .

j.l. ..

P ..

p .•..........

<1' ..

.tt, <1'2· 0'3· •..

T •••••••••••

.p .

t········· .

.6J •••••••..

A .

<1> .


Surface energy (work for creating new surface per unit area); angle; coefficient of linear expansion.

Longitudinal stress intensification factor; angle.

Shear strain, transverse stress intensification

factor, ratio of specific heats.

Translatory displacement; deflection. Normal (tensile or compressive) strain. Logarithmic strain.

Principal strains.

Viscous damping coefficient (damping ratio). Coefficient of viscosity.


Wave length.

Acoustic conductivity. Poisson's ratio. Density.

Normal (tensile or compressive) true stress. Principal stresses (true).

Shear stress.



Angular frequency.

Restrained linear thermal expansion. Angle.



Strength and Failure of Materials*

IN the simplest cases, the failure of a structural part occurs when a certain function of the stress or strain components reaches a critical value. The designer must know, then: (a) how the stresses and strains can be calculated from the applied load; (b) what are the critical combinations of stress and strain at which failure occurs.

The first question belongs to the field of applied mechanics (elasticity, mathematical theory of the plastic field, and mathematical rheology). In relation to piping systems, it will be treated in detail in subsequent chapters of this book.

The second question is concerned with the mechanical properties of solids, which is a chapter of the physics of solids. It is a relatively new field of science; until about 30 years ago, the mechanisms of fracture and of plastic deformation were almost unknown. Since 1920, however, the progress in this field has been rapid; at the same time, the demands on the designer's understanding of the mechanical behavior of materials have gone far beyond what is generally available in the traditional textbooks. Hence, it is appropriate to introduce the treatment of piping system design in this book with a brief but up-to-date sketch of the mechanical properties of solids.

Failure of a structural part can occur by (a) excessive elastic deformation,

(b) excessive non-elastic (plastic or viscous) deformation, or

(c) fracture.

The calculation of elastic deformations and of the conditions of elastic instability is the main subject of books dealing with applied elasticity (tradition-

"Prepared by Dr. Egon Orowan, George Westinghouse Professor of Mechanical Engineering, Mnssachusetts Institute of Technology.

ally, though inappropriately, entitled "Strength of Materials"). In the present chapter, only the conditions of failure by non-elastic deformation or fracture will be considered in detail. Failure by excessive deformation will be discussed in the first four sections, and failure by fracture in subsequent parts of the chapter.

1.1 Stable and Unstable Dcformutfons

A structure ceases to be serviceable if it suffers excessive deformation. The deformations leading to its failure may be elastic (i.e., deformations that disappear when the stress is removed), or nonelastic; the latter may be plastic (i.e., depending only on the deforming stress but not on the duration of its action), or they may represent a creep (i.e., they may increase or decrease with time at constant stress).

Moderate deformations (elastic or non-elastic) !TIay be beneficial in that they can redistribute the stress in a structural part or between several structural parts and so prevent its rise to levels at which fracture can occur.

In many cases, the deformation leads to changes of the shape of the body that cause an increase of the stresses produced by a given load. The simplest examples of this are elastic buckling, and the plastic extension of a rod in the course of which its cross section diminishes and the stress for a given load increases; if this increase is not counterbalanced by strain hardening, it leads to accelerated disruption. Such phenomena represent an elastic instability if the deformation is elastic, and a plastic instability if it is essentially plastic. Plastic instabilities are of great importance in the design of tubes and pressure vessels.

In what follows, failure by plastic instability will be treated separately, after the section dealing with





FIG. 1.1 Yield stress-strain curvc of copper in compression.

After Cook and Larke [I],

plastic failure without instability. Subsequently, failure by creep will be considered.

1.2 Plasticity

A. Plastic Deformation under Uniaxial Stress. As mentioned above, pure plasticity is defined as a non-elastic type of deformation without time influence. In uniaxial deformation, the plastic strain ~ is determined by the value of the stress 11 at which the deformation takes place


Elastic deformations also obey a law of this form; however, they are reversible, while in plastic deformation the relationship (eq. 1.1) is valid only for increasing stress. When the stress is reduced the


plastic strain remains approximately unaltered.

By its definition, pure plasticity means the absence of creep. No material satisfying this requirement is known; however, the behavior of ductile metals and other crystalline materials at not too high temperatures (compared with their melting point) can be described approximately as plastic.

The stress required for plastic deformation (often denoted by Y) is the yield stress. 1 Its dependence Ceq. 1.1) upon the preceding plastic strain is represented graphically by the "stress-strain curve" (more accurately, it would be called the yield stress-strain curve). The stress-strain curves of metals cannot be represented by a simple mathematical expression. For strains that are neither too small nor too large, they can often be approximated by a parabola

11 = constant X fn

At small plastic strains, as well as at very large ones, however, the stress-strain curve is usually quite different from the parabola representing it for moderate strains. In addition, the stress-strain curves of different metals are, as a rule, different in character.

lIn the treatment of plasticity, the term "yield stress" means the stress required for (initiating cr continuing) plastic deformation; owing to the presence of strain hardening, it changes with the plastic strain.


FIG. 1.2 Stress-strain curve of the "ideally plastic" material.

A familiar type, the stress-strain curve of copper, is shown in Fig. 1. L

For the calculation of the distribution of stress and strain in plastically deformed bodies, drastically simplified types of stress-strain curve must be used. Except in a few of the simplest cases, it is usually assumed for this purpose that yielding starts suddenly when a critical stress value is reached and that it progresses thereafter at a constant stre;s-in other words, that there is no strain hardening. Figure 1.2 shows the corresponding stress-strain curve of the "ideally plastic" material. It must be kept in mind that such a curve represents a sensible. though rough, approximation only if the plastic strain is large compared with the elastic strain. In the initial part of the stress-strain curve of a typical metal (compare Fig. 1.1), the deviation from the elastic line increases gradually and the idealized curve (Fig. 1.2) does not represent an approximation.

A few materials (notably, low-carbon steels) show the so-called "yield phenomenon": plastic deformation starts suddenly when the stress reaches the value of the "yield point." After its start, the stress required for further deformation may remain constant for a time, or drop immediately to a lower value (the "lower yield point"), as shown in Fig. 1.3. If such a stress drop occurs, the initial yield point is called the "upper yield point."

Of particular interest to the designer is the stress at which the plastic strain (or the total strain)

FIG. 1.3 Yield stress-strain curve of an annealed low-carbon steel.


reaches the maximum permissible value. If the stress-strain curve is of the character shown in Fig. 1.1, the value of the yield-stress at which the strain reaches some specified permissible amount (e.g., 0.2% or 0.02%) is called the 0.2% (or 0.02%) "yield strength" or "proof stress." Since the word "strength" is reserved in scientific usage for the fracture stress, the term "proof stress" will be used in the present chapter. If the yielding is discontinuous, as in Fig. 1.3, the entire range of commonly permissible strains, up to 1% or even 3%, lies on the horizontal part of the curve; in this case, the lower yield point takes the place of the proof stress. The upper yield point is a capricious quantity which can be obliterated by relatively small stress concentrations or small plastic deformations, so that the designer cannot rely on it.

Naturally, the proof stress is altered by preceding plastic deformation ("cold work"). Let OBD be the stress-strain curve of an annealed metal and OE the elastic line (Fig. 1.4); A is the point at which a critical strain of, say, 0.2% is reached. After straining in tension to B and removing the load (point e), a material is obtained of which the stress-strain curve in tension is eFD. The point F at which the permissible strain of 0.2% is reached is now higher than A, owing to the preceding strain hardening. On the other hand, if the same material, prestrained in tension to B, is subjected to compression, the microscopic residual stresses remaining in it give rise to perceptible plastic deformation even at very low compressive stresses, and the stress-strain curve in compression eG deviates from the elastic line strongly from the beginning. This softening of the material to reverse deformation is called the "Bauschinger effect." The hysteresis loop BeF observed when the stress is removed and then applied again is essentially the same phenomenon, due to directional microscopic residual stresses in a plastically strained material.

A mild heating (stress-relieving) after the deformation removes the residual stresses responsible for the Bauschinger effect and restores the proof stress for reverse deformation more or less to the increased level of the proof stress for deformation continuing in the initial direction.

B. Triaxial Stress: Yield Conditions. So far, only uniaxial stressing has been considered. If a general (triaxial) state of stress is present, with principal stresses 0'1 ~ 0'2 ~ 0'3, yielding in a material without a sharp yield point occurs when a certain mathematical expression containing the principal stresses reaches a critical value. Of several "yield




G FlO. 1.4 Increase of the proof stress by cold work; the Bauschinger effect.

conditions" suggested, only two have been found compatible with observations and at the same time simple enough for practical use: the Tresca (maximum shear stress) condition, and the von Mises (maximum octahedral stress) condition.

The Tresca yield condition {2] assumes that yielding occurs when the maximum shear stress. equal to one-half of the difference between the algebraically greatest and smallest principal stresses, reaches a critical value. It is expressed by

0"1 - 0"3 = Y (1.2)

where Y is the yield stress in uniaxial tension or com-. pression. With the Tresca condition, the inter mediate principal stress has no effect on yielding.

The Mises yield condition [3] assumes that yield. ing occurs when the "effective" shear stress2

1 ~I 2 2 2

Tcff= . ,r;:;:V (0'1-0'2) +(0'2-0"3) +(0"3-0')) (1.3)


reaches the critical value of the yield stress in pure shear, i.e., one-half of the yield stress Y in tension Expressed in terms of the uniaxial yield stress Y, it can be written as

1 ~I 2 2 2

Y= ..j2V (0"1-0"2) +(0"2-0"3) +(0'3-0'r) (1.4)

2The "octahedral" shear atress differs from the right-hand aide of eq. 1.3 by having the factor!, instead of 1/2V2, before the square root. The factor 1/2V2 has the convenience that it makes the right-hand side of eq. 1.3 equal to the maximum shear stress in the case of a uniaxial stress, i.e., for 0"2 = 0"3 '" O.



The Mises condition is often called the "shear strain energy condition," since, in an isotropic material, the right-hand side of eq, 1.3 or ·4A is proportional to that portion of the total energy which corresponds to the shear deformations. For anisotropic materials, however, the shear strain energy depends in general upon the hydrostatic component (pressure or tension) of the state of stress [4]. The attainment of a critical value of the shear strain energy, therefore, cannot be a condition of plastic yielding, which, except at extreme pressures, is not influenced by the hydrostatic component of the stress.

A characteristic feature of the Mises condition it; that the intermediate principal stress has an influence on the oceurrenee of yielding. Only if 0'2 is equal to the highest or lowest principal stress does eq. 1.4 coincide with the Tresca condition (eq. 1.2). The greatest divergence between the two conditions is present when the intermediate principal stress 0'2 is the mean value of the extreme ones

0'2 = !(0'1 + 0'3) In this case, eq. 1.4 becomes


vS Y = 0'1 - 0'3 "" 1.15Y (1.5)

That is to say, the maximum principal stress difference at yielding is about 15% higher according to the Mises condition than that given by the Tresca condition.

Experiments indicate that the behavior of metals with no sharp yield point, as a rule, is intermediate between the Tresca and the Mises yield conditions, usually somewhat closer to the latter. For mathematical investigations of stress and strain distribution in plastically deformed bodies, the Mises condition is often simpler to handle.

For materials with an upper and a lower yield point there is no reliable criterion for the onset of yielding at the upper yield point, since this quantity is extremely sensitive to slight non-uniformities of stress distribution and to the size of the specimen [5]. As mentioned, however, the upper yield point is of little importance to the designer, since the allowable stress must be based on the lower yield point, which is the stress required for the first Luders' bands to widen. From this it follows at once that the yield condition in this case cannot be the Mises condition. Since the Luders' bands are sheared layers embedded between still rigid blocks of the material, only the shear stress acting in their plane can cause them to become thicker, and the intermediate principal stress which is parallel to the Luders' layer and

perpendicular to the direction of shear in the layer must be ineffective. Consequently, the appropriate yield condition in this ease must be closer to the Tresca condition.

C. Plastic Stress-strain Relationships for Triaxial Stress. In the preceding section, the conditions of plastic yielding were considered. If they are satisfied and yielding occurs, the question of importance to the designer is how the resulting strains are determined by the applied state of stress. The difficulties of this problem become evident if one considers the facts that the resulting deformation depends on the sequence in which the stress components are applied, and that, owing to the Bausehinger effect, the slightest deformation destroys the initial isotropy of the material and makes reverse deformation easier than continued deformation. A plausible solution has been given only for the simple case of an ideally plastic isotropic material (strain hardening and the Bauschinger effect being ignored). According to this solution, a given triad of principal stresses 11'I, 0'2, 11'3 is related to the increment of plastic strain arising during its application; this increment is to be added to the plastic strains created by preceding actions of stresses. According to Levy and Mises,

OEl = OX[O'I - ! (0'2 + 0'3)1 O£2 = 0A[0'2 - !(0'3 + 0'.)]

OE3 = oA[0'3 - !(0'1 + 0'2)] (1.6)

where OElt OE2, 0£3 are simultaneous increments of the principal strains, and oX is a parameter determining the extent of the deformation. The Levy-Mises equations determine only the ratios of the principal strain increments; the absolute amounts depend on how long the straining is continued at the constant principal stresses 0'1. 0'2, 0'3·

In the literature, occasionally the stress-strain relationship

£1 = X[O'I - ! (0'2 + 0'3)] £2 = X[0'2 - ! (0'3 + O'dJ

f3 = X[0'3 - !(0'1 + 0'2)] (1.7)

is used. If the principal stresses remain invariant during the deformation, these equations represent simply the integrated form of the Levy-Mises equations; if not, they are incorrect. These equations are sometimes referred to as the "deformation theory," as contrasted with the Levy-Mises "incremental theory."

For strain-hardening materials, several authors



have suggested the generalized stress-strain relationship


where Tcff is the effective shear stress defined by eq, 1.3, and 'Y~f£ the effective shear strain defined by the analogous equation

'Y~[f= ~v (El -E2)2+ (E2-E3)2+ (E3 -El)2 (1.9) vZ

Equation 1.8 has not yet received sufficient experimental verification; it can be a satisfactory approximation only if the anisotropy due to preceding plastic deformation can be neglected.

1.3 Failure by Plastic Instability

A. Instability of Plastic Extension: the Ultimate Tensile Strength. Like elastic, so plastic or viscous deformation may also lead to buckling, e.g., of a compressed column, or of a thin-walled tube under external pressure. The treatment of such cases is analogous to that of elastic buckling, but the literature of plastic and viscous buckling is relatively small. For details, reference should be made to the published literature [6].

A case of plastic instability of great historical and practical importance is that occurring in the tensile test. Initially, the extension is uniform; unless fracture intervenes, however, the tensile load reaches a maximum in the course of the test, and at the same time a neck begins to develop. Further extension is then concentrated in the neck and ceases everywhere else in the specimen. The maximum load, divided by the initial cross-sectional area, is called the "ultimate tensile strength" or "ultimate tensile stress"; its significance for engineering design will be discussed in detail in Part C of the present section.

Let a = U(E) be the equation of the (true) yield stress-strain curve of a purely plastic material in uniaxial tension; the strain used is the linear strain defined as

E = (l - lo)llo


where l is the current length of the tensile specimen and io its initial length. Since the volume V docs not change significantly during plastic deformation, the product of length I and cross-sectional area A in the range of uniform extension remains constant:

LA = 1oAo = V

The load F = uA reaches a maximum when dF = a dA + A dr1 = 0




dr1/rr = -dAIA





FIG. 1.5 Considere's geometrical construction of the maximum load point and of the ultimate tensile stress.

Differentiation of eq. 1.11 gives dill = -dAIA

Combination of this equation with eq. 1.13 leads to

dulu = dIll Equation 1.10 can be written as I = lo(l + E)



from which

dl = 10 de

From the last two equations

dIll = de] (1 + E) Introduced into eq. 1.14, this results in da I de = a I (1 + E)

(1. Iii)

Equation 1.15, representing the condition for the load to reach a maximum during the tensile test, has a simple geometrical meaning. Let the stressstrain curve O'(E) be plotted in Fig. 1.5, and let the point P on the negative strain axis have the distance 1 from the origin; i.e., the same distance as the point Q on the positive strain axis representing E = 1 = 100% extension. For any point of the stress-strain curve, duldE is the gradient of the tangent line, and a 1(1 + E) the gradient of the line connecting the point (0', E) with the point P. The condition for the load maximum is equality of these gradients; i.e., the maximum occurs at the point M in which a line drawn from P is tangent to the stressstrain curve. The ordinate AM of the point of contact is the (true) stress at maximum load; OA is the tensile strain E" at maximum load. This theory of the maximum load point was given by Considere in 1885 [7}.

The ultimate stress,3 defined as the maximum load divided by the initial cross-sectional area,


3Since in the scientific treatment of this field the word "strength" ought to be reserved to !I. fracture stress, the ultimate strength will henceforth be called "ultimate stress."





Fro. 1.6 Determination of the instability stress on the true stress logarithmic strain curve in tension.

is not identical with the true stress at maximum load

c1m = Fmnx/A


The relation between them is S,,/urn = A/Ao which can be written as

S,,/Um = loll

in view of eq. 1.11. According to eq. tlOa, ~/l = 1/(1 + E)



SOl = Urn 1 + E"

where E" is the "uniform strain" at the moment of the load maximum, In Fig. 1.5, PO = 1; P A = 1 + E, and AM = O"m; from the similarity of the triangles PMA and PUO it follows, therefore, that the intercept OU of the ordinate axis between the origin and the tangent PM drawn from P to the stress-strain curve is the ultimate stress.

A similar graphical construction can be obtained if the logarithmic strain is used instead of the linear strain. The relationship between logarithmic strain E* and linear strain E is


E* = log. (1 + E)




dE* =--

1+ E

Substitution of this in eq. 1.15 'gives da lde" = 11


Figure 1.6 shows the corresponding graphical determination of the maximum load point from the logarithmic stress-strain curve: the subtangent P A at the maximum load point is unity.

n. Instability of the Plastic Expansion of Tubes, Vessels, and Plates. Plastic instability

due to the decrease of the load-carrying cross section occurs also when a tube or a hollow sphere is subjected to internal pressure [8, 9}. It is remarkable that the instability condition in these cases is not identical with that for the rod under tension, and the maximum pressure withstood by the tube or the spherical shell cannot be derived from the knowledge of the ultimate tensile stress, In view of the practical importance of these cases, their characteristic features should be pointed out.

For a hollow sphere of radius r and (small) wall thickness i, under an internal pressure p, the tensile stress 11 is given by

pr21f = 21rTtu (1.21)

The volume of the shell is

V = 41rT2t





Substituting eq. 1.23 into eq. 1.21 and observing that the volume remains constant during plastic deformation,

p = 4v::/V thu = CltHO" For a thin-walled closed tube,

2rp = 2lO"

V = 21fTt per unit of length


(1.25 )



For a square plate of edge length l and thickness t, extended uniformly in all directions in its plane by tensile forces F acting upon its edges,

F = ltl1 (1.27)




For the tensile specimen under uniaxial tension, already considered, the corresponding relationship would be

(1. 29)

where t is the thickness of the (round) rod.

It is seen that the pressure p or the force F as a function of the thickness of the specimen is given in all cases by an expression of the type

p (or F) = Ct:« (1.30)

where n = 2 for the tensile rod and the thin-walled tube, ,~- for the thin-walled hollow sphere, and! for the uniform-biaxially extended plate.


The maximum load or maximum pressure at which the extension becomes unstable is obtained from

dp (or dF) =''0

In view of eq. 1.30, this means

ntl1-1u dt + in dO' = 0


n(dt/t) = -du/u

For the hollow sphere, the tube, and the plate, dt/t = -d€*, where E* is the logarithmic strain perpendicular to the wall or the plate. Thus, the condition of instability is

du/dE* = nO' For the sphere, this is

da /d€* = {3/2)u

(1.31 )

for the tube

du/d€* = 2cr

and for the plate

de /d€* = (1/2)0'

For the tensile rod, dt/t is the increment of the transverse logarithmic strain; since the volume is constant, this is - (1/2)d€*, where dE* is the increment of the longitudinal logarithmic strain. Thus,

du/de* = a as before (cf. eq. 1.20).

Figure 1.7 shows the corresponding graphical construction, quite analogous to that in Fig. 1.6, carried out for the four cases. It shows that the instability point on the stress-strain curve (true maximum stress vs. greatest logarithmic strain) is different for each.

Particularly interesting is the practically important case of the thick-walled cylinder under internal pressure. The solution of this problem has first been published by Manning [10]; see also MacGregor, Coffin, and Fisher [l l]. The relatively simple calculation shows that here, too, the pressure reaches a maximum as the tube expands plastically, and then drops. The maximum pressure (often called "bursting pressure") can be calculated successfully from the stress-strain curve of the material. It is remarkable, however, that it cannot be derived from a single point of the curve and the corresponding tangent. In the thick-walled tube, the strain depends on the distance from the axis; at any moment during plastic deformation, states of stress and strain extending over a more or less wide region of the stress-strain curve are present. As a consequence, the maximum pressure cannot be

Yield SI reu a

(Thin-walled hollow .ph.r.

~ 1 ~

(Rod under unlaxiol Icn,ion)


Plal~ und., two equal mutually perpendicular ton,ion,


Logarithmic ~.


FIG. L 7 Graphical construction of maximum load or maximum pressure in various cases of tensile loading.

calculated without the knowledge of the entire stressstrain curve, or at least a substantial part of it. In other words, the maximum pressure withstood by the thick-walled tube cannot be derived from any single "working stress."

C. Ultfmute Stress and Working Stress. The ultimate tensile stress has served in the past generally, and still serves in many cases, as a basis for deriving design (working) stresses; for this purpose, it is divided by a so-called safety factor. Has this conventional procedure a realistic basis? From the preceding considerations, the answer can be easily recognized.

There are two types of failure by plastic deformation. In the first, the structure becomes unserviceable by suffering an inadmissible amount of distortion; in the second, it is destroyed by plastic disruption. In many practical cases, the second possibility either cannot occur (e.g., if the loading is flexural or compressive), or is of minor importance because the consequences of failure by excessive distortion are not significantly aggravated by subsequent disruption. In the design of pipes and pressure vessels, on the other hand, a moderate plastic deformation may be no more than a nuisance; the danger that must be excluded is disruption (bursting).

If the practically important type of failure is due to distortion, the design must be based on the stress at which plastic deformation reaches the maximum permissible value, i.e., on the "yield strength" or "proof stress." As is seen from the Considere construction of the maximum load and of the ultimate strength (Figs. 1.5 and 1.6), there is no general relationship between the ultimate strength and the proof stress (or, in the case of the annealed





FIG. 1.8 Uniform extension (strain outside region of neck) for different types of materials.

low-carbon steels, the lower yield point); the old practice of deriving the working stress from the ultimate strength by means of a fictitious safety factor has then no justification. A certain exception to this is the case in which different batches of the same type of material are compared (e.g., different deliveries of a low-carbon steel) ; the proof stress, or the lower yield point, may (but need not) be then approximately proportional to the ultimate strength.

If the only practically important type of failure is plastic disruption (bursting), the working stress should be derived, as a rule, from the load or pressure at which plastic instability leading to rupture sets in (the possibility of brittle or fatigue fracture should be disregarded in this section; it will be treated further below). The structure is then dimensioned so that the design load or design pressure is a certain fraction of the rupture load or bursting pressure. For a rod under uniaxial tension, the corresponding working stress is the ultimate tensile strength divided by an appropriate safety factor (which, in this case, is not a fictitious one).

It is to be kept in mind that the maximum load is given by the ultimate tensile stress only in the case of a structural part under uniaxial tension. For a tube, or a pressure vessel, the maximum pressure occurs at a (conventional or true) stress that may be very different from the ultimate stress, as will be discussed in more detail in Chapter 2. In exacting cases, therefore, the maximum load or maximum pressure cannot be derived from the ultimate tensile stress but must be obtained by accurate calculation based on the stress-strain curve, or from a model experiment. Often, however, this is not necessary. If the ultimate stresses for tension and for the plastic expansion of a tube differ by only 10% to 20%, and the safety factor may be anything between 3 and 6 according to tradition or code regulations, it may not be worth .carrying out an accurate design stress determination for a structural part of subordinate importance.

The Considers construction shows that the

ultimate stress is fundamentally unrelated not only to the behavior of the material at small, but also to that at large, strains. In particular, the knowledge of the ratio between the ultimate and the proof stress gives no indication of the fracture strain: fracture may occur immediately after the maximum load point, or at strains 10 or 50 times higher than the maximum load strain. The simple tensile test. in which only the maximum load but not the stressstrain curve is measured, however, may give a quantity that is extremely useful for judging the ductility of the material for certain uses. This quantity is the uniform extension, i.e., the strain at which thc load. maximum is reached and necking starts (OA in Fig. 1.5). Since practically no further extension takes place outside the neck after this has been initiated, the uniform extension can easily be measured on the fractured tensile specimen if this is long enough to contain parts sufficiently removed both from the neck and from the heads? of the specimen. A material with small uniform extension (a few per cent) is disrupted easily ill tension and is therefore unsuitable for drawing operations (wire or deep drawing). At the same time, however, it may show a high ductility (i.e., reduction of area at fracture), so that it may be eminently suitable to operations involving large plastic strains without tension. Thus, pure nickel, tin, or lead are very unsuitable for drawing, but extremely good for operations like bending or cold extrusion i austenitic chromium-nickel steels, on the other hand, have much less ductility but they are, owing to their large uniform extension, very suitable for drawing. Figure 1.8 shows how the shape of the stress-strain curve is related to the uniform strain. Materials with a fairly sudden yield and little strain hardening afterwards, like pure nickel, lead, or tin. have sharply bent stress-strain curves of the type A; the tangent construction gives for them a small uniform strain. On the other hand, materials that strain-harden slowly but steadily in the initial part of the stress-strain curve, like copper, brass, or 18/8 Cr-Ni steel (type B in Fig. 1.8), have a large uniform strain, independent of whether fracture occurs soon after necking 01' is preceded by a large reduction of area

1.4 Creep

A. The Andrade Analysis of the Creep Curve.

If a material can undergo progressive deformation

"I'he U.S.A. standard specimen is not long enough for this purpose; a useful specimen can be obtained, however, by increasing its gage length from 211 to 4".


a.t constant stress, it is said to show creep. The simplest type of deformation that corresponds to t~is defi~ition is viscosity: a material is called purely VISCOUS If the rate of straining, d'Y/dt is a function of the stresa.j'(r) and does not depend on the deformation already undergone

d'Y/dt = fer)


If. the functional relationship is simple proportionahty (Newton's law of viscosity),


T = 1)- (L33)


the material is. said to show Newtonian viscosity; the constant n IS the coefficient of .viscosity. Most of the common liquids are of the Newtonian type.

Th~ creep behavior of metals, particularly at not too high temperatures, is markedly different from pure viscosity. If a constant load is applied to a te~ile specimen (as is usual in technological creep testing) and the strain plotted as a function of time usually curves of type A in Fig. 1.9 are obtained:

S?lid sol~tions with a tendency to develop a sharp Yield ~omt (a-brass, Monel metal, Nickel silver) ~ay g~ve curves of the type C; other alloys show an induction period, as seen in curve D. However, ~Urve A can be regarded as the pure type observed if no structural changes occur during creep. It shows that the rapid, almost sudden, extension that follows the application of the load is followed by a period of deceleration; before fracture occurs there is a period of acceleration, and between the periods of deceleration and acceleration there is an interval of constant creep rate which may be quite long, or may be merely a point of inflexion.

In his analysis of creep, Andrade [12J found that the final acceleration is usually a trivial consequence of the increase of stress due to the decrease of crosssectional area in the course of the constant-lord tension test. If the experiment is carried out at constant tensile stress, the acceleration disappears in




rome FIG. 1.9 Types of creep curves for various materlals.



Timo Pur~~.r PICitic: Tr-ans.jent ViKOtJI

StrC-tn Croep Creep

FIG. 1.10 Andrade's analysis of the creop curve.

many cases and curves of type B are obtained. A period of final acceleration is frequently observed even at constant stress; however, it is always due to structural changes taking place during creep, and so curve B can be regarded as representing the pure and simple type of creep curve.

In his pioneering experiments, Andrade has observed that the slope of the straight parts towards which the creep curve tends asymptotically depends strongly on the temperature. At sufficiently low temperature, the asymptote becomes horizontal and the creep rate vanishes in the course of time. The period of deceleration, on the other hand, is always present, even in the neighborhood of absolute zero. From this, Andrade concluded that the creep curve (B in Fig. 1.9) represents the superposition of two essentially different creep processes, which follow the sudden straining after the application of the load. The first component is the decelerating one, the rate of which disappears with time; this is at present called transient creep. Superposed to this, at least if the temperature is not too low, is a constant-rate creep process, usually called viscous creep because its rate depends, roughly speaking, only on the applied stress and not on the preceding amount of strain. Figure 1.10 shows Andrade's analysis of the creep process: the observed creep strain is the sum of the purely plastic (plus elastic) strain which follows immediately the application of the stress, the transient creep strain, and the viscous creep strain.

B. Transient Creep. At low temperatures (below, say, one-third of the absolute melting point) viscous creep is insignificant and transient creep dominates; hence its alternative name "cold creep." At temperatures (in the hot-creep range), the transient component is often negligible beside the viscous one; hence the name "hot creep" for the latter.

In Andrade's original experiments, which were of relatively short duration, the transient creep curve could be represented by the expression

'Y = 'Yo + c4i (t = time). (1.34)

At lower temperatures, however, the logarithmic expression [13]

'Y = 'Yo + Clog t


It seems certain that no such simple expression can represent generally a process depending strongly on complicated structural features of the material, However, one of the above expressions, or perhaps another simple relationship, may well be found accurate enough for practical purposes in the case of an individual material.

The temperature dependence of viscous creep shows a similar picture. Like all thermal reactions, it is ultimately governed by the Boltzmann expression for the frequency of thermal activations; without further structural complications, this would lead approximately to an exponential dependence of the creep rate upon the reciprocal absolute temperature:

dyldl = Ce-AlkT (1.39)

where A is the "activation energy" for the creep process, k is Boltzmann's constant = 1.37 X 10-16 ergr C, and T is the absolute temperature. It can be shown {17] that,. in N ewtonian viscous flow, A is practically independent of the applied stress whereas C is proportional to the stress; on the other hand, in plastic deformation based on crystalline slip, the increase of the strain rate dy [dt with the increase of the applied stress is due mainly to the decrease of the activation energy A with increasing stress [18, 19}. In the case of crystalline plasticity. C may be regarded as a constant because its dependence upon the stress is small relative to that of the exponential. That this is true for the creep of metals can be seen in the following way: dy I dt is the strain per unit of time; its reciprocal is the time required for unit increase of the creep strain. Now creep fracture (see subsection G below) takes place after a strain of f% = f 1100 j the time t elapsing between the application of the load and fracture is related to the mean creep rate (ly I dt by

fll00l = dy/dt Introduction of this into eq. 1.39 gives

!/100t = Ce-AlkT



Vi....,,,, Ct~ep RaID


Fro. 1.11 Stress dependence of the viscous creep raw of lead wires at 17 C. After Andrade.

fits the curve better. All transitional types between the Andrade formula and the logarithmic formula. can be observed, as well as curves which represent a more-than-logarithmic decrease of the creep rate.

C. Viscous Creep. The viscous component is often represented by a reasonably straight curve, as shown schematically in Fig. 1.10, if the duration of the test is not very long. Otherwise structural changes (recrystallization, precipitation, etc.) are almost invariably present, and then the rate of viscous creep may increase, decrease, or irregularly fluctuate in the course of time. This is the basic factor that makes the extrapolation and practical use of creep tests difficult.

The experiments of Andrade [12] have shown that viscous creep in metals is far from being Newtonian (eq. 1.33); it is vanishingly small up to a certain stress region and then increases very rapidly with the stress. Figure 1.11 shows the curve given by Andrade for the viscous creep rate of lead wires at 17 C as a function of the applied stress. The character of the curve resembles that of the "Bingham material," an idealized material often referred to in rheology (Fig. 1.12, in which the stress is plotted as ordinate according to convention). The Bingham material is assumed to have a sharp yield point, and to show linear increase of the strain rate with the stress above the yield point. The behavior of metals at high temperatures differs from that of the Bingham material in that the increase of the viscous creep rate with the stress, as shown in Fig. 1.11, is much more rapid than a linear increase. Expressions suggested for its dependence are, e.g., the following ones:

dyldt = AT"

dyldt = A(ea~ - 1)

Norton [14] (1.36) Soderberg [151 (1.37)


d-r/dt = A sinh (ar)

Nadai [16}


where A, n, and a are constants.


Shear St, ...

Yie!d SI,on

flow Rotc

FlO. 1.12 Definition of the Bingham material.



or, if the logarithm of base 10 is taken,

log(100t/f) + log C = O>l-434A/kT (1.3gb) According to Larson and Miller {20], the dependence of the fracture time upon the temperature for various stresses is often satisfactorily represented by eq. 1.3gb with values of log C that vary, for different materials and experimental conditions, between 15 and 23 if t is counted in hours. Thus, log C is in fact almost constant. Its order of magnitude can be derived theoretically in a simple way. It is well known that, for processes of this kind, the activation energy is always around 1 electron volt (ev) at room temperature. If it were significantly higher (say, 2 ev), thermal activation would be so sluggish that the creep rate would become too small to be observable; if it were somewhat lower (say, 0.5 ev), the creep rate would be too high to be followed experimentally. At room temperature, kT is -lo ev, so that A/kT = 40. As a representative example, let it be assumed that the fracture strain f is 4% and that fracture occurs after 1000 hours. With these values, eq. 1.3gb gives

log C = 0.434 X 40 - log (25,000) = 13

For A = 1.5 ev, A/kT = 60, log C would be 22.6. The observed values of C, therefore, correspond to a range of activation energies between about 1 and 1.5 ev.

It should be remarked that, however narrow the range of the observed values of log C is, it would be dangerous to use eq. 1.3gb for extrapolating creep test results to times exceeding the duration of the test by a factor of 10 or more, because during the extrapolated time interval structural changes (e.g., precipitation, grain boundary oxidation) may occur and the permissible stress for a given service time may be reduced far below the extrapolated value (see Subsection a, "Creep Fracture").

D. Creep under Triaxial Stress. The problem of how to obtain the principal creep rates for general triaxial states of stress has been treated by Soderberg {15]. His solution is a rational extension of the treatment of three-dimensional cases in the theory of plasticity, and is in fair accord with the available experience. According to Soderberg, the basic viscous stress-creep rate relationship is a functional relationship between the effective shear-stress and the effective shear-strain rate, where the former is

and the latter

. 1 _I •• 2 •• 2 •• 2

'YeH = v'2 V (~l - E2) + (E2 - ~3) + (f3 - El)


f., (:2, and E3 being the principal strain rates; volume constancy demands that


Thus, the general viscous creep law would be


analogous to the three-dimensional stress-strain relationship suggested for purely plastic materials (ef. eq. 1.8). The relative magnitudes of the principal creep rates are assumed to be given by the LevyMises equations

fl = Clul - t(U2 + (3)] (:2 = C{U2 - t(U3 + uIll

(:3 = Clu3 - t(Ul + (2)1


The common factor C on the right-hand side is no longer indeterminate as in the case of ideal plasticity: it is determined by the condition that, if the principal strain rates are substituted on the right-hand side of eq. 1.43, the correct value of Teff must result. Details of practical calculations are found in Soderberg's paper.

E. The Mechanism of Creep. Although the details of the mechanism of transient creep are far from being clear, there is no doubt that it is a consequence of thermal vibrations enforcing slip when superposed to a sufficiently high applied stress. In the course of the creep process, the material hardens and thermal vibrations are then less and less frequently able to produce local slip; this is the cause of the gradual disappearance of transient creep. The fact that transient creep can be observed down to the lowest temperatures is due to the circumstance that the applied stress must always be high enough to cause at least a small amount of sudden plastic strain before transient creep can be observed. If it is sufficient to cause slip without any thermal help, very slight thermal fluctuation should be capable of producing local slip at the points where the applied stress is nearly high enough to induce slip without thermal help.

It has been found that viscous creep itself is a compound process. At least two different mechanisms can produce it, and often the two act simultaneously. The first type of viscous creep is called recovery creep. After the application of the load,



the rapid plastic deformation produces strain hardening which raises the yield stress to the level at which it equals the applied stress 1tnd thus can resist the load. If the temperature is high enough, however, thermal recovery or even recrystallization gradually reduce the strain hardening. In order to carry the applied load, therefore, the material must strain-harden further until the amount of strain hardening lost by recovery is replaced. This means that, in every unit of time, additional plastic strain arises, the amount of which is just sufficient to make up for the strain hardening removed by recovery.

The second important type of viscous creep is due to sliding between the grains of a polycrystalline metal when a stress acts at a sufficiently high temperature. At low temperatures, the grain boundary is a strong part of the structure: it resists the slip in the grains. At a high temperature, however, the boundary becomes soft and viscous and is an element of weakness. The tungsten filaments of incandescent lamps, which work at the highest temperature used in engineering, can be preserved from gradual deformation by their own weight only by being made of single crystals, without grain boundaries present.

F. Evaluation and Engineering Use of Creep Tests. Transient (cold) creep is of great practical importance, e.v., in prestressed reinforced concrete design. However, since its evaluation does not involve complex problems to the engineer, and since the problems in which it plays a role are somewhat specialized, it will not be treated here.

In many high-temperature applications of metals, the viscous creep strain during the lifetime of th« equipment is so much greater than the initial transient creep strain that the latter is frequently neglected (sometimes with no sufficient justification). In such cases, the usual practical rule is to assume that the long-time creep rate on which the design should be based is equal to the "minimum creep rate" observed in a constant-load tension creep test, i.e., to the creep rate in the straight part of curve A in Fig. 1.9. Although in the hands of the experienced creep practitioner this prescription usually work:'> fairly well, strictly speaking it is fundamentally wrong. When the minimum creep rate occurs, transient creep mayor may not have disappeared. If it has not, the minimum creep rate is not that of the viscous component alone, but the sum of the viscous and the residual transient creep rates. In extreme cases, solely the acceleration of transient creep, due to the decrease of the cross-sectional area, may give rise to curves of type A, Fig. 1.9, at low temperatures where no trace of viscous creep can be present. The

common rule, therefore, has to be supplemented by the condition that the constant-rate part of the creep curve must extend over a long time, sufficient for the disappearance of the transient component, in order that the minimum creep rate can be identified with that of the viscous creep.

Since structural parts must often have a service life of 10 or 20 years, whereas creep tests cannot be extended in engineering practice beyond about one year (often they must be obtained within a few weeks), the extrapolation of creep test results to the service life is the central problem of creep testing. Some of the extreme short-time testing methods suggested between the two wars failed because their authors were unaware of the compound nature of creep. Unless the test is extended long enough for the transient component to become relatively small, it cannot give even an approximate idea of the magnitude of the viscous component. The present conventional methods of creep testing usually avoid this pitfall; they can be subdivided into the following three classes:

1. Abridged tests. The creep strain is measured as a function of time.for a few stresses around the probable service stress, at the service temperature, and extrapolated to the service life.

2. Mechanically accelerated tests. The maximum permissible creep strain is enforced within the time available for the test by a suitably increased stress. From several such tests at different stresses, the stress is plotted as a function of the time after which the permissible strain is reached, and the curve extrapolated to the service life to give the permissible service stress.

3. Thermally accelerated tests. The maximum permissible creep strain is enforced within the time available for the test by a suitably mcreased temperature. From such tests at a few different stresses and temperatures, the stress is plotted as a function of the test temperature and of the time required for reaching the permissible strain, and extrapolated to the service life and service temperature.

The abridged test would give a correct extrapolation if structural changes taking place in the material during its service life could be discounted. Thermally and mechanically accelerated tests are in principle more likely to lead to errors because they take place under stress and temperature conditions different from those in service. However, occasionally certain structural changes that would occur during the service life but do not take place during the abridged test may be observed in the mechanically or thermally accelerated test. Then these



tests, although less correct in principle, may lead to a better extrapolation. No general extrapolation method can take into account the highly individual reactions of materials to stress and temperature, and the likelihood of grossly erroneous results can only be reduced by an intimate knowledge of the metallurgical, structural, and plastic properties of the material.

G. Creep Fracture. The grain boundaries of poly crystalline metals, being places of atomic disorder, behave like a two-dimensional glass. They have a softening range of temperature (roughly identical with the "equicohesive temperature") in which they change from being .a hard structural component to being the softest. At very high temperatures their effective viscosity is so low that, at low stresses, most of the deformation is localized in them: the grains slide almost as rigid units on their neighbors. This leads to the opening up of gaps between the grains, and finally to the type of fracture peculiar to high temperature creep: at first sight, it appears almost brittle.

The strain at which creep fracture occurs depends on the stress and the temperature. At low stress and high temperature the deformation within the grains is insignificant compared with the effect of sliding of the grains upon their neighbors, and thus the fracture strain is smalL However, the variation of the fracture strain in a given range of stress and temperature is always very small compared with the simultaneous variation of the creep rate. The latter may change in the ratio 10,000,000 to 1 while the fracture strain increases, for instance, from 2% or 3% to 10% or 15%. Consequently, the fracture time is usually inversely proportional to the mean creep rate, to a fair approximation.

The creep fracture test5 consists ill applying to the specimen a constant tensile load and recording the time elapsing to fracture. This test is simpler and easier to perform than the standard creep test because strain measurements are omitted. It is required for design whenever the material has such poor ductility under creep conditions that fracture may occur before the maximum permissible creep strain is reached. Since creep strains exceeding 1% are not often permitted (pressure vessels and pipes are an exception), and fracture occurring after less than 1% strain is infrequent, the creep fracture test is usually unnecessary. It is nevertheless widely used because it can be interpreted as a crude creep test. As mentioned above, the fracture strain varies within relatively narrow limits, so that the

~rn the creep teeter's vernacular, "stress rupture" test.

creep fracture test represents a creep test in which the time required for a certain strain (the fracture strain) is measured for various stresses and temperatures. The great shortcoming of the test is not so much the variation of the fracture strain as the fact that it is always performed at high stress levels in order to obtain fracture within 1000 or, at most, 10,000 hours. It has been shown by many experimenters, particularly by Grant and his collaborators [21], that the creep rate may change abruptly even after 10,000 hours owing to some structural change (e.g., coarsening of a precipitate, or oxidation). For this reason, extrapolation from highstress short-time tests to the long-time service behavior is impossible, unless it is known (from a thorough investigation of the material extending over years) that no structural changes may be expected in the time interval between the duration of the routine creep test and the service life.

1.5 Types of Fracture; Molecular Cohesion; the Griffith Theory

Fracture is the disintegration of a body into fragments under mechanical stresses. If a certain type of fracture occurs in a given material when a stress component reaches a critical value, this is called the strength or fracture stress. Many types of fracture, however, do not take place at a characteristic value of a stress component.

Until about 20 years ago it was not realized that there are many fundamentally different types of fracture obeying quite different laws. They can be classified into two main groups: brittle fractures and ductile fractures. The former occur with little or no plastic (or other non-elastic) deformation; the mechanism of the latter essentially involves plastic deformation. The mechanism of brittle fracture was elucidated long before that of ductile fractures. mainly by the work of A. A. Griffith in 1920 [221- Griffith's effort was directed to the explanation of the extraordinary discrepancy between the very high values of strength inferred from the magnitude of the intermolecular and interatomic forces, and the observed values of the tensile strength, which are usually hundreds or thousands of times lower.

The way in which the tensile cohesion of a material is determined by the attractive and repulsive forces between its molecules is illustrated in Fig. 1.13. Suppose that a crystal contains atomic planes with the spacing b perpendicular to the direction of tension. As the tension is raised, the spacing b increases. The net interatomic force acting between two parts of the crystal across the gap between two atomic

raise the applied stress to high local values. It was Griffith, however, who calculated the critical value of the applied tensile stress (J at which a crack of atomic sharpness and of length c, starts to propagate, He used the following approach. When the crack extends, the surface area of its walls increases and this requires energy for overcoming the attractive forces between the atoms separated by the crack. If the grips between which the specimen is pulled do not move during the crack propagation process, the only source from which the necessary surface energy can be obtained is the elastic energy released as the crack extends. Let dS be the surface energy needed for enlarging the crack by an infinitesimal amount, and dW the elastic energy released simultaneously. The crack can propagate only if dW is at least as large as dS; thus,

AttrooLve force






RDpul'ive Force

FIG. 1.13 The dependence of the intermolecular forces upon the molecular spacing.

planes vanishes if no tension is applied; in this case, the attractive and repulsive forces cancel. If a tension is applied and the atomic spacing increases, the repulsive forces diminish more rapidly than the attractive ones; the excess of the attractive forces over the repulsive ones balances the applied tension. As the atomic spacing in the direction of tension increases, the repulsive forces become insignificant, and the tensile force transmitted through the crystal lattice must then start to diminish with increasing strain owing to the decrease of the attractive forces with increasing separation of the atoms. Consequently, the net atomic force transmitted through a cross section must have a maximum, equal to the highest external force the material can withstand i.e., its strength. From the general knowledge of the atomic forces it can be estimated that the maximum must occur when the spacing of the atomic planes has increased by a large fraction of its initial value; for an order-of-magnitude estimate, it may be assumed to occur when the atomic spacing has increased by some 25% or 50%. If Hooke's law were applicable for such large strains, the tensile strain would be between 0.25 and 0.5 and the corresponding tensile stress, i.e., the molecular strength of the material,

(Jm = 0.25E to 0.5E


where E is Young's modulus. Instead of approaching the order of magnitude indicated by eq. 1.45, the measured tensile strengths are extremely low. The strength of ordinary sheet glass is about 1/1000 of its Young's modulus; that of rock salt crystals, less than 1/10,000. It was known to physicists before Griffith that the most likely cause of the discrepancy was the presence of invisibly small cracks or other flaws which produce stress concentrations and thus

dW = dS


is the condition for the crack being able to propagate under the tensile stress. It will be seen that the stress needed for propagating a crack decreases as the length of the crack increases; once condition 1.46 is satisfied, therefore, the crack will extend rapidly, and fracture will occur.

Griffith carried out this idea in the simple case of a plate containing an internal crack of length 2c (Fig. 1.14). It can be shown that the effect of such a crack upon the fracture stress of the plate is equal to that of an external crack of length (depth) c in one of the side edges of the plate. A sharp and flat internal crack of length 2c can be regarded as an elliptical hole of major axis 2c and an extremely short minor axis; the stress distribution around it when the plate is put under a tensile stress (J was calculated by Inglis in 1913 [23]. From this the excess energy in the plate, due to the presence of the



FIG. 1.14 Plate with a flat elliptical hole (=crack).


crack, is obtained as

TV = 1{(lc2jE


per unit thickness of the plate, where E is Young"s modulus; if c increases by de, the released elastic energy is

On the other hand, the increase of the length of the crack is 2dc, and the increase of its wall surface area is 4dc per unit thickness of the plate; consequently, if a is the work required for creating a new surface of unit area, the increase of the total surface energy is

dS = 4adc


Equating dW and dS gives

a = f2;;E \/-;;


This is the famous Griffith equation for the tensile strength of a brittle material containing an internal crack of length 2c, or a surface crack of depth c. In the calculation, it has been assumed that the problem is two-dimensional, and that the plate is very large in both directions, but at the same time thin compared with the length of the crack; if it is thick, the factor (1 - y2) has to be applied to the denominator under the square root, v being Poisson's ratio.

For glasses of the ordinary types, the crack length c necessary to explain the observed tensile strength is of the order of 1 micron. In glasses, the dangerous cracks are almost always at the surface; tensile stresses confined to the interior are relatively harmless. This is the explanation of the high strength of "tempered glass," obtained by quenching glass from the softening temperature by an air blast. By the time the interior has become rigid, the surface has cooled down considerably; when subsequently the rigid interior cools, it puts the surface layers under a tangential compressive stress. Any tensile stress produced by external forces is diminished at the surface by the residual compression. In the interior, the residual stress is tensile, but this is of no consequence because there are no sharp cracks present from which fracture may start. Thus, the strength of the glass is strongly increased,

The Griffith theory explains very satisfactorily the strength properties of completely brittle materials such as glass i for detailed treatment, reference should be made to the literature [24].

An interesting feature of the theory is the answer it gives to the question of strength under triaxial


stresses. The discussion of the complete answer is beyond the scope of this chapter (241; the result is that, so long as the highest compressive principal stress is less than three times the highest tensile principal stress, fracture should occur when the greatest tensile principal stress reaches the value of the tensile strength deduced for uniaxial tension (eq, 1,48) ; the algebraically smaller principal stresses have no influence. According to the theory, the compressive strength should be eight times the tensile strength if the material is isotropic and contains cracks randomly distributed in all directions.

Thus, the theory confirms partially a well-known statement found in textbooks on the strength of materials concerning the condition of brittle failure: in the essentially tensile region of principal stresses, failure does obey the maximum tensile stress criterion. However, the maximum tensile stress condition cannot be valid for any state of stress. If it were, the compressive strength of brittle materials would be infinitely high. This shortcoming of the textbook rule has been corrected by the Griffith theory, in the way just mentioned.

One of the most important results of the work of Griffith is the realization that the strength of a brittle material is determined by the flaws it contains. This is strikingly illustrated by glass, the strength of which can be made a hundred times higher than normal, if by a special design (fibre glass) the worst cracks are made ineffective.

1.6 Ductile Fractures

The Griffith theory and the fracture condition (eq. 1,48) are applicable only to fracture of the cleavage type ("brittle fracture"). In addition to this, there is a large group of fractures in which separation into fragments occurs as a consequence of certain plastic deformation processes; these are the "ductile" fractures. The simplest ductile fractures are straightforward geometrical consequences of plastic deformation; a wire of gold, e.g., breaks in tension by the formation of a neck which becomes thinner and thinner until it is drawn out to two needle points in contact. Similarly, single crystals of zinc or cadmium may break, after slow extension at a high temperature, when one part of the crystal slips off the other along a slip plane in which the deformation has become concentrated.

The nature of the fracture process is less obvious in the common fibrous fracture of ductile metals, which produces the bottom of the cup in the cupand-cone fracture. However, it seems to be fundamentally the same type of geometrical attenuation

certain precipitation hardened alloys, can be sheared off during tightening after a small amount of plastic twist. Another instance is that of extremely creepresistant alloys which may fail by creep fracture at high temperatures after a relatively small creep strain.

Constrained yi.ld .tr... <I Y



Ten,il .. SlrH' f1

0,,<111. (libra",) !ra<lure

Pla,tie Tonsilo Strotn E

FlO. 1.115 Scheme of the classical triaxial tension theory of notch brittleness, after Mesnager [25], Ludwik [26}, and Orowan [27}.

as in the preceding examples, repeated many times on a microscopic scale in the surface of fracture. Shear fracture, which forms the sides of the cup and the cone, is a somewhat different phenomenon. The plastic deformation leads here to the propagation of a crack at the tip of which there is a high concentration of strain, destroying locally the cohesion of the material.

A ductile fracture cannot obey the Griffith condition (eq. 1.48). This can be realized in the following simple way: The plastic deformation mechanism which leads to ductile fracture is not essentially dependent on the elastic moduli of the material; it could take place even if Young's modulus were infinitely high. On the other hand, cleavage fracture of the Griffith type would be impossible in a perfectly rigid material; an infinitely high value of E in eq: 1.48 would give an infinitely high tensile strength.

One of the conditions governing ductile fracture can be easily recognized: it coincides with the condition of the particular type of plastic deformation which is responsible for the fracture. Thus, in the tensile fracture by neck attenuation the only fracture condition is that the tensile load must reach the value of the yield stress in the neck, multiplied by the cross-sectional area of the neck and by the plastic constraint factor. In shear fracture, too, this is a necessary condition for the propagation of the shear crack. Another condition, however, must also be satisfied: the shear strain at the tip of the crack must reach the critical value at which the cohesion disappears.

Ductile fractures usually occur after the structure has become unserviceable by excessive plastic deformation. However, if the material has a low ductility, shear fracture or other types of ductile fracture may occur after very little deformation. A threaded bolt of a low-ductility material, such as

1.7 The Brittle Fracture of Steel ("Notch Brittleness" )

Low-carbon and medium-carbon steels behave in a maimer that is not a mere intermediate case between glassy brittleness and high ductility. A common structural steel can be very ductile in the ordinary tensile test, with no sign of a potential brittleness, but it can break with little or no visible> plastic deformation if it contains a crack or a notch.

The classical triaxial-tension theory of notch brittleness was put forward by Mesnager [251 and, independently, by Ludwik {261. In a form modified according to the present state of knowledge [271, its principle is illustrated by Fig. 1.15. The abscissa in this figure is the tensile strain and the ordinate the tensile stress j Y represents the ordinary tensile yield stress-strain curve. The theory assumes that a material suffers brittle (cleavage) fracture when the tensile stress reaches a critical value B ("brittle strength") which, in its dependence upon the plastic strain, is given schematically by the curve B. In the ordinary tensile test, ductile fracture occurs at the point D on the curve Y, before the tensile stress reaches the value of the brittle strength. However, if the specimen contains a notch or a crack, plastic constraint raises the value of the tensile stress reached during plastic yielding to q Y, where q, the "constraint factor," is greater than 1. The curve qY may intersect the curve of the brittle strength B

1 Compressive force

fridionat conlfraint
upon spedmen
-~ _-
--I I I_Ton
- -
m doncy 10 spread

FIG. 1.16 The ongm of plastic constraint in a notched tensile specimen illustrated by the frictional constraint acting upon a tlat compression specimen.



before the plastic strain is high enough to produce ductile fracture, and so brittle fracture may occur

at F. . ...

The way in which plastic constraint arises is illustrated in Fig. 1.16. Suppose that I is a coin compressed plastically between two hard cylinders, I I and II I. The necessary mean compressive stress is higher than the yield stress Y in uniaxial compression: it has to overcome, not only the resistance Y of the material to plastic deformation, but also the frictional resistance of the compression blocks (indicated by the arrows) to the lateral spread of the coin. The radial frictional forces, together with the axial pressure, create a state of triaxial compression (a hydrostatic pressure superposed to an axial pressure). The mean axial stress required for plastic compression is then not Y but q Y > Y; of this, Y is required for the plastic deformation itself, and (q - 1) Y for overcoming the friction.

Figure 1.16 can also be regarded as representing a circumferentially notched cylindrical specimen, I being the notch core and II, III the full sections of the specimen. If the specimen is plastically extended, the conditions are similar to the case of the compressed coin, with the shear cohesion between the core and the adjacent parts of the specimen replacing the friction. As before, the axial stress required for producing plastic deformation in the core must be higher than the yield stress Y.

Plastic constraint is fundamentally different from elastic stress concentration. It cannot arise without some preceding plastic deformation; moreover, its magnitude depends on the depth and sharpness of the notch in a very different way. In pure elasticity, the stress concentration at the tip of a notch becomes infinitely high as the radius of curvature of the tip converges towards zero. In contrast to this, the plastic constraint factor of a circumferential notch such as is illustrated in Fig. 1.16 increases only to a value of the order of 3, instead of rising towards infinity, as the tip radius is reduced to zero [271. This is the reason why so many ductile metals cannot be made to fracture in a brittle manner by the application of a sharp crack; if, for any value of

I ! I











I I 1


Work of


FIG. 1.17 Extreme types of transition curves,

Yilll'd s.freu it'l fen:l-ton Y :


Temperature of comp[oie embriHt-ement


Trun1,ilion tern_percturo betwoon nOld> b,iHlonqu and f,,11 dudilily

FIG. 1.18 Davidenkov-Wittman Theory of the transition between brittle and ductile fracture, as modified by the author.

the plastic strain, the brittle strength is more than about 3 times higher than the yield stress, plastic constraint alone cannot raise the tensile stress to the fracture level.

An important feature of notch brittleness is the existence of a transition temperature between notchbrittle, and purely ductile, behavior. Figure 1.17 shows the dependence of the work of fracture, as measured with a Charpy or Izod pendulum hammer, on the temperature in low-carbon steels. Above a certain temperature region it has a high value, and the fracture of the notched specimen is entirely of the fibrous type. At low temperatures, the fracture work is extremely small, and the fracture is entirely of the cleavage ("crystalline") type. Between these two temperature regions, there is a transition zone in which the fracture work drops rapidly with decreasing temperature; at the same time, the area of cleavage in the surface of fracture increases towards 100 per cent. With some materials, the transition zone is so narrow that one call speak of a "transition temperature"; in other cases, e.g., of many 10\\"alloy ferritic steels, it is spread over hundreds of degrees F.

Figure 1.18 shows schematically how the classical theory interpreted the transition phenomenon [28]. Y is the curve giving the temperature dependence of the yield stress; the curve q Y (== 2 or 3 times Y), therefore, represents the highest tensile stress that an atomically sharp crack can produce during plastic yielding. Experiments and theory show that the temperature dependence of the brittle strength B must be less strong than that of Y or q Y; this is schematically indicated in the figure. It is seen that the tensile fracture is entirely brittle below the temperature Tit, even in the absence of any notch. If a notch or crack of maximum sharpness is present, brittle fracture is possible below the temperature '1'1, but not above it.



Recent investigations [291 have shown that the fundamental cause of brittle fracture in normally ductile steels is not plastic~onstraint but the abnormally high velocity-dependence of the yield stress of ferritic steels. The experiment from which this can be recognized is as follows: The edge of a lowcarbon steel plate is provided with a brittle crack by forcing a chisel into a notch at a low temperature. If the plate is subjected to tension at room temperature, it is found that the brittle crack is unable to propagate as a brittle crack. Instead, large plastic deformations arise around its tip, accompanied by some fibrous crack propagation; after this, the fracture suddenly reverts from the ductile to the cleavage type and the newly created brittle crack runs across the plate. This shows that, at low rates of straining, plastic deformation in microscopically small regions around the tip of a brittle crack cannot create the degree of triaxiality of tension necessary for brittle fracture; quite large deformations, such as can be seen with the naked eye and felt with the fingers, are required. However, once cleavage cracking starts again, it runs at high speed and without large plastic deformations.

The simplest interpretation of these t.bservations is that in the brittle fracture of steel the stress is raised to the level of the brittle strength by the high rate of plastic deformation around the tip of a running crack rather than by plastic constraint. Without a sufficiently high velocity of the crack, the production of the plastic constraint necessary for cleavage fracture requires such extensive plastic deformations that the fracture, though of the cleavage type, is far from being brittle, i.e., of low energy consumption. Triaxiality of tension, then, is probably no more than one of several ways of initialing cleavage fracture; the cleavage fracture is then transformed into brittle cleavage fracture by the velocity effect upon the yield stress as the crack gathers speed.

The rather exceptional combination of ductility with potential brittleness in steel may be understood now as being a consequence of another exceptional property of low-carbon steels, the unusually strong dependence of their yield stress upon the rate of straining [30, 31]. The yield stress of copper or aluminum increases only some 10 to 20 per cent between "static" and ballistic testing speeds; for lowcarbon steels, however, increases of 100 and 200 per cent have been recorded.

Why such large deformations are needed for starting cleavage fracture at the tip of a crack under slow tension is a question not yet answered. It has been suggested that plastic constraint alone cannot raise

the tensile stress to the fracture level in typical cases of notch brittleness under static loading; it must be aided by strain hardening, and this requires considerable plastic deformation. However, brittle fracture can start in a welded structure with very little plastic deformation. The plastic strains produced by thermal expansion and contraction during welding and the corresponding strain hardening can hardly be made responsible for this, because the thermal strains seem too small to take the material beyond the region of yield into that of strain hardening.

The final question is this: What is the condition under which the cleavage crack arising from the intermediate ductile crack in static loading becomes a rapidly running crack, in which the velocity-increase of the yield stress can replace the heavy plastic deformation necessary around a slowly extending crack to produce cleavage? A crack can run rapidly under static load only if the work required for its propagation is obtained from the elastic energy stored in the specimen. It was seen in Section 1.5 that the Griffith equation (1.48), by virtue of its derivation, is the condition for the crack propagation work to be supplied from the released elastic energy; however, it cannot be applied directly to brittle fracture in steel, It has been found [27] that cleavage fracture in low-carbon steel around room temperature is not quite brittle; there is a thin cold-worked layer at the surface of fracture, representing an energy of cold work of about 2 X 106 ergs/ern", This is around 1000 times greater than the surface energy of steel; the work of crack propagation per unit area of the crack walls, therefore, is given by the plastic surface work p, beside which the surface energy is negligible. If the plastic surface work per unit area of the cleavage fracture can be treated on the same footing as the surface energy, the condition for the work of propagation of a brittle crack in steel to be supplied by the simultaneously released elastic energy is [24, 32]


instead of the Griffith equation (1.48). In eq. 1.49 the factor V2/rr has been omitted to indicate that the equation does not pretend to be accurate enough for this factor to matter.

Brittle cleavage fracture in steel, therefore, requires the fulfilment of two conditions:

1. Thc temperature must be below the transition range;

2. The applied stress must satisfy the crack propagation equation (1.49).


i i~


. ~. !


The first condition is satisfied by most structural steels, at least at low winter temperatures. The designer, therefore, can avoid the possibility of brittle fracture only by taking care that the crack propagation condition should not be satisfied. The simplest, although practically not always easy or even feasible, way to do this is to avoid the presence of cracks exceeding in length a certain limit. The smaller the crack length c, the higher is the (mean) tensile stress (J" in the plate at which crack propagation is possible. Since the stress cannot rise above the yield stress Y, the length of the smallest crack that can start brittle fracture is obtained from eq. 1.49 as

Co = Ep/y2


Oracks below this length are harmless (unless, of course, they can grow by a non-brittle mechanism which does not require the fulfilment of the crack propagation condition, eq. 1.49). If, therefore, the possibility of cracks exceeding in length the critical value Co can be eliminated by careful fabrication or inspection, brittle fracture cannot occur even below the transition temperature. With E=3X107 psi= 2 X 1012 dyne/cm'', p = 2 X 106 erg/cm'', and Y = 6 X 104 psi = 4.1 X 109 dyne/cm'' for the strain-hardened steel, the critical minimum crack length Co is obtained from eq. 1.50 as

Co = 0.25 em = 0.1 in.

To avoid any crack exceeding this length is difficult and costly, but not impossible, as is shown by the occasional use of non-aging low-carbon steels for pressure vessels at liquid-air temperatures.

Alternatively, the designer may attempt to keep the stress level so low that eq. 1.49 is not satisfied even though the longest cracks unavoidably present might exceed the critical length Co. If, e.g., the presence of cracks of 0.4 in. length cannot be excluded, the tensile stress must be kept below 30,000 psi; cracks of 1 in. length would set an upper safe limit of about 19,000 psi to the stress, and so on.

Naturally, the propagation condition (1.49) may not be the only condition that must be satisfied before brittle fracture can occur. If eq. 1.49 is correct, brittle fracture cannot occur below the stresses derived from it; however, some other, more exacting condition may in some cases set a higher limit, so that fracture in fact may not occur at stresses M low as correspond to eq. 1.49. A simple example of this is the case of a steel plate containing a brittle crack and subjected to slowly applied tensile stress, as in the experiments described above. Although the stress given by eq. 1.49 may be quite low, the crack


cannot start to propagate before considerable plastic deformation takes place around its tip, and the stress required for this may be quite dose to the yield stress of the plate. In other words, in this case an initiation condition must be satisfied besides the propagation condition, and the former is more exacting.

In recent experiments [33] in which the difficulty of crack initiation was overcome by a wedge hammered into the crack by the impact of a bullet, fracture could not be provoked below a fairly clearly recognizable stress level which depended on the conditions of the experiment (notch angle, plate size, etc.). Since the mechanics of the crack initiation by wedge impact is very complex, it is difficult to recognize the significance of this result. The observed stress threshold is probably due to the necessity to satisfy some crack initiation condition; whether this condition is of more general significance, or a particular characteristic of the wedge impact experiment, is an open question.

The practical importance of brittle fractures in steel structures has rapidly increased in recent times, owing to the widespread use of welding and of hightensile steels. Welding results in high residual tensile stresses adjacent to the seam, and it may also cause structural damage (e.g., grain boundary oxidation). This may lead to the formation of cracks which can run across the weld seam and wreck the entire structure in a fraction of a second. The high yield stress of many modern steels, obtained by alloying additions, cold work, or heat treatment, may lure the designer to the use of working stresses under which spontaneous crack propagation becomes possible (cf. eq. 1.49). Clearly, an uncritical raising of the design stresses on the ground of the increased yield stress is entirely unjustified, unless the transition range is also considerably lowered. If the latter condition is not satisfied, higher yield stress may merely mean that the working stress is no longer determined by the yield stress but by the necessity of avoiding brittle fracture.

Good ductility (high fracture strain, reduction of area) in the ordinary tensile test ending with ductile fracture does not mean increased immunity to brittle fracture in the case of ferritic steels. The possibility of brittle fracture can be assessed only by determining the transition curve of the steel and estimating the size of the most dangerous crack that may be present. For low-carbon steels, it appears that a fracture work of 15 ft-lb in the V-notch Charpy test at the lowest service temperature gives a high degree of protection against brittle fracture even if

Table 1.1



Static .~ Cyclic
Mechanical Creep fracture Ordinary cyclic
Chemical Delayed fracture Corrosion fatigue
of glass; stress
corrosion cracks cannot be avoided. This figure, however, does not apply to harder steels. If a heat-treated high-tensile steel of 160,000 psi yield stress gives a V-notch Charpy value of 15 ft-Ib; the deformation of the notch-bend specimen is only about one-quarter of that of a plain low-carbon steel with the same Charpy value but a yield stress of only 40,000 psi. The 15 ft-lb high-tensile steel, therefore, has a tendency to brittle fracture comparable to that of a hot-rolled low-carbon steel with a Charpy value of 4 It-lb.

If a steel is to be used in the brittle-fracture danger zone of temperature and stress, careful design and workmanship are of the greatest importance. Sharp stress concentrations, such as abrupt cross-sectional changes, sharp thread profiles, or blind root welds, must be avoided, and the formation of cracks during fabrication and heat treatment prevented. On important equipment, or where failure may endanger lives, particular attention must be given to careful inspection and to the removal of internal stresses.

1.8 Fatigue

A. General Features. The term "fatigue" is used if a specimen breaks under a load which it has previously withstood for a length of time, or during a load cycle which it has previously withstood a number of times. There is a remarkably sharp distinction between those cases of fatigue in which only the total duration of loading matters while it is of secondary importance whether the load is steady or interrupted, and those where only the number of load cycles matters and the duration of the cycles is of a subordinate importance. The first type of fatigue is called static, the second cyclic.

Purely elastic deformation cannot cause fatigue; all it does is to strain atomic bonds, and these cannot wear out. Fatigue can be the consequence either of non-elastic deformations (i.e., of lattice injuries or intergranular displacements it produces), or of chemical or physicochemical processes accelerated by the applied load. Thus, fatigue phe-

nomena can be classified according to their physical cause as mechanical or chemical.

In this way, a twofold subdivision of fatigue phenomena is obtained, as illustrated in Table 1.1.

An example of static mechanical fatigue is creep fracture, already discussed in Section 1.4. A littleknown case of static fatigue is that observed in the brittle fracture of steels which may occur suddenly after prolonged steady loading. The time delay between the application of the stress and the occurrence of fracture must be due to a slowly progressing deformation process; the rate of this process may be determined by the rate at which carbon atoms diffuse in the iron lattice. Thus, the delayed brittle fracture of steel may possibly represent a case of physicochemical static fatigue.

The cause of the static fatigue of glass is undoubtedly physicochemical [34J. It is known that air (probably mainly its moisture content) reduces the surface energy of mica by a factor of 10 or 12. It must also reduce the surface energy of glass; consequently, the Griffith crack propagation condition (eq. 1.48) rimy be fulfilled for a given stress II and crack length c in the presence of air (i.e., when a has a lowered value), but not in vacuum. In this case, the crack can only propagate at the rate at which air or moisture can diffuse to its tip. After a period of slow propagation with the help of absorbed moisture, the crack length may increase to the value at which the applied stress can propagate the crack even without the reduction of the surface energy by moisture i fracture then occurs suddenly. The physicochemical nature of the delayed fracture in glasses is verified by the observation that static fatigue is absent in vacuum.

The best known type of static fatigue due to chemical action is stress corrosion, of which the "season cracking" of cold-worked brass and the "caustic embrittlernent" of steel are familiar examples. In some cases, its cause is the precipitation of a phase in the grain boundary which deprives the adjacent parts of the grains of an element that increases the resistance to chemical attack [35]. In the case of some austenitic Cr-Ni steels, for instance. chromium carbide may segregate in the boundarv during heating in a certain temperature region, and the boundary regions of the grains are then depleted in chromium. Crack propagation by solution of the more easily attacked (more anodic) boundary layers cannot progress, however, without the presence of a tensile stress which opens up the crack and provides space for the corrosion products. Under the applied stress plastic deformation occurs at the tip of the



crack; this may disrupt protective layers, and the increased free energy of the deformed region makes it more susceptible to attack (mote anodic). Whether these two effects represent important causes of stress corrosion is not certain.

Stress-corrosion cracking can progress not only along the grain boundaries but also across the grains; brass single crystals crack under tension in the presence of ammonia much like polycrystalline brass [36, 37]. This suggests the possibility of a stress-corrosion mechanism similar to that of the static fatigue of glass [38). The effective surface energy of the crack walls which enters into the Griffith equation (1,48) can be lowered not only by adsorption but even more radically by chemical combination between the corrosive agent and the metal atoms; .consequently, a crack may propagate in the presence of a corrosive medium by cleavage under a relatively low tensile stress while, in the absence of corrosion, the propagating stress demanded by eq. 1.48 may be higher than the yield stress so that crack propagation by cleavage is impossible. Obviously, the effect of the adsorptive or corrosive is to cut the cohesive bonds between the atoms of the crack walls at an early stage of the cleavage process, by converting -them into chemical or van der Waals bonds between the atoms of the crack walls and the atoms, molecules, or ions of the adsorptive or corrosive agent.

In accordance with its chemical origin, the susceptibility of metals to stress corrosion is extremely specific. Thus, for instance, the caustic embrittlement of Cr-Ni-Mo low-alloy steels apparently can be avoided by omitting anyone of the three alloying elements.

Corrosion fatigue differs from stress corrosion in that it occurs only if the stress varies cyclically. It is fairly insensitive to the duration of the cycles (i.e., to the total duration of stress application). Corrosion fatigue starts with the appearance of surface pits which then spread and join up to form surface grooves not unlike the cracks on the bark of a birch tree. These pits and blunt cracks apparently develop because they give rise to stress concentrations



FIG. 1.19 Typical "tress cycle,

tog 5

(SI'M' omplilude)

o.g., Fomlic molerl,,!.

Ge~rQI type, light olloy>

log N (Numbor of qdM 10 f,odu,e)

FIG. 1.:1'0 Representative fatigue fracture stress curves for metals,

where the increased elastic energy or plastic deformation locally raises the free energy; at these spots the material is electrolytically more soluble in the corrosive solution (more anodic) than its surroundings. Another possible reason for the local attack is that the plastic deformation at the pits or cracks may prevent the formation of protective (passive) layers.

Those features of corrosion fatigue which are of quantitative interest to the designer will be mentioned briefly after the treatment of ordinary mechanical cyclic fatigue. The chemical mechanisms of corrosion fatigue, like those of stress corrosion, are too specific to allow any general treatment. In what follows. therefore, the main emphasis will be laid on common mechanical fatigue, which is the most important fatigue phenomenon from the point of view of the engineer

The existence of mechanical fatigue of materials under cyclic stressing was established by Rankine in 1843, and the basic laws of the phenomenon were investigated experimentally by L. Wohler between 1852 and 1869. To describe it in clear terms, a simple terminology should first be introduced.

Generally, a cyclic stress is the superposition of a steady stress s and an alternating stress of amplitude 8 and range 28 (Fig. 1.19). The stress amplitude that causes fracture after N cycles will be called the fatigue strength for N cycles; if it tends towards a finite value for infinitely increasing N, this will be called the limiting fatigue strength or, briefly, the fatigue limit. In the literature, the fatigue strength is usually called fatigue endurance; however, there is no reason why the correct technical term "strength" for a fracture stress should not be used in this case also. The fatigue strength depends, in general, on the steady stress superposed upon the purely alternating stress.

If the logarithm of S (the stress amplitude) is plotted as a function of the logarithm of N (the number of cycles to fracture), curves of the type shown in Fig. 1.20 are obtained. Plain carbon steels



FIG. 1.21 Effect of high-amplitude fatigue on silver chloride sheet.

usually have a clearly defined fatigue limit; recent experiments indicate that this may be a consequence of the phenomenon of strain aging shown by such steels. Nonferrous materials may also give curves showing, 'more or less clearly, two straight parts connected by a curved transition region; however, the second straight part is usually not quite horizontal but slightly descending. In such cases, there is no clear fatigue limit within the experimentally accessible values of N. The fatigue strength on which the design must be based is then that for the number of cycles which the structure must withstand during its intended life.

n. The Mechanism of Fatigue. A revealing observation about the mechanism of fatigue is that the fatigue crack, in general, seems to run along slip planes, not cleavage planes [39, 40]. This can be recognized without ambiguity in iron where slip planes and cleavage planes never coincide.

It seems that alternating slip can lead to the development of high tensile stresses in the slip planes due to a progressive warping of the slip "packets" in the course of cyclic straining. Figure 1.21 shows the waviness developed during a high-amplitude fatigue test in some of the large grains in a polycrystalline silver chloride sheet. The development of tensile stresses during the warping of slip planes may be understood by means of the dislocation theory of plastic deformation [41, 42]; if the stress is high enough, it can cause local fracture.

However, the local tensile stresses which arise in the course of prolonged alternating slip do not provide a sufficient explanation of mechanical fatigue. If the material strain-hardens with plastic deformation, the first stress cycle ought to harden it so that no further slip can occur unless the stress amplitude of the following cycles is progressively increased; how, then, can alternating slip continue in tests-at constant stress amplitude? On the other hand, observations show that alternating slip continues, with gradually decreasing amplitude, even in safe ranges of stress; how can it then he explained that, in such cases, even hundreds of millions of nonelastic strain cycles are insufficient for accumulating the amount of internal stress and lattice damage necessary for fracture? Thus, the basic questions of fatigue are (1) How is progressive slip and structural damage possible under cycles of constant stress amplitude; and (2) How are safe ranges of stress possible?

The answer to these questions is given by the general theory of fatigue {42, 24J, which is concerned with those typical features of the fatigue phenomenon which are largely independent of the individual molecular mechanism of the fatigue damage. A quantitative description of the theory would require too much space to be presented in this chapter; however, a qualitative outline of the main points can be given briefly.

The salient point is that in cyclic stressing progressive plastic deformation soon becomes confined to relatively small regions (e.g., at the tips of small cracks, or in particularly unfavorably situated grains) which are then surrounded by more or less purely elastic material. Now it is easily seen that. if a largely elastic specimen is subjected to cycles of constant stress amplitude, a small plastic region embedded in it will experience stress cycles of increasing and strain cycles of decreasing amplitude. This is a consequence of progressive strain hardening: as the yield stress of the plastic region rises, its elastic surroundings have to exert upon it increasing stress amplitudes to enforce further plastic deformation. By Hooke's law the clastic surroundings must then suffer increasing strain amplitudes, and so the strain amplitude in the plastic region decreases because the sum of the two strain amplitudes must remain constant for a given amplitude of stress applied to the specimens as a whole.

The gradual decrease of the plastic strain amplitude explains why safe ranges of stress are possible. It can be shown [42, 43J that the total (integrated absolute) amount of plastic strain in an elastically



embedded strain-hardening plastic region always converges towards a finite value as the number of cycles increases toward infinity. This value of the total plastic strain decreases with the decrease of the stress amplitude applied to the specimen. Below a certain stress amplitude the total plastic strain can never reach the critical value necessary for producing that combination of strain hardening (i.e., of the local stress amplitude) and structural damage at which fracture occurs. On the other hand, if the local plastic region fractures, a small crack arises and gives rise to a region of stress concentrations in which plastic deformations may now begin. A repetition of the above process may lead to the extension of the crack and finally to the fracture of the specimen.

An interesting point emerging from the theory is that a fatigue fracture can arise without any reduction of the cohesion (strength) by structural damage. Strain hardening alone may raise the stress in plastic regions gradually to the fracture level even if the initial strength of the material is not reduced in the course of the alternating plastic straining. In most real cases, however, increase of the local stress by strain hardening and reduction of the strength by structural injuries probably go hand in hand.

Observations indicate that, in reality, the last traces of alternating slip never disappear; there is apparently a minimum value of the plastic strain amplitude below which no strain hardening is produced. This can be recognized most directly from the fact that the width of the hysteresis loop decreases but does not vanish during cyclic stressing.

It may be mentioned that the general theory of fatigue leads to a semiquantitative derivation of the typical shape of the log S-log N curve, and it also


CUNn 1; 0.11:

explains the remarkable fact (see below) that the influence of the steady stress upon the fatigue strength is, as a rule, very small and sometimes imperceptible up to the value of the static yield stress.

To sum up, it can be said that the typical features of fatigue under cycles of constant stress amplitude follow directly from the fact that plastic deformation is not uniformly distributed but, after an initial deformation that may possibly extend over most of the specimen, becomes confined to a few local regions. Once plastic flow becomes locally concentrated, the conditions governing the development of fatigue cracks can be investigated by a general consideration of the change of stress and strain amplitudes in plastic regions embedded in elastic surroundings subjected to cycles of constant stress amplitude. As far as the general features of the fatigue phenomenon are concerned, the molecular nature of the fatigue process is of secondary importance; in particular, fatigue fracture might conceivably occur without any decrease of the cohesion, solely by the rise of the local stress by strain hardening to the fracture level.

C. Influence of a Superposed Steady Stress.

Figure 1.22 shows the dependence of the fatigue strength (limiting stress amplitude) of three plaincarbon steels on the steady stress (mean stress of the cycle) according to the experiments of Pomp and Hempel {44, 45, 46]; the dash-dotted lines at 450 to the coordinate axes are the loci of the points at which the maximum stress of the cycle (including the steady stress) reaches the conventional elastic limit (in the present case, the 0.2% proof stress). The curves reflect, first of all, a general feature of the dependence of the fatigue strength upon the steady stress: up to the elastic limit, they represent straight

~ 50
'" OL-------~~~~~--*_--~--~--~----~------~--~

o Yl 150 200

" Meon SI,,,,, of Cycle. I 03 psi

FIG. 1.22 Dependence of fatigue strength on steady stress in plain carbon steels.







~ so


:;: .,;



\ \

o+- ~------~------~-------~--~u






s. Mean Siren Clf Cycle. 103 p1i

FlG. 1.23 Dependence of fatigue strength on steady stress in patented (0.62% C) steel wire.

lines which slope downwards only slightly with increasing steady stress. Occasionally this line is horizontal; in all cases, the influence of the mean stress on the fatigue strength is small.

Another feature of Fig. 1.22 is the rapid change of the character of the curve at the elastic limit. The slope changes abruptly with the onset of significant plastic deformations; the curves show a distinct increase of the fatigue strength (limiting safe stress amplitude superposed to the steady stress) at the end of the elastic region. This "step" at the elastic limit is followed by a second abrupt change of slope, during which the fatigue strength declines with further increase of the steady stress.

With strongly cold-worked metals, proof stress and ultimate stress nearly coincide. In such cases only the first part of the "step" seen in Fig. 1.22 can be observed. An example is shown in Fig. 1.23 ("patented" steel wire, O.G2% C) [45].

If a rod is circumferentially notched (e.g., if it is threaded), its static yield and ultimate stresses referred to the smallest cross section arc higher than

Plcln bcrs

,"" end l:r WhllwO(th tbreeded bon, cveroqe volut!:

Mt"on Sireu. s, 01 Cyde. 103 -pt.

FIG. 1.24 Comparison of fatigue strengths of plain lind threaded bars of 0.1 % C steels.

those for the smooth rod (owing to "plastic constraint" exerted by the adjacent larger sections); its fatigue strength, however, is reduced by the stress concentration present. Figure 1.24 shows fatigue strength curves for 1 in. and lk in. Whitworth threaded rods of the carbon steel which, in the form of smooth cylindrical specimens, gives curve 1 in Fig. 1.22 (this curve is repeated in Fig. 1.24) [46].

The designer is mainly interested in stresses within the elastic limit; for this reason, the present considerations will be confined to the first part of the curves in Fig. 1.22. This can be represented schematically as a straight line connecting the point P of the fatigue strength in purely alternating stressing with a point Q on the abscissa axis (cf. Fig. 1.25); as before, the dash-dotted 45" line represents the elastic limit beyond which curve deviates from the line PQ. In the stress range of interest to the designer, the effect o( the steady stress is therefore given by the equation

S = So (1 - s:)


where S is the fatigue strength at the steady stress s, So its value for s = 0, and Sp = OQ, a stress parameter that determines the position of the line PQ.

For many decades in the past, the dependence of the fatigue strength upon the steady stress was usually represented by the Goodman diagram in which the assumption was made that the stress parameter Sp = OQ can be identified with the ultimate stress; the Goodman diagram is indicated in Figure 1.25 by the dashed line PU where OU is the ultimate stress. Goodman's idea was that the line would have to go through the point at which "failure" would occur in purely static tension. As can be seen from the discussion in Section 1.3, this argument is invalid: the ultimate stress is not a stress at which fracture occurs but merely the

FlO. 1.25 The influence or steady (mean) stresa upon the fatigue limit,



(conventional) stress at which the maximum load is reached and necking begins in the static tensile test. For this reason, the ultimate stress point U has no place on any curve showing the dependence of the fatigue strength upon the steady stress, and much less on the straight line that forms the initial elastic part of such curves. In the experimental curves shown in Fig. 1.22, for instance, the extension of the initial straight part may intersect the abscissa axis quite far from the point U of the ultimate stress; the stress parameter Sp in eq. 1.51 and the position of the point Q can only be derived from fatigue tests. The only point that can be made in defense of the Goodman line is that its errors, however large, usually lie in the safe direction.

D. Influence of a Compound State of Stress.

Relatively little is known about the condition of fatigue fracture for cyclically varying triaxial states of stress. However, a practically important case, that of a shaft subjected to cyclic torsion and bending simultaneously, has been investigated in detail by Gough and Pollard [47]. They found that for a given (large) number of cycles, those corresponding values Sand T, respectively, of the tensilestress amplitude due to bending and of the shearstress amplitude due to torsion at which fracture occurs are determined approximately by the relationship


where So is the fatigue strength for the same number of cycles in pure bending, and To the fatigue (shear) strength in pure tomion.

E. Influence of Notches and of Surface Flaws.

Stress raisers are relatively unimportant in ductile metals under static stress, because plastic flow levels down the stress at the stress concentrations. In cyclic stressing, the situation is different: local cyclic straining produces progressive strain hardening with consequent rise of the local stress. If the strain hardening could continue with cyclic plastic deformation at a finite rate, no matter how small the plastic strain amplitude, it would finally raise the yield stress until no plastic deformation could occur. The local alternating stress amplitude and the effective stress concentration factor would then be the same as in a purely elastic body of the same geometry.

Experience shows that this is not the case in fatigue. The effect of notches, cracks, and surface flaws is usually much greater than in static stressing, but it is still far below what it would be for a purely

elastic material. The simplest explanation of this remarkable fact is to assume that cyclic straining ceases to produce strain hardening when the strain amplitude becomes too small (see above); if this is the case, the material at the tip of the crack never becomes quite elastic and the stress can never reach the level of the elastic stress concentration. Different materials have different "notch sensitivities"? (not to be confused with the notch sensitivity for static brittle fracture). Some of them, like grey lamellar cast iron or certain bronzes, are almost insensitive to the presence of small sharp cracks or notches; their q value will therefore be close to zero. Others, like hard steels, are very sensitive, with q in the neighborhood of 1-

Similarly, the surface quality has an influence upon the fatigue strength of ductile metals that is between those for a completely brittle material like glass and for a ductile metal under static stress. Occasionally, the fatigue strength of extruded lightalloy rods with the extrusion skin has been found to be as low as one-half of the fatigue strength of a machined specimen of the same rod. In some cases, the fatigue strength can be raised" considerably by surface rolling or shot blasting (e.g., for heat-treated spring steels); in others, such a treatment has no significant beneficial influence (e.g., with many light alloys). Excessive surface rolling or shot blasting in materials of limited ductility may even reduce the fatigue strength by producing surface cracks.

There is a difference of great importance between the fatigue strength of a ductile metal and the (static) strength of a brittle material like glass. In the latter case, the strength can be raised sometimes by a factor of 10 or even 100 if surface cracks are very carefully avoided. In ductile metals, it is relatively easy to improve the quality of the surface so that any remaining flaws have no influence on the fatigue strength. However, this does not raise the fatigue strength spectacularly because plastic deformations set in as soon as the elastic limit is exceeded, and they produce cracks after sufficiently prolonged cyclic stressing in a way that is now more or less understood. For this reason, there is no hope that the fatigue strength may be raised much above the elastic limit. If, on the other hand, the elastic limit is raised by strain hardening, precipitation harden-

6A conventional quantitative definition of the relative notch sensitivity q in cyclic stress is q = (k, - l)/(kc - 1) where kc is the elastic stress concentration factor for a given notch and k, is the factor by which the fatigue strength is reduced by the presence of the notch. or course, q depends in general on the size and shape of the notch.



ing, or in any other way, a decrease of ductility is unavoidably associated with the increase of the

fatigue strength. .~

F. Fatigue Tests on Specimens vs. Fatigue Tests on Structural Parts. The strength of structural parts under static load can usually be calculated with reasonable accuracy on the basis of tests performed on specimens. Stress concentrations either do not matter (in very ductile materials), or they can be calculated by methods given in the theory of elasticity. The situation, however, is very different in cyclic stressing. The effective stress concentration factors depend here not only on the geometry and On the elastic constants, but in the first line on the "notch sensitivity" of the material, which depends On the size of the notch. Whenever a structural part has a strongly non-uniform stress distribution, therefore, its fatigue properties cannot be calculated from tests on specimens with any reasonable accuracy. If the structural part cannot be overdimensioned so as to exclude any danger, it is necessary to carry out full-scale fatigue tests on it [48]. This is particularly important, of course, in the case of aircraft structures. As already mentioned, attention must be given in any case to possible differences between the fatigue behavior of specimens with carefully machined surfaces and specimens or structural parts with surfaces as they will be present in the structure.

There is a more trivial reason why so often conclusions drawn from experiments with specimens are not fulfilled by structures. Fatigue tests are usually constant stress tests, occasionally constant strain tests. On the other hand, if a structure is subjected to cycles of constant load or deformation amplitude, some of its elements (for instance, regions of stress concentrations) will be under cycles of increasing stress amplitude and decreasing strain amplitude, for the reason explained above in connection with the general theory of fatigue. It follows, then, that the results of constant amplitude tests cannot be applied directly to the calculation of the fatigue strength of structures with non-uniform stress distribution. A general method of calculation in such cases has been given [43]; for the present, however, lack of experimental data prevents the practical use of this method except in the simplest cases.

G. Periodically Varying Thermal Stresses.

If a body is rigidly clamped at two points, increase Or decrease of its temperature gives rise to thermal stresses in it. The magnitude of these stresses depends not only on the temperature change and on the material, but also on the shape of the body; in a

multiply bent, relatively thin tube the stresses arc much lower than in a straight bar fixed at two cross sections.

Obviously, the action of a stress upon a material is quite independent of how it is produced; COnsequently, the fatigue effect of a thermal stress cycle is identical with that of a mechanical load cycle involving the same stresses at the same temperatures. Compared with the ordinary fatigue test, the only new factor introduced by the thermal cycling of a rigidly supported specimen is that, together with the stress, the temperature also varies during the cycle. If the temperature amplitude is relatively small, the fatigue effects of a thermal cycle will be the same as those of an isothermal load cycle involving the same stresses at a constant temperature equal to a suitably chosen mean temperature of the thermal cycle. That this can be so even for cycles of considerable temperature amplitude is indicated by recent experiments of Coffin [49J. The equivalent mean temperature of the cycle, however, is not necessarily the mean value of its highest and lowest temperatures. If the magnitude of fatigue damage is determined mainly by the amount of plastic deformation, the temperature of the equivalent isothermal cycle will lie nearer to the maximum than to the mean temperature of the thermal cycle because the material is softest, and plastic deformation greatest, in the high temperature part of the cycle. The opposite behavior (the 10\'1'temperature part of the cycle being of dominating importance) may conceivably also occur. If the specimen is a straight rod or tube with fixed ends, it is always in tension during the low-temperature part of the cycle. If now the tensile part of the cycle is more likely to produce fatigue damage than the compressive part, the effect of the thermal cycle may be closer to that of an isothermal cycle with the same stress range taking place near the lowest temperature of the stress cycle.

A new factor appears (both in thermal and in purely mechanical cycling) if the temperature is so high that the strain hardening and the structural damage due to plastic deformation arc currently removed during the cyclic straining. In this case, the progressive changes which represent cyclic fatigue cannot develop. Nevertheless, fracture may occur owing to a different phenomenon which has been treated already under the heading of creep fracture. At very high temperatures (in the hot creep range), the grain boundaries become soft, and the consequent relative displacements between the neighboring grains open up cracks which finally can lead to fracture ("static fatigue"). At first sight, it might


seem that this cannot occur under purely cyclic stress because the displacements produced by the tensile part of the cycle are reversed ''by the compressive part. However, the compressive part cannot undo all damage done by the tensile part, and so fatigue fracture can also occur under purely cyclic stress, although much more slowly than under a steady tensile stress.

Lazan and Westberg [50] have carried out experiments in the interesting transition region just below the hot-creep range j they applied both purely cyclic and purely static stresses and intermediate types of loading with a static stress superposed upon a steady stress. Figure 1.26 illustrates some of their results. As in room temperature experiments, a relatively low mean stress has only a slight influence upon the fatigue strength if the duration of the test is not too long. If the time to fracture is 150 hours or longer, creep predominates over cyclic fatigue, and the steady component of the cycle becomes important from the beginning. The vertical parts of the curves show that the static fatigue strength is almost uninfluenced by the cyclic component until the cyclic stress amplitude becomes higher than about one-half of the static stress. The observed curves, therefore, consist essentially of a nearly horizontal part representing cyclic fatigue (except in very prolonged tests, as mentioned above), and of a vertical part representing creep fracture. The transition between the horizontal and the vertical part is the region in which cyclic and static fatigue are of comparable importance.

If a material has been cold worked and then subjected to plastic deformation at a higher temperature, it may soften more than if it had been subjected to the effect of the higher temperature alone without


Speed gf Cyding :214,000 rev.rsals/hr.

';0 Q.

Time to fradure



.; -0 ,_

u 20 '0

§ ;;;


~ 10 £



i, Moon Stre .. 01 Cycle, 103 psi

FIG. 1.26 Fatigue-creep rupture interaction curves for N-155 at. 1500 F. After Lasan and Westberg.


Oomogo Areo

fatiguo Froduro Slro .. Curve

Log N (N = Numbar of .tre .. cycl",) FlO. 1.27 The damage area in fatigue.

further deformation [19b, 51]. In the course of the deformation, its relatively highly hardened structure changes to the less hardened structure characteristic of deformation at the higher temperature. A similar strain-softening effect can also be observed in fatigue tests with previously strongly cold-worked materials [52]. This, however, does not mean that strain hardening is not an important factor in fatigue. Local regions of stress concentration, e.g. at the tip of a fatigue crack, may well harden under cyclic stressing, while the static yield stress of the prestrained bulk material decreases by thermal recovery with or without strain softening.

H. Thermal Fatigue. The most severe case of cyclic thermal stressing takes place when the surface of a metal is rapidly heated to a high temperature and then cooled again. This occurs in hot rolls, gun barrels, etc.; if the temperature amplitude is high, the usual effect is the formation of surface cracks ("crazing") which gradually spread inwards. Frequently this cannot be prevented; the life of the body can be prolonged, however, if the surface is machined off before the cracks become too deep. In other cases, thermal cracking would occur with most materials but can be avoided by the use of special metals, such as, e.g., the 12% Cr steel used for rolls in continuous sheet glass manufacture.

Anisotropic metals such as zinc, or metals that suffer phase transformations in the temperature range to which they are subjected, can suffer plastic deformations on a microscopic scale within the grains which are confined and distorted by their neighbors, even if there is no significant temperature gradient present. This may result in progressive structural damage during thermal cycling.

J. Damage by Overstress. If a material is subjected to stress amplitudes above the fatigue limit,


it may suffer permanent damage which reduces its fatigue strength for subsequently applied cycles of lower stress amplitude. It seems that those combinations of stress amplitude and number of cycles above which permanent damage occurs lie in the area D (Fig. 1.27) between the high-stress part of the log S-log N curve and a line below it which joins the curve at the bend [53]. This line, shown dashed in Fig. 1.27, is the "damage line." The permanent damage suffered in the damage area D consists probably in the formation of small cracks.

K. Corrosion Fatigue. If the cyclically stressed material is in a chemically active solution, its fatigue strength may be SUbstantially lowered. Whether in this case an approximate fatigue limit exists is not certain; as in stress corrosion, the phenomenon is so strongly influenced by the individuality of the metal and of the surrounding solution that the only general statement that can be made about it is a warning against premature extrapolations to even slightly different metals and solutions.


1. M. Cook and E. C. Larke, "Resistance of Copper and Copper Alloys to Homogeneous Deformation in Compression," J. Inst. Metals, Vol. 11, p. 371 (1945).

2. M. Tresca, "Memoire sur le poinconnage et la theorie meeanique de la deformation des metaux," Campi. rend., Vol. 68, pp. 1197-1201 (1869).

3. R. von Mises, "Mechanik der festen Kerper im plastiachdeformablen Zustand," Nachr, kgl. Ges. Wiss. Math.-Phys. Klasse, 1913, pp. 582-592.

4. R. Hill, The Mathematirol Theory of Plasticity, p. 20, The Clarendon Press, Oxford, 1950.

5. J. L. M. Morrison, "The Yield of Mild Steel with Particular Reference to Effect of Size of Specimen," Proc., Inst, Meeh. Enqre, (London), Vol. 142, pp. 193-223 (1940).

6. O. Hoffman and G. Sachs, Introduction to the Theory of Plasticity for Engineers, McGraw-Hill Book Co., New York, 1953.

7. M. Considere, "L'emploi du fer et de I'acier dans les constructions," Ann. Pants et Chaussees, 6th Series, Vol. 9, pp. 574-775 (1885).

8. G. Sachs and J. D. Lubahn, "Failure of Ductile Metals in Tension," Trans. ASME, Vol. 68, pp. 277-279 (1946).

9. H. W. Swift, "Plastic Instability Under Plane Stress," Journal of Mech, & Pilys. oj Solids, Vol. I, No. I, pp. 1-18 (October, 1952).

10. W. R. D. Manning, "The Overstrain of Tubes by Internal Pressure," Engineering, Vol. 159, pp. 101-102, 183-184 (1945); also "The Design of Compound Cylinders for High Pressure Service," Engineering, Vol. 161, pp. 349- 352 (1947).

11. C. W. MacGregor, L. F. Coffin, Jr., and J. C. Fisher, "The Plastic Flow of Thick-Walled Tubes with Large Strains," J. Appl. Pluu., Vol. 19, pp. 291-297 (1948); also "Partially Plastic Thick-Walled Tubes," J. Franklin Inst., Vol. 245, pp. 135-158 (1948).

12. E. N. cia C. Andrade, "The Flow in Metals Under Large Constant Stresses," Proc. Roy. Soc., Series A, Vol. 90, pp. 329-342 (1914).

13. P. Phillips, "The Slow Stretch in Indiarubber, Glasa, and Metal Wires when Subjected to a Constant Pull," Phil. Maq., 6th Series, Vol. 9, pp. 513-531 (1905).

14. F. H. Norton, Creep of Steel at High Temperatures, McGraw-Hili Book Co., New York, 1929.

15. C. R. Soderberg, "The Interpretation of Creep Tests for Machine Design," Trans. ASME, Vol. 58, pp. 733-743 (1936).

16. A. Nadal, "The Influence of Time upon Creep. The Hyperbolic Sine Creep Law," S. Timoshenko 60th Anniversary Vol., Macmillan Co., New York, 1938.

17. E. Orowan, "Discussion on Plastic Flow in Metals," Proc. Roy Soc., Series A, Vol. 168, p. 307 (1938); also Proc. First Nat. Conqr. App!. Meek., June, 1951, p. 453. J. W. Edwards, Ann Arbor, Mich., 1952.

18. R. Becker, "Uber die Plastlzitat amorpher und kristalliner fester Korper," Physik. Z., Vol. 26, p, 919 (1925); also Z. Tech. Physik., Vol. 7, p, 547 (1926).

19. E. Orowan, "The Creep of Metals," Z. Physik., Vol. 98, p. 382 (1935); also "The Creep of Metals," Trans. West oj Scotland Iron Steel Insi., pp. 45-96 (1947).

20. F. R. Larson and J. Miller, "A Time-Temperature Relationship for Rupture and Creep Stresses," Trans. ASME, Vol. 74, pp. 765-771 (1952).

21. N. J. Grant and A. G. Bucklin, On the Extrapolation of Short-Time Stress-Rupture Data, ASM Preprint No. 18, 1949.

22. A. A. Griffith, "The Phenomena of Rupture and Flow in Solids," Trans. Roy. Soc., Series A, Vol. 221, pp, 163-198 (1920-21); also First Internat. Conar. Appl. Meeh., p. 55, Delft, 1924.

23. C. E. Inglis, "Stresses in a Plate due to the Presence of Cracks and Sharp Corners," Trans. Insi. Naval Archil., Vol. 55, Part I, pp. 210-230 (1913).

24. E. Orowan, "Fracture and Strength of Solids," Reports on Progress in Physics, Vol. 12, pp. 185-232 (1949).

25. A. Mesnager, Reunion dell Membres Froneaie et Belges de l'Associatwn Internationale dell Methode d'Essais, pp. 395- 405, December, 1902.

26. P. Ludwik and R. Scheu, "Uber Kerbwirkungen bei Flusseisen," Siahl und Eisen, Vol. 43, pp. 999-1001 (1923).

27. E. Orowan, "Notch Brittleness and the Strength of Metals," Trans. Ins!. Er!grs. Shipbuilders Scot., Paper No. 1063, pp. 165-215, December, 1945.

28. N. N. Dnvidenkov and F. Wittman, "Mechanical Analysis of Impact Brittleness," Phys.-TecJm. Inst. (U.S.S.R), Vol. 4, p. 308 (1937).

2!l. D. K. Felbeek and E. Orowan, "Experiments on Brittle Fracture of Steel Plates," Welding J. (N.Y.), Res. Supp/., Vol. 20, No.7 (Hl55).

30. M. J. Manjoine, "Influence of Rate of Strain and Temperature on Yield Stresses of Mild Steel," J. Appl. .lfechanics, Vol. 11, pp. A2U-218 (1944).

31. G. 1. Taylor, "Testing of Materials at High Rates of Loading." J. Insi. Cio. Ertgrs., Vol. 26, pp. 486-519. (1946).

32. E. Orowan, "Fundamentals of Brittle Behavior in Metals," in William M. Murray, ed., Fatigue and Fracture of Metals: A Symposium, pp. 139-167, John WjJey & Sons, New York, 1952.


33. F. J. Feely, Jr., and M. S. Northup, "Study of Brittle Failure in Tank Steels," presented at the Midyear Mtg., Am. Petro Inst., in Houston, Texas, May, 1954.

34. E. Orowan, "A Type of Plastic Deformation New in Metals," Nature, London, VoL 149, p. 643 (1942).

35. G. Akimow, "Eine neue Theorie der Struklurkarrosion," Korrosion U. MefaUsckutz, Vol. 8, p. 197 (1932).

36. G. Wassermann, "Untersuchungen tiber den Vorgang der Spannungakorrosion," Z. Meiallkunde, Vol. 34, p. 297 (1942).

37. G. Edmunds, "Season Cracking of Brass," ASTM Symp. on Stre8s-CtJ1T. CracA:ing in Mela.!.s, p. 67 (1944).

38. E. Orowan, in a paper presented before The Elcctrochem.

Soc., Boston, Oct. 4, 1954.

39. J. A. Ewing and J. C. W. Humfrey, "The Fracture of Metals Under Repeated Alternating Stress," Trans. Roy Soc., Series A, Vol. 200, pp. 241-250 (1903).

40. F. A. McClintock, "On Direction of Fatigue Cracks in Polycryetalline Ingot Iron," J. Appl. Mechanies, Vol. 19, pp. 54-56 (1952).

41. E. Orowan, "Dislocations and Mechanical Properties" in DisCocations in Meta!.s, AIME, New York, 1954.

42. E. Orowan, "Theory of the Fatigue of Metals," Proc.

Roy. Soc., Series A, Vol. 171, pp. 79-1013 (1939).

43. E. Orowan, "Stress Concentrations in Steel under Cyclic Load," WeldingJ., (N.Y.), Res. Suppl., Vol. 17, pp. 273s- 282s, June, 1952.

44. A. Pomp and M. Hempel, "Dauerfestigkeitsscbaubilder von Stii.h1en bei verschiedenen Zugmittelspannugen unter BerUcksichtigung der Prufstabform," Mitt. KaiserWilhelm-Inst. Eisen/orsch. Dusseldorf., Vol. 18, pp. 1-14 (1936).


45. A. Pomp and M. Hempel, "Dauerprtifung von Stahldrehten unter wechselnder Zugbeansprunohung," !ffiU. KaiserWilhelm-Inst. Eisenforsch, Dusseldorf, Vol. 19, pp. 237- 246 (1937).

413. A. Pomp and M. Hempel, "Dauerfestigkeitsschaubilder von Gekerbten und Kaltverformten Stahlen sowia von 1"_ und li"- Schrauben bci Verschiedenen Zugmittelspannungen," MiU. Kaiser-Wilhelm-Insl. Eisenforsch:

Dusseldorf, Vol. 18, pp. 205-215 (1936).

47. H. J. Gough and H. V. Pollard, "The Strength of Metals under Combined Alternating Stresses," Proc. Insi. Med». EngTs. (London), Vol. 131, pp. 3-54 (1935).

48. R. L. Templin, "Designing for Fatigue" in William M.

Murray, ed., Fatigue and Fracture of Metals; A Symposium, pp. 131-138, John Wiley & Sons, New York, 1950.

49. L. F. Coffin, Jr., "A Study of the Effects of Cyclic Thermal Stresses on a Ductile Metal," Trans. ASME, Vol. 76, No.6, pp. 931-950 (1954).

50. B. J. Lazan and E. Westberg, "Effect of Tensile and Compressive Fatigue Stress on Creep, RUpture and Ductility Properties of Temperature-Resistant Materials," Proc:

ASTM, Vol. 52, pp. 837-855 (1953).

51. J. E. Darn, A. Goldberg, and T. E. Teitz, "The Effect of Thermal-mechanical History on the Strain Hardening of Metals," AIME Tech. Pub. No. 2445, 1948.

52. N. II. Polakowski, "Softening of Certain Cold-worked Metals Under the Action of Fatigue Loads," ASTM Preprint No. 74, 1954.

53. H. J. French, "Fatigue and Hardening of Steels," TraM.

Am. Soc. Steel Treating, Vol. 21, pp. 899-946 (1933).


. .,.


Design Assumptions, Stress Evaluation, and Design Limits

THE previous chapter passed over the problem of calculating stresses and strains from the applied load in order to concentrate on certain fundamental knowledge from the physics of solids which, it was pointed out, is relatively new and as yet largely unformulated for use in routine design engineering. The present chapter offers a general examination of the factors which enter into the evaluation of stresses in piping systems due to various external and internal loadings, their association with design limits and Code rules, and finally, their significance and application to practical design.

With the increasing complexity, size, and economic significance' of piping installations, it is necessary to look beyond the limits of ordinary piping design practice and to give attention to the experiences of designers in related fields, particularly that of pressure vessel design. Indeed, there is often no logical distinction between pressure vessels and piping. Therefore, appropriate comments relative to comparative piping and pressure vessel design approaches are given frequently in the discussions which follow.

In further consequence of the economic importance of present-day piping installations it is necessary, just as in the design of structures and pressure equipment, to effect a careful and realistic compromise between design features (not overlooking materials, fabrication, and inspection requirements) and the overall plant economics (first cost plus maintenance and contingency for damages to property and personnel in event of failure). Safety of operating

lPiping is II. major item in process plants, running from 50 to 75 per cent of the total plant cost, Similar significant expenditures arc incurred in power generation and marine propulsion installa tiona,

personnel and the interests of the general public dictate that all feasible precautions be exercised. Maximum assurance of safety, however, would require complete examination of all materials and fabrication by the best available means and with duplicate independent inspection. Even so, absolute assurance of safety could not be attained due to personnel fallibility and the limitations in sensitivity of available methods of nondestructive examination. With this realization, in the practical approach of achieving adequate safety economically, lower levels of quality are accepted on the basis of including compensating safety factors in design, which are the combined result of experience and reasoning. Many inconsistencies still exist in current practice relative to quality requirements of materials and fabrication, and in the value placed on various degrees of inspection, tests, and nondestructive examination.

It should be appreciated that Codes and Standards can establish only a level of minimum requirements for average service, based on knowledge, experience, and the consensus of qualified individuals. Many circumstances relating to service operation, materials and fabrication, inspection limitations, or to unusual design deserve special consideration if the resulting piping systems are to be reasonably frce from maintenance, and provide satisfactory length of life with safe operation. To assist the piping engineer in the exercise of good judgment on these special problems, this chapter offers approaches which largely depend on well-established practical experience.

2.1 Codes and Standards

The objective of Code rules and Standards (apart from fixing dimensional values) is to achieve mini-




mum requirements for safe construction; in other words, to provide public protection by defining those material, design, fabrication, and inspection requirements whose omission may radically increase operating hazards. Absolute assurance of safety would require perfect design, materials, and fabrication; this is seldom, if ever, achieved. On the other hand, experience with Code rules has demonstrated that the probability of disastrous failure can be reduced to the extremely low level necessary to protect life and property by suitable minimum requirements and safety factors. Obviously, it is impossible for general rules to anticipate other than conventional service, and it would be uneconomic for them to provide for corrosion, erosion, fatigue, shock, or potential brittle fracture, except to the degree that such conditions are known to be present. Suitable precautions are, therefore, entirely the responsibility of the design engineer guided by the needs and specifications of the user.

A listing of all Standards and Specifications concerning piping design, together with their mandatory effective edition references, appears in an appendix of the Code for Pressure Piping (ASA B31.1). Those which affect the mechanical design of piping are briefly commented on in the following paragraphs, relative to their basic approach and significant details.

One of the difficulties which often confronts designers of vessels and piping, as related to Code requirements and particularly local governmental regulations, is the proper classification of borderline pressure equipment. Currently (1955), neither the ASME nor the ASA Code contains definitions for vessels or piping which are helpful in this respect. While the Code Committees have considered this matter, no common agreement has been reached. Some items in piping systems often considered and fabricated as part of the piping, e.g. pulsation dampeners, are classed as pressure vessels in some States. In doubtful cases it is advisable for the user to check with the local authorities, especially in localities having regional pressure vessel laws.

ASME Boiler and Pressure Vessel Code. Section I, Power Boilers, contains rules for the pressure design of boiler piping within the specified boiler limits which are associated with appropriate steam and feed-water stop valves. The design, fabrication, and inspection requirements of the ASME Unfired Pressure Vessel Code, Section VIII, arc often used by reference in company specifications to supplement the Piping Code. Section IX of the ASME Code is the universal baaia for qualification of welding procedures and operators of all pressure equipment.

ASA B31.1: Code for Pressure Piping. This is the standard "Piping Code" which includes sections on Power,

Gas & Air, Oil, District Heating, Refrigeration, Oil Transmission, Gas Transmission and Distribution Systems (ASA B3l.l.8-l955), and Chemical Piping. Its basic or general supporting sections deal with requirements for internal pressure, flexibility, materials, fabrication, and testing. At the present writing (1955), this Code is in the process of evolution from a Design Practice to a Safety Code. The Gas Transmission and Distribution Section has been adopted by several States and is under consideration by others; the entire Code is used as a basis of enforcement in several U. S. cities and in the Provinces of Canada. In recognition of this trend, a Conference Committee similar to that of the ASME Boiler Code and composed of the Chief Inspection Authority of each State and each Canadian Province which has adopted the (Piping) Code, has been appointed. At the same time a procedure was established to provide interpretations in the form of Cases, which again parallels the ASME Boiler Code procedure.

This transition is largely due to recognition by public authorities that pipe line failures associated with a sudden release of stored energy are potentially as dangerous us pressure vessel failures. Experience with piping systems also demanded a change in the former attitude that thermal expansion strains could not be responsible for a major failure. Although this type of failure is due to fatigue rather than to a single application of strain loading it can be a definite hazard in most services.

ASA 89: Safety Code for Mechanical Refrigeration.

This Code contains, in Section 9, brief rules for pressure and general design of piping for this specific service.

Piping for Ships. Such piping requires special consideration because of added strains from the motions of the ship. Naval vessels are subject to added shock due to Budden maneuvering, gunfire, explosions, etc. Requirements for merchant and naval vessels are contained in the following Standards;

U. S. Navy, Bureau of Ships; Genernl Machinery Specifications; General Specifications for Building Naval Vessels.

American Bureau of Shipping; Rules for Building and Classing Vessels.

United States Coast Guard; Marine Engineering Regulations and Material Specifications.

Lloyd's Register of Shipping Rules

Flange and Filling Standards. The B1G group of ASA Standards apply to pipe-fitting details. Although their significance is primarily dimensional, they involve design factors which should be appreciated. These are summarized in the following sections:

Steel Flanges. The proportions of separate flanges and those integrated with fittings were established many years ago, based on simplified cantilever analysis. In 1953 the steel flanges were reinvestigated according to present ASME Boiler Code formulas, New ratings were established for two general classes of gaskets and facing details. These appear in ASA Standard, BI6.5-l953, and also in the AS ME Codes. The basis of the new ratings is recorded in Appendix D of the BIG.5 Standard. The calculated stress in the flanges shows appreciable variation with size, series, and facings. A stress of 8750 psi at the primary pressure rating was selected for the purpose of establishing Class A ratings. Class B ratings arc approximately 83% of Class A ratings. In the creep range at or above the primary rating temperature, ASME Power Boiler Code stresses are adhered to. For temperatures up to

higher series are required to have increased physical properties and accordingly are assigned ratings about 33% higher.

Other Standards: Other Standards which contribute to piping design are those of the Manufacturers Standardization Society of the Valve and Fittings Industry (MSS Standard Practices), American Water Works Association (AWWA), American Gas Association (AGA), Federal Specifications Board (FSB), and Association of American Railroads (AAR). These are for the most part dimensional standards and rating tables for specific piping and fittings.



GOO F the ratings are based on allowable stresses, which are approximately 60% of the yield strength. This is similar to the allowable stress basis of the Piping Code, Section 3, Oil Piping. Ratings between G50 F a"lld the primary service temperature are established by a straight line transition. In general, the bolting, particularly when alloy steel, is of substantially greater strength than the flanges, which can be distorted by overtightening. This excess bolt strength is significant in the ability of ASA flanges to transmit line moments, as discussed later in Chapter 3.

Steel Flanged FiUin'g Thickness. Fitting thicknesses were originally established for east-carbon steel by application of the Barlow (outside diameter) formula with an allowable stress of 7000 psi at the primary pressure rating, and applying a 50% increase in thickness as a "shnpe Iactor ," Thisnpproach was later extended to other cast and forged materials. The fitting thicknesses in the 1953 issue of BIG.5 are based on this same allowable stress, which is 80% of the value used for rating Class A flanges, using the primary service pressure and the modified Lame formula now common to the Codes. An excess thickness of 50% is provided for 0.11 flanged fittings in recognition of the reinforcment required at the side outlets of tees, bonnet necks of valves and similar branch connections, as well as for elbows, whether or not they have side branch connections.

Steel Butt IV eIding Filling Thicknese. For cast- or wroughtbutt welding fittings the thickness required by ASA Standard B16.9 at the welding ends is the same as that of the pipe size and schedule with which they are intended to be used. Instead of establishing minimum wall thicknesses or "shape factors" as is done for flanged fittings, this Standard requires only that the bursting strength be not less than that of a pipe of the corresponding material, size, and schedule number; the pressure-temperature rating then becomes identical with that of the intact pipe.

API-ASME Code for Unfired Pressure Vessels. This pressure vessel Code is sometimes used as a reference in company specifications. Except (or the absence of mandatory random examination requirements, its provisions are essentially the same as Section VIII of the ASl',IE Boiler and Pressure Vessel Code.

API Standards. In addition to material specifications for line pipe, threads, etc., the American Petroleum Institute has standards for certain types of iron or steel valves for refinery or drilling and production service (API Standards 600, 6C and 6D) and for ring-joint flanges (API Standard 6B). The flanges and ratings utilized in Standard 600 are based on ASA standards. Standards 6C and 6D assign separate pressure ratings for "pipe line service" and "drilling and production service" at 100 F. In addition to utilizing ASA Standard flanges, API Standard 6B includes a special "2900 Ib" series. This is similar to the original assignment of a 4000 lb rating to 1500 lb series flanges, drilled one size smaller, which was advanced and used by The M. W. Kellogg Company, except that the design was refined, in accordance with calculations using ASME Code formulas, by Messrs. Petrie and Watts of the Crane Company and Standard Oil Company (Indinna), respectively. The API Standard assigns a 100 F rating of 7500 psi for pipe line service, and 10,000 psi (or drilling and production service. For the latter service, materials with higher tensile and yield strengths are required. The API ratings for ASA flanges are the same as ASA ratings for pipe line service; for drilling and production service the 600 lb and


2.2 Design Considerations: Loadings

A piping system constitutes an irregular space frame into which strain and attendant stress may be introduced by the initial fabrication and erection, and also may exist due to various circumstances during operation, standby, or shutdown. In its erected position, a piping system is subject to loads due to dead weights (pipe, fittings, insulation), snow or ice, contents of the line, wind load for exposed piping, and earthquake or other shock loading in special situations. Internal (or external) pressure loads may be imposed in service or off stream. The restraint of thermal expansion provided by terminal and intermediate anchors, guides, and stops introduces thermal stresses in piping due to temperature changes. Further stress may be introduced by the rnovement'' of terminal equipment, foundations, or buildings under temperature changes or other loading, or from any influence affecting the relative position of the line, anchors, or intermediate restraints.

The dead load effects, except contents, are usually maintained at all times, while wind or earthquake effects will be variable and reach maximum design values infrequently, if ever. Pressure and temperature changes usually occur simultaneously, but may be independent or have a variably dependent relationship. They may be relatively uniform for entire service periods, or involve swings of variable duration.

Dead load and wind or earthquake effects all piping are no different than for conventional structures, while pressure effects are essentially the same as those encountered on pressure vessels or boilers. Overall expansion effects differ from those on structures exposed to ambient temperature changes, in that the range of temperature variation on piping is much greater.

For many problems, the designer must consider more than one service condition, as well as start-up, shutdown, and emergency conditions; for example, a specific plant may involve more than one feed

2Frequently termed "extraneous" movement by piping designers.

allowed while a safety valve is blowing; for oil piping a 33~% increase is often used, compared to 10% on pressure vessels, except under exposure to external fire when 20% is allowed. This situation will probably be rectified when adequate rules for protection against overpressure are provided in the Piping Code.

The static effect of individual loadings forms only one phase of the broad subject of the design of piping systems. It is equally important to consider the duration, frequency, and manner of application of each loading, and their mutual occurrence. Both pressure and temperature stress, if applied in a sufficient number of repetitive applications, may result in fracture by fatigue. Failure may be accelerated by the dynamic influence of very sudden changes of pressure or temperature. Dynamic effects may also introduce the possibility of direct shock failure, apart from the brittle fractures associated with metallurgical considerations or ferritic steels at temperatures below the transition range. While failure due to corrosion or metallurgical changes is not a subject for this book, it should be mentioned that the level of stress in the piping or the occurrence of plastic flow may be a contributing factor in some cases. Failure by stress corrosion is an important example.

The loadings' which have been discussed can be segregated for design purposes into two categories: 1. Those representing the application of external forces which, if excessive, would cause failure independent of strain.

2. Those representing the application of a finite external or internal strain. These are generally introduced through temperature change.

The design consideration of individual loadings may be approached on the basis of the duration, frequency, nature, and probability of their occurrence. Individual loadings may be:

a. Present during extended normal operation but not during off-stream condition.

b. Maintained throughout the service life.

c. Occasional and of short duration as well as low cumulative duration (including start-up and shutdown conditions).

d. Emergency Or abnormal conditions of short duration.

For proper establishment of design assumptions, it is necessary to have an adequate appreciation of all direct and contingent requirements to which the piping system will be subjected, and also to understand the interrelations between the behavior of structures and materials, according to our present state of knowledge. It is the aim of this chapter to provide useful assistance toward the first objective.


stock or several alternate products which may require different processing pressure and temperatures. Many plants involve-highly inflammable, toxic, or otherwise unusual fluids, or specialized machinery and equipment which must be carefully isolated from air or contaminants. Start-up and shutdown may require protracted periods of warming up, cooling off, or operations such as purging, washing down, pickling or passivating, solvent cleaning, air-steam decoking, etc., each of which may introduce entirely different combinations of temperature and pressure over given portions of piping systerns. Temperature differences, or other loading more· severe than normal service conditions, may result where circumstances dictate that parts of a system be heated successively. A proper appreciation of these various possibilities requires an adequate knowledge of the process design, operation, instrumentation, and control of the connected equipment or entire plant. It is not unusual for start-up and shutdown procedure to be governed by mechanical design limitations rather than to suit process only.

For exhaust steam vacuum service, opinion differs as to whether the design temperature for thermal expansion effects should be based on the normal operating temperature under vacuum conditions plus an occasional rise to 212 F, which temperature would be approached with loss of vacuum, or on 212 F, as though it were the normal operating temperature. The first approach is consistent with the handling of other operating upsets, it being recognized that at reduced capacity or after lengthy periods of operation or with abnormally high cooling water temperatures, higher absolute pressures and corresponding temperatures may occur. It is therefore concluded that design considering the 212 F case as an abnormal short duration (not an operating) temperature is reasonably logical.

The Piping Code (ASA B31.1-1955) is deficient in adequate rules for protection against overpressure. The requirements of the ASME Boiler Code, Section VIII, for safety valves, etc., are a useful guide but require modification to suit common piping practice. Pipe wall thickness is generally established for a design pressure equal to the maximum (nonshock) service pressure, without provision for a margin between service and design pressure, and safety valves are generally set to relieve at about 10% above the design pressure. This is in contrast with pressure vessel practice, where at least one valve must be set to open at or below the design pressure. Differenoce.also exist in the maximum overpressure


pipe subject to bending, discussed in Chapter 3. Secondary stresses are not a source of direct failure in ductile materials upon single load application. If above the yield strength they merely effect local deformation which results in a redistribution of the loading and a reduction of the stress in the operating condition. If the applied loading is cyclic, however, they establish a local strain range corresponding essentially to their full original magnitude. They thus constitute a potential source of fatigue failure.

Localized stresses are those which die away rapidly within a short distance from their origin. Examples are the bending stresses in the hub of a flange, at a sharp cone-to-cylinder junction, or at the inside diameter of a branch connection. Localized bending stresses can be considered equivalent in significance to secondary stresses. It is possible in some cases for the plastic flow which may result from an initial overstress to alter the contour of the pipe to a stronger shape. This would lower the local strain range during subsequent applications of the loading and the fatigue resistance would be raised accordingly. Allowing large initial amounts of localized deformation carries the risk, however, of propagating flaws in the base material, particularly in welds, and of initiating cracks in less ductile heat-affected zones adjacent to welds.

The Pressure Vessel and Piping Codes contain tables of allowable stresses at various temperatures which are related only to the primary static-loading stresses (categories (a) and (b) of Section 2.2). The level of localized stresses at nozzles, branch connections, in heads, etc., is only loosely and indirectly controlled by formula and shape requirements and may easily be 100% or more above that of the primary circumferential pressure membrane stress. Due to the lack of adequate analyses or to the difficulty attendant to their evaluation, many secondary and localized stresses are neglected by the Codes, such as the bending stresses in vessel or pipe shells due to piping reactions, although the Code may warn the designer to consider such loadings,

Two criteria are associated with piping stresses.

One is the so-called "Code allowable stress" at the operating temperature, familiar to all designers of pressure equipment j the other one is the somewhat less known "allowable stress range," which is derived from Code allowable stresses and which has appeared in the Piping Code since 1942 as the basis for expansion and flexibility design. The application of each of these criteria is covered later in this section in connection with specific loadings.

The allowable stress is a function of the material


Chapter 1, together with the references cited, should prove valuable in establishing a reasonable and broad fundamental understanding ... of the flow and fracture of metallic materials.

2.3 Design Limits, Allowable Stresses. and Allowable Stress Ranges

In the preceding section of this chapter, piping system loadings have been grouped into two categories: external effects which, if excessive, might cause direct failure, and strain effects attendant to temperature change. Categories for individual loadings were also suggested, depending on the duration, frequency, and nature of the loading. This section is devoted to the discussion of the nature of stresses for the various forms of loading common to piping, as well as to a consideration of allowable stresses and an examination of the design limits which are not directly provided for by conventional allowable stresses and nominal safety factors.

When considering basic allowable stress values, it is appropriate to distinguish between primary, secondary, and localized stresses. Although there is probably no accepted definition of primary and secondary stresses in piping systems, the following criteria will be advanced for purposes of this discussion:

Primary stresses are the direct, shear, or bending stresses generated by the imposed loading which are necessary to satisfy the simple laws of equilibrium of internal and external forces and moments. Among the primary stresses due to external effects are the direct longitudinal and circumferential stresses due to internal pressure and the bending and torsional stresses due to dead load, snow and ice, wind, or earthquake. In addition there are the direct, bending, and torsional stresses due to restrained thermal loading, the external forces being supplied in this case by the line anchors or other restraints. In general the level of primary stresses directly measures the ability of a piping system to withstand the imposed loadings safely. Accordingly, those stresses due to sustained external loading (categories (a) and (b) of Section 2.2) are controlled to the Code allowable stress value for the operating temperature. Some overstress is allowed for temporary external loadings (categories (c) and (d».

Secondary stresses are usually of a bending nature, varying from positive to negative across the pipewall thickness and arising generally because of differential radial deflection of the pipe wall. A most important example of secondary stresses is that of the circumferential bending stresses in a curved


properties and safety factors as associated with specific design, fabrication, and inspection requirements. Experience with the pressure vessel Codes as presently constituted has shown that pressure and other maintained loading can be sustained by average equipment within this allowable stress limit for an indefinite period. Also, it is not uncommon to allow moderate short durations of overload or overtemperature due to abnormal or emergency circumstances. In a more precise approach, however, such overloads should properly be assessed on an integrated basis with respect to duration and frequency. In the following pages, the various considerations influencing the serviceability or safety of piping systems are summarized and augmented by current opinion as to advisable limits of stress, or other design criteria.

For Pressure Loading: In the 1952 ASME Boiler and Pressure Vessel Code, the basis for the allowable stresses for ferrous materials in both Section I, Power Boilers, and Section VIII, Unfired Pressure Vessels is given in Appendix P of Section VIII. This appendix is important as a general reference not only for its explanation of the basis of allowable stresses given in the Code but also for its guidance in setting stress values for similar materials. For nonferrous materials Appendix Q (Section VIII, Unfired Pressure Vessels) similarly establishes the basis of allowable stresses.

The allowable stresses for Section 1, Power Piping, of the ASA B31.1-1955 Code for Pressure Piping are identical with those of the ASME Power Boiler Code; those of Section 3, Oil Piping, within refinery limits, are in agreement in the creep range with Section VIII of the ASME Code. At lower temperatures, the safety factor on tensile strength is lower than that of the Unfired Pressure Vessel Code, allowable stresses being limited to one-third of the minimum tensile strength or 60% of the minimum yield strength. The other sections of the Code for Pressure Piping are intended for either ambient or relatively moderate temperature service, with allowable stresses in varying percentages of the yield strength Sv or tensile strength Su as indicated below.

Section 2. Gas and air piping: 0.6 to 0.72 Sv Section 3. Oil transmission lines outside refinery

limits: 0.85 s,

Section 4. District heating systems: 0.25 Su Section 5. Refrigeration piping systems: 0.25 S .. Section 8. Gas transmission and distribution pip-

ing systems: 0.72 SII max.

The assignment of higher allowable stresses for high

yield-strength materials operating at temperatures below the creep range, and recognition of yield strength enhanced by cold work and/or heat treatment, reduces the margin of safety provided by the Piping Code for unassessed stresses and for fatigue life under cyclic conditions. In addition, Sections 2 and 8 use nominal rather than minimum pipe-wall thickness, which further diminishes safety margins.

The dependence of fracture (and bursting) stress upon the shape of the part is quite properly recognized in Chapter 1. This effect, however, is one that is commonly ignored in ordinary design practice and in the Codes which represent such practice. Hence, the Code safety factors against bursting, related only to fracture of conventional tensile test specimens, must be regarded as nominal values which are not necessarily the actual safety factors for the bursting of a cylindrical vessel under pressure, or for any other general shape. While an exact evaluation of the disparity between safety factors for a tensile test specimen and those for a tube requires a complete knowledge of the plastic stress-strain properties of the material, a general evaluation for a wide range of materials is made possible by certain reasonable assumptions.

At first, the material under consideration is considered to obey the effective stress-strain relationship of eq. 1.8, stresses being dependent upon strains in accordance with the deformation theory of HenckyMises (eq. 1.7). Further, it is assumed that a function of the type


where ql = true stress in uniaxial tension

E*Z = logarithmic strain in uniaxial tension

Band n = assumed material constants,

can adequately describe the stress-strain curve in uniaxial tension. The types of stress-strain curves obtainable from eq. 2.1 through a variation of the constant n (sometimes referred to as the strainhardening exponent) are shown in Fig. 2.1.3

From the foregoing assumptions, it can be shown that the engineering (conventional) stress in a tensile bar, at the instant of attaining the maximum load, is given by

Su = B(n/c)n


where Su = ultimate (conventional) tensile stress e = 2.71828 = base of natural logarithms Band n are as previously defined.

aB, also called the "hardness factor," is simply the true stress value at a logarithmic axial strain of 1.0.



loga,ithm;c St,a;n, ft

I~IG. 2.1 Analytical representation of the tensile stress-strain curve for various values of n.

This instability stress value is identical with the conventional "ultimate tensile stress."

In a thin-walled cylindrical pressure vessel the conventional circumferential stress at instability (at the maximum sustainable pressure) can be expressed as


For a structure in uniaxial tension a design based on 11k of the ultimate tensile stress (as given by eq. 2.2) represents a true safety factor of k. For pressure vessels the safety factor should appropriately be applied to eq. 2.3. If, instead, safety factors are related to the ultimate tensile stress for pressure vessel design, then the quotient

Q = St/SO' = 1.155(0.577)n (2.4)

will indicate whether the real safety factor against bursting, on a single application of overpressure, is larger (Q > 1) or smaller (Q < 1) than the nominal or presumed value of k, i.e.,

(S.F.,· = Q X (S.F.tcnsion) (2.5)

A plot of eq. 2.4 in Fig. 2.2 gives values of Q for values of n ranging from 0 to 0.5 and shows that (for materials behaving as assumed) the safety factor for bursting of thin cylindrical vessels will be larger than the tensile safety factor when n is less than 0.263 and smaller when n exceeds this value.

In commonly encountered materials the strainhardening exponent n varies from about 0.05 to 0.15 for greatly cold-worked or tempered materials and is within the range of 0.2 to 0.45 for soft annealed metals. Carbon and low-alloy steels generally have n values from 0.15 to 0.25. Within this range Q has a value barely exceeding 1.0. Thus, if t of the ultimate tensile stress is used as a basis for design, an actual safety factor equal to or somewhat higher

than 4.0 on bursting will apply to cylindrical pressure vessels (of carbon or low-alloy steel), as proved by numerous static destruction tests. Similar comments apply to Codes using a different fraction of the ultimate tensile stress as a design basis. Thus, for the ASA B3l.1 Code, Section 3, which limits design stresses to t of the ultimate tensile stress, a safety factor of around 3.0 will be available against bursting of thin-walled cylinders. With other materials or with departures from the simple cylindrical tube, however, it would appear that the shape effect may bear investigation for more accurate assessment of bursting conditions.

In the creep range a similar safety factor does not exist. That is, if creep continues while the pressure is maintained, fracture will inevitably take place after a sufficiently long time. Hence, the design stress is selected to avoid failure within the service life period.

For the case where 100% of the extrapolated 105 hour creep fracture stress is allowed by the Code, and if this value governs the design stress (i.e. it is lower than the stress causing 1% creep extension in 105 hours), it would appear that the "life factor" (actual vs. desired life) may be no more than 1.0. In other words, if the desired design life is also 105 hours (about 11.4 years), fracture should follow when the design life is exhausted. Admittedly, there are only a few ferrous metals whose extrapolated stress value for creep fracture is less than the stress producing 1% creep in 105 hours. However, even for these metals, no case of fracture following intended life is known in the annals of the industry, although many pressure vessels have operated in the creep range for periods considerably exceeding 11.4 years.

One reason for this lies in the fact that the allowable long-time design stress values (for both creep and creep rupture) are obtained by extrapolation



Vol". of n









FIG. 2.2 The "safety factor ratio" Q as a function of n.


from short-time tests. Although not strictly admissible, this extrapolation generally leads to acceptable results for the creep values as shown in Fig. 2.3. On the other hand, in the very short-time creep rupture tests comparatively high stresses are used. As mentioned in Chapter 1, this tends to promote intracrystalline deformation, with an ensuing high ductility. At the longest commercial testing periods (generally 104 hours) the stresses are much lower; intracrystalline deformation is largely absent, and the ductility is considerably lower, although the stresses are still higher than those producing 1% elongation in the same time. The respective position of these stress values does not change even when the loading period is increased to 105 hours. However, the conventional log-log extrapolated value based upon test results up to 104 hours in duration may in some cases yield a fictitious rupture strength at 105 hours which is below the 1% creep stress value, as shown in Fig. 2.3. The unrealistic aspect of this extrapolation partially explains why pressure vessels do not fracture after 11-12 years even if extrapolated test data would predict this in cases where the creep fracture value governs design.

Structural Effects. The Piping Code rules ASA B3 1. 1-1955 require that primary stresses due to weight of pipe, fittings and valves, contained fluid and insulation, and other sustained external loadings be maintained within the hot allowable stress Sh. Occasional effects such as wind and earthquake should have little influence on the fatigue life of the piping system or creep at high temperature. Therefore, they can be treated more liberally, similar to AISC (American Institute of Steel Construction) practices, where 33i% higher stress is allowed for the separate effects of wind or earthquake superimposed on the basic loading.

In average piping systems, structural loading is not investigated in an overall fashion; instead it is controlled by standardized practices and details. In extreme cases of large or stiff piping it is advisable to evaluate the complete loading. Attention should be directed to those loadings which can occur simultaneously, so as to obtain an integrated equivalent cyclic strain as discussed in Section 2.6 and under "Temporary Loadings" in this section.

Structural instability or collapse of piping under longitudinal loading, such as is encountered in columns, is possible only under unusual circumstances. Collapse by circumferential buckling is more likely to occur, although the thickness-to-radius ratios ordinarily used in piping applications are usually high enough to prevent this. As a design


o Creep Fracture Te,., 0010 X Cre.p Role



Creep Sir ... cu~.... E.lrcpolcled

,., 1 % C'oop Ral. ~C'O.p-I'OdU'. Cu",.

~ _ r': Creep ... Iroctvre

--.......~_ Curve

.......... --

SI, ... '0' 1 % Creep Rete '~

in 10' hC<lr. ~ ---

E.I,opolalod SI, .. , '0' F,odu,. in 10' hours _!

(design :sofren_,)



Tim e, hOUr> (log. ,,01.)



FIG. 2.3 Comparison of extrapolated and actual creep .. fracture curves for a typical material at constant temperature.

criterion to guard against circumferential buckling, it is suggested that primary longitudinal compressive stresses should not be permitted to exceed 0.07 Ellr, where E is Young's modulus of elasticity, t is the wall thickness, and r is the radius.

The allowable stress range was suggested initially by Rossheim and Markl [11 as a measure of the permissible strain range in a cycle of load application to guard against the possibility of a fatigue failure after a given number of cycles. It is selected so that it will be applicable to ductile materials and to average commercial pipe surface conditions at the location of highest stress (strain) range. The principal cyclic loadings are restrained thermal expansion and pressure, although weight of contents and occasional effects such as wind and earthquake are also repetitive in nature. A cycle of external loading usually varies from the full presence of the loading during operation to its complete removal under offstream conditions; the distribution of the associated internal strain between the cold and hot ends of the cycle may on the other hand vary due to the dependence of strains on the material properties at each temperature and the presence of initial fabrication stresses or residual stresses set up as the result of plastic flow.

With the erection and completion of the final joint of each leg of a piping system, internal stress may be introduced by cold pull, weld shrinkage, or flange makeup. This establishes an initial state of stress, limited only by the yield point of the material. With temperature change on the first period of operation, expansion strain is superimposed on the residual fabrication strains. If the total exceeds the elastic limit at any point, yielding occurs, leading to relaxation of the initial fabrication stresses and a redistribution of the thermal strain. Prolonged elevated temperature will serve to further reduce the



~A' A2,
L ~
.. ,
t=l 12 :1 FIG. 2.4 Representation of bar for calculation of plastic strain concentration factor.

hot stresses by creep, at a rate proportional to the combined stress (expansion, pressure, weight, etc.). The reduction of the stress due to thermal strain loading by plastic flow or creep at the operating temperature is termed "relaxation." The relaxed strain reappears at the cold end of the temperature cycle with reversed sign.

For moderate-temperature piping, the division of thermal strain between the hot and cold condition is adjusted during the initial cycle in an amount dictated by the initial residual fabrication stress and the thermal-stress magnitude. The imposition of a temporary overload during operation can effect a further strain shift from the hot to the cold condition. For higher temperatures, where creep occurs, strain adjustment continues until the combined stress at the operating temperature is reduced to the relaxation limit. For convenience in design this is generally assumed to be the Code allowable stress level. Although such adjustment takes place, it is important to grasp the fact that the strain range per cycle does not change and that the ability of the pipe material to sustain the range is a function of both its hot and cold properties. The process wherein the pipe line seeks an equilibrium condition, and the resulting self-adjustment accomplished by yielding and creep, is termed "self-springing."

Self-adjustment may be minimized by prespringing (cold springing) which consists of incorporating prestress during erection. Since this practice is particularly useful in controlling initial reactions so as to protect connected equipment it will be discussed in that regard under the heading of Piping and Equipment Intereffects in Section 3.14.

As to whether prespringing offers advantages beyond controlling the initial hot reaction, a general answer cannot readily be given. In the 1942 edition of the Piping Code, the allowable stress range could in effect be increased when 50% or more prespring was provided by the permissible reduction in the expansion loading to two-thirds. The 1955 edition provides a uniform stress range regardless of the initial strain condition. This is based on the reasoning that fatigue life is primarily dependent on the range of strain which is unaffected by prestress, and

that the piping system seeks an equilibrium condition by self-springing. Credit for prespring is, however, still permitted when estimating maximum hot and cold reactions on terminal equipment. By prespringing, the plastic flow which the line may have to undergo on the first, or first few cycles, in order to effectively self-spring itself, can be avoided entirely or appreciably reduced. This is sometimes considered advantageous in minimizing the risk of an early failure due to "follow-up elasticity" effects should there be a highly localized weak link in the system. However, from a fatigue standpoint, no benefits arc attributed to cold springing once selfspring has been effected. The advantage of prespring in this respect is more important for piping which is to operate at temperatures in the creep range. The proposition has also been advanced that the hot plastic flow associated with self-springing will detract from the final available ductility under high temperature "creep" conditions; in reality, the mechanism of self-springing is probably more nearly akin to fabrication hot forming operations. In this light, the only clearcut conclusion that can be drawn is that prespringing can have only advantageous and no deleterious effects, especially as concerns initial terminal reactions. Therefore, it is a desirable practice when economically justified and effectively carried out.

The 1955 Piping Code rules call attention to the possibility of an undesirable amount of creep in areas of reduced strength, such as short runs of reduced size in highly stressed zones under certain conditions. The possibility of the unit strain in local highly stressed areas being magnified under conditions of plastic flow by reason of the follow-up elasticity of the more lowly stressed areas is not generally appreciated. In order to gain a better understanding, it is of interest to study a simple analogue consisting of a bar having a section of reduced area, as shown in Fig. 2.4, restrained at the ends and subjected to cyclic heating and cooling. The bar will be assumed to be made of an ideally elastic-plastic material (non-strai nhardening).

Let this bar now be subjected to cyclic heating and cooling of constant amplitude, to a level which causes plastic flow in member 1 on each cycle. It can be shown then that during any thermal halfcycle (from heating to cooling or vice versa), other than the first heating operation, the total (elastic plus plastic) unit strain in member 1 is given by





All valu .. col",l"ted lor A,/A}'" 0.5

Ee = elastic strain range limit

Bile s.;

=-+-. -

e, s,

e = unit linear thermal expansion for a temperature rise of !!.T.

L = total length.

Lb L2 = lengths of members 1 and 2.

Ab A2 = cross-sectional areas of members 1 and 2.

SlIe, 8lJh = yield strength at the cold and hot temperatures, respectively.

Ee, Eh = Young's modulus of elasticity at the cold and hot temperatures respectively.


Had this bar been analyzed on the assumption that all strains remained elastic, the calculated unit strain range in member 1 would be given by:




Ee = --~-

1 + AlL2


The strain given by eq. 2.6 is higher than that indicated by eq. 2.8, and the ratio of the two can be termed the strain magnification factor f3e which is given by the following equation, valid for Eo ;:::: Eo


f3c = 1 + AlL2 [1 - ~J


This is an extremely interesting result, since Eo is the maximum unit strain calculated by application of elastic theory and Ec is the maximum unit-strain range which the material can accept without allowing plastic flow on each cycle. So long as Ee does not exceed Eo there is no magnification factor. The magnification factor for Eo greater than Eo is given by eq. 2.9. Figure 2.5 is a plot of this equation for a specific ratio of Ad A2 = 0.5 and shows the magnification factor as a function of Ee/E. and L2/L1; high values can be reached which would materially reduce the fatigue life of such a bar. The magnification factors increase as area A. approaches area A2. At first thought this might be unexpected; the explanation is that, as Ad A2 approaches unity, the portion of the calculated strain in member 2 which is never developed, but instead causes plastic flow in member 1, increases as a direct function of A'; A:>.

From this simple analogue it can be generalized that, in any system which is stressed so that plastic flow occurs over a portion of the total length only, the unit strain is magnified in the portion undergoing



OL_-L __ i-~ __ ~~ __ ~~ __ ~~L--L __ LF~

1 2 3 " 5 6 7 8 9 10 11 12

Ratio 01 C"kvlaled Ela,li. Sirain Range to Avai!able [I",'i. Slr"I' per Hall Cyde

FIG. 2.5 Strain magnification in a locally weakened bar.

such flow by the follow-up elasticity of the more lowly stressed portion. It is not necessary that the area of the critical portion be less than the remainder. All that is necessary is that plastic flow occurs prefer-

. entially in the critical portion rather than over the rest of the system. Lower mechanical properties can have the same effect as reduced area. Systems stressed in bending are subject to this effect even when of uniform properties and size due to the nonuniform stress distribution which prevails. Strain magnification will occur whether the plastic flow is due to exceeding the elastic limit or is due to operation at high temperature where the plastic flow and strain magnification factor would be a function of time per cycle.

Similar conclusions were obtained in a recent paper by Robinson [2}. Analyzing a few selected piping systems operating at elevated temperatures (in the creep range), he found that severe strain concentrations can exist in layouts where the. maximum stress is limited to a very short length of the piping, and where the follow-up elasticity of the remainder of the system is great. These findings are in agreement with those of the previously presented analysis for strain concentrations under plastic flow conditions.

The allowable stress range limits established by the Piping Code are such that plastic flow due to expansion effects is not permitted to occur with each cycle. Both yielding and creep effects have been considered in basing the hot portion of the allowable range on the hot yield or creep strength, whichever governs. Repetitive strain magnification over substantial lengths of the piping should, therefore, not occur. For lines which are not presprung, it is, however, possible for some such strain magnification to occur during the initial operating period, while the


minimum of 7000 cycles of operation without failure. Local and secondary stresses are kept within this limit by the stress-intensification factors. For a number of cycles greater than 7000 the stress range is reduced by a factor relating the allowable stress range to the number of cycles as determined by ambient temperature fatigue tests on carbon-steel pipe. The reduction factor has a lower limit of 0.5. Some adjustment of these factors, particularly for materials other than carbon steel, will undoubtedly be necessary as further fatigue information is obtained.

The possibility of fatigue failure under the cyclic straining conditions present in piping systems has been questioned by many individuals. The propositions were variously advanced that the internal strain loading associated with thermal cycling cannot initiate fatigue cracks, or that the stress-relieving and annealing effects at elevated temperatures would prevent the propagation of such cracks. As indicated in Chapter 1, reasoning should lead to the opposite conclusion; furthermore, experimental verifieation that fatigue under constrained thermal loading does occur. is provided by the work of L. F. Coffin, Jr. l3, 4,], who demonstrated that fatigue failure is primarily associated with the range of cyclic plastic strain, while stress or strain relief is of a secondary order of influence.

The Code allowable stress range cited above assumes that longitudinal stresses due to pressure and other sustained external loadings are not over the basic hot allowable stress, Sh. For hot lines the expansion stresses at operating temperatures are assumed to be gradually lowered by yielding and creep, so as to be carried essentially as an off-stream or cold stress. If the longitudinal stress due to sustained loadings is less than Sh, the Code permits the unused portion to be applied to extend the stress range available for expansion effects. Therefore the Code, in effect, permits a total maximum allowable stress range equal to 1.25(80 + 8h), for thermal expansion stress combined with stresses from other sustained loadings. For service temperatures below the occurrence of significant creep, the total permissible longitudinal stress (both bending and direct) is equivalent to approximately 1.25 times the yield strength for power piping and 1.38 to 1.5 times the yield strength for oil piping.

In general, Code design is simplified for general use; at best it considers only average static conditions and establishes minimum design requirements, placing dependence on the safety factor to take care of unassessed stress conditions. The cyclic nature


line is undergoing self-springing. Since this occurs only once it must be considered in an entirely different light and would have no influence on fatigue life.

The bar analogue presented above was used to derive magnification factors assuming that the weak area was initially known and that an elastic analysis of stress conditions was made. The analogue could be readily modified to show the extremely high local magnification factor which would exist at a defect in a bar of uniform area, which is sufficiently serious to cause local plastic flow. It is well known that fatigue failure follows rapidly in the presence of such a defect.

The allowable stress range, as associated with the various types of repeated loading, is discussed in detail in the following treatment of specific loadings.

Expansion Stresses. Since thermal expansion occurs as a finite strain load associated predominantly with bending effects, fracture on initial application is unlikely to occur in ductile materials. Fractures resulting from repeated applications of thermal strain loading are similar to fatigue failure under mechanical loading. Therefore, the allowable stress or strain range must be related to the number of cycles anticipated during the life of the piping system. Failure will occur in the zone of highest cyclic strain, whether primary, localized, or secondary. For this reason it is necessary to apply stress intensification factors for any individual piping component wherever stresses above the level of the primary stresses are introduced. Due to the importance of such stresses from a fatigue standpoint, Chapter 3 is entirely devoted to recording present knowledge of stress intensification in various components of piping systems as well as their influence on flexibility.

Overall design is based on the stress range for the critical component, as established by its intensification factor and the nominal primary stress at its location."

The basic allowable stress range established for thermal expansion stresses in the 1955 Piping Code

1.258c + O.258h

where Be = allowable stress at ambient temperature Sh = allowable stress at operating temperature,

has been selected with the objective of providing a

~Since the pressure vessel codes do not provide rules for thermal expansion loading, it is desirable to check the effect of comparatively stiff piping on vessel shells of low thickness/ radius ratio. This is accomplished in the manner outlined in Chapter 3 for terminal connections.


Earthquake loading is not normally assumed in design unless it is specifically required for the locality concerned. Some consideration has been given to requiring that all structures be checked for some minimum lateral thrust of this type, lower than in recognized earthquake zones, but this is not the practice at present.

Gun Fire. Piping on warships is sometimes checked for the dynamic effect due to the firing of guns.

Water Hammer or Flow Surge Effects. The Piping Code contains water hammer allowances for cast iron pipe, in the form of a required increase in design pressure. On steel pipes no standard allowance is made for flow surge or hammer, and allowances are usually made only on high-head water flow lines, such as penstocks. The shock pressure due to sudden stopping of a liquid is a function of its velocity, stoppage time, and the elasticity of the pipe. Pressure surge effects are present wherever reciprocating pumps or compressors are used. The accompanying mechanical vibrations may in certain cases be sufficient to result in fatigue failure, if not promptly corrected. This subject is treated in more detail in Chapter 9.

Brittle Fracture in Ferritic Steel. The potential dangers of the brittle fracture of steel structures were made clear during W arid War II and after by the numerous failures of merchant ships, and by occasional partial or complete failures of bridges, pressure spheres, gas-transmission piping, and storage tanks. The phenomenon and conditions under which fracture may occur were discussed in Chapter 1. From the practical design standpoint it has been realized for a long time that, as ambient temperatures are reduced, the hazard of brittle fracture in ferritic steels is increased. As a result, the Pressure Vessel Codes have required for many years that for services below -20 F (excluding applications for service at prevailing ambient temperature, such as outdoor pressure storage tanks), ferritic materials have an impact value of at least 15 ft-Ib, at the lowest intended service temperature as determined by keyhole or U-notch Charpy specimens.

The numerous fractures of ships and other structures have resulted in extensive investigations for the causes underlying brittle behavior. While no complete practical remedy for avoidance of brittle fracture has resulted, several factors have been recognized to have important influence. Although individual impact or equivalent testing of each plate, bar, or tube at the lowest service temperature still provides .the best assurance as to its transition tern-


of loading and the possibility of fatigue failure are not specifically considered, except in the Piping Code's treatment of piping flexibility for thermal expansion. It might be asked why the fatigue design approach is currently limited to piping expansion analysis. This is due to the fact that the Unfired Pressure Vessel Code rules limit primary pressure stresses in f'erritic materials to 62~% of the yield stress and 25% of the tensile strength. This provides a reasonable margin against the possibility of fatigue due to localized and secondary stresses, which may be 100% or more above this allowable stress, for the type of cyclic conditions normally encountered in most pressure vessel services. By comparison, thermal strains playa greater role in the design of piping, which would be seriously affected economically (and would be virtually impractical in the case of large stiff systems) if total stress including expansion effects were to be held within the Code allowable stress at the operating temperature. Spurred by this necessity, experience and analytical work have led to the Piping Code's more advanced treatment of thermal strains, and to rules which recognize the influence of number of cycles, hot and cold material properties, and local stress intensifications.

It remains for the piping engineer and designer to recognize any unusual demands imposed by the design or service on piping systems. The following topics, in particular, are not at present adequately covered by the minimum Code design.

Shock or Dynamic Loading. Shock or dynamic loading conditions warrant special consideration because of the added stress which can be introduced by the rate of application of the motivating influence and the fact that the yield point of steel can be appreciably raised by very rapid loading. Localized yielding at points of stress concentration may be inhibited under such conditions and fracture more readily initiated, The general subject of vibrations which are a source of concern from a fatigue standpoint is treated in Chaptel' 9. The more significant dynamic loadings which enter into piping design can be listed as follows:

Earthquake. The accelerations associated with earth tremors are generally of the order of 1 to 8 ft/sec2• These values represent about 3% to 25% of the 32.2 ft/sec2 acceleration of gravity. For this reason, earthquake design is commonly approached by applying a horizontal force acting at the center of gravity of the structures; this force is 10% to 20% of the structure weight, depending on the maximum accelerations recorded for the locality considered.


joints. For low-temperature underground lines expansion provision is usually not necessary.

Temporary Loadings. An allowance of 33i% above the basic allowable hot stresses established for oil piping in the Piping Code has been suggested for temporary loadings due to wind or earthquake. Stresses due to occasional brief overloads in operation can be similarly treated i such might be occasioned by minor upsets in operating conditions or by starting-up or shutdown conditions. For power piping applications the ASME and ASA Codes specifically recognize occasional operating variations in pressure and temperature, allowing the following increase in the calculated stress due to internal pressure:

1. 15% during 10% of the operating period.

2. 20% during 1 % of the operating period.

This permissible overstress is intended to cover the surges expected to occur due to the heat lag of large boilers when the output is suddenly decreased. It is not recommended as a general design practice for normal operation variations in pressure or temperature as it is better to design for the maximum pressure and temperature conditions expected to occur in regular operation. However, brief temperature or pressure upsets may be treated on this basis, provided they are such as to require quick remedial adjustments in operation to restore normal conditions.

Severe upset or emergency loadings sometimes call for immediate drastic corrective measures and may require shutting down the unit. Wherever practicable the same limit as proposed for temporary loadings should be observed, but the nature and probability of the emergency often requires special consideration. In the case of piping where design is controlled by creep and stress-rupture properties. analysis of the ability of the system to sustain an occasional short duration emergency can be based on the short-time properties of the material or, if more frequent, on the permissible creep stresses for the shorter time period involved, by evaluation of the cumulative creep for service and unusual conditions. No standard guide can be given. More study and tests are desirable to assess the cumulative effect of short-duration high overloads and long-duration normal loads. It is known that, for a given total period of overloading, the number of times the loading is applied has a significant effect, being more damaging as the frequency of application increases for a constant total duration of the overload.

Where basic allowable stresses are set higher or are established by cold-worked properties (e.g, gas


perature, there is definite evidence that average transition temperatures are lowered and the incidence of failures significantly reduced, within the range of ambient temperatures, by using openhearth or electric-furnace steels, controlling the manganese-carbon ratio of plates over ! in. thick, and by employing killed steels made to fine-grain practice, particularly for thicknesses over 1 in. (see ASTM Spec. A131-53T for example). Normalizing is also desirable for important plate applications over 1 in. although none of the ASTM Specifications for structural steel at present requires this in any thickness; however, ASTM Specification A131-53T in paragraph 4 (b) mentions that plates over 1 i in. may be required to be produced to special specifications. The ASTM Specification A373-54T covers structural steel for welding and is similar to A131 except that it makes no reference to fine-grain practice for plates over 1 in, or to special requirements over 1 i in. The development of these specifications and their gradually more widespread use in the construction of ships, tankage, and other structures at insignificant increase in cost is an encouraging trend. Though it represents only a modest start it indicates that much more could be accomplished by economic steel specification control and that its extension to all pressure services is a necessary undertaking.

The experimental work also showed that a significant improvement in performance can be achieved through careful design by the avoidance of high stress concentration or areas of high local restraint (e.g., ship hatch corner design). Significantly, all such failures have been triggered off by a relatively minor flaw or notch, the majority of which were associated with welds. Apparently, in addition to the possibility of welds containing small cracks, the local residual stress pattern associated with them is a factor. The latter plays a significant role, not only in initiating crack propagation, but in accelerating the crack propagation speed to a level where it can continue as a spontaneous process through a much lower stress field. This is in keeping with the theory given in Chapter 1.

Non-ductile Materials. Cast iron and other non-ductile materials are usually confined to relatively low temperature service when used for pressure parts. Bending stresses for these materials must be kept within well-defined allowable values (for cast iron, usually I! times the allowable stress for tension). Bell-and-spigot or packed joints of a design incapable of taking longitudinal stress are provided with anchors at the end of each run, with expansion absorbed by movement at the packed


the cyclic or fatigue life under thermal expansion is taken into account through so-called stress intensification factors. The following discussion presents background information and comments to aid understanding of the current approach in treating various

loadings. .

2.4« Internal Pressure up to 3000 psi Maximum. In their present status, the Pressure Vessel Codes already mentioned are stated to be applicable when the pressure does not exceed 3000 psi. Pressures above this may require special attention to design and fabrication details, closures, branch connections, etc., in view of the heavier wall and thickness/diameter ratio involved. Actually, any such limit is strictly arbitrary and should more properly be established as a pressure/stress limit so that the influence of different materials and the effect of temperature would be included.

For the most common surface of revolution; the cylinder, the so-called inside diameter (or membrane) and outside diameter (or Barlow) formulas were first used for thickness/diameter below and above 0.1 respectively. These were later supplanted by the mean diameter formula and, more recently, by the universally adopted formula approximating the results of the Lame formula. All these formulas may be expressed in a common manner as follows:


S = (prt!t) + Kp

(2.11 )

transmission line piping), overstress due to temporary loading should be avoided.

Abnormal temperature differences may occur due to upsets or during start-up operations, which can cause thermal expansion stresses higher than assumed for the normal design condition. When infrequent compared to the normal design condition, some increase in the permissible stress range can be justified. For example, when working to the rules of the 1951 ASA Piping Code, The M. W. Kellogg Company designed for emergency thermal expansion conditions using a 50% increase over the basic allowable stress range. A more appropriate design approach would be one which would determine the number of cycles at the Code allowable stress range which would be equivalent to the number of cycles under the diverse conditions actually anticipated. Assuming a basic relation between number of cycles Nand stress range SR of the form

the equivalent number of cycles N. at a stress SA can be established roughly as

(SI)" (S2)n (SD)n

No = SA Nt + SA N2 + . .. SA N D


where K and n are constants for the material.

Nl is the number of cycles producing an overload stress SI.

N2 is the number of cycles producing an overload stress S2, etc.

N D is the number of expected operating cycles on the normal design basis.

SD is the corresponding calculated stress.

SA is the Code allowable stress range for

7000 cycles.

Since the Code stress range is intended to provide for a minimum of 7000 cycles at a stress SA, if No does not exceed 7000, the design may be considered equivalent to a Code design. Tests on carbon-steel pipes [5] indicated that n can be taken equal to 5. Without similar test data, the use of n = 5 for other materials is open to some question.

2.4 Stress Evaluation

Stress evaluation is commonly limited to primary direct, bending, and torsional stresses which, in piping, result from the effect of pressure, weight, and thermal expansion. Localized and secondary stresses which do not affect the overall system are not ordinarily evaluated directly although their influence on

where p = internal pressure. r i = inside radius.

t = wall thickness.

K = constant having values between 0 and 1.

If K is given the value of 0, the inside diameter formula is obtained; for J( == 0.5, the mean diameter; for K = 1.0, the outside diameter. When the value of 0.6 is used, stresses are obtained which correlate reasonably well for values of t up to about 0.5r; with the recognized inside circumferential stress formula of Lame. This approximation, discovered by H. C. Boardman, was rapidly adopted for moderatetemperature piping by both Pressure Vessel and Piping Codes, while for piping in the creep range it is considered applicable if a further adjustment of K is made as covered later in this section. Similar relationships, which approximate the direct circumferential pressure stress at the inner-wall surface for other shapes of revolution, are presented in Table 2.1. For dished heads it may be noted that the Code also relates the design of torispherical and ellipsoidal heads to the sphere formula, which is suitably modified by a correction factor to correspond with the



Table 2.1 Internal Pressure-Circumferential Stress Formulas for Elastic Conditions




SE - 0.6p


p SEt

T; + 0.6i


E- [T; + 0.6tl t


Use (_!2_) in place of Ti in the cylinder formulas.


(1 - 2~) Tj + Kt

where Ti = inside radius (use meridional radius in general formula, i.e., radius from axis of revolution and normal to surface, see Fig. 2.6).

E = weld joint efficiency.

R = torus center line bend radius.

R; = actual radius of curvature in meridional plane at the point in question (positive if concave to pressure)

(Fig. 2.6).

a "'" ~ cone included (apex) angle.

K = 0.6 C + ~T;/R) (use absolute value). S = circumferential stress.

p = internal pressure.



Torus (pressure inside l]

2SE - 0.2p

prj [R - 0.5T'J

SE - pK R - T;

General shape of revolution]

"Not covered by Piping Code at present. tNot given in any code at present.

membrane stresses associated with their contour. The pressure design of shell openings for nozzles, manholes, and branch connections is based on the simple maintenance of the original cross-sectional area, by replacement of the removed metal by reinforcement immediately adjacent to the weakened area. Flanges and cover plates involve primarily bending stresses; the direct stresses in these components are commonly neglected due to their lesser magnitude. Specific formulas are given in the Codes for their pressure design.

2.4b Internal Pressure over 3000 psi. The Codes at present (1955) do not cover the design of high-pressure vessels, although this subject has received considerable attention in the last two decades. Many problems arise at high pressure for which conventional code details arc either totally unsuited or present an undesirable choice. Examples are: nozzle reinforcements which, within Code limits for reinforcement, entail extremely abrupt changes in section, cones, etc., involving inside corner radii which are small in comparison with the wall thickness. As the pressure is increased, practical limits are reached for design as covered by Code rules. In the following it is attempted to summarize the practices which


2~t [r; + 0.2tJ

E... [(R - 0.5r,) Ti + Kt]

Et R -"i

T; + 0.2t SEt

(R - 0.5ri) K

R T;+ t - Ti


1?_ [(1 - ~) r. + KtJ

Et 2 RI '

FIG. 2.6 The meridional radius of curvature for shells of revolution.

are followed in the design of shells, heads, closures, and connections of high-pressure piping.

The Lame formula and the Rankine (Maximum Principal Stress) criterion, on which the ASME Boiler Code and ASA Code for Pressure Piping are based, no longer predict general yielding or rupture within reasonable limits when the thickness/diameter ratio exceeds approximately 0.20. Although the error is on the safe side, the deviation becomes greater the more the thickness/diameter ratio is increased. For initiation of yielding the Maximum Shear or Maximum Shear-Strain Energy Theories are in good agreement with experimental evidence,

Circumfcrflntiat Sire ..


lomite 1 laro





Axial I-I-I--+--l--l Streu

FIG. 2.7 Typical stress variation in a pipe under elastic or creep conditions.

For a severely cold-working material the assumption that the strain is the sum of an elastic strain obeying Hooke's law and a plastic strain can be considerably in error. Special analyses have also been worked out. for strain-hardening materials.

Plasticity analyses are generally based on the assumptions that (1) elastic strains are negligible in comparison with plastic strains; (2) the volume of the material remains constant during deformation; and (3) the length of the pipe is unchanged under the application of pressure. The distribution of circumferential stresses changes completely from the elastic results, the maximum in the plastic range occurring at the outside fiber. The shear stress also tends to be constant through the wall thickness, but. remains a maximum at the inner fiber. Figure 2.7 illustrates the difference in stress distribution. For a thick-walled cylinder of an ideally plastic (non-workhardening) material, Nadai {5} gives the following formulas at the onset of general yielding:

Se::: = p[1 - log. (ro/r)} log. (r o/r i)

-p[loge (ro/r)]

Sr::: = ---='-'-..:::.;;__;;_::.:...._.:_::

log. (ro/r,)




2S. :::: Se::: - SrI = 1 (/) (independent of r)

og, To r,


as mentioned in Chapter 1. Either of these theories may be used to practical advantage as general yielding or bursting criteria when applietl in conjunction with plastic stress analysis.

For thick cylinders, yielding of the inside fibers leads to compressive residual stresses in the plastically deformed portion of the wall when pressure is removed, increased stress in the outer fibers under pressure loading, and greater uniformity of shear stresses throughout the wall thickness. This redistribution of stresses due to plastic flow is termed "auto-frettage"; it was first employed for casting guns in the early nineteenth century. Later, greater control and uniformity of stress distribution was attained by shrinking successive closely machined shell layers on to each other, thus producing a thickwalled cylinder, whose inner layers are in a state of precom pression.

The fact that initial yielding of the inner fibers occurs at only a fraction of the pressure corresponding to general yielding distinguishes thick-walled vessels from thin-walled shells. Since the pressure to produce failure in thick-walled vessels is more properly associated with plastic rather than elastic criteria, a valid design of these structures can be based on plastic analyses, and related to the general yielding and bursting conditions. The various approaches which have been suggested are discussed in the following paragraphs.

M edified Elasticity. This approximate solution assumes that a safety factor of 4 on bursting is maintained so long as yielding of the inside fibers is avoided at the design pressure. This approach also requires that the stress at the mean wall thickness, as calculated by the Lame formula, does not exceed the usual allowable (0.258,,) value. The safety factor assumed by this analysis is likely to be in error on the unsafe side.

A uto-freUage. The wall is assumed to be in two layers with the inner layer taken to be in a state of precompression, attained by applying a suitable overpressure and yielding the inner fibers. The stress is then calculated by the Lame formula considering the initial prestressed condition. The results will be similar to the preceding approximate approach for the same safety factor.

Partial and Complete Plasticity. Stress analyses of cylinders having an inner plastic-elastic zone and an outer elastic zone are available in many text books dealing with plasticity. These solutions are generally based on the assumption of an idealized material which is elastic up to the yield stress and plastic (non-work-hardening) at the yield value.



where Scx = circumferential stress at any radius r.

Sr7: = radial stress at any point T. B. = shear stress.....

r 0 = outside radius.

r, = inside, radius.

r = radius at point in question.

The value of 28. is equal to Be at the outside radius. If this is accepted as a suitable criterion of general yielding or bursting, it is interesting to know that eq. 2.14 can be closely approximated by the simple mean diameter formula.

Spurred on by an interesting paper by Burrows and Buxton [71 on available formulas for cylinders under internal pressure, the ASA B31 Committee appointed a special task group to study the subject and recommend a simple appropriate formula for the design of heavy-walled piping in the creep range. This task group recommended that the value of K in the simple formula of eq. 2.11 be gradually modified from 0.6 to 0.3 at temperatures over 900 F for ferritic steels and over 1050 F for austenitic steel. This recommendation was approved and the formulas for piping in the ASA Piping Code, Sections 1 and 3, and the ASME Power Boiler Code now include this provision.

The formulas given in eqs, 2.12 to 2.14 will provide a reasonably good answer for the behavior of thickwalled cylinders made of materials with only a mild strain-hardening tendency. Where a more exact evaluation of probable performance is desired,' the stress distribution should be evaluated from the actual stress-strain characteristics of the material [8,9, 10]. An analysis of thick-walled cylinders under internal pressure in the creep range has also been advanced by Bailey [11].

Concerning the practical design details of thick shells, an effort should be made to avoid stress raisers in the form of abrupt changes of section at the location of openings, nozzles, and intersections. The observance of these rules, coupled with careful control of materials and fabrication, and with adequate testing, may permit a reduction in the overall nominal safety factor without diminishing (and possibly improving) the real safety factor. With the trend to higher pressures and temperatures, more adequate use of material is imperative. Lower safety factors for simple surfaces of revolution or for construction of controlled low stress intensification is also necessary [12J.

2.4c External Pressures. External pressure loading involves, in addition to control of direct stresses, the consideration of stability. Direct

stresses for external pressures are governed by the same formulas as for internal pressure, except that the signs of all of the equations containing the pressure p have to be reversed, indicating compression stress.

Stability of cylinders -against collapse is well covered by the rules of the ASME Boiler Code, Section VIII, which provide for the design of both unstiffened and stiffened cylinders of all Code materials. For an explanation of the Code charts, reference should be made to a paper by E. O. Bergman [13]. This paper also contains an extensive bibliography on this subject. Similar to columns, the limiting compressive load which a cylinder will sustain is related to its equivalent slenderness, end conditions, and deviations from true contour. In the case of long unstiffened cylinders (length/ diameter over about 10), the collapsed contour approximately follows a figure 8 outline, consisting of two complete lobes. Consequently, an unstiffened cylinder may be compared with a fixed-end column whose length equals one-half of its circumference. For stiffened cylinders, the number of lobes increases as the length-between-stiffeners/diameter is decreased, with a corresponding increase in collapse pressure. The Code design of a stiffened shell establishes a shell thickness and combined moment of inertia for the stiffener and shell to assure the stability of the entire shell section. This results in heavier stiffeners than would be obtained by a design approach wherein the stiffener loading is based on division of load between the connected shell and stiffener under pressure, and the elastic conditions up to the point of collapse. The collapsing pressure of heads (which in early Code editions involved a flat reduction in allowable external pressure to 60% of that allowed for internal pressure) is now predicated on the collapse pressure of a complete sphere having a radius equal to that of the spherical part of the head.

The ASME rules attempt to maintain the same nominal safety factor of 4 against collapse under external pressure as is used against bursting under internal pressure. There is some reason to question whether this is entirely logical, since the effect of localized stresses or stress concentrations, such as at branch connections, may be entirely different. Also, the degree of hazard in the event of failure will generally be appreciably less for external pressure, although hazard must still be judged independently for individual applications. In addition, the Code rules maintain the same safety factor for failure by elastic instability as for failure by plastic yielding,

indirectly controlled in a standardized way (e.g., support standards) or individually estimated and controlled so that the sum of all effects will approximately meet the same combined stress criterion. For large-diameter or otherwise stiff piping systems, particularly where expensive materials are involved or where the increased space for additional flexibility would require enlarged buildings or other considerable expense, every contribution to the overall strain should be evaluated by a simultaneous solution.

Weight effects are conveniently minimized by the provision of adequate supports. Where such supports just balance the weight reaction, they can be validly ignored in the expansion analysis. This condition is seldom achieved even with elaborate compensating spring hangers. However, average piping is sufficiently stiff so that the local restriction due to some support friction or unbalance is not a serious factor. For separate estimation, conventional column and beam analysis of individual critical. members, or frame analysis of combinations of members is recommended. Wind and dynamic effects can be similarly treated. Unbalanced pressure effects are resisted wherever possible by rigid stops or ties which are taken into account in the flexibility analysis, unless such provisions would adversely affect the behavior of the line. In the latter case a careful analysis may be made to determine whether the pipe itself can be designed to carry the loads. If not, the unbalanced pressure effect must be handled by special design arrangements.

2.5 Combination of Stress: Stress Intensification and Flexibility Factors

The 1955 Piping Code rules for flexibility contain the following equation for the combination of stresses due to thermal expansion:



except for small tubes where a variable lower safety factor is recognized. The practice of the Structural Steel Codes in reducing the safe't"y factor on columns as the length/radius-of-gyration is reduced appears logical. For vessels or pipes a similar practice could be followed by lowering the safety factor to 2 on the yield point as a suitable function of diameter/thickness, but this practice is not yet recognized.

2.M Expansion. The evaluation of external reactions at terminal points and intermediate restraints of piping systems is given in detail in Chapters 4 and 5. The expansion forces in space systems will generally result in 3 force and 3 bendingmoment components at each terminal point. The number of such components is reduced with partial end fixation.

The evaluation of the terminal reactions permits the calculation of the three moments (2 bending and 1 torsional) at any point in the pipe line by the application of statics. These moments, in turn, permit the designer to calculate the stresses by utilizing the section moduli of the pipe. The contribution of direct forces for the expansion stresses in piping systems is generally insignificant, unless the piping layout is extremely stiff.

For simplicity the Piping Code provides that expansion stresses be calculated with the cold (ambient temperature) modulus of elasticity. The design values of Poisson's Ratio and the torsional modulus for expansion stresses likewise refer to this temperature. The Code also provides thermal expansion data for evaluating the change in length over any temperature range. This use of roomtemperature data avoids the necessity of using elevated-temperature properties, which may be less accurately determined. With the principal strain generally present at atmospheric temperature due to pre- or self-springing, the Code practice of using the "cold" values of mechanical properties is entirely sound.

2.4e Other Loading. Other loading which may act on piping systems includes: the weight loads of the piping, including structural members; the weight of the insulation, and contents; snow and ice loading; wind loading if exposed; loading due to acceleration imparted by earth tremors; special shock loading, such as gun fire or moving vehicles; and unbalanced static pressure or flow effects.

It is possible to include any or all of these loads in a complete solution, following the methods of Chapter 5. Ordinarily, these effects are not sufficiently critical to warrant the extra engineering cost of this more precise approach. Instead, they are

where SE = equivalent stress to be compared with the allowable thermal expansion stress range, psi.

Sb = resultant longitudinal bending stress psi = f3il'h/Z.

SI = resultant torsional shear stress psi =


M b = resultant bending moment, lb-in, M, = resultant torsional moment, lb-in, Z = section modulus of pipe, in.3

{3 <= stress intensification factor.

This equation is based on the Maximum Shear Theory and for convenient comparison with Code

, ~"~"~~~~- ... -------------


Then the resultant principal stresses at the outside fiber can be written as


allowable stress range, eq. 2.15 represents two times the maximum shear stress due to expansion loading. As stated in Section 2.3, the Pipitlg Code establishes a separate limit of Sh for the maximum longitudinal stress due to pressure, weight, and other external sustained loadings, with the provision that, if such loadings do not add up to Sh, the difference may be used to increase the allowable stress range for expansion effects. This approach has been adopted for convenience in practical design calculations. It is obvious that, when using combined-stress formulas and a specific yield criterion, stresses from all loadings should be included to determine the principal stresses before combining them. On the other hand, from a fatigue failure standpoint, the loadings which cause cyclic stresses are the most significant. There is, therefore, reasonable logic in combining these separately for comparison with an allowable stress range. Actually, so long as the allowable stress range is adjusted to suit the methods of calculation and stress combination which will be used, designs arrived at by various approaches can be made substantially the same. Simplicity of application has been the objective of the Code.

The Code's use of the maximum shear-stress criterion for expansion stresses represents a departure from the evaluation of stresses elsewhere in the Code, where only principal stresses are considered. While a uniform criterion would be preferable to avoid confusion and permit better assessment of safety factors, there is greater need for closer evaluation of cyclic strain loadings which may lead to a fatigue failure.

The approach laid down above is recommended for ordinary practice, in view of the mandatory requirements of the Code and the relative simplicity of handling expansion stresses separately. For critical applications, or where loadings are simultaneously analyzed, it is more appropriate to evaluate all stresses prior to combining them and compare them to the total allowable stress range 1.25 (Sc + Sh).

The additional provision that the principal stress due to long-time sustained loadings other than expansion should not exceed Sh, must also be observed.

For convenient reference the following formulas are given:

Let SL =: maximum longitudinal stress due to pressure, weight, and other sustained loading plus expansion stress Sh as defined above.

Sp = circumferential pressure stress.

S, = shear stress due to torsion as previously defined.

Sl = 0.5[SL + Sp + V4St2 + (SL - Sp)2)

S2 = 0.5[SL + Sp - V4S? + (SL - Sp)2) (2.16) S3 = 0

and the combined "equivalent" stress for the respective yield condition becomes

Maximum Shear Theory (Tresca)

The greater of SI as given above or


Distortion-Energy Theory (Mises) v3Sl + Sr} + Sp2 - SLSp


Use of the maximum shear theory is favored for consistency with the Piping Code.

In the sample calculations in Chapters 4 and 5 the Piping Code rules are followed. The examples in Chapter 4 involve expansion alone; in Chapter 5, Sample Calculations 5.14, 5.15, and 5.16 include weight or wind effects.

In the General Analytical Method, the influence of localized effects on deflections and rotations is provided for by the inclusion of flexibility factors with the shape constants. In effect, this compensates for the additional displacements by providing an increase of the length of the member to a so-called virtual length, producing the desired relative deflection. The net influence of this increased flexibility is to decrease reactions and nominal primary stresses.

This greater flexibility of local components, such as bends, is the result of localized stresses whose magnitude above the nominal primary stress level is expressed by a stress-intensification factor, whose use is mandatory in the new Piping Code rules. These rules contain suggested flexibility and stress factors for usual piping components, with the provision of allowing the alternate use of experimentally determined factors.

2.6 Evaluation of Deflections and Reactions

Line movements or deflections are of interest in the design of yielding supports, such as spring hangers, and in establishing clearances for the free expansion movement of a large-diameter or complex line. Sample Calculation 5.10 in Chapter 5 illustrates that the evaluation of deflections by the Kellogg General Analytical Method requires little extra effort after

whichever is greater, and with the further condition that


s, s.,

-S X -E IS less than 1

E 'h

where C = cold spring factor varying from 0 for no cold spring to 1 for 100% cold spring.

SE = maximum computed equivalent expansion stress (per eq. 2.15).

E; = modulus of elasticity in the cold condition.

Eh = modulus of elasticity in the hot condition. R; = range of reactions corresponding to the full expansion range based on Ec.

Rc and Rh represent the maximum reactions estimated to occur in the cold and hot conditions, respectively.

Obviously, the Code formulas for reactions, based upon a division of strains between the ambient and service temperatures, are somewhat arbitrary. Equation 2.19 attempts only to establish the initial magnitude of the hot reaction for purposes of checking the capacity of equipment to take such effects. Equations 2.20 and 2.21, in turn, are aimed at establishing the maximum value of cold reactions, either as obtained through initial cold springing, or due to subsequent self-springing under service conditions. The signs (directions) of the hot and cold reactions are always opposed to each other. For temperatures in the creep range, the hot reaction will eventually be lowered to a value roughly corresponding to the design allowable creep stress Sh. This value approxi-


mately corresponds to Rh = S RT, whereas the cold


reaction increases to the value given by eq. 2.2l.

Equations 2.19 and 2.20 are applicable to a multiplane system only when the prespring is applied as a uniform percentage in each direction. In practice there may be instances where prespringing in a preferred direction only may be sufficient and be utilized because it is simpler to carry out. For such a case the reactions for the actual prespring to be applied should be calculated by an appropriate analytical method in place of eq. 2.20. For the most complete control of prespring, an analysis of the type shown in Sample Calculation 5.13 is recommended.

When prespring is not specified, or is not adequately controlled, the reactions due to fabrication may in exceptional cases correspond to yield-point stress in the system, unless thermal unloading has been employed. Fabrication residual strains will be reduced when the piping system is first heated if the combined

the reactions have been determined. Line movements at any point are also readily determined by Model Test for any condition of ld'liding, as discussed in Chapter 6.

It must be appreciated that calculated deflections establish only a range of movement; the absolute position of any point at a given time is, in addition, dependent upon the combined effects of initial fabrication stress, relaxation and creep, changes in dead load, adjustment of hangers, and local temperature differences at the cross section. Except for temporary overload of terminal equipment, etc., a line may be adjusted to any desired initial position so that the movement range occurs over the desired location. Equipment may be protected against erection overload by thermal unloading (controlled local stress relief) as discussed in Chapter 3.

Since maintained loads, such as piping weight and insulation, are essentially constant, deflection calculations are ordinarily confined to expansion effects. In general, the effect of maintained loads (such as piping weight and insulation) and transient loads (such as contents, snow, and wind loads) are effectively limited by properly placed and designed supports, guides, or ties. The significant movements will then be associated only with thermal expansion, and deflection calculations can be confined to this effect.

The calculations in Chapters 4 and 5 and the model tests in Chapter 6 give, as their first result, the reactions of the supports on the piping system. These forces and moments are determined on the basis of a strain equivalent to the total expansion and using the modulus of elasticity and Poisson's ratio at atmospheric temperature. They do not include the influence of initial stresses due to fabrication. The resulting reaction range will be immediately realized in its full magnitude only for piping systems subjected to 100% cold spring. Beyond this consideration, it is important to know the maximum reactions to be expected in the hot and cold conditions for the purpose of examining their effect on terminal equipment. The Piping Code provides the following rules on this subject:


R, = CRr or



The value of R; is taken from equations 2.20 or 2.21,



expansion and residual stresses exceed the yield strength. The fabrication strain so relieved is not reestablished during subsequent=service, nor will it affect the fatigue life of the piping system. Its significance lies largely in the load which it introduces on equipment or foundations as long as it lasts.

The individual hot and cold reaction values are of interest mainly for judging their effect on sensitive equipment, such as pumps and turbines which involve maintaining close clearances and alignment; they are also of interest in connection with foundation design. In regard to localized stresses in the shell of terminal equipment, however, the reaction range rather than the magnitude of the individual hot and cold reactions is the significant factor. This aspect is dicusssed in more detail in Chapter 3.

2.7 Design Significance of Inspection and Tests

Wall thickness calculations, when dealing with pipe or a cylindrical shell, have always included a so-called "j oint efficiency" for welded seams which has usually been applied only to circumferential pressure stresses. For structural loading the joint efficiency is sometimes neglected. It is also generally disregarded in flexibility calculations or compressive loading and most situations where only bending stress is involved.

The term "joint efficiency" is a holdover from riveted construction, where a definite breaking strength could be associated with a specific design. On welded joints, where weakening effects such as rivet holes are absent, it is not difficult to provide design strength equal to the base material as evidenced by procedure tests of sample welds, even for lapped joints. Better criteria of the reliability and performance of a welded joint are its capacity to take deformation as a measure of its safety against cracking, the absence of weakening defects as assured by examination, pressure tests, and mechanical tests of occasional complete joints or specimens; for hightemperature service the tests should be carried out both at room and service temperature, but this is not current practice except for special applications. Assessment of performance tests and degree of examination would lead to establishment of a "quality factor," rather than a "joint efficiency." Where repetitive loading is involved, the potential influence of the design details, fabrication quality, and basic structure of the weld- and heat-affected zone of the parent metal can apparently be accurately evaluated only by full-scale fatigue tests under combined loadings and temperature cycling. However,

scale tests and the use of small specimens have proven valuable in investigations of certain aspects of this problem, particularly for establishing general trends or for the quality control of procedures and actual fabrication.

The level of quality of design, materials, and fabrication attained, as assured by adequate inspection and tests, is at maximum economic effectiveness when the individual factors are controlled to the same degree [12]. Overemphasis on any aspect does not ordinarily lessen the hazards attendant to the neglected factors, so that the probability of failure is not proportionately reduced.

There is some opinion that pipe girth joints are less critical than longitudinal welds. This view stems from the fact that longitudinal pressure stresses are approximately only half the circumferential stress. It ignores the fact that expansion and structural effects usually make longitudinal stresses the criterion of design, and that weakness in a longitudinal direction causes a local weakness circumferentially. In addition, initial flaws, in propagating, tend to change orientation for maximum influence from the maximum stresses present.

Adequate pressure testing, as practiced on pressure vessels, often presents economic problems in piping. Shop tests of irregular or large-diameter runs require special fittihg~ or else extra welds for closures. Adequate field. pressure tests require the installation of blinds and often extra flanged joints, in order to protect lower pressure vessels and terminal equipment; they sometimes require special rigging for inspection access, and temporary supports. Such cases require individual treatment. When the field test is adequate the shop test can be waived by mutual agreement. In judging the adequacy of the field test, the degree of inspection and level of test stress should be jointly considered.

A water- or liquid-pressure test fulfills dual functions. The design, materials, and fabrication are checked to a reasonable minimum extent by a pressure test based on 1.5 times the design pressure increased by the ratio of cold to hot allowable stresses (SclSh)' During testing, there is an opportunity to detect leaks due to cracks, porosity, or other flaws which extend through the wall. These objectives are accomplished at minimum hazard when the testing fluid is essentially incompressible, thereby limiting the stored energy. During such a test, should a break initiate, there is immediate loss of pressure, usually before extensive damage is done or fragments detached and propelled through space.


The detection of leaks can be accomplished with equal or greater effectiveness at lower pressures by using liquids of lower surface tension properties, or by reducing the surface tension by additives, or by the use of air or other gas. Air pressures of 5 to 10 psi usually suffice for the detection of leaks with equal or better effectiveness than water at full test pressure. The Vessel Codes permit air tests as a substitute for water tests at a reduced stress level (83.3% of hydro test for ASME Code and 73% for . \PI-ASME Code), while Section 3 of the Piping Code limits air tests to 50 psi. The Vessel Codes require that pressure be applied in successive stages to minimize high-energy rupture hazards. The precautions exercised should be in step with the size, volume, stored energy, test stress level, and quality of inspection.

The effectiveness of a test in proving the soundness of a structure decreases rapidly as the pressure recedes from l~ times the equivalent cold working pressure. It is doubtful that air tests, at the level prescribed in the Codes, accomplish much in this direction other than detection of gross omissions or deficiencies. These, for the most part, should have been revealed by careful visual inspection. In addition, one application of pressure at or near normal design stress will often not reveal poor welds or even lengthy cracks, unless already extending through the wall. Higher pressure tests are more effective as a result of greater overstress in weak areas, and the initiation of plastic flow in distorted or poorly fit-up areas. However, there is certainly no complete assurance that a structure is safe as a result of successfully passing a single pressure test.

Equal or greater assurance of soundness can be obtained by radiographic examination of all welds coupled with a pressure test at the design pressure. For magnetic materials where thicknesses do not exceed! in., a magnetic powder examination inside and outside in lieu of radiographic examination can also be considered acceptable. This is not intended to imply that weld inspection and tests are interchangeable. Instead, they must be considered


simultaneously in evaluating safety. For heavywalled or critical-service piping, all practicable inspection procedures and tests are desirable and necessary for adequate safety.


1. D. B. Rossheim and A. R. C. Markl, "The Significance of, and Suggested Limits for, the Stress in Pipe Lines due to the Combined Effects of Pressure and Expansion," Trans. ASME, Vol. 62, No 5 (Hl40) .

2. E. L. Robinson, "Steam Piping Design to Minimize Creep Concentrations," presented at Annual Meg. of AS1·1E, New York, 1954.

S. L. F, Coffin, Jr., "A Study of the Effects of Cyclic Thermal Stresses on a Ductile Metal," ASME Paper No. 5S-A-76, presented in December, 1953.

4. L. F. Coffin, Jr., "The Problem of Thermal Stress Fatigue in Austenitic Steels at High Temperature," presented at ASTM meeting, Chicago, June, HI54.

5. A. R. C. Markl, "Fatigue Tests of Piping Components," Trans. ASME, Vol. 74, No. S, pp. 287-303 (1951).

6. A. Nidai, Plasticily, McGraw-Hili Book Co., New York, 1931.

7. W. J. Buxton and W. P. Burrows, "Formula for Pipe Thickness," Trans. ASME, Vol. 73, pp. 575-587 (July, 1951 ).

8. W. R. D. Manning, "The Overstrain of Tubes by Internal Pressure," Engineering, Vol. 159, pp. 101-102, 18S-18i (1945).

9. C. W. MacGregor, L. F. Coffin, Jr., and J. C. Fisher, "Partially Plastic Thick-Walled Tubes," J. Franklin Insi., Vol. 245, pp. 135-158 (1948).

10. C. W. MacGregor, L. F. Coffin, Jr., and J. C. Fisher, "The Plastic Flow of Thick-Walled Tubes with Large Strains," J. Appl. Pllys., Vol. 19, pp. 291-297 (19·18).

11. R. W. Bailey, "Creep Relationships and Their Application to Pipes, Tubes, and Cylindrical Parts Under Internal Pressure," Proc. Insi. Mech. Engrs. (London), Vol. 164, pp. 425-4S1 (1951).

12. J. J. Murphy, C. R. Soderberg, Jr., and D. B. Rossheim, "Considerations Affecting More Economic but Equally Safe Pressure Vessel Construction Utilizing Either Present-Day Ductile or New High-Strength Less-Ductile Materials," API Paper presented at St. Louis, May 10, 1955.

13. E. O. Bergman, "The New-Type Code Chart for the Design of Vessels Under External Pressure," ASME Paper No. 51-A-137, presented at Atlantic City, November, 1951.



Local Components

THIS chapter will consider important components of a piping system other than straight pipe, including flanges, bends, miters, corrugated pipe, branch connections, and terminal connections, all of which are designated herein as "local components" since individually they usually occupy a limited length of the total pipe run. The localized stress pattern which they introduce often significantly increases the flexibility of the entire piping system at the expense of stress intensification or strain concentration at their location. It is the intent of this chapter to offer a digest of current knowledge about each local component, and discuss practical application to the design of piping. Accurate evaluation of stress and deflection for localized effects is often complex, or even impossible with present knowledge; as a result simplifying assumptions and shortcut solutions are resorted to, some of which will be discussed herein.

3.1 Pipe Bends: Structural Loading (Static and Cyclic)

Pipe bends are curved bars with an annular cross section, whose reaction to external loading is complex. Visual observation, as well as scattered tests, established quite early that the elementary theory of elasticity is inadequate to account for the peculiar properties of tubular bends. Despite this fact, considerable time passed before a satisfactory analysis was undertaken. While theories are sufficiently advanced today to account for the major aspects of the behavior of pipe bends, many refinements of this problem still demand clarification and a further extension of theoretical inquiry.

Systematic investigation of pipe bends began in 1910, when Bantlin [1] observed and reported on the phenomenon of ovalization, and on the fact that it

leads to greater flexibility than could be accounted for by bar theories. A year later, the first theoretical treatment of the subject was published by Vall Karman [2], who investigated the stress distribution in curved tubes subjected to in-plane bending.' At about the same time Lorenz [3] and Marbec [4] independently furnished a solution of this problem, using Castigliano's theorem in their work instead of the principle of minimum potential energy as used by Karman. Hovgaard continued Karman's work and arrived at an identical solution through a different approach {5] while Karl [6] refined the solution by considering more terms in the series expansion for the basic variables. In 1943 Vigness {7] extended the theory to include the case of out-of-plane bending of curved pipes. These theoretical investigations readily establish the following points:

1. The elementary bending theory for bars, which assumes a linear variation of longitudinal stresses, cannot account for the actual stress distribution in curved tubes under external bending loads. In reality, the longitudinal bending stresses in the extreme fibers are greatly relieved by the ovalization (flattening) of the cross section, which, under different loading conditions, takes the forms shown in Fig. 3.1. At the same time the maximum stresses are shifted nearer the neutral axis, as shown in Fig. 3.2.

2. This altered bending-stress distribution, in turn, decreases the bending-moment resistance of the section. The ratio of the resulting increased

lIn-plane bending refers to the case in which the pipe is subject to bending by forces or moments applied in the plane of the bend. Out-of-plane bending designates the case in which the forces or moments act perpendicularly to the plane of the bend. Obviously, these two cases can be combined to give a solution for forces or moments acting in any arbitrary plane.




deflection to that predicted by conventional beam theory is termed the "flexibility factor" for that

member. .~

3. The maximum longitudinal stresses in pipe bends will differ from those generated in straight tubing of equal dimensions. High circumferential bending stresses are set up as well. For pure (inplane) bending, theory indicates that the peak stresses will be the circumferential stresses near the neutral axis (or = 0) of the pipe. The ratio of the maximum stress in the curved pipe bend to that which would exist in straight pipe subjected to the same moment is termed "stress intensification factor. "

These findings were subsequently reexamined by Beskin [8], who found that the previously established results were applicable only when the bend characteristio/ was comparatively large; as the characteristic diminished, the results became increasingly divergent. Instead of a maximum flexibility factor for in-plane bending of 10, and a maximum stress intensification of about 3.5, as implied by earlier analyses for the mathematical limit of h = 0, Beskin found that both flexibility and stress intensification factors become infinite at this extreme value. Further investigation showed that Karman's solution would have yielded results identical with those of Deskin, had the Fourier expansion been carried to more terms than one.

Treating the problem of in-plane bending of curved tubes by means of the theory of thin shells, Clark and Reissner [9, tOJ found that the Lorenz, Karman, Karl, and Deskin solutions merely represented higher order approximations (in the order mentioned), and confirmed Karl's findings {61 that alter-

~The bend characteristic is h = tR/r",2, where t = wall thickness of pipe, R = radius of bend, and Tm = mean radius of pipe.


(0) In-plone Bending {tongenh 10,<<<1 logethe,)

(b) In-plane (e} Out-of-plene

Ben&ng Bendin9




(a) elemenlary Bending Thoory

(b) Thoory of Curved Pip'"

Max. 'ongitud~nal ,trm.s occurs at angl. al {see Fig. 3.8)

1 _j


FIG. 3.2 Distribution of longitudinal stresses in curved pipes.

nate solutions by the principle of minimum potential energy (used by Karman) and the principle of least work (adopted by Lorenz, Karl, and Baskin), establish upper and lower limits for the true rigidity of the tube. The Clark-Reissner solution is obtained in terms of a trigonometric series expansion for the stress function and meridional angle change. By retaining only two terms of each series expansion, and limiting the range to h > 0.5, the Clark-Reissner approach becomes equivalent to Karl's solution. For h < 0.5, the number of terms needed for satisfactory accuracy increases rapidly; therefore, an asymptotic solution was investigated. Making assumptions which hold true when h is much smaller than 1, closed-form solutions were obtained which are startlingly simple. All analyses dealing with the problem of bending of curved tubes predict equal flexibility factors for in-plane or out-of-plane bending.

Karman's original solution (first approximation) for the flexibility factor, k, is


k= 1+ 12h2+1

Second, third, and nth approximations Ill] have the form

k 9+0.255/h2

-1 + --=---'---...:....._--~

- 12h2+ 1.3400+0.00750/h2

k= 1 9+0.3003/h2+O.OO10587/h·1

+ 12h2+ 1.4004+0.013946/h2+O.00001276/h4 (3.3)3

k=l+ 9 12h2+ I-j

In eq. 3.4, j is a function of h; for known values of h

3In eqs. 3.1 to 3.4 the rigorous mathematical analysis would demand that h(l - ,,2)-h be used instead of h.

FIG. 3.1 Ovalization (flattening) of pipe bends under external bending moments. EXaggerated.



the magnitude of j can be obtained by interpolation from the following table:


h 0 0.05 0.1 0.2 0.3 0.5 0.75 1.0

j 1 0.7625 0.5684 0.3074 0.1764 0.07488 0.03526 0.02026

Beskin's solution for the flexibility factor cannot be expressed in closed form; his numerical results, which merge with Karman's nth approximation, are plotted in Fig. 3.3. Clark and Reissner's asymptotic solution yields the following expression, valid for small values of h:

k = 1.65/h

{3.5 )"1

Stress intensification factors, unfortunately, differ for in-plane and out-of-plane bending. In general, in-plane bending leads to higher circumferential stress maxima than out-of-plane bending for identical pipe bends subjected to equal bending moments. For longitudinal stresses exactly the opposite of this statement holds true, as witnessed by eqs. 3.6 to 3.9. The stress intensification factors at the outer surfaces, valid only for small values of the bend characteristic (h < 0.5), have the following expressions:

In-pl~ne {LOngitudinal f3i = 0.84/h%

bending Circumferential r« = 1.80/h% (3.7)5


4" = 0.3 is assumed in eqs. 3.5, 3.6, 3.7, and 3.10. 5Rigorously, the correct value in eq. 3.7 should be:

1.80 2"

'Yi = - (1 - " )-» i»


.04 .06 .08 1

Out-?f-plane {LOngitudinal f30 = l.08/h % bending Circumferential 'Yo = L50/h%

(3.8) (3.9)

In this same range the variation of angle al (pertaining to the largest longitudinal stress, as shown in Fig. 3.2) with the characteristic h can be given as

al = O.82hH


Equations 3.6, 3.7, and 3.10 are obtained from the asymptotic analysis of Clark and Reissner. Equations 3.8 and 3.9 represent empirical proposals made by The M. W. Kellogg Company, and Markl (12). respectively. The various stress intensification Iactors are charted in Figs. 3.4 to 3.7, whereas at it; plotted in Fig. 3.8.

These results convey that for either in-plane 01" out-of-plane loading, the circumferential.stress in the neighborhood of a = 00 will first exceed the yield point. This stress is a pure bending stress (excluding internal pressure effects), varying from a positive maximum at the outer surface to a negative maximum at the inner wall. A slight amount of yielding, leaving the elasticity of the pipe wholly unimpaired, will materially relieve this stress, as has been observed in experiments (13). Similar deductions can be made concerning the maximum longitudinal stress at the outer surface. Pronounced yielding will ensue only when the maximum longitudinal stress at the middle surface also exceeds the yield point. Therefore, in

(1) - - A;ymploti. 50lulion 1 .~s (Clark _ R.i.,""r)

(2)----B .. ~in'. (large.radiu; bend.)

(3) - - -Symood. and Pardue'. (Smoll-rodiu; bend,)

(4) --Von Karmcln'. nth Approximation

(5)----Approximolo, l.~o (Vissol- D.IBuono) (For small·radiu. thick-walled bend.)


FIG. 3.3 Flexibility factors for in-plane or out-of-plane bending.


,1 m

described above reveals that, in addition to dealing only with pipes having a constant curvature of the center line, constant cross-sectional properties, and being made of an isotropic and homogeneous material obeying Hooke's law, the analyses are based on the following assumptions:

1. Plane sections remain plane and the neutral axis retains its original length after loading.

2. Longitudinal and circumferential stresses are principal stresses.


O.S. (l)--A>ymploli.SoMion h~1l (Clark - Roi"nor)

(2) - - So.kin'. (lorll" ... "diu; b.nd.)

(J)- .. -Symond. and Parduo', (Smoll -r adiu. i><Ind,)


(4) -- - - Appro.imale, hVi (Vi""! - Del Buono)

(for lnIoll-radiu. lhick-walled b.nd.)




.6 .8 1.0


the opinion of many investigators the stress intensification factor of greatest practical significance is the one pertaining to maximum longitudinal stress existing at the middle surface of the pipe wall thickness.P

Closer scrutiny of the theoretical developments 6Fatigue tests do not support this. In fatigue, cracks open up perpendicular to the actual maximum stress which is the circumferential stress at the inner pipe wall. This is to be expected, since under reversed strain loading beyond the elastic range, as applied in a fatigue test, initial plastic flow is of little help in alleviating the range of strain at each point.

10 8



.06 .OS .1

FIG. 3.4 In-plane bending: outer surface longitudinal stress intensificntion factor.


- ,


~ 4
.~ :2
.03 .04 .06 .08 .1

(l- . 1.S0

1 A.ymploll. Solulion h'/l (Clark - Rei" ne r)

(2) - - So,kin'. (large-radiu. bend.)

(3)----Symond. and Pardue'. (Short-radiu. bond.)

{4}- •. -Approximalo, ~~s {Vi.sal - 0.1 Buono} (for .mall·radiu. !hick-walled bonds)



.4 h=!!!. r~

.6 ,8 1.0


FlO. 3.5 In-plane bending: outer surface circumferential stress intensification factor.



3. The bending moment has a constant value for the entire length of the bend.

4. Radial and longitudinal strains are uniform through the wall thickness.

5. Circumferential strains produce pure bending, and thus vanish at the middle surface of the pipe wall.

6. The radius of the bend is much greater and the wall thickness is small compared with the diameter of the pipe.

Assumption I is fundamental to the theory of elasticity and can be accepted as being true. The second and third conditions will be satisfied only if

10 8

the curved pipe is acted upon by pure bending moments. According to St. Venant's principle, local disturbances imposed at the boundaries will cancel a short distance therefrom. In this light, assumptions 2 and 3 can also be adopted as having reasonable validity.

The remaining assumptions deserve closer scrutiny.

Assumptions 4 and 5, dealing with the strains developed under loading, are idealized simplifications of the actual strain distribution, and will be in accord with the actual strains only when RIT", > 10 (i.e., larger than a "five diameter" bend). For shortradius bends, characterized by I < Rlrm < 10, it

(1)- - ao.kin', (largo.radius "".d,)

(2) - - -- Symonds and Pardu.', (Smell-radiu. bond.) (3)--Appr"xim<rto, ~.~~ (Woil)

FJG.3.6 Out-of-plane bending; outer surface longitudinal stress intensification factor.

h"'!.& r' m

10 e






. 1.50

(1)--Approxima10, h,/J (Mark!)

(2)- -Be;kin', (Largo redius bond.)

(3)----Symonds and Pardue's \~mall radius bonds)

.O~ .06 .08 .1


,2 m


.6 .8 1.0


FJO,3.7 Out-of-plane bending: outer surface circumferential stress intensification factor.



has been shown [7, 13] that under in-plane bending (reducing the radius of curvature) the circumferential stresses do not vanish at the middle layer (see assumption 5).

The last assumption plainly limits the accuracy of the foregoing theories to thin-walled, large-radius tube bends; the generally accepted view is that these analyses are proper only if both conditions, namely that R/rm and rm/t be greater than 10, are simultaneously satisfied. Since Beskin's derivation indicates that at h > 1.0 the flexibility and stress intensification factors become generally negligible, it is of interest to note that this development is not conditioned upon the above-stated limitation on wall thickness.

To extend the validity of previous analyses, Symonds and Pardue [14] undertook to investigate the effect of R/rm ratios considerably less than 10, (2':::; R/rm ~ 3). It may be pointed out that under these conditions the wall thickness ratio assumes a much greater importance; the fact that the "shortradius" development is based on thin-shell theory plainly limits the range of accuracy to about h = 0.2 for R/rm = 2, or h = 0.3 for R/rm = 3. The Symonds-Pardue theory represents a first-order approximation to the influence of R/rm, and shows that for short-radius bends (long- and short-radius welding elbows), the flexibility factor suffers little change, but stress intensification factors are generally higher than for large-radius pipe bends, as seen in Figs. 3.3 to 3.7. As might be expected, the results of this work merge with Baskin's solution, as R/rm increases to 10.

Lastly, all of the theories described apply rigorously only to endless toroidal sections. If the curved tube is not endless, the theory is accurate only if the end conditions allow the development of idealized strains

(I) --Clark - Rei"n.,

I:: 50° o 0,400 ..




(2)----Asymplolic Solulion 0.82h'/l (Clark and Roi ..... r)

and displacements (flattening or ovalization). End restraints tending to oppose ovalization, (straight pipe tangents to a minor degree, flanges or terminal connections to a severe degree) will lower flexibility and stress intensification factors; in these cases the theory will give higher values than those actually operative. Thus deviations between theory and actual behavior will be greater the more severe the end restraint, or for a given end restraint, the lesser the subtended arc of the pipe bend.

Having elaborated on the underlying assumptions and results obtained from an analytical approach, it is enlightening to examine how the theories compare with results obtained from experimental work. Investigations must be separated into tests performed under static conditions and those relating to fatigue conditions, since they represent fundamentally different types of loading.

The first significant static tests were made by Hovgaard [5, 13, 15, 16], who proved that experiments were in close agreement with theoretical predictions for the flexibility, distortion and stresses of a given system, although calculated stresses showed smaller extremes than those actually observed. It must be added that Hovgaard's tests were performed mostly within the limitations of his theory: stress distribution measurements were confined to sections remote from the disturbing effects of type-of-loading or end-fixity conditions, and most of the experiments were limited to large-radius bends (R/rm > 10).

Similar observations were made by other investigators [17, 18, 19, 20, 21, 22], who again found the longitudinal stresses to be slightly in excess of theoretical values. It was also observed [21] that the flexibility of pipe bends for in-plane bending was greater than that predicted by theory. This deviation was small, but consistent, and was ascribed to


.04 .06 .08 .1


.2 h= !! <;'


.6 .8 1.0



FJO. 3.8 In-plane bending: angle C(l corresponding to location of maximum longitudinal fiber ~trCB8.



the fact that the theory of curved tubes did not take into account secondary influences predicted by the theory of curved bars. Tests conducted under outof-plane bending [7}, in turn, showed that the rigidity of pipes was greater (i.e., the flexibility less) than indicated by analysis. This was attributed to the restraining effect of straight tangents applied to the ends of the quarter pipe bend. Stresses meanwhile were smaller than was anticipated from theoretical research.

A thorough investigation on the effect of end conditions was carried out by Pardue and Vigness [23]. Dealing first with flexibility factors, they found that even the most detailed theory [14] was capable of predicting flexibility factors for out-of-plane bending only if the pipe bend merged with a straight tangent of sufficient length. Substituting a flange for the tangent at either end resulted in a drastic drop of flexibility; when both ends were flanged, flexibility dropped even further. Right-angle bends were subject to these reductions in a greater degree than U-bends, confirming the logical expectation that the smaller the sub tended angle of a pipe bend the greater will be its sensitivity to disturbances caused by end restraints.

Almost identical statements apply to the stress intensification factors. The theory is in agreement with actual behavior only insofar as the bend is furnished with sufficiently long straight tangents, the experimental values being generally a shade on the high side. With an increased degree of end fixity, this correlation breaks down. Applying flanges to both ends of the bend initiates a much greater reduction of the stress intensification factor than using one flange and one straight tangent; and again, right-angle bends were subject to these modifying effects to a greater degree than U-bends.

Additional confirmation of theoretical results was provided by a series of tests carried out by Gross and Ford {24, 25, 26]. These tests proved that, in line with theoretical predictions, the circumferential stress in the vicinity of a = 0 was the largest absolute stress; in the tests carried to failure, the cracks always ran along the side of the bend at about the location of the neutral axis. Contrary to assumption 5 of the theory, however, the circumferential stress at the middle layer did not vanish. Application of strain gages to the external and internal faces proved that the maximum stresses were always situated on the inner surface of the bend, which may explain the observation [25, 27] that cracks in pipe bends are initiated on the inner face and penetrate outwards. To account for this defi-

ciency in the theory, Gross [25] suggested that the transverse compression, ignored in the analytical work, be taken into account. He also presented an approximate derivation for this quantity, and proved that adding this stress component to the others included in the theory actually brings experiments and analysis into good accord.

The examination of test results also confirmed that the theory of maximum distortion energy predicted quite accurately the load and location at which incipient yielding occurs. No such criterion could be advanced for the ultimate load-carrying capacity of the bends, except for noting that failure (which took place by collapse under a moment shortening the chord of the bend) occurred at a load which was generally twice as large as that required to initiate yielding.

Vissat and DelBuono [28] describe the results of tests on welding elbows with a ratio of R/rm equal to 2 or 3. Restricting the tests to in-plane bending and fitting experimental points by analytical expressions, the following relations were proposed for flexibility and stress intensification factors:

Flexibility factor

k = 1.40/h


Stress intensification factors (in-plane bending)

{3i = 1.2/h%

'Yi = 1.07/ho.78

(3.12) (3.13)

As seen in Figs. 3.3 to 3.5, these proposals result in lower flexibility and circumferential stress intensification factors, but higher longitudinal stress intensification factors as compared with the data of all other theories. While these results can be used on welding elbows having the characteristics investigated, some reservation should be exercised, since on the thick-walled short-radius bends used in this work the restraining influence of straight tangents or flanges has not received sufficient evaluation.

The next point of interest is to examine whether the conclusions drawn above remain valid if the bend is acted upon by repeated cyclic loading rather than a single static load. Obviously, fatigue conditions will hardly modify the flexibility of a sound local component. What is sought through a fatigue test, therefore, is the practical effect of stress intensification on the number of cycles to failure.

In addition to manner of loading, fatigue tests differ in two aspects from static tests: in the manner of measuring stress intensification, and in the weight given to plastic flow. In static tests, the stress



intensification factor denotes the ratio of actual peak stresses to those developed in a straight member of identical dimensions (for pure bending, the reference stress is M/Z).7 In fatigue, the effective stress intensification factor relates the stresses causing failure over a given number of cycles in a straight pipe tangent (or polished bar) to those initiating fracture in the test piece subjected to an equal amount of stress cycles.

As regards the significance of plastic strains, static-stress measurements are strongly dependent upon the presence of plastic flow with its attendant redistribution of loading and stress-mitigating effect. In the fatigue test the local strain range per cycle is the significant value determining performance; therefore, a redistribution of stresses due to plastic strains has only a minor significance. While it is common practice to use and establish design practices in terms of stresses based on elastic theory, it should be appreciated when dealing with fatigue that in reality these stresses are being used as a suitable index of the strains involved.

Fatigue tests on piping components were initiated by Rossheim and Markl [29J, followed by a detailed research program carried out by Markl [12, 30]. Since it was felt that the stress peaks developed in local components as compared with straight runs of pipe constituted the desired fundamental information, stress intensification factors were based on a comparison with S-N diagrams obtained for straight commercial finish pipe, containing butt welds, a clamped edge or similar stress raisers. A stress intensification factor of unity was assigned to the latter for practical reasons.

The first finding of interest was that, while the 8-N curves for both straight pipes and welding elbows of carbon steel seemed to reach no endurance limit within the number of cycles employed (2 X lOll cycles max), both curves were approximately straight and parallel to each other on a log-log plot. This indicated that the stress intensification factor could be given as 11 constant, regardless of the number of "tress cycles involved.

In comparing test results with theory, it was found that Beskin's or the Symonds-Pardue theory predicted quite accurately the flexibility of the bend or elbow, as well as the type and location of failure. The agreement between tests and theory for stress intensification factors was less satisfactory; however, 11 reasonably good correlation was obtained if the test results were drawn into comparison with only one-half of the maximum theoretical stress intensi-

7Whero Z = I IT. is the section modulus of the cross section.

fication factor (referring to circumferential stresses for both in-plane and out-of-plane bending).

When considering the significance of this finding, it is important to note that Markl's reference point of unity for the test results is not a theoretical but a practical, one. In the first place, Markl [30] found that the clamped edge used in the earlier tests involved a stress intensification factor of about 1.5, as compared to pipes with a tapered end. The remaining factor of about 1.4 needed to bring experiments and theory into agreement may perhaps be attributed to the stress raisers inherent in commercial finish pipe as compared to the theoretically considered smooth homogeneous tube. Markl could have changed his reference point of unity and assigned different test factors; however, he found that butt welds involved the same stress intensification as the clamped edge. Therefore, he reasoned that a base line which would include the effect of such normally encountered stress raisers would be much more satisfactory for practical design. This reasoning has been supported by practical designers and by the ABA Code for Pressure Piping Committee. It is, nonetheless, an important point which should be kept in mind, particularly in connection with the practical application of any theoretically derived factors for other piping components.

Recourse to eqs. 3.7 and 3.9 indicates that the experimentally found reduction factor of 2.0 leads to design stress intensification factors (when referred to the aforementioned base line) of the following magnitude:

'Yi = O.90/h% for in-plane bending,

'Yo = O.75/h'H for out-of-plane bending

It happens that these stress intensification factors are very close to the theoretical value of the longitudinal factor for in-plane bending, which has long been in customary use instead of the more proper circumferential factor. The seeming justification of this latter practice stemmed from Hovgaard's findings that permanent overall deformation of the pipe bend occurred only when the longitudinal stress at the middle surface exceeded the yield stress.

The recommendations of the revised ASA B31.1 Code for Pressure Piping are derived principally from these experimental observations. For both in-plane and out-of-plane bending, the Code recommends that the stress intensification factor

{:J = 'Y = O.90/h'H ~ 1.0


be used, the choice of a single factor having been



2R+, ..


FIG. 3.9 Distribution of circumferential stresses in pipe bend subjected to internal prCS8L1re.

accepted for practical reasons only. The flexibility factor, as proven by theory and experiments, is given as

k = 1.65/h


Additional fatigue tests again proved the restraint of straight tangents upon the full development of the flexibility and stress magnification factors; as in static tests, this influence was increasingly accentuated with reduction of the subtended angle of the pipe bend. As a rough measure, it could be stated that both flexibilty and stress intensification factors were reduced from their full value for a quarter-bend elbow to unity, as the subtended angle of the bend approached zero. This rule was upset only at the very small arc bends, where the disturbing effect of closely spaced welds obliterated the restraining influence of the straight tangents, causing a concomitant rise in the stress intensification factor.

3.2 Pipe Bends: Internal Pressure

The foregoing theories and experiments dealt solely with pipe bends subj ected to external loadings. In addition to this effect the pipe wall will be stressed by the pressure of the fluid in the system. For a curved pipe subjected to a pressure p, the longitudinal and circumferential membrane stresses are given approximately by [31}


2R + r m sin a r mP

(Jc= -

2{R + rm sin a) t


8Notic!! that these formulas are identical with the equations for straight pipes, except for the first fraction appearing in eq. 3.16.

The resulting stress distribution is shown in Fig. ~.9. A more elaborate investigation [32] and tests earned out on curved pipes under internal pressure [25] confirmed the general validity of eqs. 3.15 and 3.16, and showed that maximum stresses will be reached at the line of the bend having the least radius of curvature (crotch), as predicted by eq. 3.16. Yielding will first occur at this point. Despite this, it is not normal practice to apply these formulas to the design of pipe bends.

When external loading and internal pressure are imposed simultaneously on a pipe bend, experiment~l results [26] show (as should be expected) that maximum circumferential stresses occurring for external moment loading alone will be reduced by the presence of internal pressure. While the presence of internal pressure will slightly reduce the flexibility of the bend [13, 241, the stress, whether referring to principal stresses or combined stress, will also be mitigated [26].

3.3 Miter Bends

Particularly for the less severe services, changes in direction are not infrequently made by mitering straight pipe (Fig. 3.10). Yet miter "bends" have received much less attention in the literature than curved pipes. Nevertheless, it was shown by Z~no [33], who investigated the flexibility of a five-section right-angle miter bend (h = 0.0158), that the theoretical flexibility values of curved pipes were ap-

• '" miter tpacing 'I' =miter anglo

R'" equivalent bend fadi", = t cot 'I' Tm'" m""n radi'" 01 pipe

FIG. 3.10 Geometry of miter bends.



proached as the tangents were made sufficiently long." Similar indications were obtained by Gross and Ford [26], who measured stri:!&ses and flexibility on a miter bend of h = 0.0483, and found them reasonably well predicted by the theory of curved tubes.

Little additional information is available concerning the properties of welded miter joints under static loads, since in addition to difficulties encountered with plain pipe bends, miter joints arc subject to variations introduced by differences in fit-up and welding as well as in the arrangement of segments. Available evidence, however, seems to indicate that the flexibility is less and stress intensification is greater for miters than for plain bends of the same major dimensions.

Miter bends have also been subjected to intensive cyclic testing [12] with the finding that their behavior could be predicted with reasonable accuracy through analogy with curved pipes when the proper characteristic variables were included. From geometry (see Fig. 3.10) the radius of the tangent arc of a bend can be expressed as R = ts cot 4>, where s = miter spacing at center line, and 4> = miter angle. If there is but a single miter or if the miter spacing becomes large, however, this radius loses its significance and an effective radius was suggested, empirically expressed as R = Tm (1 + cot 4»/2. Thus the bend characteristic assumes the following form:

1 tR cot q, ts f . .

! = 2 = -- --2 or small miter spacing,

Tm 2 Tm


- - tan 4> < 1 (3.17)


I + cot q, t

h = 2 - for large miter spacing, Tm



- tan 4> > 1 Tm

By using this bend characteristic with thc expression derived for curved pipe, eq. 3. H, values of stress intensification are obtained which show a good correlation with the tests. The flexibility factor was somewhat smaller than for plain curved pipes, and resembled that for welding elbows with one flange and one plane tangent. The flexibility factor for

9Without tangents, i.e., with flat plates welded directly to the end of the Iast miter segments, the flexibility for in-plane bending was found to be reduced to only 3% of the theoretically predicted value for a bend.

miter bends can hence be given as

k = 1.52/h% 2. 1.0


with h taken as the lesser of the values obtained from eqs. 3.17 or 3.18. These results are incorporated in the recommendations of the ASA B3l.1 Code for Pressure Piping.

3.4 Bends and Miters: Summary

Pipe bends depart from conventional beam theory chiefly as a result of distortion (ovalization) of the cross section under bending. Under static loading, theories predict the flexibility, maximum stresses, and occurrence of incipient yielding with good accuracy for bends with plain tangents whose subtended angle is larger than 90°. At present theories do not consider the restraining influence of straight tangents (particularly significant for curved pipes whose bend angle is less than 90°), nor can they effectively deal with the inhibiting tendency of severe end restraints, such as flanges or terminal connections. To evaluate the characteristics of components falling into this category, reference must be made to such test data as are available.

In actual service, idealized static loading conditions are seldom encountered. A certain amount of plastic flow will always take place, enabling the bend to carry loads in excess of those predicted by the classical elasticity theory. The significance of theoretically calculated stress values is further reduced by the fact that even straight piping with a commercial finish carries an inherent stress-raising factor, and that the performance of bends is, for practical reasons, referred to teat of butt-welded or clampedend pipes rather than polished test specimens. Not only experimental evidence but also a long history of successful design practice support these facts.

These considerations will hardly affect the flexibility factor. Therefore, it is sound practice to use the theoretically derived value of this factor, as given by eq. 3.5. When considering stress intensification factors, however, it is sufficient to base calculations on only one-half of the theoretically predicted value; as supported by Markl's fatigue tests, the appropriate equation for this factor is given by eq. 3.14.

The increase of membrane stresses in pipe bends subjected to internal pressure loading (as compared to straight pipes), is generally not significant. In fact, tests demonstrated that a static load alone will lead to higher localized stresses than a combination of this static land and a moderate internal pressure. The effect of internal pressure on bends can, there-




FIG, 3,11 Types of reinforcement for branch connections.

fore, be ignored in normal applications, An investigation of this effect, in line with the principles laid down in the text, is warranted only for very critical service.

The stress intensification and flexibility factors of short bends (subtended angle less than 900) are known to be less than those indicated above, Despite this fact, it is recommended that no reduction for either factor be used on short bends, since the experimental evidence on this subject is not conclusive.

Miter bends, as a rule, have lower flexibility and higher maximum stresses than those pertaining to curved pipes of similar dimensions, By this token, the appropriate design value for the flexibility factor of miter bends can bc obtained from eq, 3,19. The stress-raising factor will be given by eq. 3.14, with the bend characteristic given by the smaller value obtained from eqs, 3.17 or 3.18. These design criteria for miter bends originate from tests conducted on 4 in. miters only. For large-diameter miter bends, fit-up and fabrication difficulties are likely to lead to more severe conditions than would be indicated by the design rules stated above.

3.5 Branch Connections: Static Pressure Loading

The junction of a branch with a header, usually referred to as a branch connection, is inherently a point of structural weakness in piping. Not only the absence of metal in the header opening but also the abrupt directional changes and oftentimes sharp variations in cross section give rise to severe stress intensification. While this handicap of a branch connection can be overcome to a large extent by reinforcement and by the use of favorable contours. it is difficult to achieve the ideal of developing a strength equal to that of the intact pipe.

Branch connections must be designed first in regard to their ability to resist static loads. This will be the concern of the present section, while the effect of repeated loads will be considered in Section 3.6. S~ction 3.7 will present a short review of various pertinent Code rules, and Section 3.8 will give a summary together with practical design recommendations.

An involved geometrical shape and the strong influence of certain secondary effects'? make the analytical investigation of branch connections subjected to pressure or structural loading prohibitively difficult. Consequently, investigations dealing with this subject are largely confined to experimental research. The salient information on the action of branch connections subjected to internal pressure is summarized in Table 3.1 although mention should also be made of a few tests reported in references [34, 35, 36, 37, 38, 39]. Figure 3.11 shows various types of reinforcements which have been proposed.

From Table 3.1 and its underlying tests, the following conclusions can be drawn: Unreinforced fullsized connections are deficient in both yielding and bursting strength. This deficiency decreases as the branch becomes smaller in comparison to the header. The limited number of tests seems to indicate that an unreinforced 900 intersection develops the full bursting strength when the ratio of branch to header diameter does not exceed i. Unpublished tests on 30 in. diameter pipe and the general experience with pressure vessels, however, show that this rule cannot be extended beyond the commonly available sizes of commercial pipes.

The addition of a pad reinforcement is beneficial in that it permits the fabricated connection to develop almost the full bursting pressure of the header.

Lu,such as the existence of longitudinal bending in the header due to removal of part of its wall, and the interplay of radial displacements of both header and branch under internal. pressure.


. ..



Pad reinforcements, however, afford little restraint against plastic flow and are, therefore, ineffective in raising the yielding pressure of ><the intersection to the desired value [42, 43].

Several alternatives have been advanced to eliminate the shortcomings of unreinforced or pad-reinforced intersections. The reinforcing saddle [44], shown in Fig. 3.l1b, adds reinforcement around the highly stressed areas of the crotch and shoulder. The complete encirclement pad is pictured in Fig. This proposal [45] extends the reinforcement

beyond the region where most failures of the padreinforced branch connections originated. An extension of this concept (46] supplements the encircling band with shoulder pads, as shown in Fig. 3. lid. While no test results are presented, the authors of these proposals have stated that the performance of full-sized branch connections reinforced in accordance with these alternate details was entirely adequate under pressure loading. On the other hand, the horseshoe-and-gusset reinforcement (Fig. 3.lIe), due to its extreme rigidity, led to stress concentra-

Table 3.1 Summation of Internal Pressure Test Results on Piping Branch Connections

Pressure as per cent of Pressure as per cent
Angle of that supported by
Size of of Type that supported by right-angle
Ref. of intact header
No. Authors Inter- Reinforce- intersection sin a Remarks
section, ment"
Header Branch debfccs at Proper- at at Propor- at
in. in. tional limit Bursting tional limit Bursting
--- --
35 Everett
McCutchan 8 8 90 None 38.5 69.6
40 Crane 8 4 90 None 76.9 10).1
Co. 12 e 90 None - 91.4
8 8 90 Pad 61.5 98.9
12 12 90 Pad - 93.0
--- -- ---
41 Seabloom 24 24 90 Horseshoe - 38.5
and gusset
-- --- -
- - 90 None -70.0 -70.0 Averaged
test values
11.9 6.2 90 None }SI.O }82.5 {ValUes aver-
10 4.2 90 None aged for
48 Blair two tests
7.5 7.5 90 Pad 74.5 96.0
11.9 7.5 80 Collar 79.0 90.0
11.9 7.5 SO Gusset & pad S5.0 90.0
11.9 7.5 80 Unbalanced 110.0 >100.0 Averaged
triform values
6 6 90 Balanced 121.0 >100.0
--- --
49 N. Gr039 -S -8 90 Welding - -96.0 Averaged
tee value
_-- ---
- - 90 None 70.0 70.0 ll.o Averaged
us 10 90 None 79.0 79.0
6 ·1 90 None 83.0 86.0 value
- - 60 None GO.O W.O 85.6 85.6 }0.866 Averaged
48 Blair 11.9 10 GO None 56.0 66.0 70.9 83.5 value
6 4 W None - 66.0 - 76.7
- - 45 None 50.0 50.0 71.4 71.4 } 0.707 Averaged
n.s 10 45 None 43.0 44.0 54.5 55.7 value
6 ·1 45 None 51.0 61.0 61.5 70.9
- - 60 None 63.0 65.0 90.0 92.8 0.880 Y -connectionj
- - 30 None 3·1.0 40.0 18.6 57.1 0.502 OSee Fig. 3.11 [or identification of types of roiafcrcements.

tLast column deuotee I/(co~ a - 0.5 cot a) instead of sin a for equal-sized V-intersections.



tions of such magnitude that the intersection sustained only 38.5% of the bursting pressure of the

intact header. . ....

Collar reinforcements of the type shown in Fig. 3.1 If were pioneered by the Swiss firm of Sulzer Brothers, Ltd. [47], as early as 1928. Experiments [48] indicated that this method had characteristics similar to the pad-type reinforcement. As a further improvement, Blair [48] suggested that the stiffening collar be supplemented by a third horseshoe encircling the bottom of the header. He gave the name "triforrn" to the resulting arrangement, shown in Fig. 3.11g. As Table 3.1 shows, triforms performed very satisfactorily, considering both yield and bursting pressures.

While tests confirm the effectiveness of triforms, this type of reinforcement requires intricate fitting and welding which does not lend itself to radiographic examination. In high-temperature service the ribbed construction leads to thermal gradients. Furthermore, the sharp re-entrant corners suggest high stress concentrations which may not be revealed in static-pressure tests but would become critical under repeated loading. American experience with the triform is quite limited, hence in the United States it is regarded as a novel approach until its performance is more adequately assessed.

Welding tees, Fig. 3.11h, are preferred structurally to fabricated welded intersections, especially where the size of the branch is equal to or approximates the size of the run. Recently, cast tees proven to be sound by radiographic and magnetic particle examination and by hydrostatic test, arc finding increased acceptance. Only a few articles [49] are published concerning the design and strength properties of drawn tees subjected to internal pressure. The reason for this lies in the requirements of ASA Standard BI6.9, which prescribes that welding tees must be able to withstand the full bursting pressure of straight pipe in sizes for which they are intended. On the other hand, the Standard makes no demands regarding the pressure to be supported by drawn tees at their yield strength.

Despite the presence of high stresses at the internal surface of the crotch [49], welding tees, in general, involve lesser fabrication difficulties and stress concentrations than those associated with welded intersections. Their performance with regard to bursting, based on the Standard and the meager test data that are available, :is also satisfactory.!' Welding tees

lIThe cylindrical we tested by Gross [49], which failed at 96% of the pipe bursting pressure, would not comply with American Standard requirements.

having a "spheroidal" intersection zone are claimed to develop an increased resistance to yielding under internal pressure. Branch connections subject to extremely high internal pressures arc usually forged and bored [50].

Test results dealing with acute-angle (inclined) branch connections are even more scarce than those for right-angle intersections, being confined to those reported by Blair [481, and assembled here in Table 3.1. This evidence indicates that within the limits of the experiments (i.e., branch angles from 30° to 90°), the strength of such intersections is roughly proportional to the sine of the branch angle for both full and reducing sizes, which is equivalent (as Blair proposed) to basing reinforcement simply on the area removed from the sidewall of the header. The quantity of test data, however, is hardly sufficient to support any such conclusion. Intuitively it would seem that acute angle branch connections are further weakened by the increased stress concentration at the crotch due to the elliptical shape of the cut-out, and by the pressure load transfer through the reentrant crotch corner. Recognition of these effects has led to the current ASA Code requirements which will be discussed in a subsequent section.

The foregoing material deals with branch connections in pipes subjected to internal pressure loading. Closely related to this field is the subject of nozzles and openings in pressure vessels. While no weIldrawn division exists between these two fields, two criteria may be mentioned, which help to separate these problems. First, in pressure vessels the diameter of the branch (nozzle) is usually small as compared to that of the header (vessel). This fact. diminishes some of the secondary effects to lower levels. Secondly, the wall thickness to diameter ratio is also exceedingly small in large-sized pressure vessels. This permits the investigator to study the effects of openings in pressure vessels by means of flat-plate analogies, which would otherwise be of little use or validity to the designer of piping branch connections. It should be remembered that even on pressure vessels, the theoretical flat-plate analogies cannot assess the effect of the hydrostatic end pull exerted by the branch on the vessel. This effect can be evaluated on the basis of recent contributions by Bijlaard [51] and Hoff [52].

The theoretical approach to the problem of stresses around nozzles in pressure vessels has traditionally consisted of the investigation of flat plates with a circular opening reinforced in the manner shown in Fig. 3.12. Among these analytical studies [53, 54, 55, 56] Beskin's work [56} is the most



Rim (pipe collar)

FIG. 3.12 Edge reinforcement of circular cut-outs in flat plates.

(doubler plate only) is less efficient in reducing stress peaks. The dimensions incorporated in Table 3.2 are characterized by Fig. 3.12.

The plastic behavior of flat plates having a circular cut-out reinforced by a pipe collar was also investigated [57J. The analysis was restricted to ideally plastic materials (no strain hardening) which obey the maximum shear-stress flow condition. It was found that for a "full-strength" reinforcement (load at fully plastic condition in reinforced plate equal to or greater than that referring to intact plate), the dimensions of the pipe collar, as shown in Fig. 3.12, must satisfy the equations:

H 1 + tnlR for Rlin ~ ~
H 1 + (niR for Rlin ~ ~
Vl + 2(tnIR)2 - 1 These equations bear resemblance to the results of the elastic analysis, indicating that, for a given amount of reinforcement (inH = constant), maximum effectiveness is achieved when the reinforcement is concentrated near the opening (small in and large H). Naturally these results can be considered to retain their validity only within consistent limits.'? Furthermore, for strain-hardening materials or large plastic strains eqs. 3.20 can be used, at best, only as a rough guide. This plastic analysis was recently extended [58] to reinforcements of various cross sections.

Experimental work has corroborated the basic findings and, in some respects, the numerical results of theoretical work. It was found [591 that unreinforced circular openings in either heads or cylindrical vessels led to stress concentrations in excess of those

12For instance, the assumption of a very high and very slender rim would violate the fundamental condition that tho stress distribution be constant over the height of the rim.

complete. In this study, combinations of "rim-type" and "flat ring-type" reinforcements, applied symmetrically on both sides of the plate, are investigated. Based on the principle that the distortion energy governs yielding, stress intensification factors are given in terms of the "effective stress" rather than any one of the principal stresses.

The results of Beskin's investigation are shown in condensed form in Table 3.2. Since the idealized stress condition in shells under internal pressure can be decomposed into a hydrostatic and uniaxial circumferential stress of equal mangitudes, the last column ofstress intensification factors in Table 3.2, headed by "Average" reflects upon the conditions prevailing around nozzles in pressure vessels. As can be seen, both the rim-type reinforcement and the doubler-rim combination (with at least 50% of the reinforcing area supplied to the rim) are quite effective in diminishing the peak stresses existing around unreinforccd openings. In both cases, best results are obtained when the ratio of total reinforcement to hole area (reinforcement ratio) is in the neighborhood of 0.8-1.0; at this ratio the average stress intensification factor is reduced to a level of aboutl.Bfi, By contrast, the pad-type reinforcement

Table 3.2 "Effeetil'e Stress" Concentrations around Circular Holes in Flat Plates, Reinforced by Various Methods

Stress Intensification Factor Area of Reinforcement
HI. t' R' Area of Cut-Out
., - Binxial 1 uniaxial] A
Rt I R Rim I Doublerl Total
Tension Tension verage Rim Reinforcement only

0 1.0 1.0 2.0 I 3.0 2,50 - - a
0.4 1.0 1.0 1.32 1.52 1.42 0.4 - 0.4
0.8 1.0 1.0 I.OS i 1.63 1.35 0,8 - 0.8
1.2 1.0 I 1.0 1.01 1.16 1.38 1.2 - 1.2
- ........... -- Doubler Plate Reinforcement only

- 2.0 1,4 1.52 I 2.30 I 1.91 - 0.4 0.4
- 3.0 1.2 1.43 2.04 1.74 - 0.4 0,4
- 2.0 1.8 1.37 ! 2.17 1.77 - 0.8 0,8
, 0.8 0.8
- 3.0 1.4 1.22 I 1.90 1.56 -
- 5.0 1.2 1.03 I 1.75 1.39 - 0.8 0.8
- 5.0 1.3 1.00 1.85 I 1.42 - 1.2 1.2 Doubler and Rim Combined

0.2 2.0 1.2 1.36 1.78 1.57 0.2 0,2 I 0.4
0.4 2.0 1.4 1.01 1.61 1.31 0.4 0.4 0.8
0.4 3.0 1.2 1.02 1.73 1.38 0.4 0.4 0.8
0.4 5.0 1.1 1.04 1.72 1.38 0.4 0.4 I 0.8
0.6 5.0 1.15 1.00 1.85 1.43 0.6 0.6 1.2 66

conventional manner. Strain readings, however, were taken on both faces of the vessel. For reinforcement ratios!" of roughly 0.23, 0.615, and 1.0, the maximum stress concentration factors on the nozzle side were about 2.8, 2.3, and 1.8, respectively. These results are in reasonable agreement with previous experimental and theoretical work. Significantly different values were, however, found on the internal face, the maximum stress concentration values here being equal to 3.7, 2.8, and 2.4. This showed that an increase in thickness alone cannot bring about the desired reduction of peak stresses. and that only a moderate improvement in the stress distribution on the unreinforced face can accrue from reinforcement applied to the opposite side.

Actually, the circular opening is not the ideal shape for a cut-out in pressure vessels. Stress concentrations in a plate are minimized if the shape of the opening is an ellipse with an axis ratio equal to the "ratio of biaxiality," the major axis of the ellipse being aligned with the direction of the greatest principal stress. In pressure vessels, this would call for an elliptical cut-out with a majorto-minor-axis ratio .of 2 with the minor axis being in the longitudinal direction; the stress concentration associated with the unreinforced opening then becomes 1.5 (as contrasted to 2.5 for the circular hole). Both an analytical investigation [63] and experimental work [61J verified the desirable qualities of reinforced elliptical openings. In the tests, an increase of the reinforcement ratio from 0.16 to 1.13 lowered the maximum stress intensification factor from 1.40 to 1.19. While these results establish the sound concept of elliptical nozzles, it must be added that the fabrication and preparation of nozzles of this type would be beset by severe difficulties. These may overshadow the desirable aspects by increasing the cost of elliptical pipe attachments to a prohibitively high level.

3.6 Branch Connections: Repeated Loading

Having considered the performance of branch connections under internal pressure, attention will now be focused on their behavior under repeated external loads, such as imposed by thermal expansion of the pipe line. Although this subject received some consideration in Blair's paper, the most detailed information is found in Markl's work [12, 64].

These tests produced the following results: Failures of, full-size unreinforced intersections occurred at locations similar to those of curved pipes. The

HEffective height of reinforcement taken equal to radius of finished opening.


predicted by flat-plate theory. 13 These peak stresses diminished, in general, with a reduction of the ratio of the diameter of the opening to that of the vessel.

Full-scale tests also indicated that excessive stiffening led to greatly increased bending moments just beyond the toe of the weld attaching the nozzle. Optimum conditions were obtained [55, 59, 60J by concentrating the reinforcing metal near the opening. With appropriately reinforced openings the stress concentration factors were successfully limited to the theoretically predicted value of about 1.35. For small openings these optimum results were achieved by a pipe collar whose height-to-thickness ratio varied between 3 and 4. Flat doubler-type reinforcements were found to be ineffective, as predicted by theory; the stress concentration in these cases was in the vicinity of 3.0 at the longitudinal axis of the opening.

Further proof of these results was offered by Schoessow and Brooks [61]. Heavy rim-type reinforcements (reinforcement ratio 1.06-1.15) were only moderately effective, reducing stress concentrations from 2.50 for an unreinforced hole to about 2.05. Heavy doubler and thin rim combinations, however, effectively reduced stress intensifications to between 1.26-1.51, roughly in line with the predictions of the flat-plate analogies.

The tests described above were all conducted with the reinforcement being applied only to one side of the vessel. Stress measurements were likewise limited to the external surface. In contrast, theoretical predictions are bused on the assumption that the reinforcement is applied in equal proportion to both faces of the plate. The question, therefore, arose: if reinforcement is applied to one face only, what conditions will prevail on the unreinforced side?

An experimental answer to this question was soon forthcoming. It was shown [62] that the analogue prediction indicated the correct trend only as long as the reinforcement was applied symmetrically, as assumed by theory. Reinforcement applied to one face benefited only the surface onto which it was attached [55]; the stress pattern in the other face, however, remained essentially the same as it had been in the unreinforced opening. Further proof came from tests conducted more recently by Gross [49]. The reinforcement was applied to one side only, by welding nozzles to the vessel opening in the

13Stress concentration maxima for the hoop stresses occurred at the ends of the hole diameters parallel to the axis of the vessel, with some values as high as 5.5 in contrast to the theoretical prediction of 2.5.


\. A_


stress intensification factor could be correlated reasonably well with that for a single miter bend (see eq. 3.14) if the characteristid'variable was taken to be

h = tlrm



where t, = effective thickness = average of crotch and side wall thicknesses.

rc = crotch radius.

Experimental results for three different 4 in. commercial welding tees were in reasonable agreement with stress intensification factors obtained from eq. 3.14, if eq. 3.23 was adopted for determining the characteristic variable.

The flexibility factors associated with unreinforced and reinforced fabricated intersections or welding tees have not received sufficient attention. Rough tests seem to indicate that the added flexibility of full-size branch connections is small; that is to say, the branch will act as if it were fixed at the header, whereas the header will retain the flexibility of an intact pipe. These results, however, are open to question since full-sized intersections should approach single miter bends in flexibility. In addition, flexibility of the branch would be expected to increase for reducing-size intersections (see, e.g., eq. 3.27 in Section 3.14). Lacking specific theoretical or experimental results, and in order to remain on the safe side, it is suggested that a value of 1.0 be assumed for the flexibility of all types of branch connections.

3.7 Branch Conncctions: Comparison with Codc Requ'irements

It is of interest to compare now the experimental data with established design practice as expressed by Code requirements for 900 (perpendicular) branch connections. The Code for Pressure Piping, ASA B3l.l, Section 6, utilizes the area replacement method, requiring that the area removed from the wall of header (referring to the required minimum wall thickness times the diameter of the finished opening) be replaced by the excess thickness available in the header or nozzle wall plus any metal applied to the intersection in the form of reinforcement. This reinforcement is considered to have value only within the rectangular "reinforcement zone," the length and height of which is limited as shown in Fig. 3.13. In the subsequent derivation, the following nomenclature is used:

tIl = minimum thickness of header} less corrosion tB = minimum thickness of branch allowance.

RIl = radius of header} tsid

R d· f b h OU 81 e.

B = ra lUS 0 ranc

w = leg of fillet weld.

teristic variable of welding tees has been simplified to h = 4.4 tlr by making assumptions for t. and Tc which conservatively reflect customary proportions.

Reinforced fabricated intersections cannot be categorized with equal facility, since the amount of metal incorporated in the reinforcement and the manner of its distribution will affect the stress intensification and flexibility factors. In an attempt to formulate a rule which would correlate reasonably well with limited tests on 4 in. size pipe and be applicable to most reinforced branch connections, Mark! {12] proposed that the average thickness of the header and branch at the crotch, te, be assumed as the governing factor. Assuming that reinforced intersections otherwise behave like unreinforced ones, the characteristic variable would then become

h = (~)2.5 _!_

t Tm


while the stress intensification factor is again obtained from eq. 3.14,17

These results refer to tests where the assemblies were loaded through the branch; loading straight through the header proved to be less severe in all cases. Furthermore, it was shown that the direction of bending (in- or out-of-plane) did not seriously influence these results, so that one factor can be used in practical design. While Markl's work represents a marked advance in practical design approach, it must be conceded that the experimental data are rather limited. More work would certainly be desirable to eheck its applicability to large-diameter piping and to reducing-size branch connections.

Data regarding the performance of full-size ASA standard welding tees under repeated external loading can again be found in Markl's papers. Assuming that the metal thickness available in the crotch zone and the crotch radius are the controlling variables, the characteristic variable was expressed in the form:

h = (~)2.5 '. (1 + ro) (3.23)18

t rm rm

l~his equation is obtained by simply substituting ¢ = 45° in eq. 3.18 for single miter bends.

16The design formula of the Code is given in a. modified form of eq, 3.22.

17The quantity tin cq, 3.22 is the thickness of the pipe used in the stress calculation. The intensification factor of eq, 3.14 is again applied to this pipe. These results refer to fullsized intersections and should be used with discretion for other cases.

181n the Code, the recommended formula for the charae-

unreinforced branch connection as compared to the intact header of the same size. As an example assume standard pipe sizes, w = t in. for 4 in. branches or smaller, w = i in. for larger branch sizes, and a corrosion allowance of 0.1 in. The "pressure reduction ratios" may then be obtained for various header and branch sizes from eq. 3.24. The results of these calculations for the specific case are tabulated in Table 3.3 except that branch connections not exceeding 2 in. or 25% of the header size are shown with a 100% rating since the Code permits this arbitrarily.

An examination of this table indicates the following trends:

1. For equal-size intersections (proceeding along the diagonal of Table 3.3), the pressure reduction ratio decreases with increasing pipe sizes to a limiting value of 50%. This is in reasonable accord with experimental evidence, although tests carried out on full-size intersections up to 12 in. did not show bursting-pressure reduction ratios below 65%.

2. Increasing the size of the branch connection for a given header (moving from left to right in a given row of Table 3.3) decreases the pressure reduction ratio. Although this trend is borne out by tests, the Code reduction ratio appears conservative for small-size headers, since it permits only 56% of the "intact header pressure" to be applied to halfsize intersections with an 8 in. or 12 in. header, as contrasted to the 90-100% obtained in experiments. More complete and more searching experimental data would, however, be necessary to justify closer evaluation of certain sizes and proportions.

3. An increase of header size for a given branch size over 2 in. (traversing Table 3.3 from top to bottom of a specific column) results in a reduced allowable pressure. This is contrary to the limited experimental evidence, which shows that the bursting pressure developed by a branch connection increases as the diameter ratio between the branch and header pipes becomes smaller. Again more searching experimental data are desirable.

4. The arbitrary Code provision assigning 100% for welded branch pipes not exceeding 2 in. or 25% of the header size, while reasonable from test results, introduces abrupt breaks in allowable ratings. A smoother transition is desirable.

The ASA B3Ll Code rules applicable to oblique branch connections are mandatory for branch angles not less than 45° and when the branch/header diameter ratio is not less than 1/4. These rules, which recognize the higher stress intensification in the acute crotch, require that the replacement area be


Rcinfotcame-01' len.,

The smeller of 2.S', +'.

FIG.3.13 "Area of reinforcement," as specified by the Piping Code, ASA B31.1, Section 6.

PH = maximum service pressure permissible for intact header.

S = allowable stress at operating temperature. t1 = thickness of header required by Code for given size, service pressure, and operating temperature.

t2 = thickness of branch required by Code for given size, service pressure, and operating temperature.

P = allowable pressure permitted by Code for the completed manifold.

tp = thickness of reinforcement pad (if used).

According to the Code, the required thicknesses can be expressed as


t1 = S + OAp ;

pRn Rn

{z = = - ll'

S + OAp Rll '

t PJlRll


S + OAPll

The "area to be replaced" is A = 2t1 (Rn - tn)

For an unreinforced intersection with a branch not heavier than the header, the available excess metal within the "zone of reinforcement" can be given by

AR = 2(tll - td(Rn - tn) + 5in(tn - t2) + w2

Equating AR to A, and using the wall-thickness expressions, yields the following result:

Rlf _ 0.4 tIl




0.8Rll(Rn - tn) + Rntn 0

2 2 - .4

O.4tH(Rn - tn) + in + O.2w

The "pressure reduction ratio," p/PII, expresses the decrease in allowable pressure for the completed


equal to the area removed from the header multiplied by a factor of (2 - sin a) where a is the branch angle. The rules (for branch/header ratios of 1/4 and larger) make no distinction between full-sized and reducing branches, a practice which appears somewhat contrary to experience.

3.8 Branch Connections: Practieal Considerations and Summary

In the foregoing sections giving the highlights of available test and analytical data, it has been noted that stress concentrations can be expected to be present around all circular openings and branch connections, and that even for the most carefully designed reinforcement, the factor' is· not likely to subside below 1.3. The question naturally arises as to the practical significance of such effects. The answer at present must be sought primarily in experience. Service experience using nominal design allowable stress values (as established by Section 3, Oil Piping, of the ASA B3Ll Code for Pressure Piping, and the concept of the simple replacementof-area method) has been reasonably good despite the fact that design and attachment details and fabrication quality used have not always been as good as they should be. This fact may be accounted for by the following considerations:


1. The design stress used provides a considerable margin for local overstress.

2. Highly localized stress can be relieved by local yielding. Such yielding induces local residual stresses of the opposite sign in the off-stream condition, so that the area can operate on a "stress range" basis in the same manner that thermal expansion strains may be absorbed in piping systems.

3. Most applications do not involve a very large number of cycles. Therefore, the design need not insure that stresses be kept at all times below the endurance limit of the material.

4. Experience to date is largely confined to steel, which normally acts in a ductile manner.

Thus, although this experience has been generally satisfactory, those service failures (and all of the laboratory fractures) that occurred in pressure vessels and pipe lines to date have, almost without exception, been shown to originate at branch connections 01 local attachments. Therefore, good engineering demands that careful judgment be exercised when selecting designs and fabrication details, and that fabrication quality be adequately controlled. Poor fit-up, welding, and lack of root penetration on welded branches, can easily furnish added stress-raising effects which can lead to failure.

Table 3.3 Pressure Rcduction Ratios in Per Cent for Unreinforced Intersections*
~ e' Il" 2" 2l" 3'1 4" 6" S" 10" 12'1 14" 16" lS'1 2011 2411
111 100
Ilil 100 100
2'1 100 100 89
2JJI 80 71 67 66
s« 77 69 65 64 63
4" 100 65 62 62 61 59
6" 100 100 59 60 58 57 60
8" 100 100 100 58 57 56 58 57
10" 100 100 100 56 55 54 56 55 55
12/1 100 100 100 56 55 54 56 55 55 54
1411 100 100 IOO 56 55 55 57 56 55 55 54
16" 100 100 100 56 55 54 57 56 55 54 54 53
18" 100 100 100 56 55 55 57 56 56 55 54 54 53
20" 100 100 100 56 55 55 57 56 56 55 55 54 53 53
24" 100 100 100 56 55 55 57 56 56 55 55 54 53 53 52 "Bnsed on the Code for Pressure Piping, ASA B3l.1 for; standard weight pipe with 0,1 corrosion allowance; leg of fillet weld « 1" {or branches 4" or smaller, and ~II for larger branch sizes.



As design stress levels and temperatures increase, greater attention must be given to reinforcement details. The same is true when design stresses are raised in proportion to enhanced physical properties of material obtained by cold work, since the effect of localized stresses becomes much more serious.

Full-size 900 branch connections are difficult to fabricate by welding without appreciable distortion, particularly when a pad-type reinforcement is used. This difficulty increases with the size of the header. It is best to avoid such connections wherever it is economically justified. In critical service, welding tees, when available, should always be used in preference to fabricated welded intersections. Integral reinforcement obtained by using a heavier pipe for the header (or for both header and branch) is generally preferred to built-up construction and is satisfactory for most applications. Sharp corners at the intersection should be avoided by the use of concave weld fillets. Fabrication presents considerably greater problems as the size of the branch relative to the header is increased and must receive special care when this ratio exceeds 50%, particularly for header sizes above 12 in. On headers of large size with small openings (branch to header diameter ratio less than 50%), the method of reinforcement should be guided by the principles established for the reinforcement of nozzles on pressure vessels.

The greatest benefit from a given amount of reinforcing metal will be obtained by concentrating the reinforcement near the finished opening. Flow considerations permitting, the effectiveness of the reinforcement can be increased by application of the reinforcing metal to the inside, as well as outside, surface of the header. The use of elliptical nozzles may be considered for extremely severe service conditions, since they extend the possibility of reducing stress concentrations to the limiting value of l.

For the design of special heavy-walled . fittings in critical service The M. W. Kellogg Company has found the rather simple design approach given in Fig. 3.14 satisfactory. This is, in effect, an analysis designed to control the average membrane stress within the chosen limits, and includes a correction for non-uniform stress through the wall thickness equivalent to using the mean diameter cylindrical hoop stress formula instead of the inside diameter formula. Therefore, it assures a fitting strength roughly equal to the connecting pipe. The regions over which the pressure area and metal areas are averaged are arbitrarily selected as being in reasonable accord with experience.

The use of gusset or rib stiffeners is not recommended, due to the high stress concentrations likely to exist at their ends or adjacent to the attachment welds. They are even more objectionable on, hot piping, since the ribs act as cooling fins and local thermal stresses are imposed; if such stiffeners are used on hot piping the thermal effects should be minimized by the application of heavy insulation.

The effect of structural loadings other than pressure and cyclic loadings must be given due consideration. Markl's work in establishing suggested stress intensification factors for piping flexibility analyses is a good start, but more work is necessary on other sizes and reducing branches. Where an individual flexibility analysis is not warranted, yet expansion stresses are expected to be at or near Code levels at the branch location (with the moment loading being carried through the branch), it is recommended that branch connections be reinforced to develop the full strength of the header, even if the operating pressure may not require it.

The selection of design and fabrication details as well as the methods and extent of inspection must be in line with the expected severity of service. Weld details which minimize distortion and promote best root-welding conditions are to be favored. As an example, setting a branch on a pipe and welding it before the hole in the header is cut will reduce distortion when the branch pipe is large compared to the header; set-on construction also permits the use of a backing ring.

Regarding inspection methods, a magnetic particle examination should be favored for magnetic materials; for non-magnetic materials, a penetrant oil examination is quite practical and is recommended for important services. Radiographic examinations of branch attachments are being increasingly used as a quality control check; although they are useful in controlling the general quality level of an individual operator's work, such radiographs cannot be interpreted to assure the absence of cracks unless many angled shots are taken. An indiscriminate appraisal of radiographic examination may create an unwarranted degree of assurance regarding absence of harmful defects.

3.9 Corrugated Pipe

As pointed out in the introduction and in Chapter 7, straight corrugated pipe provides intermediate flexibility' between a rigid piping system and an p.xpansion joint system . Its use may be advantageous where acute space limitations exist, or where reactions on equipment attendant to stiff or large-size






p{E+tA) A


p(f+i:B) Sa;; ---8--










FIG. 3.14 Special heavy wall fittings: check of reinforcement for internal pressure.



pipe must be reduced without further addition to pressure drop, process problems or similar factors. The fact that corrugations greatly increase the flexibility of a straight cylindrical tube has long been appreciated. However, it is less well known that this reduction in stiffness is obtained by the introduction of bending stresses, the level of which must be controlled for satisfactory service; also, it is not always appreciated that a corrugated bend may be less flexible than a plain pipe bend due to the fact that the corrugations resist ovalization.

Corrugations were initially obtained by hot rollforming processes such as are used for shaping the flues of Scotch Marine boilers. This imposed limitations on the depth and pitch so that resort was made to uniform localized heating and controlled collapsing. This process resulted in some upsetting and a sharper radius of curvature at the crown. More recently, equipment has been designed which provides rolled corrugations of greater depth. The mean diameter of rolled corrugations is usually that of the original pipe, while those formed by controlled collapsing have a mean diameter roughly equal to the initial outside diameter of the pipe. Corrugating can only be accomplished on straight pipe, so that bends must be formed afterward. Creased bend construction is usually limited to sharp radii bends, commonly 2 to 3 diameters in radius; these are formed by heating plain pipe on one side and bending so as to bulge out the corrugations on the inside of the bend.

Early tests [21, 65] showed that, for bends having a five-diameter radius, corrugated construction provided no greater flexibility than plain bends, and that for smaller radii corrugated or creased curved pipes are usually less flexible than plain bends; also, that the torsional stiffness of corrugated pipe was slightly greater than that of straight pipe of the same nominal diameter.

In the foregoing tests, as well as later ones [66], it was established that a corrugated bend derives its flexibility in bending or direct loading mainly from a change in axial length (through an increase on the tension and a decrease on the compression side), as compared to a plain bend, which derives its increased deflection from ovalization of the cross section, and the attendant modified stress distribution. A creased pipe bend takes an intermediate position between these extremes, deriving its flexibility on the plain portion by ovalization, and on the creased portion by change in length.

Consistent flexibility values for corrugated and creased 6 in. diameter pipes were obtained from static

tests by Dennison {57], who related test results to the calculated values for plain pipes of the same dimensions.I'' as given by elementary beam theory. Corrugated bends were found to have higher flexibility factors than creased bends, although a good approximation for both configurations was 6.0. Nominal stress intensification factors (denoting the ratio between the endurance limits20 of small polished specimens to that of the actual piping component) were obtained from fatigue tests, indicating an average factor of 8 for both corrugated and creased pipes of the type tested. With one exception, incipient cracks in corrugated pipes originated on the inside surface and penetrated outward, while on creased bends the cracks always initiated on the external surface.

The foregoing tests have led to the acceptance of an assumption of uniform flexibility factors for commercial corrugated or creased pipes. For the stress intensification factors, however, it was felt that unduly conservative values resulted by basing the comparison on the endurance limit of polished bars. Therefore, further tests were carried out by Rossheim and Markl {29], in which the fatigue strength'" of plain tangents was taken as a basis of comparison for stress intensification factors. Based on these experiments a stress intensification factor of 2.5 was suggested as reasonable for "non-cyclic" service (i.e, less than 20,000 stress cycles), whereas for "cyclic" service (up to 500,000 reversals) a stress intensification factor of 5.0 was proposed. A flexibility factor of 5.0 was suggested as a conservative value for average commercial creased or corrugated pipe. These values form the basis of the current suggested values in the ASA B31.1 Code, viz: a flexibility factor of 5.0 and a stress intensification factor of 2.5 for usual commercial corrugated or creased components under bending or direct axial loading. The Code also suggests a flexibility factor of 0.9 and stress factor of unity for torsional loading.

In contrast to the uniform values recommended by the Code for all sizes and shapes of corrugated pipes, the actual flexibility and stress factors are a function of the size and wall thickness of the pipe

19The term "plain pipe of the same dimensions" refers to a pipe having the same diameter and wall thickness as that used for making the corrugated (creased) pipe in consideration.

20 "Endurance limit" denotes the alternating stress which a specimen can infinitely sustain in a fatigue test. In nctuul tests 2 X 106 cycles are taken to be equivalent to "an infinite" number of cycles.

21"Fatigue strength," as opposed to "endurance limit," denotes the average maximum alternating stress which :L specimen can sustain for a given number of streas cycles.



and particularly the depth and pitch of the corrugations. Test results show that an increase in the depth of the corrugation will improve its flexibility, but increase the stress intensification factor.

The greatly simplified case of a curved beam shown on Fig. 3.15, which is obtained after segmenting a corrugated pipe into strips of unit width, can be analyzed readily. This analogue will indicate higher flexibility and lower stresses than those existing in the actual structure. Nonetheless, it is useful in roughly predicting the influence of dimensional changes, and for comparison with established service. The results obtained from this analogue, assuming that v = 0.3 and the corrugation pitch is 4r, are:

Flexibility factor

= 1r[(3r/t) + .09]

Theoretical stress intensification

factor = [(6r/t) + 1]

Stress intensification factor com-

pared to plain pipe (Code basis) = O.5[(6r/t) + 1]

These relations are plotted in Fig. 3.16, which shows the strong. dependence of both of these design factors on the ratio of r/t.22 An increase in the pitch of the corrugations with other items unchanged would decrease the flexibility factor; likewise, a change in the shape of the corrugations to a more rigid shape would decrease the flexibility, but also decrease the stress intensification factor.

A detailed analytical evaluation of stresses in corrugated components is extremely difficult. Moreover, if the manufacturing tolerances and variations ill shape or between successive corrugations are considered, the theoretical treatise becomes impractical. An approximate solution was developed by Donnell

22Due to the simplifications assumed here, the ratio of pipe radius to pipe wall thickness R/t does not enter the solution.

r~ 4,

.z, ( I M

M, FIG. 3.15 Analogue representation for analysis of corrugated pipes.

100 \000
70 700
010 400
20 :100
10 100
7.0 ~70
4.0 ~~
2.0 20
1.0 10
0.7 7.0
0.4 A.O
0.2 2.0
0 Piping Codto A •• ,bl;,y Fed«

1JwoI ~ng. 01 Com ... ,clol Co~Hons





2 ,/~ Ralio of Comrg<rt;"" 00pII. 10 p;~ Wall Thkk,,1U (Pitch c Tw'''' Oop'h)

FIG. 3.16 Stress intensification' "and flexibility factors In "analogue" solution for corrugated pipes.

{68] for v-shaped and semicircular corrugations under concentric axial loads, and the results were extended by inference to corrugations of elliptical and sinusoidal cross sections. Tests on thin-walled corrugated pipes, having corrugations in reasonable accord with the specific shapes analyzed, were in good agreement with theory. For corrugation shapes as normally produced in heavier-walled pipe, the analysis can be accepted only as a rough approximation.

A more recent theoretical approach [91 treats the effects of both axial loads and internal pressure, but restricts the analysis to thin-walled cylindrical bodies of relatively large diameter, so that the results are applicable to corrugated light-gage expansion joints rather than pipe. This analysis is not readily adapted to design, since the final results are given in terms of unfamiliar functions, whose numerical values are not tabulated in standard mathematical tables. For a more detailed treatment of the expansion joint bellows, reference should be made to Chapter 7.

In summation, corrugated pipes have practical application when added flexibility for stress or endreaction reduction must be obtained in extremely limited space, thereby permitting the retention of



a rigid piping system and avoiding the use of expansion joints. Its use is best confined to straight lengths, since it will have little, if any, advantage over plain pipe when used for bends. The same may be said for creased bends, which offer no significant advantage over plain bends in flexibility and may involve higher stress intensification. Corrugated pipe, properly designed, is capable of carrying the axial internal pressure thrust in common with straight pipe, but it is important to note that the stress intensification factor applies to the longitudinal pressure load, as well as other loadings; yielding, or creep, will result if the combined static or dynamic loadings exceed established limits. For applications in the creep range an accurate evaluation of the stress intensification factor for the particular corrugation used is desirable. Occasionally, corrugated pipe is used to localize plastic deformation which might occur during extreme upset conditions; for such service the range of local unit plastic strain determines the number of cycles which can be sustained (see Chapter 7). It is advisable that limit stops or equivalent means be provided to limit overall yielding.

3.10 Bolted Flanged Connections: General Background

Flanged connections provide for the ready joining or separating of portions of a piping system to facilitate inspection or cleaning, or to avoid in-position welding or heat treatment. Their influence on the performance of a piping system involves evaluation of (1) the effect of the flange as a local component, and (2) the effect of the forces and moments transmitted through a flange on its ability to maintain a tight seal.

Analysis of flanged joints was limited to cantilever approximations until the advance made in 1927 by Waters and Taylor [69}. Combining the elastic behavior of a flat plate with a cylinder treated as a beam on an elastic foundation, they obtained expressions for the circumferential, radial, and axial stresses in flanges with short cylindrical hubs of constant thickness. These theoretical results were reasonably well substantiated by tests, and offered a reliable basis for the evaluation of loose-hubbed flanges (Vain Stone, threaded, lapped) within the ASA range of dimensions [70, 71, 72). In subsequent years this derivation was extended by Holmberg and Axelson {73) to flanges integral with the pipe wall.

These analyses were limited to hubs of uniform thickness, although the desirability of increased hub

thickness at the flange-hub intersection was generally recognized. An exhaustive theoretical analysis of this problem [74, 75} was undertaken by Waters, Rossheim, Wesstrom, and Williams. It included both bolt-moment and pressure effects, and could be applied to straight, single, or double taper combinations of hub contour, with fillets simulated by tangent tapers. A complete analysis, including direct pressure and pressure discontinuity stress, is complex for other than a straight hub; the considerable effort involved in this approach is justifiable only on high-pressure, large-diameter flanges in critical service. For usual services and flange proportions it has been found that the direct-pressure and pressure-discontinuity effects can be neglected, and the flange subjected to a uniformly distributed external moment equal to the product of the bolt load, gasket load, end pressure load, and their respective lever arms. The junction of the ring and hub is assumed to undergo zero radial displacement, the bolt load to be unaffected by changes in pressure, and ideal elasticity to be maintained without yielding or creep. Despite these simplifying assumptions, the analysis has proven adequate for most problems when coupled with.' a suitable choice of design stresses, gasket factors, etc. to provide ample margin for these effects. Originally introduced into the ASME Unfired Pressure Vessel Code on a -permissive basis, its general acceptance soon led to its adoption as a mandatory requirement. One widespread usage is on exchanger flanges, for which it has been approved by the Tubular Exchanger Manufacturers Association (TEMA) and applied to the standard flanges in their rules; it has also been widely used in connection with rerating ASA standard flanges.

Experimental work [76, 77, 78J has shown that the theoretical formulas closely predicted the stresses developed under various loading conditions. It has also been shown that a reasonably uniform gasketload distribution can be expected only when the maximum bolt pitch is a function of the bolt diameter and the flange thickness, and that, within the normal range of flange dimensions, the width of the flange has no appreciable effect upon the load distribution. Taylor Forge & Pipe Works' Modern Flange Design {79} presents the formula in terms of bolt diameter, flange thickness, and gasket factor. This formula originated in the M. W. Kellogg Company and has been widely and successfully used in practical design.

Present ASME flange-stress formulas are stated for flange and hub dimensions assumed in advance, leading to time-consuming trial-and-error solution.



Simplified methods have been developed [80, 81] to aid the designer in quickly arriving at well-proportioned economical flanges. Oth~ practical suggestions are contained in Modern Flange Design [79].

For certain relatively mild low-pressure services, such as water works, thin and relatively wide flanges with soft gaskets located inside the bolts have been successfully used. Such flanges usually cannot be justified by the Code design approach. The explanation of their satisfactory use must be sought in recognition of higher stresses, use of soft gaskets, and the possibility of the flanges contacting each other at the OD, establishing thereby a limiting countermoment before excessive strains are developed. For considerations involved in such special service, reference can be made to the paper by Waters and Williams{82] and the discussion thereof.

The design assumption that the initial bolt load remains constant for any magnitude of the internal pressure has also been explored. It has been shown [74, 76, 83, and others] that a hydrostatic end force may either increase or decrease the initial bolt load, depending upon the relative position of the gasket reaction and the elastic properties of the assembly. With customary flanges, exemplified by ABA Standard BI6.5, the bolt load decreases slightly with application of internal pressure, since the net moment on the flange ring increases, which in turn causes increased flange rotation and a decrease in the distance between flanges at the bolt circle. The bolt stress is at all times a function of the summation of the strains of the entire assembly, and their individual moduli of elasticity. Since the modulus decreases with temperature rise, the bolt load will likewise decrease as the temperature of the assembly is uniformly raised. If the temperature of the components is not uniform, the differential strains will alter the bolt stress in proportion. Except in unusual cases these effects are not of practical significance, since the flange design permits pretightening to a level sufficient to compensate for bolt-load reduction, local yielding, or creep over the period of time established by the material properties and temperature.

An understanding of flange leakage may be obtained by idealizing the assembly as two elastically coupled bodies, the bolts on one hand, and the flanges, including the gasket, on the other. The complete joint is then represented schematically as two springs with different initial lengths and stiffnesses. When the joint is tightened, this initial length difference is eliminated by submitting the bolt spring to tension and the stiffer flange spring to compression. A

superimposed hydrostatic load would have the simple effect of subjecting both springs to equal amounts of added tensile strain. In this simplified representation; leakage would occur when the tensile strain imparted to the flange due to the hydrostatic load offsets the compressive strain set up by tightening the bolts (i.e., the spring representing the flange is under no force and returns to its original length). This concept was presented by Dolan [84] who gave pictorial representation to this interpretation by means of a simple force-extension diagram. Needless to say, the elastic coupling concept is a considerable oversimplification of actual conditions in a flange which must include all components, changing moment and rotational effects, temperature, creep, etc. as treated, for example, in [82J.

Code rules establish two criteria which must be satisfied to maintain a gasketed joint free from leakage. The one establishes a minimum initial unit. gasket seating load, and the other a ratio of gasket load to internal pressure for operating conditions; both are related to the gasket material and construction. As to the effective width of gasket, arbitrary assumptions are made which are related to the flange-facing details arid relative concentration of loading; double this width is used for application of the gasket operating pressure factor.

Actually, the performance of a gasketed surface depends not only upon the elastic properties of the gasket material as influenced by its design details, but also upon the bolt load and deflection of the flanges (initially and under pressure); the maximum gasket load (at the inner and outer edge or at projections); the gasket thickness and physical properties; and the surface finishes of gasket and flange, which determine the elastic and plastic deformation attendant to an initial seal. In actual installations, the varieties of gasket stiffness, surface finish, and imperfections of gasket and flanges, as well as the properties of the contained fluid or gas, can cause wide variation in the minimum load and gasket pressure to maintain tightness. The Code rules for minimum load and gasket factor are, therefore, approximations of average conditions for flanges of usual proportions. For this reason, these rules are not mandatory but merely tentative.

Where very stiff flanges or minimum-width gaskets are involved, the entire width of the gasket may be at essentially the same unit load; conversely, for thin flexible flanges or wide gaskets, the gasket load may be much lower than predicted by Code rules. Extremely soft gaskets, such as gum rubber, deflect directly under the internal pressure, and flow later-

..... _--_._---_._--



ally to equalize and distribute the gasket load. Due to this behavior, they can often be made tight under very low unit gasket. loading or, at tImes, even in the absence of a net gasket load from the bolts. Special facing details or gasket designs to promote selfsealing tendencies are successfully used where the service permits the use of soft gasket material, or where extremely high working pressure can be utilized to provide sufficient load to seal harder gaskets. Some facing details are based on mechanical concepts, such as ring-type joints; others, as for instance, lens rings, benefit from reduced or line contact.

Where elastic conditions are maintained, leakage of a properly fabricated and assembled flanged joint should not occur if the initial bolt load is sufficient to maintain the required gasket load above the longitudinal loads developed by pressure and structural effects, and to compensate for the expected reductions in bolt load due to flange deflection and change in elastic modulus. The influence of structural loading is treated further in the next section. Flange bolts are ordinarily made up at ambient temperature; as the temperature is raised in service, the temperature of the bolts, flanges, and pipe may no longer be the same, either during the transient heating period or in the equilibrium service condition. Since the bolts receive heat through limited contact with the flanges they will respond more slowly to changes; similarly, non-integral flanges, such as the Van Stone type, will lag behind the pipe under temperature change. In the absence of insulation, these temperature differences will be much greater. Where flow temperatures fluctuate rapidly or where external influences (such as rain on exposed flanges) upset the equilibrium between bolts and flanges, joints may leak due to loss of gasket sealing load. This reduction in sealing pressure is traceable either to expansion differences, or to yielding resulting from temporary overload. If serious yielding does not occur the joint will eventually re-establish the same gasket load, although it is possible that the temporary leakage will have caused wire-drawing and prevent re-establishment of a seal,

In actual installations, completely elastic conditions are almost never realized j aside from localized yielding, creep will be present in high-temperature service and to some extent at all temperatures, particularly for non-metallic, non-ferrous, or highly stressed gaskets. Under repeated load applications the degree of yielding or creep with respect to time is increased.

In order to deal with the effects of plastic flow or

creep due to temperature effects, it is necessary to consider both transient and constant thermal conditions. Of these two, the transient thermal state is often the more important, since it will generally lead to higher stresses and a greater amount of yielding. This case was investigated by Bailey {85] for loosering and integrally welded flanges. Making a few simplifying assumptions, he showed that stresses in welded integral flanges were only 60% of those in the less intimately connected loose flanges under transient temperatures. As a continuation of this work, Bailey undertook to investigate the effect of creep upon the elastic stress distribution. Again both the loose and welded integral flanges were considered for various creep strength ratios of bolt material to flange material. The effect of bolt holes was taken into consideration through an analogy with the tensile creep-relaxation properties of a solid strip of metal versus a strip of metal with a series of holes of varying pitch and diameter. The effect of thermal bending moment acting on a joint due to pipe expansion was omitted, on the assumption that under creep conditions external forces would in time be reduced to negligible magnitude. The analysis showed that the tightness duration of the flanged joint (as defined by the time at which the stresses would fall below a permissible level or permit leakage) is a function of flange thickness for given material properties of flange and bolts. It was also found that an optimum flange thickness exists for each joint, which increases as the ratio between the creep resistance of the bolts and flanges becomes greater. These optima were generally greater than required by the Code. As would be expected the analysis indicated the desirability of having high elastic strains in bolts and flanges combined with high creep resistance. These properties are in opposition, since high stresses cause accelerated creep. For given materials certain changes in design will, however, provide increased elasticity without increased stresses, e.g. increasing the effective bolt length; similarly, changing the material of any component to one of equal elasticity, but greater creep resistance, will improve high temperature performance.

Bailey's analytical work was not followed by experiments of sufficient extent to prove or disqualify the conclusions reached. However, an interesting report, including some test data on various aspects of flange design, was published by Gough [86] on this subject.

The fatigue characteristics of various types of steel flanges subject to repeated bending strains received attention in an investigation carried out at



atmospheric temperature by Markl and George [87]. Using a constant-displacement type fatigue-testing machine, on 4 in. standard weight and 0.080 in. wall pipe with 300 lb ASA standard RF flanges, fatigue failure occurred almost invariably in the pipe proper adjacent to the flange attachment, where there is a marked change in contour, and not in the flange or bolts. A few tests were made with 600 psi internal pressure; in these tests leakage well in advance of failure was noted only on threaded joints. Gasket leakage was not experienced when bolts were pretightened to 40,000 psi, although it was encountered when they were tightened to only 20,000 psi. The S-N diagrams of all types of flanges investigated were represented by straight lines on a log-log plot, which were parallel among themselves and with the lines obtained for straight tangents or butt-welded pipes. This made it possible to assign single stress intensification factors to each of the various types of flanges investigated; these results are listed in Table 3.4. The superiority of the welding neck flange is in line with service experience with regard to suitability for critical service.

The relatively poor performance of the lap joint flanges was rather surprising since such flanges have a fairly good service record; the lap thickness used was the same as the pipe wall and the poor results were attributed to inadequate strength of the lap to carry the high bending moments imposed, the lap apparently rocking back and forth on the gasket. In general, stress intensification factors increase with increasing abruptness of cross-sectional changes in the flange at the pipe connection. The welding neck flange with its smooth transition exhibits no perceptible stress-raising tendency, whereas threaded flanges, due to stress concentrations present in the threads, carry an intensification factor of about 2.30. The effect of a seal weld covering all exposed threads, as used in some services, was not investigated. It should also be kept in mind that in elevated-temperature service the load distribution on flange details involving double welds would be less favorable, and that additional thermal stresses would result from temperature differences between the pipe and flange.

For services where creep or severe cyclic effects are present, greater attention must be paid to the reduction or elimination of stress raisers. Fillet radii should be generous, and sharp corners should be avoided. Stud bolts with continuous threads or with unthreaded portions machined to the root diameter should be used in preference to headed bolts, which involve sharp fillets under the heads and the

Table 3.4 Stress Intensification Factors for

Various Flanges

Welding neck flange 1.00

Socket welding flange (double welded) 1.15

Slip-on or forged ring flanges (double welded) 1.25

Slip-on or socket welding flanges (single welded) 1.30

Lap joint flanges 1.60

Threaded flanges 2.30

thread runout. For satisfactory performance, bending in studs should also be held to a minimum.

3.11 Bolted .Flanged Connections: Practical Considerations

Experience indicates that the design rules of the ASME Unfired Pressure Vessel Code are generally entirely adequate for the design of special flanges, with gaskets located inside the bolts, for service under internal pressure. For flanges having full face gaskets, or for any design which permits the development of a counter-moment reaction outside the bolt circle, there is no recognized standard design approach; a special ASME Code Committee is currently (1955) working on this problem.

For piping applications it is necessary to consider the effect of other loadings in combination with internal pressure. These are usually longitudinal forces and bending or torsional moments due to weight, wind, or thermal expansion of the pipe line. By far the majority of piping flange applications in the United States utilize ASA Standard B16.5 flanges. The ASA Standard gives allowable pressure-temperature ratings but offers no guidance as to permissible bending loadings. These flanges have been customarily ·used up to the allowable ratings without any check of their capacity to carry additional loadings. While occasional difficulties due to such loads have been encountered, their service record must be considered very good. Unsatisfactory performance occurred generally only with pipes having an aPI?reciable excess strength or corrosion allowance and a high value of thermal expansion. An example is the use of 150 lb standard flanges with low-pressure, high-temperature heavy-walled pipe. A discussion of the influence of bending and torsional moments on ASA flanges is included in the paper by Rossheim and Markl [29]; results of tests on 4 in. 300 lb standard flanges are given in the paper by Markl and George [87]. ASA flange strengths, when judged by ASME Code analytical methods, are by no means uniform, and the piping designer should be aware that there is greater reserve strength in the smaller sizes and lower-pressure classes than in the larger sizes and higher-pressure classes. Good design prac-



tice indicates the desirability of keeping flanged joints at a minimum for operating and maintenance conditions and, insofar as possible, locating needed flanges in the most advantageous location with respect to applied moments.

For routine design investigations of the effects of loading other than internal pressure on flanges covered by ASME Code rules, The M. W. Kellogg Company has found if is satisfactory to calculate first the maximum load per inch of gasket circumference due to the applied longitudinal bending moment and force. Then the internal pressure equivalent to this loading is determined. Finally the Code design approach is applied, assuming an internal pressure equal to the design internal pres;" sure plus the calculated pressure equivalent to the other loads. The equivalent pressure, expressed in formula form, is: p; = (16MhrG3) + 4F/7rG2, where M = longitudinal oending moment in-Ib: G= diameter of effective gasket reaction as defined by Code, in. ; and F = longitudinal force, lb. The flange is checked for a pressure of P = p; + Pd, where Pd = design operating pressure. Taking the moment on the gasket center line is consistent with analysis and experience which indicates that, with a properly pretightened flange, the bolt load changes very little when a moment is applied, whereas the gasket loading changes appreciably.

The most important point for practical design is to establish a proper allowable stress-for such checks. For steady loading other than thermal expansion the same allowable stress as for internal pressure alone should be used; for temporary (short duration) loading, an increase of 33.3% in the basic design stresses is suggested .. Loading due to thermal expansion can be treated on a "stress range" basis similar to the treatment of thermal stresses in the pipe itself; allowable stresses for both bolts and flange can be established accordingly. From a stress standpoint there should be no question about this procedure. From a flange leakage standpoint the validity of this approach is somewhat questionable, particularly under creep conditions. Nonetheless, it is only by such an approach that the demonstrated capacity' of flanges to take rather sizeable moment effects can be reasonably justified. The M. W. Kellogg Company's satisfactory experience in checking thermal moment effects has been based on an allowable stress for both flange and bolts of t (Se + Sh) as prescribed by the 1951 ASA Standard B31.1 Code for piping

With adequately pretightened flange bolts the thermal moment appears during the first heating

cycle. If the operating temperature is sufficiently high, both the bolt stresses and thermal flange moments will gradually relax due to creep in the pipe line; leakage while hot would then depend on the relative relaxation rates. Assuming substantial hot relaxation, a sizeable thermal moment of opposite sign would develop when the line returns to atmospheric temperature, and leakage may occur if pressure is applied in the cold condition. Thus there are probably factors in the problem not adequately assessed, and whether the increased stress range now permitted by the Code can be applied to flange design without affecting tightness is not established.

The influence of torsional moments may be investigated as indicated in the Rossheim-Markl paper [29]; as shown therein the capacity of ASA flanges to take torsional moment is less than for longitudinal moment. This is generally true of all flanges, unless special mechanical means such as dowels or keys for transmitting torsion are provided; caution is therefore in order when high torsional moments may be imposed. Generally, however, flange leakage is not as much of a problem under torsion as it is under bending.

Occasionally, special applications may warrant a more extensive study of stress-strain relations in the flanged joint. This may be done by adapting the approach presented by Wesstrom and Bergh {S3], and Blick [SS, S91.

When selecting gasket dimensions for hot flanges it is well to make the gasket as wide as can be satisfactorily seated by the initial ASME Code design bolt load, rather than use a narrow width which will just avoid extensive initial yielding or "crushing." This will insure maximum resistance to creep under operating conditions.

In petroleum service applications the tightness performance of high-temperature flanges is usually improved by leaving the flanges uninsulated and providing a weather shield only. The flange and bolts then operate at a lower metal temperature and relaxation is slowed. Where heat loss requires it, the shield can be lightly insulated. In power piping, however, heat loss is much more a matter of concern, making full insulation generally a necessity.

By -Iar the greatest leakage troubles with flanged joints arise due to rapid temperature changes or quenches which create sizeable temperature differences within the flange components. Where these conditions can be anticipated, flanges are preferably avoided; if used, great care is warranted in their selection and location. The mating of dissimilar types of flanges, such as Van Stone and integral



types, generally tends to exaggerate difficulties arising from temperature differences.

The problem of dissimilar flanged joints is briefly discussed in Section 3.12.

3.12 Joints Between Dissimilar Materials

Individual piping systems may involve more than one material, or may be connected to equipment of different metal analysis, so that the influence of intermediate or terminal joints between materials of different physical properties must be considered. A principal factor influencing these dissimilar joints is the difference in expansion characteristics; others are variations in hardness, electrolytic potential, structure, ductility, and stiffness.

The transition in material may occur at a bolted flange, a threaded coupling or union, a rolled connection, a weld, or at a special transition piece. Potential difficulties in the form of leakage or failure due to thermal cycling are related to the range and frequency of temperature change, the differences in material properties, and the details of the dissimilar joint.

For low-temperature service, all of these types of connections are successfully used. As the temperature is raised, threaded and rolled joints require that the higher expansion material be inside. With further increasing temperatures, the elastic stress interaction is no longer maintained; this leads to leakage and finally structural failure. Seal welding does not permit appreciably higher temperatures, as structural strength is not usually improved. All of these forms of connections involve significant stress raisers and localized areas of high deformations. Hence, where repetitive cycles of wide temperature range are involved, the probability of fatigue failure dictates against their use.

For flanged joints consisting of integral buttwelded flanges of dissimilar material, (but each of like analysis with the pipe to which it is attached), leakage is dictated by the initial bolt stress and gasket sealing properties, as compared with the differential expansion between the bolts on the one hand, and the flange and gasket on the other. With certain types of gaskets, leakage may also depend on the differential radial expansion at the gasket center line. By using lapped or Van Stone flanges, the flange and bolt temperatures are reduced, and the flanges and bolts can be made of the same type of material, with differential expansion limited to the Van Stone lips and gasket. The gasket load must be sufficient to restrain the relative radial movement of the lips, or the gasket must be of a

design which will maintain a seal under differential expansion movement. While severe local stresses proportional to the restraint are caused in the lips (which must also carry the bending due to pressure and structural loading) , these stresses can and should be confined to wrought, carefully contoured material, so that reasonable fatigue life and satisfactory tightness can be obtained for high temperatures. Dissimilar flanged joints of special design using bellowstype gaskets or pressure sealing have been utilized in power plant service [90J. For joints between ferritic and austenitic steel, the design selected should be such as to permit the use of ferritic alloy bolts; austenitic bolts are unsatisfactory because of their low yield strength and high coefficient of expansion.

Bolted joints generally are avoided for extremetemperature service, a construction not involving gasket sealing being preferred. With welded construction, leakage is no longer a factor. Satisfactory service in this case depends upon the local strainrange differential, on the number of applications of this strain range, and finally, upon the metallurgical factors associated with presence of a weld.

Using the most common material combination as an example, the austenitic steels have a coefficient of expansion of about lOA X lO-6 in./in./F (within the temperature range 70-1200 F), whereas the low-chrome-alloy steels have a coefficient of about 8. X 1O-ll in./in./F. When joined with a sharp interface, this difference will induce stresses when the joint is heated or cooled. For a butt joint between pipes of equal thickness the maximum thermal stress will be the circumferential (hoop) stress developing at the junction, which has a magnitude of

IJ = ~E AT Ao:


Here E represents the modulus of elasticity (assumed identical for both metals), AT the temperature change, and Ll.o: designates the difference between the coefficients of expansion for the two metals. Applying a stress-relief at 1200 F to the junction and cooling the pipe then to 100 F, Ll.T becomes 1100 F, Ll.a:: is equal to 2.4 X lO-ll/F. Taking now E = 29 X 106 psi, a circumferential stress of 38,300 psi is calculated (tensile in austenitic material and compressive in ferritic material).

The foregoing analysis assumes a thin cylinder and evaluates only the differential radial deflection at the mean radius. In addition to this effect, there is also a discontinuity at the interface due to the differential change in thickness of the two materials which will introduce further local radial stresses of equal magnitude, tensile in the one material and


AI1!.lenilk Steel




Ir'lsid!!! and outside- grQl,llId s.mooth and wb'$.lon.'itilly flLtsh alter ..... e-!-di:tlg

FIG.3.17 Dissimilar weld joint, 150 interface.

compressive in the other. The resulting state of equal biaxial stress will raise the calculated stresses by a factor of 1/ (1 - v). A more detailed analysis would disclose some edge-bending stresses across the junction.

A simple butt weld approaches the above assumptions and appears at least the equal of other possible details from a standpoint of stress magnitude. Stress is, however, only one factor determining performance; its influence must be weighed along with that of other factors. One of these factors is the particular detail of the joint. For example, connecting an austenitic pipe branch to a ferritic header would force almost all of the differential strain into the austenitic part. This would increase the maximum stress by a factor of about 3.G, <1S compared to the simple butt weld.

A1, in the case of overall thermal expansion strains in piping systems, the performance of dissimilar joints would be dependent on the number of cycles and the strain range per cycle. While such joints generally give satisfactory service in constant temperature operation with relatively few temperature cycles and an absence of sudden quenching conditions, many joints subjected to more severe conditions have failed. Investigations conducted on this subject [27, 91, 92, 93], and experience indicate that metallurgical factors and flaws seriously affect performance when associated with plastic deformation due to yielding or creep. The heataffected zone on the ferritic side of the dissimilar weld has been shown to be the most critical zone, due possibly to reduced ductility of the mixed analysis in the fusion zone and metallurgical changes during the course of the test or service which result in strain concentration at this location. From a stress standpoint the superposition of internal pressure loading and external longitudinal loading reduced the number of cycles which could be withstood in a hot fatigue test at constant temperature. While all welds withstood a large number of cycles, a somewhat improved performance was obtained by

changing the weld bevel so that the heat-affected zone on the ferritic side was more inclined (Fig. 3.17). These tests did not, however, include the effect of differential expansion stresses, so that improved performance resulting from inclining the ferritic heat-affected zone may in this case be attributed largely to the lower axial stress component in this critical zone.

When Carpenter et al. [92] changed from a hot fatigue test to a thermal quenching test, the number of cycles which could be withstood dropped drastically, failures being experienced after7G to 318 cycles, as compared to 89,000 or more for the hot test. Weisberg and Soldan {931t in a separate series of tests, obtained no failures after 100 quench cycles. Under such drastic quenching appreciable additional thermal stresses are introduced because of transient temperature differences across the wall thickness.

The possibility of improving the performance of dissimilar joints by inclining the bond line originated with The M. W. Kellogg Company and was aimed at a largely longitudinal interface rather than one transverse to the pipe axis. This method has been incorporated in the design of special joints made by the Kelcaloy process (Fig. 3.18). Tests [91, 93] and service experience over a period of six years show excellent performance. Whereas differential expansion stresses are likely to be somewhat higher than in a simple butt joint, the primary advantages over a conventional weld or a weld simulating the construction lie in the essentially longitudinal interface and in the unique manufacturing process. 23 The physical properties of metal deposited by this process are consistently superior to those obtained

23Described by Blumberg and Bunn in their discussion of Weisberg's paper [91].

tnterfcce AIJ51cnitic SI-001

In,.rface Au.t.nitit SI •• I

FIG. 3.18 Kelcaloy transition pieces.


by ordinary casting or welding methods. Austenitic sections, for example, are significantly free of microfissuring. In addition, progressive=and rapid solidification around the entire circumference of the bond zone occurs simultaneously, resulting in greater uniformity, minimum residual stresses, and less acute material transition and heat-affected zones.

The Kelcaloy process is also being used to produce joints with a simple butt bond substantially transverse to the pipe axis. Their principal advantage over welded joints again lies in the metallurgical superiority and relative soundness inherent in the process of manufacture. Additional advantages are: only one heat-affected zone compared to two in a conventional weld, and adaptability of the process to produce and closely control special chemical analyses. As an example of the latter advantage, carbon migration at the interface (which has been experienced at ferritic-austenitic junctions and which hastened some failures) can be combated by introducing a carbide stabilizer, such as columbium, into the chrome-molybdenum steel, leading to an analysis which is not generally available.

While this discussion has emphasized that metallurgical aspects greatly influence dissimilar weld performance, detailed discussion of this subject is not within the scope of this book. Principal factors, however, can be listed as follows:

1. Carbon migration resulting in a carbon-depleted zone in the ferritin steel near the austenitic weld interface.

2. Formation of sigma phase in the austenitic material near the interface.

3. Abrupt change in structure and physical properties of weld metal and heat-affected zone resulting in a "metallurgical notch."

4. Tendency of austenitic weld deposits toward microfissuring.

5. Oxidation or other corrosive notching at the ferritic material junction accelerated by local strain.

The discussion in reference [92] will be found interesting and instructive. There is still a great deal to be learned about austenitic and dissimilar-joint welding and the service performance of austenitic welds. The same can be said in general about the high-temperature performance of the heat-affected zones of all types of welds under plastic deformation and creep conditions. Where weld difficulties have been encountered in service, the preponderance of cracking has been associated with heat-affected zones.

In important practical applications of dissimilar joints for high-temperature service, no matter what


type of design and fabrication technique is decided upon, every effort should be made to eliminate mechanical stress raisers. Weld reinforcement should be built up to effect an anneal of the preceding layer and then should be removed without notches or other surface stress raisers near the weld; machining is preferred where practicable. Backing rings should be avoided, and controlled inside contour welds (K-Weld) should be used where back welding cannot be accomplished. The joint should be examined for soundness by the best nondestructive methods applicable.

3.13 Other Components

Various components other than those specifically mentioned in individual sections of this chapter may be encountered, but insofar as their influence on the flexibility and fatigue performance of the system is concerned, the principles outlined in this chapter usually can be applied as necessary. Some deserve at least brief additional comment.

A valve should be able to carry loading from attached piping similar to a standard tee of comparable pressure-temperature rating, but as a further consideration should be sufficiently free from warpage or distortion to permit operation and tight shutoff. These problems are within the province of the individual manufacturer; their engineering has advanced considerably in recent years, particularly for high-pressure, high-temperature service. As in the case of flanges, care should be exercised in using low pressure rating valves with relatively heavy walled pipe, since the imposed moment as erected or due to expansion may be beyond their structural capacity. For large-diameter piping, valves are sometimes used with venturi ports (particularly when motor operated) or standard valves one or two sizes smaller are used with reducers in conventional lines. The bodies of venturi valves are usually designed with consideration for the moment which may be applied by the larger piping. Similarly, when using standard smaller valves, such piping moments must be given consideration.

Flanged fillings of either cast or wrought steel (ASA Standard BIG.5) are capable of developing the full structural strength of their flanges; for further comments on the moment capacity of flanges see Section 3.11 of this chapter. This applies to tees, crosses and elbows. The pressure rating of such fittings is given in the ASA Standard and commented on in Chapter 2.

Screwed and socket weldingjittings (ells, tees, crosses, unions, and couplings), whether cast or wrought



M = Applied Moment (in Un.) ---


p interflOl preuur .. psi


F .. EJdomal axial end force (lb s. ) (po>itive in diredion shown)

(Total end lorce;: F pl.; hydro'talic end load due to pres.ur. p)

FlO. 3.19 Conical transition.

steel, are similarly furnished to nominal pressure ratings and are limited to small sizes and generally to moderate service conditions. Screwed joints are obviously limited in their capacity for transmittng torsional moment. Actually, in plumbing practice they are often relied upon to relieve thermal expansion by permitting a small angular rotation of one thread upon the other, a practice, however, which not infrequently results in leakage. In tension or bending, screwed fittings can be depended upon to be equal in strength to unthreaded pipe of the same rating, but due allowance for metal removed in threading must be made in determining the wall thickness of the pipe. For cyclic effects (mechanical or thermal), screwed joints involve the stress-raiser effect of the threads, an effect not entirely eliminated even with heavy seal welds. At the higher temperatures, seal welding is usually necessary to prevent leakage. Socket welding fittings also involve the stress intensification effect of fillet welds, but are superior to threaded joints if the welds are adequate. Threaded and socket-welded fitting joints would be expected to involve stress intensifications in line with the fatigue test results obtained for flanges of threaded and single-welded socket types.

The thicknesses of blind flanges in ASA Standard B16.5 are the same as those of functional flanges. For nonstandard cover plates, the rules of the ASME Unfired Pressure Vessel Code may be used; it should be mentioned, however, that the ASA blind flanges are of lesser thickness than those resulting from the application of these rules. EUipsoidal

welding caps are commonly used, as presently covered by ASA Standard BI6.9. Individual applications for larger sizes or special shapes, including flat heads, may be checked using the rules of the ASME Unfired Pressure Vessel Code.

Markl [12] has fatigue-tested smoothly contoured commercial reducers and finds a stress intensification factor of unity justified. For the design of special conical reducers reference can be made to the interpretive report of the work of the Design Division of the Pressure Vessel Research Committee {94] on pressure vessel heads. For the particular case of a sharp cone-to-cylinder junction the local stresses at the intersection can be closely approximated by using the familiar beam on an elastic foundation analysis and treating the cone as though it were a cylinder having a radius equal to the meridional radius of the cone at the junction. The resulting stresses due to an internal pressure p and external loads F and M are given by the following formulas (with Poisson's ratio 1', taken as 0.3):

Outside intersection (Point (1) in Fig. 3.19)

{Sll - =Fl.816C3 + (pR/2tn cos a) Cone SCI: -C1 + (pR/tn cos a) =F 0.546C3

{Sl = =F1.816nZC3 + (pR/2t) Cy1.

se = - C2 + (pR/t) =F 0.546n2C3

Re-entrant intersection (Point (2) in Fig. 3.19\

{Sl = ±l.816C3 + (pR/2t2n cos a) Cone Sc: == C1 + (pR/t2n cos a) ± 0.546C3

{S12 = ± 1. 8 16n2C3 + (pRz/2tz)

C~.. 2

Sc2 = C2 + (pRZ/t2) ± O.546n C3

In these equations upper signs refer to the stresses in the outer fibers and lower signs to those in the inner fibers. The subscripts land c refer to longitudinal and circumferential stresses, respectively; a positive sign denotes tension. The constants appearing above are given by the following expressions:

C1 = ~'1 [C5( vncosa + ~2)

- c, (2vn cos a + 1 + 1~2)]

C2 = ~4 [ Co ( vn cos a + ~z)

+ CIl (n2 + 1 + V: 2 )]

n cos a



C3 = }c [CS(Vn cos a + 1) + C6(nZ - 1)] n "


C" = nZ + 1z + 2 (vn cos a + 1 + Vi)

n n cos a


Cs = 2.57 -- -- + - + - tan a for lU-

t1.5 2 211"R 1I"Rz '

tersection (1)

v'R;[PRz F M ] .

= 2.57 --"'l5 - + -- + --Z tana,form-

t2 . 2 211"R2 1I"Rz

tersection (2)

C6= 0.85 pR (1 - __ 1_) for intersection (1)

t n cos a

= -0.85 pR2 (1 __ 1_) for intersection (2)

t2 n cos a

n = ttlt for intersection (1) = It/tz for intersection (2)

It has been assumed in the above that the intersec-

tions are far enough apart (about 2 fii'i': min) so


that their local effects do not influence each other significantly. The maximum fiber load due to the external moment is taken as though it were uniform around the circumference; this approximation is considered to be on the safe side.

The special case when il cos a = t is of interest for intersection (1). When n = l/cos a the stress formulas for this intersection reduce to:

{Sit = =t=3.63C7 eos2 a + (pR/2t) Cone

Set = - (1 + cos2 a ± 1.089 cos2 a)C7 + (pR/t)

{SI = =t=3.63C7 + (pR/2t) Cyl.

S( = - (1 + cos2 a ± 1.089)C7 + (pR/t)

where C7

2.57 sin a cos a [VRJ[PR F M ]

= 1+6cos2a+cos'la F 2+211"R+ 'lfR2

For consistent treatment with other stress intensifications in the ASA Code for Pressure Piping, rolledpipe data should be chosen as a basis of comparison. Therefore, the calculated maximum stress as given by the above formulas should be divided by two when comparing with the usual expansion stress limits.

3.14 Piping and Equipment Irrtereffects

In the over-all picture a piping system and the mutually connected equipment, structures, Iounda-

tions, and soil constitute an integrated structural system with equilibrium of interloading effects. Each part of this structural system is influenced by its individual environment, e.g. pressure, temperature, weight, etc., as well as by the effects transmitted from attached parts of the system.

Ordinarily, supporting structures, foundations, and soil are subject only to ambient temperatures, and are sufficiently rigid so that deflections under pipe expansion, etc., are small enough to be neglected. Sometimes, however, temperature rise is unavoidable in . steel structures; slender or high structures may also, in combination with their foundations, involve significant deflections under even moderate reactions. Connected equipment will undergo dimensional changes which may augment or decrease the thermal expansion loading. The fabrication and assembly of such an integrated structural system necessarily involves deviations from nominal dimensions. Hence the fitting of piping, in combination with weld shrinkage, sets up initial internal stresses which at the weakest location of the system may equal the yield strength. All such conditions must be recognized and provided for by the piping designer.

For simplification in analysis, the ends of a piping system are usually considered fixed at the equipment connections. Obviously, this is a limiting condition for the maximum reaction; localized bending or direct loading of the equipment, by causing deflection or rotation, serve to reduce the piping reaction. The result will be an intermediate fixity between fixedand hinged-end conditions. While the conventional assumption of complete fixation may seem unnecessarily severe, it must not be inferred that excessive additional safety results. It is possible to deviate from fixed-end assumptions without increased risk of fatigue damage to connection equipment only when analysis is made of the bending stresses in the equipment whose localized deflections are being utilized. When dealing with rotating or other equipment where alignment is sensitive to distortion, piping can seldom be permitted to exceed the stiffness obtained with fixed-end assumptions.

Considerable misunderstanding on the part of equipment designers relative to piping reactions has existed in the past. Manufacturers sometimes have made it a condition of their warranty that no piping reactions be transmitted to their equipment. In other cases, forces have been limited to unreasonably low values, while completely ignoring the more important effects of bending moments. Such impractical criteria, however, are detrimental to all concerned,



including the equipment manufacturer, as the piping designer is left without a usable guide.

At present, the more progressive manufacturers attempt to provide more realistic load carrying capacities or offer to check their equipment, on large critical units, for the reactions of the proposed piping. The problems which they face should be appreciated by the piping designer. With moving parts and the need for close clearances, all strain must be carefully controlled to avoid misalignment, rubbing, binding, excessive wear, or other maloperation, At the same time, the involved and discontinuous contours usually required are not amenable to a reliable evaluation of either deflections or stresses. Thus, the manufacturer is often forced to rely on the judicious use of experience or the projection of occasional test data to individual equipment.

The magnitude of effects at sensitive equipment should be kept low particularly when cost is not significantly increased by so doing. This objective may be approached by providing local pipe flexibility in complex systems to favor the equipment, using local restraints to take reactions directly or to force deflection into other portions of the system (discussed further in Chapter 8); by an over-all increase in the flexibility of the system; or by a favorable relative positioning of the equipment. The potential influence of effects on equipment may often be moderated by the selection of types which are relatively insensitive to distortion and misalignment. Also, the location of equipment should be such as to keep pipe sizes to a minimum insofar as practicable, a point of particular concern in regard to the larger lines (e.g. pump suction and driver exhaust lines).

In the absence of suitable manufacturer's data or applicable experience, approximations for limiting

reactions on such equipment have been suggested. These are no more than rules of thumb which represent experience alone without supporting analysis. Various factors are used as indices, either individually or in combinations, such as weight, cross section. suction or discharge nozzle sizes (and rating), cubic volume, and pressure shell thickness-to-diameter ratio. For many years piping reaction limits of 1000 to 3000 pounds were specified regardless of size or details of the equipment involved. Such limits are now generally considered meaningless.

Other approaches relate metal cross section to different unit values for resultant forces and moments. Pressure shell thickness/radius ratios are also employed sometimes to establish the potential maximum pressure which that portion of the shell can withstand. A more accurate evaluation of local moment capacity of surfaces of revolution is given elsewhere in this section.

Cubic volume is not a significant parameter. It has been used largely in the absence of equipment weight, by assuming an overall density 2 to 5 times that of water. Weight, where obtainable, is a more suitable parameter and is usually increased by the estimated weight of the contents. In either approach, the weight is considered as the maximum value which the resultant force may attain.

Suction or discharge nozzle sizes provide a more comprehensive index of rotating equipment design, since they reflect equipment size. Pressure may be assumed to maintain a rough balance between pipe and equipment stiffness. Based on a survey of acceptable piping designs for equipment piping, Rossheim and Markl [29] proposed the cube of (pipe OD plus 3 in.) as a criterion to which constants were applied to establish the maximum axial and lateral forces in pounds or bending moments in foot pounds.

Table 3.5 Allowable End Reaction Exerted by Connected Piping on Pumps, Turbine Casings, and Pressure Vessels

Type of End Wolosewick for Maximum Temperature 650 Ft
Reaction Rossheim- 4-Point Support 2-Point Supports
Forces,lb Markl* Actual Maximum Actual Maximum
Moments, in.-Ib Value Allowable Value Allowable
Radial reaction, including weight of pump riser,
etc. 3.25D3 250D 4,000 300D 2,700
Tangential reactions, any direction . 1.50D3 100D 1,500 85D 900
Longitudinal bending moment . 10,000
Circumferential bending moment 60D3 2700D 40,000 1700D 22,000
Twisting moment 18,000 -In this column D denotes the OD of the pipe increased by 3 in.

tD is equal to the sum of the nominal diameters of suction and discharge decreased by 15% lor every 50P increase over 650 F.

Sign up to vote on this title
UsefulNot useful