You are on page 1of 10

Indian Journal of Engineering & Materials Sciences

Vol. 18, October 2011, pp. 351-360

Weld residual stress prediction using artificial neural network and


Fuzzy logic modeling
J Edwin Raja Dhasa* & Somasundaram Kumananb
a
Department of Automobile Engineering, Noorul Islam University, Nagercoil 629 180, India
b
Department of Production Engineering, National Institute of Technology, Tiruchirappalli 620 015, India

Received 26 February 2010; accepted 11 October 2011

Artificial intelligent tools such as expert systems, artificial neural network and fuzzy logic support decision-making are
being used in intelligent manufacturing systems. Success of intelligent manufacturing systems depends on effective and
efficient utilization of intelligent tools. Weld residual stress depends on different process parameters and its prediction and
control is a challenge to the researchers. In this paper, intelligent predictive techniques artificial neural network (ANN) and
fuzzy logic models are developed for weld residual stress prediction. The models are developed using Matlab toolbox
functions. Data set required to train the models are obtained through finite element simulation. Results from the fuzzy model
are compared with the developed artificial neural network model, and these models are also validated.

Keywords: Weld residual stress, Artificial neural network, Fuzzy logic, Finite element analysis

Weld residual stress is a major parameter in the thermal and mechanical phenomena observed in
evaluating the quality of weldments. Quality of weld welding processes. Although the finite element
plays an important role in the performance of a method has emerged as one of the most attractive
welded product as it improves fatigue strength, approaches for computing residual stresses in welded
corrosion resistance, creep life and reduces rework joints, its application to practical analysis and design
and scrap. Due to intense concentration of heat in heat problems has been hampered by computational
source of welding, the regions near the weld line difficulties and also this method of obtaining residual
undergo severe thermal cycles, thereby generating stresses is not feasible for all welding parameters.
inhomogeneous plastic deformation and residual Lee et al.9 used multiple regression analysis for
stresses in weldment. Welding-induced residual prediction of process parameters for gas metal arc
stresses play an important role in the function of welding and Yang et al.10 used linear regression
welded structures. Different experimental methods equations for modeling the submerged arc welds.
for directly measuring welding residual stresses are Due to the inadequacy and inefficiency of the
available like X-ray diffraction1, Neutron diffraction, mathematical models to explain the nonlinear
Deep hole drilling2,3 holographic interferometry4. properties existing between the input and output
All these methods require special equipments and are parameters, intelligent systems such as ANN,
expensive. These techniques are limited in obtaining fuzzy logic and expert system have emerged. ANN
the entire picture of the residual stress distribution technique11 is used to handle problems of
in weldment. In 1971, Ueda5 applied finite element nonlinearity. Jeongick et al. 12 utilized ANN
method to analyse thermal elastic-plastics stress and technique for back-bead prediction of gas metal
strain during welding and Nomoto6 pioneered finite arc welding process. Nagesh et al.13 employed ANN
element method to analyze the thermal stress during to predict weld bead geometry in shielded metal-arc
welding. Muhammad et al.7 investigated the finite welding process. Kim et al.14 applied ANN to predict
element simulation of laser beam welding induced bead height in robotic arc welding. Edwin et al.15
residual stresses and distortions in thin sheets. Andres used ANN to predict weld bead width using
et al.8 applied finite element models to analyze artificial neural networks. Hakan Ates16 applied ANN
—————— technique for prediction of gas metal arc welding
*Corresponding author (E-mail: edwinrajadhas@rediffmail.com) parameters.
352 INDIAN J ENG. MATER. SCI., OCTOBER 2011

It is difficult to control process parameters temperature-history dependent (Fig. 2). Finite element
for welding process as the relationship between the analysis of welding is carried out in two stages as
input and the output parameters are complex and thermal analysis and stress analysis. A non-linear
interrelated. Also, it is very difficult to control the transient thermal analysis is conducted first to obtain
system in real time because the controlled system is the global temperature history generated during the
nonlinear, time-varying, uncertain and fast-response welding process. Convention and radiation boundary
controlled system. Fuzzy control17,18 systems are conditions are applied to the model. The heat flux
effective to handle uncertain, nonlinear as well as is applied over the weld bead as the load input to
dynamic time-varying processes control systems. the thermal analysis. Equation (1) gives the arc heat
Fuzzy logic model for predicting weld pool size in input.
gas metal arc welding processes was developed by
q = (ηa × V × I)/A … (1)
Boo et al.19. Fuzzy logic is applied to control gap
parameters for electro discharge machining20. Tarng where q is the arc heat flux in w/m2, ηa is the arc
et al.21 used fuzzy logic in the Taguchi method for the efficiency, V is the arc voltage in volts, I is the arc
optimization of the submerged arc welding process. current in amps and A is the weld bead area in m2.
Xue et al.22 used Fuzzy model to predict and control Area of the weld bead is calculated by the
the bead width in the robotic arc-welding process. approximations from the empirical relations for the
Neuro-fuzzy method is used to model hot extrusion corresponding welding current, arc voltage and
process23. Xixia et al.24 used SVM-based fuzzy rules welding speed25,26 . Welding speed is incorporated in
acquisition system for pulsed GTAW process. This the analysis using load step options. The governing
paper addresses the development of ANN and fuzzy differential equation for two dimensional transient
logic models to predict the weld residual stress for heat transfer during welding is given by Eq. (2).
consistent weld quality. The validity of the developed
models is verified by confirmatory experiments.  ∂q ∂q y  ∂T
−  x +  + Q = ρc p

Proposed Methodology  ∂x ∂y  ∂t
… (2)
Weld residual stress prediction (Fig. 1) undergoes where qx and qy are the components of heat flow rate
four stages of development: (i) data collection by vector per unit area in the plate (x; y), Q the heat
finite element simulation method, (ii) building ANN
and fuzzy logic models, (iii) training the developed
models and (iv) validation of the developed models.
The validated models are forwarded to predict the
residual stress of weld.
Data collection
In this work, prediction of residual stress of the butt
weld joint through simulated finite element models is
presented. The model is proposed, developed, tested
and validated for various weld conditions. Two mild
steel plates of dimensions 100 × 45 × 10 mm are
modeled to form a butt weld and simulated using
finite element analysis software. General purpose
finite element package ANSYS 5.4 version is used for
both thermal and stress analysis. The following
assumptions used during the analysis: (i) the welding
process is modeled as a single pass weld in this
analysis, (ii) the bottom surface of the weld area is
fixed to prevent body movement, (iii) no penetration
and overfill of the weldments are considered.
The material properties of the weld metal, bead
metal and heat affected zone are both temperature and Fig. 1—Developed scheme to predict weld residual stress
DHAS & KUMANAN: WELD RESIDUAL STRESS 353

Fig. 2—Material properties of mild steel

generation, ρ the density, cp the specific heat and parameters were taken from the AWS handbook.
(∂T/∂t) represents the temperature distribution with The amount of heat input to the model is found as
respect to time which is expected as the output from the product of arc efficiency, voltage and current.
thermal analysis27. A non-linear transient thermal analysis is conducted
The model is meshed with thermal element first to obtain the global temperature history generated
PLANE 55 for the weld bead region and PLANE during the welding process. Convention and radiation
35 for the work piece. The quadrilateral shape of boundary conditions are applied to the model.
PLANE 55 and the triangular shape element of In thermal analysis heat flux is given as the load
PLANE 35 make the developed finite element model (arc heat) to the work-piece incorporating welding
properly meshed with suitable thermal, structural and speed in terms of load step time. At the end of
material properties. The convection and radiation the heat load an extra load step with time equal to
surfaces are meshed with Link34 and Link31. The the cooling time is added without any heat input.
real constant for the convection and radiation In the solution phase, this part of the analysis gives
elements are specified. For the structural analysis, the temperature distribution as the output and this result
thermal elements are converted to its corresponding is stored in a separate file. The temperature
structural elements, i.e., Solid Quad 4 node 42 and distribution over time at a user-defined node is
Triangle 6 node 2. These elements are capable obtained. The defined node is in the work-piece or
of exhibiting structural property without change in in weld bead.
mesh structure used for thermal analysis. Four A stress analysis is then carried out with the
welding parameters are given as the input for this temperatures obtained from the thermal analysis
analysis and they are arc efficiency, voltage, current as the loading to the stress model. The output
and welding speed. The working ranges of the obtained out of this analysis is residual stress
354 INDIAN J ENG. MATER. SCI., OCTOBER 2011

distribution over the work-piece after cooling. Development of proposed fuzzy logic model
Residual stress is obtained at the required nodes by Fuzzy logic model for weld residual stress
using the time history postprocessor. prediction is developed in different stages (Fig. 6).
Residual stresses are recorded in the weld bead The first step in the development of fuzzy logic model
and in the work-piece/weld bead interface (Fig. 3). is to take the inputs and determine the degree to
The required results are tabulated for any node in the which they belong to each of the appropriate fuzzy
work-piece or in weld bead using the time-history sets via membership functions (Figs 7-11). In the
postprocessor (Table 1). fuzzy logic system, the input is always a crisp
numerical value limited to the universe of discourse
Development of proposed ANN model of the input variable. The input crisp variables are
ANN model is proposed, developed and validated welding current, arc voltage, arc efficiency and
to predict residual stress (Fig. 4). It is feed forward welding speed (Table 2). The output is a fuzzy degree
back propagation network trained with Levenberg- of membership in the qualifying linguistic set. The
Marquardt back propagation algorithm. The data fuzzy logic system is based on rules and each of the
required for training and testing the ANN model is rules depends on resolving the inputs into a number
obtained from finite element analysis simulation of different fuzzy linguistic sets. Before the rules
(Table 1). The learning function is gradient descent are evaluated, the inputs are fuzzified according to
algorithm with momentum weight and bias learning each of these linguistic sets. The inputs are fuzzified
function. The number of hidden layers and neurons and the degree to which each part of the antecedent
are determined through a trial and error method, is accommodated for each rule. The input to the
in order to accommodate the converged error. The fuzzy operator is two or more membership values
structure of the proposed neural network is 4-12-13-1 from fuzzified input variables. The output is a single
(4 neurons in the input layer, 12 neurons in 1st hidden truth-value.
layer and 13 neurons in 2nd hidden layer and 1 neuron Every rule has a weight (a number between 0 and 1),
in the output layer). With a learning rate of 0.55 and which is applied to the number given by the
a momentum term of 0.9, the network is trained for antecedent. Once proper weighting has been assigned
10000 iterations. The error between the desired and to each rule, the implication method is implemented.
the actual outputs is less than 0.001 at the end of the The result is a fuzzy set represented by a membership
training process and the command window shows function, which weights the linguistic characteristics
the input test data and output obtained from the that are attributed to it. The decisions are based on
developed ANN model (Fig. 5). the testing of all the rules in fuzzy inference system.
The input of the aggregation process is the list of
truncated output functions returned by the implication
process for each rule. The input for the defuzzification
process is a fuzzy set and the output is a single
number. Fuzziness helps the rule evaluation during
the intermediate steps, the final desired output
for each variable is a single number. However, the

Fig. 3—Stress distribution over butt-welded structure under study Fig. 4—Developed ANN architecture for residual stress prediction
DHAS & KUMANAN: WELD RESIDUAL STRESS 355

Table 1—Training data set obtained from finite element analysis


Arc efficiency Welding Welding Welding Heat Heat Residual stress
speed voltage current input flux Work piece/ Weld interface
% mm/s V A W W/m2 MN/m2
0.8 2 21 170 2856 14280000 67.90
0.8 2.5 21 170 2856 14280000 62.79
0.8 3.3 21 170 2856 14280000 74.50
0.8 4 21 170 2856 14280000 236.30
0.8 2 22 150 2640 13200000 63.00
0.8 2.5 22 150 2640 13200000 61.40
0.8 3.3 22 150 2640 13200000 115.00
0.8 4 22 150 2640 13200000 297.00
0.8 2 24 140 2688 13440000 64.44
0.8 2.5 24 140 2688 13440000 61.00
0.8 3.3 24 140 2688 13440000 108.50
0.8 4 24 140 2688 13440000 286.00
0.8 2 28 130 2912 14560000 69.00
0.8 2.5 28 130 2912 14560000 63.00
0.8 3.3 28 130 2912 14560000 73.60
0.8 4 28 130 2912 14560000 213.00
0.75 2 21 170 2677.5 13387500 60.90
0.75 2.5 21 170 2677.5 13387500 64.20
0.75 3.3 21 170 2677.5 13387500 110.00
0.75 4 21 170 2677.5 13387500 289.00
0.75 2 22 150 2475 12375000 60.04
0.75 2.5 22 150 2475 12375000 65.14
0.75 3.3 22 150 2475 12375000 178.00
0.75 4 22 150 2475 12375000 325.00
0.75 2 24 140 2520 12600000 60.70
0.75 2.5 24 140 2520 12600000 64.45
0.75 3.3 24 140 2520 12600000 143.20
0.75 4 24 140 2520 12600000 319.20
0.75 2 28 130 2730 13650000 65.30
0.75 2.5 28 130 2730 13650000 61.30
0.75 3.3 28 130 2730 13650000 100.10
0.75 4 28 130 2730 13650000 275.99
0.7 2 21 170 2499 12495000 60.36
0.7 2.5 21 170 2499 12495000 64.86
0.7 3.3 21 170 2499 12495000 162.80
0.7 4 21 170 2499 12495000 322.30
0.7 2 22 150 2310 11550000 58.93
0.7 2.5 22 150 2310 11550000 66.05
0.7 3.3 22 150 2310 11550000 254.00
0.7 4 22 150 2310 11550000 300.80
0.7 2 24 140 2352 11760000 58.90
0.7 2.5 24 140 2352 11760000 65.92
0.7 3.3 24 140 2352 11760000 238.70
356 INDIAN J ENG. MATER. SCI., OCTOBER 2011

Fig. 5—Command window showing the input test data and outputs obtained from the developed ANN model for weld residual stress
prediction

Fig. 6—Membership function for arc efficiency


DHAS & KUMANAN: WELD RESIDUAL STRESS 357

Fig.7—Membership function for weld speed

Fig. 8—Membership function for arc voltage


358 INDIAN J ENG. MATER. SCI., OCTOBER 2011

Fig. 9—Membership function for weld current

Fig. 10—Membership function for weld residual stress


DHAS & KUMANAN: WELD RESIDUAL STRESS 359

Fig. 11—An example from fuzzy logic modeling for prediction of weld residual stress

Table 2—Crisp input values used for fuzzy logic modeling developed fuzzy model. Residual stress after
cooling was determined by an advanced solid
Input Crisp values
state X-ray stress analyzer of type AST × 2001
Welding current (A) 140 - 170 which works on solid state X-ray camera’ principle.
Arc voltage (V) 21 – 28 The AST X2001 uses a modified ψ-inclination
Arc Efficiency (%) 0.7 – 0.8 technique28 to measure residual stresses. Stresses
were measured with Cr-Kct radiation yielding
Welding speed (mm/s) 2.0 – 3.3
Cr 21 l-reflection at an angle 2θ = 145° 8″. The
aggregate of a fuzzy set encompasses a range of X-ray voltage is 30 kV and the X-ray current is
output values and so must be defuzzified in order 5.8 mA. The calibration distance D is 49.49 mm.
to resolve a single output value from the set. Experimentation was carried out at Welding Research
Centroid method of defuzzification is used in Institute, Tiruchirappalli, India. Percentage of error
developing the model. is calculated by [(Actual value – Predicted value)/
Predicted value] × 100 (Table 3). The errors in
Results and Discussion weld residual stress prediction by fuzzy model
After developing the fuzzy logic system very rarely exceed by 5% and fuzzy model was
for the prediction of the weld bead width, able to predict with significant accuracy than ANN
it is tested by giving various input values using model. Also the time elapsed by the developed
fuzzy rule viewer (Fig. 10). If the input fuzzy logic model to predict the residual stress at
values are changed, the corresponding output is a particular node in the welded structure is 50% less
automatically obtained from the developed fuzzy than finite element simulation of welding residual
system. The configurations of the computing stress and 30% less than that of ANN prediction
machine used are Intel Pentium IV 1.8 GHz and it is extraordinarily less than the X-ray
processor, 512 MB RAM and 80 GB Hard diffraction method. The developed fuzzy logic
Disk Drive. Confirmatory experiments are done model is forwarded to predict weld residual stress
using X-ray diffraction method to validate the under different weld conditions.
360 INDIAN J ENG. MATER. SCI., OCTOBER 2011

Table 3—Percentage error from of the developed models with confirmatory experiment
Process parameters Residual stress, MN/m2 Percentage error, MN/m2
Welding Arc voltage, Arc Weld speed, Experiment FEA ANN Fuzzy FEA ANN Fuzzy
current, A V efficiency, % mm/s model model model model model model
140 24 0.7 2 58 58.9 61.2 59.05 -1.52 -5.22 -1.77
170 21 0.8 3.3 79 74.5 73.2 74.5 6.04 7.92 6.04
140 24 0.75 2.5 63 64.45 60.11 61.30 2.32 4.8 2.77

Conclusions 10 Yang L J, Bibby M J & Chandel R S, J Mater Process


This paper has explored the application of Technol, 39 (2003) 33
11 Mohamad H H Fundamentals of Artificial neural networks
intelligent techniques for weld residual stress (Prentice Hall of India., Delhi), 2002.
prediction. Data for developing the ANN and fuzzy 12 Jeongick Lee & Kiwoan Um, Opt Lasers Eng, 34 (2000)
model is obtained by finite element modeling of 149.
the residual stress. Results from the fuzzy model 13 Nagesh D S & Datta G L, J Mater Process Technol, 123
(2002) 303.
are compared with the results from the developed 14 Kim I S, Son J S, Park C E, Lee C W & Yarlagadda Prasad
ANN model. The fuzzy model predicts the weld K D V, J Mater Process Technol, 131 (2002) 229.
residual stress with good accuracy. Finally, the 15 Edwin Raja Dhas J & Kumanan S, Prediction of weld bead
models are validated. width using artificial neural networks, paper presented at the
National Conference on Factory Automation, Robotics and
Acknowledgements Soft Computing, NIT, Warangal, India, 2007.
16 Hakan Ates, Mater Des, 28 (2007) 2015.
Authors express sincere thanks to the Ministry 17 Zadeh L A , Inform Control, 8 (1965) 338.
of Human Resource Development, Government of 18 Yen J & Langari R, Fuzzy Logic Intelligence Control and
India for the sponsorship under the research and Information, (Pearson Education, Delhi), 2003.
development programme to undertake this research 19 Boo K S & Cho H S, Control Eng Pract, 2 (1994) 1007.
20 Rajurkar K P & Wang W M, J Manuf Sci Eng, 119 (1997)
work. 770.
21 Tarng Y S, Yang W H & Juang S C, Int J Adv Manuf
References Technol, 16 (2000) 688.
1 Prevey P S, in X-ray diffraction residual stress techniques, 22 Yu Xue I S, Kim J S, Son C E, Park H H, Kim B S, Sung I J,
edited by Mills K, (American Society for Metals, Metals Kim H J & Kang B Y, J Mater Process Technol,164-165
Park. Metals Handbook), 9th Ed, 10 1986, 380-392. (2005) 1134.
2 Schajer G S, J Eng Mater Technol, 110 (1998) 338. 23 Sharma Rahul Swarup, Upadhyay Vikas & Hans Raj K,
3 Weng C C & Lo S C, Mater Sci Technol, 8 (1992) 213. Indian J Eng Mater Sci, 16 (2009) 86.
4 Nelson D, Fuchs E, Makino A & Williams D, Exp Mech, 34 24 Xixia Huang, Fanhuai Shi, Wei Gu, Shanben Chen, Eng Appl
(1994) 79. Artif Intell, 22 (2009) 1245.
5 Ueda Y & Yamakawa T, Trans Jpn Weld Soc, 2 (1971) 90. 25 Mandal N R, Welding Distortion and Control, (Narosa
6 Nomoto T, Finite element analysis of thermal stress during Publishing House, New Delhi, India), 2004.
welding, Ph. D. Thesis, University of Tokyo, 1971. 26 Syngellakis A, Wu S & Mellor B G, Materials Research
7 Muhammad Zain-ul-abdein, Daniel Nélias, Jean-François Group, Computational Engineering & Design Group, School
Jullien & Dominique Deloison, Mater Sci Eng A, 527 (2010) of Engineering Sciences, University of Southampton,
3025. Highfield, Southampton. SO17 1BJ, 37.
8 Andres Anca, Alberto Cardona, José Risso & Víctor D 27 Singerland J Larry, Applied Finite Element analysis, 2nd Ed,
Fachinotti, Appl Math Modell, 35 (2011) 688. (John Wiley & Sons), 1984
9 Lee J L & Rhee S, Proc Inst Mech Eng B, 214 (B6) (2000) 28 Korhonen M A, Lindroos V K & Suominen L S , Adv X-ray
443-449. Anal, 32 (1988) 407.

You might also like