## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

at the University of Missouri-Columbia _______________________________________________________ In Partial Fulfillment of the Requirements for the Degree Master of Science _____________________________________________________ by YU-CHENG SU Dr. YUYI LIN, Thesis Supervisor DECEMBER 2009

The undersigned, appointed by the dean of the Graduate School, have examined the thesis entitled MODELING, VERIFICATION, OPTIMAL DESIGN OF NONLINEAR VALVE SPRING presented by YU-CHENG SU a candidate for the degree of master of science, and hereby certify that, in their opinion, it is worthy of acceptance.

DR. YUYI LIN

DR. ROGER FALES

DR. JAMES NOBLE

ACKNOWLEDGEMENTS

To my advisor, Yuyi Lin, Ph. D, P.E., for his patient guidance and support; To the committee members, Roger Fales, Ph.D, and James Noble, Ph.D, For their review and have time to take part in my thesis defense To my wife, Fengchen Huang, for her self-giving love and encourage; To my family, for their forever support. To Linlin Shen, for her assistance in the engine experiment To Joshua, Ashu, and Ian For the revision of my thesis And to Jiahuan He, for the life assistance

ii

These models are nonlinear partial differential equations. Numerical optimization results obtained in this study also showed that it is worthwhile to introduce more design variables to increase the flexibility in an optimal design process for obtaining better results. variable wire diameter. the construction of an accurate dynamic model is necessary. iii . To verify that the use of finite difference in the simulation process is providing stable and reliable results. predictive dynamic models for variable pitch angle. and to improve the parameters in the dynamic models. in general more complex than the well known and commonly used wave equation. In this study. In order to perform computer aided optimization. the numerical solutions are compared with solutions obtained using ABAQUS-MATLAB programs. showed very good agreements. The most important dynamic performance criterion of a helical spring is the resonance behavior. More design variables are also making the description of dynamics more complex. and variable spring radius are derived by fundamental mathematics and mechanics principles. and surge. including dynamic stress. Dynamic responses as a time domain.ABSTRACT The objective of this study is the optimal design of helical spring based on dynamic criteria. the more flexible and better designs are possible. Fast Fourier Transform (FFT) is utilized as a tool to evaluate severity of resonance in different models and in optimization process. In this study. either commercial or Finite Difference Method (FDM). Results in terms of system eigenvalue calculation obtained by different programs. finite difference method combined with moving boundary solutions are applied to obtain the dynamic response. discrete data from various models are compared with data from physical dynamic experiments to verify the accuracy of the models. The more variables are considered as design variables. coil closing. Numerical solution of these dynamic models is also called dynamic simulation.

..........3 Special Designs to Reduce Resonance .............................................................1 Finite Difference Method .........1 Motivation and Goal .............................................1 Fast Fourier Transform (FFT) ..........................................................................3............ ii ABSTRACT ............2.............................3...2............... 22 2.................. 29 2............................... 13 Variable Wire Diameter ............ 35 iv ................... 11 2.....1 Explicit Finite Difference Method ........4 Spring Seat Force..4 Evaluation of Dynamic Performance ................................................................ 21 2........... 5 Dynamic Equation ..........................................3...................................................... 29 2.........4 Constant Pitch Angle ............................ 11 Variable Pitch Angle ..........................................................4............................................................................ 24 2..........................................1........................ 7 2...............3 Crank-Nicholson Difference Method............. 2 Chapter 2 – DYNAMIC EQUATION AND RESONANCE ... 28 2...........2................................2.................................3 A New Nonlinear Wave Equation ......3 Numerical Solutions . 5 2..................2..3 2............3 Finite Element Method........................2 2.............................. 9 2........ xi Chapter 1 – INTRODUCTION ................................................................................................................................... and Coil Diameter ................................. 26 2.. 1 Approach and Organization ...1......3.........2... vii LIST OF FIGURES................2..........................................1 2............... 21 2........................................ 28 2.........4....2 Natural Frequency ......................................................................................................2......TABLE OF CONTENTS ACKNOWLEDGEMENTS ..........................................................................................................1 1.............................................................................3..35 3...............1...........................2 Implicit Finite Difference Method ......................................................... 16 2..........2...3..................................... 32 2......................................................... 17 2.....................................................2 Introduction and Geometric Parameters ............. 7 2........ Impact...........5 Variable Pitch Angle............................ Wire Diameter...........................................................2 Fourier Series.............................................. and Collision .......3......................3..........4...................................................................................................................3.......... 33 Chapter 3 – EXPERIMENTAL SETUP AND DESCRIPTION .........................2 Background and Motivation ............. 11 Variable Coil Diameter .... 1 1.................3....................................................... viii NOMENCLATURE .1 2...........3..................................................................2 Wave Equation Using Energy Terms ............. iii LIST OF TABLES ............... 15 2.....................................................................1 Wave Equation Using Force Terms.....

....................................... 35 Measurement ...........................................4 Limitation ..............................5........................... 44 4................. 61 5......3 5..................1 Displacement ......... 77 5.................2 3......4 Goal ............... 75 5... 84 1................5 Variable Pitch Angle .................................................59 5...............4 5........................1 Introduction and Program Structure ........1 4.................................... 63 Variable Coil Diameter (Conical Spring) ............... 60 Fast Fourier Transform in MATLAB ....................................................................................................................................................................................................................... Force and Frequency .......................................................................................................................................................2.......................................... 81 REFERENCE .......4.......................................................... 35 Experimental Apparatus ....... Program Code (Optimization program) ........ 59 Algorithm ......................... 36 Comparison Group ........................... 68 Variable Pitch Angle............................................................................................................................. Wire Diameter............. 78 Chapter 6 – CONCLUSION AND FUTURE WORK ........................................................................... 80 Suggested Future Works ............................................................. 75 5...........................80 6..................4 3.......................................3 3....... 71 Verification in ABAQUS-MATLAB Program ............................................................2 Variable Pitch Angle ............................................................3 5.......................................................................2 Summary ............ 42 Displacement...........................................5.........................................................................................................................................5................................................ 59 Demonstration .............................................2 Introduction ........................................................(D) 1.........1 6............................................ 66 Variable Wire Diameter . 76 5.....1 5....... 40 Chapter 4 – SIMULATION......................3............................5 Software ...........................4...........................3 Resonant Effects for the Stress ........ 96 Read the Experimental Data .................................................4...............(C) 1............84 1....................................................OPTIMIZATION ...................5....................................... 84 1.............82 APPENDIX ....................................(A) Main Program ..................................................... 46 4.......................................................................2..........1 5..............................................42 4...................(E) Read the Cam Profile .......................2 Force and Frequency .2 5.......................................................................................................4...............................(B) Calculate the Curve of the Opposite Installation Direction ................................ 95 1.......................... Coil Diameter ................... 97 v ......................................................... 95 Play Spring Animation ........3 Variable Coil Diameter .2 5......................................................... 44 4........... 56 Chapter 5 ..............

109 3...........2 The Force Figures at 1368...2...............................................................1 Normal Installation at 1368rpm...................................................................5 4............................................................................................................................................................................................. 2165.................................2. 113 4.......................(C) Objective Function .............................................................2.............. 4.......................1 GM-ISUZU Engine .................1 Installation Direction.......................8 Timing Pulley ..2................ 100 2.................. 116 4...2 Dayton Compressor Duty Motor ..................... 100 2...................................................... 115 V-Belt ..........................(I) 2................. 113 DATAQ Acquisition ..........9 KISTLER Cable ............. 111 Opposite Installation at 2372rpm .... 97 Objective Function ............. 4.......................................................2..................... 108 3.................................................. 98 Program Code (ABAQUS-MATLAB program) ..................................... 98 1......................2 Opposite Installation at 1368rpm ... 106 2........ 106 Experimental Data....................................... 113 KISTLER Charge Amplifier..................................2............ 97 The Main Optimization File . 116 5.........6 Normal Installation at 2165rpm.................................................................... 117 vi ............ 116 Software ............... 109 3........ 112 4.........(J) Natural Frequency Distribution .....4 3...................................................................................................................(A) Main File ..7 KISTLER Force Transducer .................................................................................. 109 3...........................(H) 1........................................................................................................4 4.......................................................(B) Optimization Main File................. Equipment in the Experiment ..................(G) 1..................(F) 1.............. 110 Opposite Installation at 2165rpm ...................................................................5 3.......................................................................... Calculate the Power Spectrum ............. 112 4........... 110 Normal Installation at 2372rpm ......3 Cen-Tech Photo Sensor Tachometer ....................... 106 2................ 111 3..............(D) Constraint Function...............................................1............................ and 2372 (rpm) . 112 4...6 4..... 108 3.................................................................. 98 Constraint Function....3 3.......................................................................

.......................... 40 Table 3....................LIST OF TABLES Table 3.......... 56 Table 4.... 41 Table 4..............................................2 The excited harmonics comparison table..................2 The comparison groups in the experiment ........ 79 vii .....................1 Numbers of data points read at different camshaft speeds....................................... 42 Table 4..........................................................2 Comparison of fundamental natural frequencies in different optimization cases .......... 56 Table 5............3 The amplitude comparison table .......................................................................1 The results from different optimal designs ...............1 The related parameters of the helical spring in this simulation ............... 75 Table 5..............

............ 1961] ................ 25 Fig.4.....................6 Cam velocity ......2......4 A schematic shows the moment of inertia ........ 52 Fig...... 51 Fig................... 49 Fig....................................... 46 Fig....4.... 14 Fig. 1961]..........1....5 Simplified charge amplifier model [Bishop....................................................3 The spring length varies with preload.5 The change in pitch angle on node 5 .........10 The explicit method stencil ........................................................4........2 The cam profile ... 43 Fig....9 The power spectrum with an opposite installation at 1368rpm .1 The structure of newer valve train [Fujimoto............... 39 Fig.................... 3 Fig.....................................2 Force analysis of an element of helical springs ......................................................... 20 Fig....12 The power spectrum with a normal installation at 2165rpm .... 44 Fig......2...................4..........1 The pitch angle of helical springs along arc length which is along the helix on the initial condition without the preload ............4...3 How to make the multi-spring rate [Chironis..... 2008] ... 52 Fig....................4.......... and compression at 1368 rpm.........7 An opposite installed direction at 1368 rpm .............1 The components and setup of experiment .......12 The Crank-Nicholson method stencil .2............ 1989] ...........3.... 18 Fig...4..4.............1 The helix description and parameters of helical springs.................. 45 Fig............... 29 Fig..................................... 38 Fig.........2.....2..........3........4...............2.....3..............3 The flow chart of signal for data acquisition ................................................2.................. 2007] ...................................13 The power spectrum with an opposite installation at 2165rpm ......................11 The implicit method stencil ..... 26 Fig........................ 51 Fig... 2007] .2.LIST OF FIGURES Fig..............................9 The grid on the space and time ..... 8 Fig..........4.......................8 Two spring segments collides on each other ............... 36 Fig............ 21....1......2....5 Impact of helical springs [Chironis............................ 20 Fig............3............................... 37 Fig...........................................2.........................4................6 A normal installed direction at 1368 rpm in the simulation ...8 The power spectrum with a normal installation at 1368 rpm .....14 A normal installed direction at 2372rpm ..........2.................................... 6 Fig........................................................................................4.4..................15 An opposite installed direction at 2372rpm............................10 A normal installed direction at 2165rpm .................... 22 Fig................2................ 4 Fig.................................... 53 viii ............................. and 34 varies at 2372 rpm ....................................2 The transducer under the spring seat .............. 49 Fig............... 53 Fig.............13 The system transformation for the natural frequency . 37 Fig..7 Cam acceleration .......3......... 45 Fig.................................................. 12 Fig..................4........................................4 Direct Force Measurement [Kistler......................... 50 Fig...........4............................... 20 Fig......................4 The spring length on node 8................................................11 An opposite installed direction at 2165rpm............................ 23 Fig...............2 The structure of older valve train [Fujimoto.............. 50 Fig....2..............

............... 111 Appendix 3................. 54 Fig.............5...........................................3 Force with a normal installation at 2165rpm ....5 The power spectrum comparison of the original and the optimal designs (case two) ... 77 Fig.............18 The power spectrum at different speeds .. 110 Appendix 3............................5 Force with an opposite installation at 2372rpm.............. 73 Fig...2............... 73 Fig.19 The phase comparison between the simulation at 1368 rpm and the cam profile .4 Force with an opposite installation at 2165rpm.........4..............17 The power spectrum with an opposite installation at 2372rpm .........6 The optimal variable wire diameter in optimization case three . 70 Fig.................5........................................... 108 Appendix 3...............3 The power spectrum comparison of the original and the optimal designs (case one) .......Fig.........2....................2..2..........................1 The flow chart of optimization program ........................ 65 Fig.......................... 113 ix ................................23 The second mode of excited resonance ..5................5..............6 Force with an opposite installation at 2372rpm..... 57 Fig.......1................................................. 67 Fig..................3 Photo Sensor Tachometer ............................ 74 Fig....22 The first mode of excited resonance .........................................1 Spring normal installation.... 108 Appendix 3....5.....4........4.......... 63 Fig................. 109 Appendix 3...... 55 Fig................................5............4 The optimal variable coil diameter in optimization case two ........5........................................................................................ 76 Fig............................................................... 111 Appendix 4................10 The optimal variable wire diameter in optimization four ........................ 57 Fig.............................. 70 Fig....... 68 Fig....2......12 Flow chart for comparing results from different methods ..............2 Force with an opposite installation at 1368rpm................................9 The optimal variable coil diameter in optimization case four............................5............................. 74 Fig. 58 Fig............4............................... 65 Fig....21 The stress distribution under a normal static loading ........................................... 78 Appendix 3...........................5.13 The generated fundamental natural frequency by ABAQUS-MATLAB program (Optimal parameters of optimization case one) .......14 The generated fundamental natural frequency by ABAQUS-MATLAB program (Optimal parameters of optimization case two) ............1............4.....................5.............2 Spring opposite installation ........... 109 Appendix 3............4...........................1 Force with a normal installation at 1368rpm ..................................................5............... 54 Fig....... 58 Fig.........16 The power spectrum with a normal installation at 2372rpm .................8 The optimal variable pitch angle in optimization case four ........5.... 110 Appendix 3...............4......................2 The optimal variable pitch angle in optimization case one. 55 Fig.............................................7 The power spectrum comparison of the original and the optimal designs (case three)...............20 The three dimensional helical spring ...................................................2.....................................5........................................4................5.........11 The power spectrum comparison of the original and the optimal designs (case four) ...................................

..............8 The pulleys with different size ......... 115 Appendix 4.6 DATAQ Acquisition ......... 116 x ...... 114 Appendix 4........................................................................5 The charge Amplifier..................Appendix 4...........................................................................................

a funciton of helix length. ሺs. a function of time and location. s ܦ: coil diameter of helical springs. a function of time and displacement ܭ: spring rate. a function of time and displacment ݔ: the displacemnt of spring for one end when the other end is ϐixed : pitch angle ሺݏሻ: variable pitch angle ݀ሺݏሻ: variable wire diameter ܦሺݏሻ: variable coil diameter ݎ: coil radius of helical springs.NOMENCLATURE ܨ: spring force. s ݀: wire diameter of helical springs ݊: coil turns ܩ: shear modulus ܬ: polar moment of inertia of wire cross section ܧ: Young ᇱ s Modulus ܫ: area moment of inertia ݏ: the arc length along the spring helix ݐ: time ܫ : mass moment of inertia ݉: mass per unit length ߩ: density of spring material ݈ total helix length ܫ௫ : area moment of inertia with respect to x axis ܫ௬ : area moment of inertia with respect to y axis ߬: torsion ߰: rotation of wire in radians ܷଵ : the torsional strain energy ܯ௦ : the clashed mass of wire segment ݇: curvature of the helix. tሻ ܷଶ : the ϐirst bending strain energy ܷଷ : the second bending strain energy ܶଵ : the rotary kinetic energy ܶଶ : the radial kinetic energy ܶଷ : the translatonal kinetic energy ܿ ᇱ : the damping force per unit length of the wire per unit of velocity ܽ: wave speed along the helix xi . a function of helix length.

ܽ௦௧ : the instant acceleration ߜ: the same as x ܯ: the mass of wire segment to be deformed Δ݇: curvature change Δ߬: torsion chagne Δ߰: rotation change 2π τ ݁: the coefϐicient of restitution ܾ: the viscous damping coefϐicient determined by the measurement ܽ ܽ݊݀ ܾ : the coefϐicients of Fourier Series ݓ: fundamental frequency ൌ ݊: number of harmonics ߬: period of cyclic motion xii .

H. J. Timoshenko [1963] published the Timoshenko-Beam theory. Usually. Wahl [1935] derived a stress correction factor taking into account the effects of curvature and direct shear. His three dynamic equations were derived from the Lagrange equation. Love [1944] developed more advanced dynamic equations based partly from J. Love did not solve the equations. H. the automotive engines have a “red line” speed around 5000 RPM. and applied his theory to explain the lateral buckling of the helical spring. Mitchell [1890] is probably the first person to consider spring wire motion in three dimensions. dynamic phenomena have dominant importance. Ancker and Goodier [1958] obtained a solution by using a thin slice method and expressed the pitch and curvature as basic variables. such as valve springs in any automotive engine or recoil spring in automatic firearm. However. 1935]. Therefore. Mitchell’s equations.1. The early spring research was focused on the static behavior. Ancker assumed that the cross-section in variable pitch angle spring remained circular which was not accurate. 1 . Wahl [1963] published most important book in springs for last 50 years.1 Background and Motivation The documented study of helical compression spring dated back more than 300 years. The research focus shifted from the static condition to the dynamic loading and vibration suppression.Chapter 1 – Introduction 1. when Hooke [1678] published his law of springs as: ܨൌ ݔܭ (1. Each small spring segment in his assumption has three degrees of freedom to move in Cartesian coordinate frame and one degree of freedom to twist about the wire axis. To study the elasticity of curved beam Timoshenko [1951. Due to the fact that the axial load is complex if the effects of the pitch angle as well as curvature change are taken into account. In modern applications of helical springs. the simulation and dynamic experiments in this thesis were carried out around 4800 RPM (referring to Chapter 5). 1956] modified the shear strength and included effects of curvature and direct shear according to Wahl’s factor [Wahl. Fig. with assumptions of no direct shear and no axial force.2 show the main components in the valve train of an automotive engine for two different types. combining his work from 1930s to 1950s. It is fairly accurate when the spring is simple and under static loading condition.1 and 1.1) This equation relates spring force to spring displacement in a simple format.

1. numerical and experimental results. or coefficients of a series that gives precise description of the helical spring. Since all solutions were first expressed as a discrete time domain sequence. Pisano and Freudenstein [1983] combined the spring study and cam profile into a complete valve train dynamic model. Fast Fourier Transform is used as a scope to view and measure the dynamic performance. Finite difference method is used for solving the nonlinear partial differential equations. not just pitch angle will be considered a function defined by a few design variables. Wahl [1963] used the force equilibrium to re-derive the wave equation. then using finite difference method to solve the governing dynamic equations. 1987] was published to offer a better result by using the extended Hamilton’s Principle and calculus of variations. Totally. and with additional theoretical derivation. This thesis work is based on Lin and his students’ previous work from last 20 years. 2 . It starts from the limitation of the commonly used wave equation. wire diameter and coil diameter are also variables that are defined by a set of design variables. More literature review will be included in each chapter. In this study. As a consequence. as it has been done by other researchers in the literature.2 Approach and Organization Using extended Hamilton’s Principle to derive helical spring with nonlinear effects. Chapter Two focus on mathematical modeling. and the construction of various dynamic models of helical springs. which is the simplest linear partial differential equation describing many physical phenomena to a satisfactory degree of precision. The modified wave equation [Lin. there are six chapters in this thesis. improving spring performance is the key in improving valve train performance. and optimally design helical springs are the main objective in this thesis. Chapter Three discusses methods for solving the dynamic equations. when the diverse topics are discussed. For this goal. the approach and a brief introduction for spring research. and proposed to use it to describe the resonant of helical springs. If the pitch angle of a helical spring is a function of a few design variables. then wave equation can no longer accurately describe the dynamic behavior.According to Lin [1989]. the valve spring is usually the first component to fail in valve train for dynamic loading situation. the term of “dynamic simulation” is used to present the solutions to various dynamic models. Chapter One presents the motivation. building an accurate model to get the vibration of valve spring is the starting point. These design variables can be the coefficients of a polynomial.

obtained by finite difference solution. Appendix section after Chapter Six contains the specifications of instruments used in the investigation and dynamic testing. Fig. The first one is using MATLAB programming to get difference solution as described in Chapter Three. ABAQUS. Then. variable wire diameter. Chapter Five compares eigenvalues obtained by different numerical tools. 2007] 3 . Experimental data and programs for optimization and in ABAQUS-MATLAB are also attached.1. Chapter Six is the summary of this thesis.Chapter Four describes the dynamic experiments in order to verify the derived dynamic equations and simulation are correct and accurate. with conclusion and suggestion for further study of valve springs. Experimental data can be divided into two sets. variable coil diameter.1 The structure of newer valve train [Fujimoto. and the combination of the first three types and evaluate the results. The first set is used to calibrate the quartz load washer. The second approach is to use MATLAB with the well known commercial finite element analysis program. The second set is used to compare with the numerical results obtained in Chapter Three. the charge amplifier and data acquisition system. or any helical springs that are used in high speed loading situations. there are four types of optimal designs to be demonstrated such as variable pitch angle.

1. 2007] 4 .Fig.2 The structure of older valve train [Fujimoto.

and the curvature occurs on t-n or b-t plane [Greenwood. normal and binormal. In addition to those.1. to optimize the design of a helical spring. 1988. modern differential geometry started from Feret-Serret 5 .2. This causes undesired force responses and has negative impact on the durability of the spring [Lin. the dynamic equations are expressed in as basic parameters and directions as possible.Chapter 2 .1.1 Introduction and Geometric Parameters A valve spring is an energy storage device. These basic design parameters are shown in Fig. and ‘b’ binormal. by changing the spring parameters such as p(s).Dynamic Equation and Resonance 2. Kreyszig. The symbol t represents the tangential. 1989].2. ‘n’ normal. the cross section and material properties of the wire have direct influence on the stiffness. the stiffer spring stores more energy. Its influence on the overall dynamics of the system is significant and cannot be ignored in a numerical model. The internal oscillations in the spring are commonly referred to as surge modes. which happens when external excitation generates waves in the spring with enough amplitudes for adjacent windings to touch. The contact between the windings is responsible for a large portion of the non-linearity of a spring. and d(s). tangential. In this study. To describe a helix by using curvature and torsion. In order not over-load the spring material. Another phenomenon in conjunction with this contact is called coil clash. D(s). The dynamic response of the spring is substantially different from the static response due to the internal dynamics of the spring. one has to consider several aspects of the valve spring like kinematic (force as a function of displacement) and dynamic characteristic (natural frequency). The torsion works on the b-n plane. In Fig. Therefore. The kinematics is usually specified by displacement and acceleration requirements of the cam. Dynamically. For two springs with the same displacement. since it reduces the number of active coils during spring compression. A number of elementary optimization textbooks use the weight of the spring as the objective function. energy stored per volume is a constrained by the material. stiffer springs will have a higher lowest natural frequency. The valve spring is usually the softest component with the lowest frequency in the valve train. Geometry and deformation of a helical spring can be completely described by curvature and torsion at any location of the helix. one can design the stiffness characteristic to the desired specification. helical springs can be described in a local coordinate along three directions. 2006].

1) Torsion is the main factor to affect potential energy. ߬ሺݏ. torsion rotation can be defined as ߲߰ሺݏ. Fig. 0ሻ ߲ݏ (2. ݐሻሻ ݎሺݏ. Lord Kelvin intuitively derived the curvature and torsion and expressed them in very simple formulas which are still used commonly: ݇ሺݏ. 0ሻ is initial torsion at the free height condition. Lin [1988] also pointed out that these simple curvature and torsion formulas are accurate accurate if the pitch and spring radius remain constant.3) where. and the deformation of helical springs and is expressed as ߬ ݐሻ ൌ ߬ሺݏ. ݐሻ െ ߬ሺݏ. Before the dawn of differential geometry. ݐሻ (2.2. ݐሻ (2. then more complex formulas must be used to avoid large error. ݐሻ ൌ ܿ ݏଶ ሺሺݏ.2) Another expression related to the torsion. ݐሻ ൌ ߬ሺݏ. cos൫ሺݏ. ݐሻ൯ sin ሺሺݏ. 1988].formula. ݐሻሻ ݎሺݏ.1 The he helix description and parameters of helical springs 6 . If the changes of spring radius and pitch are significant. then the expressions quickly get g quite involved [Lin.

the wave equation with the damping term was derived by static force equilibrium.2 Dynamic Equation Spring dynamic equations can be derived by force equilibrium or energy equilibrium. 1972. the damping force shall be considered as follows 1. Also. 1984]. 1927. modified wave equations were derived by inserting a virtual internal force.2. 2.6) (2.1 Wave Equation Using Force Terms The simplest wave equation is to have the second-order. For spring dynamic applications. Pisano and Freudenstein [1984] derived the dynamic wave equation using energy equilibrium. Pisano and Freudenstein. Force analysis in a valve spring is described as follows and shown in Fig. Wahl [1963] used force equilibrium in a spring element for obtaining the dynamic wave equation. and its result can describe the effects of natural frequency and resonance. The modified wave equation and the modified spring seat force equation are discussed below in detail. For more accuracy. linear relation derived by Newton’s 2nd Law and solved by d’Almbert in 1707.2. 2.5) In addition. Johnson.2. 3. The force acting to accelerate the element is ߨ݀ ଶ ݀ ߲ ߩ ݏଶ ݕ ܨ ൌ 4 ߲ ݐଶ The net force acting to accelerate the element: ܨ ൌ ߨ ݀ܩସ ߲ ଶ ݕ ݀ݏ 8ܦଶ ߲ ݏଶ (2. ߲ ଶݕ ߲ ଶݕ ଶ ൌܽ ߲ ݐଶ ߲ ݏଶ (2.4) d’Almbert solved this problem successfully by a simple difference method. and offered a more accurate dynamic equation. Internal hysteresis in the spring material Air damping Damping due to friction in the end turns Damping due to loss of energy in the supports 7 . This equation gives an accurate result.2. 4. In Wahl’s textbook [1963]. it has been derived in many technical papers [Love.

Therefore. and rigidity of support. this term varies with many factors such as material. ସ ᇲ Fig. If the damping is zero here. For most cases. design of end turns. As a matter of fact.8) is relatively accurate to describe the dynamic condition of valve springs with a constant pitch angle or a small pitch angle variation.2.8) Where. the coil-to-coil contact should be taken into consideration and it allows the end coils 8 . the damping force could be expressed as: ܨௗ ൌ ܿ ᇱ Since ܨ ൌ ܨ െ ܨௗ ߲ݕ ݀ݏ ߲ݐ (2. Equation (2. The damping term is usually determined by the experimental data and not easily determined by analytical method [Wahl. ቂగௗమ ఘቃ is the damping term and ቂଶమ ఘቃ is the wave speed term. 1963]. This approximate method for calculating spring dynamics considers only torsion of the wire and ignores the effect of closed end coils. amplitude of motion.7) This leads to a dynamic wave equation with the damping term ߲ ଶݕ 4ܿ ᇱ ߲ݕ ݀ܩଶ ߲ ଶ ݕ ቈ ൌ ቈ ߲ ݐଶ ߨ݀ ଶ ߩ ߲ݐ 2ܦଶ ߩ ߲ ݏଶ ீௗమ (2. In reality.2 Force analysis of an element of helical springs This damping term is a measure of the equivalent damping in the spring. it describes the longitudinal wave transmission for the prismatic bar.

The scale of this term is similar to the first bending energy term. for the better accuracy. Bending characterizes the behavior of a slender structural element subjected to an external load applied perpendicularly to an axis of the element. ݐሻሻଶ ݀ݏ 2 (2. Three potential energy terms in terms of basic parameters are represented as ܷଵ ൌ න ଵ ܷଶ ൌ ଶ ܫܧሺ݇ሺݏ. and the second bending produces that along the n axis.9) (2. On the other hand. 0ሻሻଶ ݀ݏ ܷଷ ൌ න 1 ߲߰ሺݏ. that is. the spring element produces torsional strain energy ܷଵ . During spring compression and release.11) Along the t axis. we take the torsional deflection of the wire into consideration. the first bending generates the deformation along the b axis. the wave equation in spring applications can be derived using energy equilibrium [Lin. it generates first bending energy ܷଶ . the wave speed term was modified using energy equilibrium by Pisano and Freudenstein [1984].2 Wave Equation Using Energy Terms The spring motion originates from the change in kinetic energy and potential energy. and rotary kinetic energy terms. it consists of torsional strain energy. In valve springs. using force equilibrium only. radial. In circular sections. When the local curvature is changed. Potential energy is energy that is stored in an object. ݐሻ ଶ ܬܩሺ ሻ ݀ݏ 2 ߲ݏ 1 ܫܧሺ ݕᇱᇱ ሺݏ. the resultant shearing stress is perpendicular to the radius.10) (2. As a result. ݐሻ െ ݇ሺݏ. Second bending energy ܷଷ results from the local change in pitch angle. kinetic energy is energy due to motion composed of translational. series of energy conversion between potential energy and kinetic energy happen.2. Torsion is the twisting of an object due to an applied torque. In the deformation of valve springs. In fact. 1987]. Let us discuss potential energy first. In addition. this term is the largest among all the three potential energy terms and occupies over ninety percent during the spring compression. 2.to leave their stops. and can be deduced from the Hamilton’s principle. and second bending strain energy. first bending strain energy. It is much smaller in comparison with the torsional strain energy. The three kinetic energy terms are defined as follows 9 .

Tଶ is the radial kinetic energy and Tଵ is the rotary kinetic energy. b is the viscous damping coefficient determined by the measurement. a deflection correction factor should be employed.15) Where. In reality.3. and rotary kinetic energy. bending stresses cannot be ignored and needs to be The first term on the right side describes the bending term. the extra clash force was considered in 10 . (2.15) are relatively close to those in Eq. By considering torsional energy. In this condition.2.615 4ܿ െ 4 ܿ less than D/4. valve springs always have the coil clash to affect spring performance and equation accuracy. (2. ݐሻ ଶ ݉ሺ ሻ ݀ݏ 2 ߲ݐ 1 ߲ݎሺݏ.15) can describe the spring with a constant pitch angle or a smaller pitch angle variation. but offer more accurate results. and the second term is the shear stress factor. That comparison would be shown in SECTION. ߲ ଶݕ ߲ݕ ܬܩ ߲ ଶݕ ܾ ൌ ߨ ݎଶ ݀ ଶ ߲ ݐଶ ߲ݐ ߲ ݏଶ ߩሺ 4 ܬሻ (2.ܶଵ ൌ න ܶଶ ൌ න ܶଷ ൌ න 1 ߲߰ሺݏ.8). Eq. hence by inserting one damping term into the above equation we have.8) ignored the bending effect. ݐሻ ଶ ܫ ሺ ሻ ݀ݏ 2 ߲ݐ 1 ߲ݕሺݏ. A smaller pitch angle means that it is less than 15° or deflection per turn is considered. ܭ௪ ൌ 4ܿ െ 1 0. In fact. a modified wave equation was derived as ߲ ଶݕ ܬܩ ߲ ଶݕ ൌ ߨ ݎଶ ݀ ଶ ߲ ݐଶ ߲ ݏଶ ߩሺ 4 ܬሻ Among these. resulting in the variation. the calculated wave speed values in Eq. (2. ݐሻ ଶ ݉ሺ ሻ ݀ݏ 2 ߲ݐ (2. translational kinetic energy Tଷ is the only energy term considered in (2. Eq. (2.13) most technical papers and research. A stress correction factor ܭ௪ was published to modify this problem by Wahl [1963].2. translational kinetic energy.14) Any physical system always has the damping effect. For better accuracy.3. For large deflections per turn.12) (2.

11 .2. In comparison of Eq.3 A New Nonlinear Wave Equation This section describes another derived modified wave equation. (2. This more accurate nonlinear wave equation was derived by Lin and Pisano [1987]. helical springs with smaller pitch angle variation can be calculated by Eq. original constant pitch angle springs were made to new variable pitch angle springs. (2. In fact.2. As a result. helical springs with design variables involving variable pitch angle. but it is not as simple as determining ∆݇ and J. 2. (2. using heat treatment and external force that is applied at open end of the spring to compress or to elongate the spring.3.8). The wave equation should be of simplest type in terms of basic design variables. Eq. Therefore.8) or (2.16) to approximate the real engine condition is derived. After quenching and tempering in a low temperature. and variable wire diameter in a polynomial expression.16) offers a better fitting considering the extra coil clash. (2. and a more accurate equation Eq.the Eq. (2.16) 2. variable coil diameter. 1961]. a simpler equation needs to be derived.15).15).3.15) also offers similar numeric solutions. and presents a variable pitch angle [Chironis. which causes a permanent deformation. it is reasonable to imagine that there is a virtual internal force formed in this spring. In contrast. it is accurate enough. Modifying the wave equation to handle nonlinear conditions has been a problem.2 Variable Pitch Angle As discussed previously. ߲ ଶݕ ߲ݕ ܾ ଶ ߲ݐ ߲ݐ ܫܧሺ∆݇ሻଶ ܬܩ ߲ ଶݕ ൌ ଶ ߨܦଶ ݀ ଶ ߨ ݎଶ ݀ ଶ ߲ݏ ߩሺ ܬሻ ߩሺ ܬሻ 4 4 (2.3. 2. We describe how to make variable pitch springs first and then introduce a virtual internal force to derive that. For most spring applications.1 Constant Pitch Angle Eq.8) derived by Wahl [1963] can describe the constant pitch spring or the smaller pitch angle variation. Eq.2. a variable pitch spring can be made by the method in Fig 2. (2. (2.

ܯis mass of wire segment .3. ߨ ݀ܩସ 8ܦܨଶ ݈ ܭൌ and ݔൌ 8ܦଶ ݈ ߨ݀ܩସ 4ܦܨଶ ݈ ܧൌ ߨ ݀ܩସ and ݔൌ ଼ி య ே ீௗర 1 ଶ ݔܭ 2 (2.2.18) From the previous discussion and Fig. a virtual internal force considered is reasonable. hence we have Internal force term = 4ܦଶ ݈ ߲ܨ 16ܦଶ ݈ ߲ܨ 16ܦଶ ݈ ߨ݀ܩସ ߲ݕ 2 ߲ݕ ߲ܧ ൌ ൌ ൌ ൌ ଶ ସ ଶ ଶ ଶ ߲݀ߨ ݀ܩߨ ܯ ߩߨ ݀ܩ ߨߩ ݏ߲ ݀ܩ8 ݀ߨߩ ݏ߲ ݈ ܦଶ ߲ݏ ߩ 4 ߲ݏ The spring height along the spring helix could be defined as 12 Where.2.17) we get: (2. substituting ݈ ൎ ܰߨ ܦinto K and x.17) And then from Eq. so that each mass element along the spring helix was drawn and deformed.Fig. (2.3 How to make the multi-spring rate [Chironis. 1961] Spring potential energy is defined as ܧൌ Where ܭൌ ଼ య ே ீௗర hence the simplified ܭand ݔare given by As helical springs are in a smaller pitch angle.

The moment of inertia is a geometrical property of a beam and depends on the reference 13 . This type of spring with variable coil diameter along the spring helix is like Fig.21) in a simpler expression.20) into virtual internal force term. Substituting Eq.3.4. When it is differentiated. elevating the fundamental natural frequency and an easy to manufacture. This type of spring is also called “conical Spring” or “beehive Spring”. Today.21) 2.1. the motion of their valve is controlled not by steel but by pneumatic springs. it is expressed as Inserting the virtual internal force term into equation (2.The curve in Eq.20) ݕሺݏሻ ൌ ଵ ݏଷ ଶ ݏଶ ଷ ݏ ସ (2.1. (2. variable pitch springs are usually limited to elevate its fundamental natural frequency. This case has been used in the valve train system for a long time and originated from the Formula One racing.16) we have ߲ ଶݕ 4ܿ ᇱ ߲ݕ 2 ሺݏሻ ൌ ߲ ݐଶ ߨ݀ ଶ ߩ ߲ ݀ߨߩ ݐଶ ܬܩ ߲ ଶݕ ߨ ݎଶ ݀ ଶ ߲ ݏଶ ߩሺ 4 ܬሻ (2.4 and is defined as ܦሺݏሻ ൌ ܦଵ ݏଷ ܦଶ ݏଶ ܦଷ ݏ ܦସ and ݎሺݏሻ ൌ ܦሺݏሻ/2 (2. we need to realize the area moment of inertia and polar moment of inertia first.4. its pitch angle along the spring helix can be defined as ሺݏሻ ൌ ߲ݕ ൌ 3ଵ ݏଶ 2ଶ ݏ ଷ ߲ݏ ߲ܧ 2 ൌ ሺݏሻ ߲ ݀ߨߩ ܯଶ (2.5.3 Variable Coil Diameter According to a textbook [Nunney. For Grand Prix cars.2. (2. 2007].19) And the pitch angle as a function of arc length is shown in Fig. Its idea is to use the valve stem plungers moving in chambers of compressed air to maintain the valves in contact with their cams. hence the spring with variable coil diameter was invented. The conical springs have many advantages such as. The area moment of inertia of a beams cross-sectional area measures the beams ability to resist bending.000 rpm are now being achieved. The larger the Moment of Inertia the less the beam will bend.19) is presented like Fig. Formula One is using the valve train with the hydraulic control instead of conventional springs. where engine speeds of around 18.22) In order to express the parameter J in Eq. (2.

The smallest moment of inertia about any axis passes through the centroid.2. The following are the moment of inertia in a mathematical expression: ܫ௫ ൌ න ݕଶ ݀ܣ ܫ௬ ൌ න ݔଶ ݀ܣ y: the distance from the x axis to an area element dA x: the distance from the y axis to an area element dA Fig. By the Perpendicular Axis Theorem. the ߨ݀ ସ ܬൌ ܫ௫ ܫ௬ ൌ 32 . (2. The larger the polar moment of inertia the less the beam will twist. the polar moment of inertia in a mathematic expression is as follows: ܬൌ ܫ௫ ܫ௬ ൌ නሺ ݔଶ ݕଶ ሻ݀ܣ For a circular section with diameter d ߨ݀ ସ ܫ௫ ൌ ܫ௬ ൌ 64 new wave speed parameter for this type of variable wire diameter is expressed as follows: 14 Substituting this expression ܬinto the wave speed parameter ܽ in Eq.axis.4 A schematic shows the moment of inertia The polar area moment Of inertia of a beams cross-sectional area measures the beams ability to resist torsion.21).

(2. and the variation is not over one percent. as a result. we can find this type of spring can provide the best 15 . Eq.8ሻ ൌ ቈ ଶ 2ߩ ܦ ଵ/ଶ ଵ/ଶ ܬܩ ܽ in eq. the original wave speed term in Eq. That comparison is shown as follows.8) is enough for calculating the dynamic conditions. ሺ2. As the spring is in a small pitch angle variation.23) The new dynamic wave equation with basic parameters is derived as ߲ ଶݕ 4ܿ ᇱ ߲ݕ 2 ܩሾ݀ሿଶ ߲ ଶݕ ሺsሻ ൌ ߲ ݐଶ ߨ݀ ଶ ߩ ߲ ݀ߨߩ ݐଶ ߩሺ2ሾܦሺݏሻሿଶ ݀ ଶ ሻ ߲ ݏଶ (2. In Optimization chapter.3. The characteristic in this spring is that the weight at each element is different. ሺ2. (2. we found that the polar moment of inertia is only related to wire diameter.16). In comparison with conventional springs. From this result. ݀ܩଶ ܽ in eq.Besides.16ሻ ൌ ଶ ߨ ݀ ݎଶ ߩሺ 4 ܬሻ ݀ܩଶ ൌቜ ቝ ߩሺ2ሾܦሺݏሻሿଶ ݀ ଶ ሻ ଵ/ଶ 2. it offers a relatively flexible design – variable wire diameter.2. (2. and we ଶ ݀ܩଶ ܽൌቜ ቝ ߩሺ8 ݎଶ ݀ ଶ ሻ ଵ/ଶ have ଶ ݀ܩଶ ܽൌቜ ቝ ߩሺ2ሾܦሺݏሻሿଶ ݀ ଶ ሻ ଵ (2. (2. similar to that of conical spring. As mentioned previously. the coil radius ݎin this wave speed could be expressed as ݎൌ .4 Variable Wire Diameter Lin’s technical paper [1993] studied a new type of spring with variable wire diameter and its features.16) with the polar moment of inertia can be simplified in terms of variable wire diameter. the wave speed term in Eq.24) From the above discussion. we can understand that torsion occupies over 95 percent of all strain energy.8) is relatively approximate to that in Eq.

performance than variable pitch springs, even conical springs, but it has a serious problem. The endured stress is reduced largely as the wire diameter is smaller; when designing this spring, we need to take care of this problem. The wire diameter along the spring helix is like Fig.5.6 and is defined as ݀ሺݏሻ ൌ ݀ଵ ݏଷ ݀ଶ ݏଶ ݀ଷ ݏ ݀ସ (2.25)

The wire diameter term in Eq. (2.24) can be substituted by Eq. (2.25) and the new expression with design variables along the spring helix can be defined:

డమ ௬ డ௧ మ

గሾௗሺ௦ሻሿమ ఘ

ସ ᇲ

డ௬ డ௧

ఘగሾௗሺ௦ሻሿమ

ଶ

ሺݏሻ ൌ

ఘሺଶሾሺ௦ሻሿమ ାሾௗሺ௦ሻሿమ ሻ

ீሾௗሺ௦ሻሿమ

డమ ௬ డ௦మ

(2.26)

if the coil diameter is constant, we can define

If the pitch angle is constant, we can define

ሾܦଵ , ܦଶ , ܦଷ , ܦସ ሿ ൌ ሾ0, 0, 0, ܦሿ ሾଵ , ଶ , ଷ , ସ ሿ ൌ ሾ0, 0, 0, ሿ

Eq. (2.26) can express any type of springs no matter it is variable pitch, variable coil diameter, variable wire diameter, or the combination of these variable designs. For the combination type, it is discussed below.

2.2.3.5

Variable Pitch Angle, Wire Diameter, and Coil Diameter

Eq. (2.26) can describe springs with variable pitch angle, variable coil diameter, and variable wire diameter. As the natural frequency at each element along the spring helix is completely different, it can offer the best performance due to its extreme flexible design. In Optimization chapter, we will verify that. For helical springs with a constant pitch angle, a constant coil diameter, and a constant wire diameter, equation (2.26) can also be used to do the computation. Coefficients in ሺݏሻ needs to be defined as follows: ሾଵ , ଶ , ଷ , ସ ሿ ൌ ሾ0, 0, 0, ሿ

As the wire diameter is constant, coefficients in ݀ሺݏሻ need to be defined as follows: As the coil diameter is constant, coefficients in ܦሺݏሻ need to be defined as follows: Later in Optimization chapter, we have Eq. (2.26) to run optimization and the

16

ሾ݀ଵ , ݀ଶ , ݀ଷ , ݀ସ ሿ ൌ ሾ0, 0, 0, ݀ሿ

ሾܦଵ , ܦଶ , ܦଷ , ܦସ ሿ ൌ ሾ0, 0, 0, ܦሿ

simulation with different designs.

2.2.4

Spring Seat Force, Impact, and Collision

During spring compression and release, mainly twist and bending effects are shown. The relation between them and force was derived as [Wahl, 1963] ܨൌ

ீ

cosሺሻ ∆߬ െ

ாூ

sin ሺሻ ∆݇

(2.27)

Eq. (2.27) takes into account the effect of the change in curvature and in torsion. In fact, the change in torsion is very large during spring compression, so it is very easy to predict the spring seat force. As discussed previously, Eq. (2.8) cannot describe a larger pitch angle variation due to the ignored bending effect [Wahl, 1963]. As a result, Eq. (2.27) was derived to describe a larger pitch variation by considering the bending term. For the more accurate result, a better fitting equation is sought by Lin [1987]. He used the Hamilton’s principle to insert another bending energy term ܷଷ , ܨൌ ܬܩ ܫܧ ߲ 1 cosሺሻ ∆߬ െ sin ሺሻ ∆݇ ሺ ܫܧሺߜ݇ሻଶ ሻ ݎ ݎ ߲ ݕ2 (2.28)

to fit more accurate results.

The final term on the right hand side originates from the change in curvature. However, it’s very difficult to solve directly and often needs Fourier series, Finite Difference Method or Finite Element Method for getting the numeric values. In Simulation chapter, we will use MATLAB to demonstrate that with Finite Difference Method – Crank-Nicholson Method. In the experiment, several interesting things were observed. For example, a sudden decrease of force happened when the cam finished the rise (corresponding to maximum spring compression). The faster the camshaft the greater the dip of the spring seat force at maximum compression. For an intuitive explanation of this phenomenon, consider sitting in an accelerating bus. When the bus accelerates suddenly, the passengers feel the inertia force to pushing back. By contrast, when it suddenly stops, the passenger will feel the force pushing forward. That relation is shown in Fig 2.5.

17

Fig.2.5 Impact of helical springs [Chironis, 1961] This figure shows how the helical spring works in the valve train. In the stage 1, an applied force F acts on the top end, and a spring seat force is generated at the bottom end. When it reaches stage 2, the spring seat force reaches the maximum. The process from the stage 2 to stage 3 is the most important one. When the spring just leaves the stage 2 to stage 3, there is a fictitious force upward. In fact, the process between stage 1 to 2 and stage 2 to 3 is similar, but in opposite direction, that can be expressed as ܨൌ െܯ ܸ݀ ݀ݐ

**Where, M is mass of wire segment, modified as ܨൌ
**

ீ

ௗ ௗ௧

is the instant acceleration. Hence Eq. (2.28) is

cosሺሻ ∆߬ െ

ாூ

sinሺሻ ∆݇ డ௬ ቀଶ ܫܧሺߜ݇ሻଶ ቁ െ

డ ଵ

ௌ ௗ௧

(2.29)

For the phenomenon, the faster the camshaft the greater the peak of spring seat force, it is seen in Fig.2.7, and can also be explained by a fictitious force. Besides, another phenomenon sometimes occurs in the spring, that is, coil clash, especially at high speeds. In fact, the idea for coil clash is similar as the impulse idea, mentioned above. The only difference is its direction due to collision. In order to understand that, it is necessary to understand displacement, velocity, and acceleration resulting from cam profile. In reality, the valve spring often cannot follow the cam profile at high speeds. That is this phenomenon to cause excited surge (referring to SECTION.2.4.2). The cam profile is shown in Simulation chapter (referring to Fig3.3). Cam velocity and acceleration are shown in Fig.2.6 and Fig.2.7. The two figures present the standard automotive characteristic cam curves [Rothbart, 2004]. The spring collision is presented in Fig.2.8. The upper spring segment has mass A, and the lower one has mass B. The upper A has higher velocity than the lower one due to its larger

18

The velocity difference is Substituting impact with ∆ܸ term in Eq. but its new velocity is on mass A is Va. and on mass B is Vb. The initial velocity on mass A is Va. as one is inertia impulse. 19 . the expression ܸܾ ᇱ above can be reduced to 1 1 ݁ ݁ ܸܾ ᇱ ൌ ܸܽ ܸܾ ܸܽ െ ܸܾ ൌ ܸܽ 2 2 2 2 ܸܽᇱ ൌ ܸܾ and After the collision. 1988] ܸܽᇱ ൌ and ܸܾ ᇱ ൌ ݉ െ ݁݉ ሺ1 ݁ሻ݉ ܸܽ ܸܾ ݉ ݉ ݉ ݉ ሺ1 ݁ሻ݉ ሺ݉ െ ݁݉ ሻ ܸܽ ܸܾ ݉ ݉ ݉ ݉ If we assume the two mass segments are the same and it’s a perfectly elastic impact as e = 1. Eq.deflection. that the last two terms on the right hand side of equation (2. and another is collided impulse.30) We need to consider one thing. the new spring force equation is modified as ܨൌ ∆ܸ ൌ ܸܽ െ ܸܾ ீ cosሺሻ ∆߬ െ ாூ sin ሺሻ ∆݇ డ௬ ଶ డ ቀ ܫܧሺߜ݇ሻଶ ቁ െ ܯ ଵ ௗ ௗ௧ െܯ௦ ∆ ∆௧ (2.30) offered a better curve approximation for computing spring force in dynamic loading situation for valve springs. their velocities are exchanged and head in the same direction.30) are different though both they are inertia force in nature.29). The initial velocity on mass B is Vb. (2. When mass A collides with mass B. (2. their new velocities could be obtained as [Greenwood.

Fig.2.6 Cam velocity

Fig.2.7 Cam acceleration

Fig.2.8 Two spring segments collides on each other

20

2.3 Numerical Solutions

2.3.1 Finite Difference Method

Because of the importance of the wave equation to a wide variety of fields, there are many analytical solutions of that equation for a wide variety of initial and boundary conditions. However, nonlinear partial differential equations are difficult to solve by analytical methods, or where the analytical solution is even more difficult to implement than a suitably accurate numerical solution. Here we will discuss one particular method for FD, as FEM, is commonly considered a numeric but not analytical solution of partial differential equations called the finite difference method. The finite difference method begins with the discretization of space and time such that there is an integer number of points in space and an integer number of times at which we calculate the field variables, in this case just the displacement. The resulting approximation is shown schematically in the simulation chapter. For

simplicity here we will assume equal spacing of the points s in one dimension with intervals of size ݄, and equal spacing of the time steps ݐat intervals of size ݇. This simplifies the system considerably, since instead of tracking a smooth function at an infinite number of points, one just deals with a finite number of displacement values at a finite number of locations and times. Basically, there are two types of finite difference to express a continuous derivative: forward difference method and difference method. The third type is a combination of these two, that is the central difference for derivative expression. The following is the brief mathematical description for the step size of three finite difference methods: Forward Difference: ∆ ݂ሺݏሻ ൌ ݂ሺ ݏ ݄ሻ െ ݂ሺݏሻ ݂ሺݏሻ ൌ ݂ሺݏሻ െ ݂ሺ ݏെ ݄ሻ

Backward Difference:

Central Difference:

Then, for solving Partial Differential Equations numerically, there are explicit method—solving one variable for the next time step in one equation, implicit

21

Where ݄: the space step shown in Fig.2.9

1 1 ߜ ݂ሺݏሻ ൌ ݂ሺ ݏ ݄ሻ െ ݂ሺ ݏെ ݄ሻ 2 2

method-solving solving all unknowns in the time step by solving coupled coupled simultaneous equations. According to the technical paper [Crank, John and Phyllis Nicolson, Nicolson 1947], Crank–Nicolson n method is second order implicit method and it is unconditionally stable solving partial differential equations. In addition, Reddy [1984] raised a family of approximation ߠ to check the stability of many methods, methods, included in Finite Element Method and Finite Difference Method. Method ߠ ൌ 0 -> > Forward Difference scheme -> conditionally stable ߠൌ

ଵ ଶ ଶ

-> Crank-Nicolson Nicolson scheme -> unconditionally stable

ߠ ൌ ଷ -> > Galerkin method -> unconditionally stable

For accuracy and stability, Crank-Nicolson Crank Finite Difference method is implemented to compute Eq. (2.26). (2.2 However, the approximate solutions can still contain (decaying) spurious ious oscillations if the ratio of time step to the square of space step is large (typically larger than 1/2) [Thomson, 1995].

ߠ ൌ 1 -> Backward ackward difference method -> conditionally stable

2.3.1.1

Using an explicit difference at time ݐ and making a space derivative at position ݏ shown in Fig.2. 2.9. The time step is ݇ and the space step is ݄. The time index is ݆, and the space index is ݊. For the convenience of expressing simple equations in Finite Difference Method, only basic parameters are shown below.

Explicit Finite Difference Method

Fig ig.2.9 The grid on the space and time Substituting the space and time derivatives (referring to Fig.2.10) Fig.2. ) into Eq. (2.26), (2.2 we have the recurrence equation as

22

Hence.31) Where.2. ݇ସ ܽ݇ ଶ ቁ ݄ ݇ଵ ൌ ܾ݇ ቀ1 2 ቁ ܾ݇ ቀ1 െ 2 ቁ ݇ଶ ൌ െ ܾ݇ ቀ1 2 ቁ ቀ (2. it is easy to solve that solution step by step till the boundary condition is satisfied. Exchange parameters on both sides. ൯ ݇ଶ ݕ. From Fig. this method is fast to obtain the solution.ିଵ ݕ. ܾ ܿ ൌ ܽ ݇ଶ 2݇ ݄ଶ ܩሾ݀ሺݏሻሿଶ ቝ ܽൌቜ ߩሺ2ሾܦሺݏሻሿଶ ݀ ଶ ሻ 4ܿ ᇱ ܾൌ ߨ݀ ଶ ߩ ܿൌ 2 ሺݏሻ ߩߨ݀ ଶ ଵ/ଶ Fig.ାଵ.2.Where.ିଵ ݇ଷ ݕ.10 we find only one unknown ݕ. the expression can be written as ݕ. ݕିଵ.ିଵ ݕ െ 2ݕ. as a result.10 The explicit method stencil It is seen that the above finite difference expression is necessary to distinguish whether terms are known or unknown parameters.ାଵ െ ݕ. ݕ. 23 . ݕ.ାଵ െ 2ݕ. ଶ ାଵ. ݕିଵ.ାଵ ൌ ݇ଵ ൫ݕାଵ. and it uses a forward difference technique.

Substituting the space and time derivatives (referring to Fig.ሻ ப୲ ൌ gሺsሻ at time t = 0 and the boundary condition However.26).1.షభ మ values of the subsequent period – backward procedure. the expression is written as ܾ ௬ೕ. each unknown value is function of known ௬ೕ. The process is similar as the previous case.షభ ଶ ܿ ൌ ܽଶ ௬ೕశభ.3.32) yሺ0.శభ ିଶ௬ೕ. in a row.2.ାଵ and ݕ.శభ ା௬ೕషభ. 0ሻ ൌ fሺsሻ and ப୷ሺୱ. (2.శభ ି௬ೕ. to guarantee it is stable to process the explicit differential method defined as [Thomson.ାଵ . tሻ ൌ hሺtሻ at s ൌ ݈ 2. but just one difference is that it has two unknowns ݕାଵ. In addition. ା௬ೕ.శభ ିଶ௬ೕ. tሻ ൌ 0 at s ൌ 0 and yሺ݈.ܽ݇ ଶ ቆ2 െ 2 ቀ ቁ ቇ ݄ ݇ଷ ൌൌ ܾ݇ ቀ1 ቁ 2 ݇ସ ൌ െ ܿ݇ ଶ ܾ݇ ቀ1 ቁ 2 with the initial condition yሺs.11) into Eq. the explicit method is known to be numerically unstable and its error is proportional to the time step and the square of the space step. 1993] ൌܽ where p is the stability factor ݇ 1 ݄ (2.శభ మ 24 . the stability factor p is very important and is not always greater than one.2 Implicit Finite Difference Method The difference between the implicit method and the explicit method is that the implicit method needs to solve a set of equations one time.

33) Where.ିଵ ݇ଷ ݕ. 0ሻ ൌ fሺsሻ and ப୷ሺୱ. with the initial condition yሺs.ାଵ ݇ସ ܾ݇ ܽ݇ ଶ 1 2 ቀ ቁ ݄ ݇ଵ ൌ ଶ ܽ݇ ቀ ቁ ݄ ܾ݇ ቀ1 െ 2 ቁ ݇ଶ ൌ ܽ݇ ଶ ቀ ቁ ݄ 2 ݇ଷ ൌ െ ܽ݇ ଶ ቀ ቁ ݄ ݇ସ ൌ ቀ ܽ݇ ଶ ቁ ݄ ܿ݇ ଶ (2. tሻ ൌ hሺtሻ at s ൌ ݈ 25 . െ ݕିଵ. it can be expressed as y୨ାଵ.୬ାଵ ൌ kଵ y୨.11 The implicit method stencil Exchange the terms on both sides.ሻ ப୲ ൌ gሺsሻ at time t = 0 and the boundary condition yሺ0. tሻ ൌ 0 at s ൌ 0 and yሺ݈.୬ାଵ ݇ଶ ݕ.2.Fig.

ା௬ೕషభ.షభ ଶ ܿ ൌ ܽଶ ሺ ଵ ଶ ௬ೕశభ.2.3 Crank-Nicholson Difference Method The well-known Crank-Nicholson implicit method for solving the modified wave equation involves taking the average of the right-hand side between the beginning and end of the time-step (referring to Fig. In other words.12).2. ା௬ೕ.3.1. ൯ ݇ସ ܾ݇ ܾ݇ ଶ 1 2 ቀ ቁ ݄ ݇ଵ ൌ ଶ 1 ܽ݇ 2ቀ ݄ ቁ ܾ݇ 1െ 2 ݇ଶ ൌ 1 ܽ݇ ଶ 2ቀ ݄ ቁ 4݄ଶ ݇ଷ ൌ 2 െ ሺܽ݇ሻଶ ݇ସ ൌ 1 ܽ݇ ଶ 2ቀ ݄ ቁ ܿ݇ ଶ with the initial condition 26 . an average of the explicit method and implicit method is written as ௬ೕ.షభ ௬ೕశభ.శభ ିଶ௬ೕ. ݕାଵ.୬ାଵ ൌ ݇ଵ ݕ.ାଵ and ݕ.2.ାଵ.ାଵ ݕାଵ. ିଶ௬ೕ.శభ ା௬ೕషభ.శభ ିଶ௬ೕ. െ ൫ݕିଵ.ିଵ ݇ଷ ݕ.శభ మ Fig. ݕିଵ.శభ ି௬ೕ. It can be rewritten as Where y୨ାଵ.12 The Crank-Nicholson method stencil The unknown in this case is the same as in the case of implicit finite difference method. మ మ ሻ ܾ ௬ೕ. that is.ାଵ ݇ଶ ݕ.

ାଵ െ ݕିଵ. so both end of valve spring doesn’t completely follows the cam profile within two plates.ାଵ ڭ ሺଵሻ ݇ହ ൌ ݇ଶ ݕ. െ ൫ݕାଵ. any displacement could be obtained fast and easy.ାଵ should be satisfied with the boundary condition at ݏൌ ݈. We need to take care that ݕାଵ. ൯ ݇ସ And we have Using the inverse of the matrix K. 27 By the matrix. the matrix Y can be obtained as ݕଵ. in a valve train system often spring jump.ାଵ ې ݇ ێሺଶሻ ۑ 1 ێ0 ێହ ۑ ێ ۑ Where. Due to the moving boundary. In reality.ሻ ப୲ ൌ gሺsሻ at the time t = 0 and the boundary condition In order to compute the solution more conveniently. tሻ ൌ hሺtሻ at s ൌ ݈ where The above equation at the time step t ୬ାଵcan be expressed as a matrix 1 െ݇ଵ ۍ 1 ێ0 ڮ ڭێ ێ0 0 ۏ0 0 1 െ݇ଵ ڰ ڮ ڮ ڮ ڮ ڮ 1 െ݇ଵ ݇ ۍହ ې 0 ۍې ݇ ێ ېሺଶሻ ۑ 0ێ ۑ ێ ۑହ ۑ ሺଷሻ ێۑ ڭ ۑൌ ݇ ێହ ۑ 0ێ ۑ ۑ ڭ ێ ۑ 1 ݕۏ ےାଵ.35) 1 െ݇ଵ ڰ ڮ ڮ ڮ ڮ ڮ 1 െ݇ଵ 0 ې 0ۑ ۑڭ 0ۑ 1ے Another problem in this computation is the moving boundary problem. 0ሻ ൌ fሺsሻ and ப୷ሺୱ. ݕିଵ. If both end of the valve spring can follow the cam profile and be in contact with the top and bottom plates. ݕାଵ.ାଵ ݇ହ (2. bounce or coil collision occurs. we simplify this expression one more step.ାଵ ݕଷ. ሾܻሿ ൌ ݕ ێଷ. no moving boundary problem exists.34) yሺ0.ିଵ ݇ଷ ݕ. .ାଵ ے ۏ0 0 ݇ۏହ ے ሺଵሻ ሾܭሿሾܻሿ ൌ ሾܨሿ ሾܻሿ ൌ ሾܭሿିଵ ሾܨሿ (2.ାଵ ێ ےሺାଵሻ ۑ ݇ۏହ ے ݕଵ. tሻ ൌ 0 at s ൌ 0 and yሺ݈.ାଵ ݇ ۍହ ې 1 െ݇ଵ ۍ ݕ ۍଶ.ାଵ ൌ ݇ଵ ݕ. ሾܨሿ ൌ ݇ ێሺଷሻ ۑ.yሺs.ାଵ ݕଶ.ାଵ ۑand ሾܭሿ ൌ ڭ ێ ڮ ହ ڭ 0 ۑ ڭ ێ ێ ۑ ێ0 ێሺାଵሻ ۑ ݕۏାଵ.

it takes almost 2.3 Finite Element Method Since FEM is a direct numerical solution. but provides a slower calculation.20) the two moving boundary conditions can be defined as Where ݆ is space index in the grid and ݄ is space step ሺݏሻ ൌ ሺ݆݄ሻ ൌ ଵ ሺ݆݄ሻଷ ଶ ሺ݆݄ሻଶ ଷ ሺ݆݄ሻ ସ . by Eq. it takes much more time to cumulate every term and not very efficient to use practically. Next step is to use the boundary condition to determine the solution. In fact. one for f(s). it is the main method to be implemented in most commercial FEA software. By the method of separating variables.5 hours to obtain the solution in a symbolic expression. the solution 28 . When element stiffness matrix is given. there are two types of computational methods . a Fourier series representation is expressed as [Kreyszig. The symbolic expression is usually very powerful and easy-using.2 Fourier Series There are other methods such as Fourier Series. using Fourier series by satisfying the initial condition and boundary condition. We could add two extra derivative boundary conditions to check the pitch angle of each element. but only takes 5 min in numeric expression. that is.3. (2. In MATLAB. Hence. We used FEA software (ABAQUS) in this thesis. Finally. The continuum dynamics is used to describe the elasto-plasticity dynamics.3. as a result. this method is not a good way to build the model.each computational node in Crank-Nicholson Finite Difference Method should have tolerable error to approximate the real condition. 2. this method is very slow in comparison with Finite Difference Method. 2. 2006] ିୠ nπ yሺs. Each element along the spring helix should have a minimum pitch angle. tሻ ൌ e ଶ ୲ ሺa୬ cosሺwtሻ b୬ sin ሺwtሻሻsin ሺ sሻ l ୬ୀଵ ஶ However. Such solution is usually regarded as close form solution or analytical solution. One is dynamic and the other is geometrical viewpoint. we obtain two ODEs.symbolic and numeric expression. and it is impossible to have the pitch angle on upper elements is lower than the pitch angle on lower elements. For Crank-Nicolson method. setting yሺs.27). and the other one for g(t). tሻ ൌ fሺsሻgሺtሻ. the matrix expression in Eq. series solution of PDE. It recognizes the object with two viewpoints. (2.

the construction of global stiffness matrix 2. the solution of the simultaneous equation In Optimization chapter. 2.4 Evaluation of Dynamic Performance 2. At high speeds. so avoiding the resonance has been an important issue in engine design. The input in FFT should be force or displacement. in order to check if those optimal parameters are accurate.2.13. In Optimization chapter. 2003]. vibration and impact effects play very important roles in valve train applications. In the modern technology. the computational speed is dependent on the chosen mesh seeds [Shimoseki. ABAQUS-MATLAB is demonstrated to compare them.will be obtained through the following process. there are many methods to avoid this phenomenon. there is a direct linear relationship between the input and the output represented in Fig.13 The system transformation for the natural frequency 29 . such as using conical spring or adding the friction surface by double springs or dampers [Nunney. the provision of the boundary condition 3. Fig. 1. In any linear system.1 Fast Fourier Transform (FFT) Surging. The ideal power spectral density in valve springs should be as small as possible and has a larger fundamental natural frequency. this power spectral density is taken as the objective function to process with other constraints such as force constraint and other constraints. The natural frequency of springs depends on the stiffness and mass. Fast Fourier Transform (FFT) has been used to evaluate spring performance for a long time.4. 2007] For determining the fundamental natural frequency. By combining the mass matrix and the stiffness matrix at each element to form a global matrix. the resonance may occur and cause some phenomenon such as valve float and surging at a certain high frequency.2. and the output is the power spectrum.

This input-output relationship in terms of the frequency-response function can be written as [Thomson. the only difference between both is that FFT is much faster. FFT provides a good evaluation for the power spectrum as well. it is necessary to use other expressions to do the . 1993] ݂ሺݐሻ ൌ ܽ ܽଵ ܿݏሺݐݓሻ ܽ ڮ ܿݏሺ݉ݐݓሻ ܾଵ sin ሺݐݓሻ ڮ ܾ ݊݅ݏሺ݉ݐݓሻ (2. and are defined as |C୫ | ൌ ටa୫ ଶ b୫ ଶ మ φ୫ ൌ tanିଵሺ 30 b୫ ሻ a୫ If this system is discrete. In addition. In optimization.36) ݂ሺݐሻ: the valve െ lift curve ܽ ܽ݊݀ ܾ : the coefϐicients of Fourier Series ݓ: fundamental frequency ൌ ݉: number of harmonics ߬: periodic motion of period 2ߨ ߬ where For a continuous system. The valve-lift curve ݕൌ ݂ሺݐሻ is assumed to consist of a large number of sinusoidal and cosine terms [Thomson. 1993] where ܨ ݁ ௪௧ is a harmonic function ݕሺݐሻ ൌ ܪሺݓሻܨ ݁ ௪௧ A Fast Fourier Transform (FFT) is a very efficient algorithm to compute the Discrete Fourier Transform (DFT). we would use that as the objective function to obtain the lowest power spectral density. the coefficients a୫ and b୫ are determined as follows ܽ ൌ 2 ఛ/ଶ ܾ ൌ න ݂ሺݐሻ݊݅ݏሺ݉ݐݓሻ݀ݐ ߬ ିఛ/ଶ 2 ఛ/ଶ න ݂ሺݐሻܿݏሺ݉ݐݓሻ݀ݐ ߬ ିఛ/ଶ (2.37) and ܥ is the amplitudes of these harmonics used to evaluate frequency performance. Another symbol ߮ is phase.

1963] Where ݂ሺݐሻ ൌ ܿ ܿଵ sinሺ ݐݓ ߮ଵ ሻ ܿଶ sinሺ2 ݐݓ ߮ଶ ሻ ڮ ܿ sin ሺ݉ ݐݓ ߮ ሻ 1 ܿ ൌ ܽ 2 And the amplitudes on these harmonics are usually plotted as the absolute values. we can use the exponential function to express the triangular functions. (2. Eq.38) 31 . Besides.36) can be simplified by the sum and difference formulas [Kreyszig. (2. we get ݂ሺݐሻ ൌ ஶ െ1 ௪௧ ݅ሺ݁ െ ݁ ି௪௧ ሻ 2 ܽ 1 1 ሺܽ െ ܾ݅ ሻ݁ ௪௧ ሺܽ ܾ݅ ሻ݁ ି௪௧ ൨ ൌ 2 2 2 ୀଵ ܿ ሾܿ ݁ ୀଵ ஶ ௪௧ כ ܿ ݁ ି௪௧ ሿ ൌ ܥ ݁ ௪௧ ୀିஶ ஶ Assuming there are N samples in this system. the amplitude is 1 ܥ ൌ ݂ሺݐሻ݁ ି௪௧ ൌ ܽ െ ܾ݅ ܰ ୀ ଶ ܾଶ |ܥ | ൌ ඥܽ మ ேିଵ (2. 2006] as ܽ cosሺ݉ݐݓሻ ܾ sinሺ݉ݐݓሻ ൌ ܿ sin ሺ݉ ݐݓ ߮ ሻ Its expression only in terms of sinusoidal function is written as [Wahl. Substituting ܿݏሺ݉ݐݓሻ ൌ 1 ௪௧ ሺ݁ ݁ ି௪௧ ሻ 2 ܿ ሺܿଵ ~ܿ ሻ: the amplitudes of these harmonics divided by two ߮ ሺ߮ଵ ~߮ ሻ: phase ܿ ൌ 1 ሺܽ െ ܾ݅ ሻ 2 ݊݅ݏሺ݉ݐݓሻ ൌ into Eq.computation.36). Fourier series can be represented in terms of the exponential function.

4. (2. a stress wave is created that travels back and forth from one end to the other exactly like the swimming-pool wave. the designer must be certain that the physical dimensions of the spring are not such as to possess natural frequency close to the frequency of the applied force. 1963]. Daniel Bernoulli.38) is like Fig. The automotive valve spring surge. Leonhard Euler. In general.8.2 Natural Frequency The earliest natural frequency phenomenon was discovered by Galileo Galilei with pendulums and musical strings in 1602.harmonics in X axis and ܥ is the amplitude in Y axis. (2. this wave will travel down the length of the pool. 1602] 32 . When helical springs are used in applications requiring a rapid reciprocating motion. its phase on the excited harmonic number presents chaos as the cam rotates at high speeds (Referring to Fig.2. the study on vibrations made great progress. If a wave is created by a disturbance at one end of a swimming pool.38) with SQP technique are used to demonstrate helical spring optimization. In Optimization chapter. The natural frequency. As a result. be reflected back at the far end.3. resulting in the surge of stresses that can have many times of the stress for static loading.26). Eq.8 is largest and then gradually Here we need to consider the figure in Eq.2).30).4.4.19 and SECTION. As a result. the amplitude on the 1st harmonic number in Fig. the phase of the compressed valve spring can follow the phase of the cam profile as the cam rotates at low speeds. and Joseph-Louis Lagrange. the vibration problem was studied by Jean le Rond d'Alembert. (2. Another interesting thing is the phase between the compressed valve spring and the cam profile.34) and (2. where n is number of decreases till it approximates the excited number of harmonics and then keeps decreasing [Wahl. (2. The numeric solution in the implemented finite difference method is discrete. Otherwise resonance may occur.4. and it is called spring surge. in the worst possible situation. and continue in the back-and-forth motion until it is finally damped out. Historically. is that the spring actually jumps out of the contact with the end plates. If one end of a compression spring is held against a flat surface and the other end is disturbed. The amplitudes on these higher harmonics decrease as the increasing harmonic number. The earliest natural frequency is defined for lumped parameter system as [Galileo Galilei. 2. By contrast. The same effect occurs in helical springs. electric field intensity and polarity were also measured by Hertz in 1887.

the outer diameter of the inner spring may be made equal to the inner diameter of the outer spring.݂ൌ ݊ ݇ ඨ 2 ݉ (2. Lain G. utilize a volute form with the smallest diameter and hence stiffens coil at 33 . 1961. 1963. 1990. 1988]: ݂ൌ ݊ܽ .3 Special Designs to Reduce Resonance To avoid resonance in any system containing a spring has been an important subject for study and research. Furthermore. an internal sleeve may be installed within the spring to reach the similar effect.4.40) It can be expressed in terms of design variables as [SMI. … ݊ 2݈ (2. Kurisu. [Chironis. less commonly. This rubbing contact promotes a friction damping effect. 1991.39) Another method to calculate that is that the fundamental frequency of a wave in a string with either both ends open or both ends closed can be found using the following equation [Lin. Various additional measures may be taken to minimize surge. The incidental advantages of using double springs include greater spring stability and less risk of engine damage in the event of a spring breaking. so that rubbing occurs between them.41) where n = 1 is the fundamental natural frequency 2. Alternatively. which suppresses surge by dissipating its energy as heat. Main. 2007] Double springs – they are arranged concentrically about the valve stem. Nunney. Springs of this type are either close coiled at their stationary end or. The valve spring with valve fingers – they are located from the stationary ends of the valve spring and press inwards to make rubbing contact with its centre coils. Wahl. each having a different natural frequency of vibration. 1987. ݊ ൌ 1. 2002] ݂ൌ ݊݀ ܩ ߨ݊݀ ܩ ඨ ൎ ඨ ଶ 2ߨ ܰ ܦ2ߩ 2 ݈ܦ2ߩ (2.2. Conical springs – this type is also commonly used to suppress surge. Paranjpe.

Variable wire diameter springs – the advantage is. and no space limitations.the moving end. which produces change in its natural frequency of vibration so that resonance is avoided. The effect is to vary the number of active coils in the spring during the valve lift. However. space limitations are the biggest problem in this type due to different coil diameters on both ends. it has original spring height. 34 . it is very difficult to manufacture and the stress may be elevated over the original design. In general.

Only if the numerical solutions agree and describe the intricate dynamic behaviors from the physical experiments. an electrically powered AC motor to drive the valve train. 3. All moving components not required for the operation of the valve train were removed from the engine. 3. the experimental apparatus and setup are described. in developing and evaluating predictive dynamic models. .5L). The file can be saved in an EXCEL file format. The modified wave equation was found accurately predicting the valve train behavior.1 Motivation and Goal Valve spring motion was investigated with computer simulation technique in Simulation chapter (Chapter 4). hence no gas forces.Chapter 3 – Experimental Setup and Description 3. The accuracy of any math model must be verified by physical model. and instrumentation fitted to the valve train system. This real time display can operate in a smooth scroll or triggered sweep mode of operation. Event markers with comments allow us to annotate our data acquisition session with descriptive information as we are recording to disk. It was identified that a few lower frequency harmonic components from the cam profile have significant effects on exciting the first one or two lowest natural frequencies of the valve spring. offers a real time display and disk streaming for the Windows environment. These findings allowed the development of a new valve spring form that can improve the performance of the valve train at high speed loading. The coil clash actually was helping to decrease the excited resonance. MATLAB was used in doing the frequency analysis and comparing with the simulation results via generated EXCEL files by WinDAQ.3. The derived equations and procedures in Simulation chapter can be used by research investigators in understanding system response. and can be scaled into any unit of measurement. 35 . combustion occurred in the engine. In this chapter. The electrically driven cam shaft and valve train is illustrated in Fig.1. WinDAQ (Appendix 5).3 Experimental Apparatus The experimental apparatus included a vehicle engine (1987 GM & Isuzu four-cylinder with a displacement of 1.2 Software The data acquisition software package. then we can say the math modeling is accurate and successful.

A data 36 .1) Dayton Compressor Duty Motor (Appendix 4.7) Motor Oil Cen-Tech Photo Sensor Tachometer (Appendix 4. The pulley was marked to measure the camshaft speed with a photo sensor (Appendix 4.4) KISTLER Charge Amplifier (Appendix 4. The function of the pressure gauge is to make sure that the oil in the engine block was at normal levels and at the regular oil pressure of around 40 psi.4 Measurement A piezoelectric force transducer installed under the valve spring seat recorded the spring force applied to the engine block.9) Oil Tube Pulley (Appendix 4.3. The following is a list of the major parts in the apparatus: GM&ISUZU Engine (Appendix 4.3) KISTLER Force Transducer (Appendix 4.Fig.2) Oil Pump Oil Storage Pressure Gauge Cable (Appendix 4.3).2.1 The components and setup of experiment Motor oil from an external pump was fed into the engine block via the oil tubes to lubricate the valve train.6) 3. as illustrated in Fig.3.8) V-Belt (Appendix 4.5) DATAQ acquisition (Appendix 4.

acquisition system was utilized to collect the dynamic data.2 The transducer under the spring seat Fig. especially its ultra 37 . Fig.3.3.3 The flow chart of signal for data acquisition Quartz (piezoelectric element) was used in this transducer.3. as illustrated in Fig.3.

3. The output voltage in this circuit can be written as [Kistler. static measurement are better with quartz than with any other deformation induced force sensing material. The force transducer generates electric charge.3. The simplified charge amplifier model is shown in Fig.3. 1989] ܸ ൌ where െܳ 1 ܥ 1 1 ሾܥ௧ ܥ ܥ ሿ ܣ ܥ (3. Then the charge was converted into voltage in a charge amplifier. The transducer in this experiment is load washer type.4 Direct Force Measurement [Kistler.1) ܥ௧ : transducer capacitance ܥ : cable capacitance ܥ : rangeሺfeedbackሻ capacitor ܴ௧ : time constant resistor ܸ : output voltage ܴ : insulation resistance of input circuit ܳ: charge generated by the transducerሺinput chargeሻ Fig. When the valve spring is compressed. the force transducer under the spring seat induces a force downward as shown in Fig. 1989] 38 .5.high insulation resistance. It uses the inverting voltage amplifier with a MOSFET or JFET as its input to form the high resistance.4.

5. 2008]. (3. Specifications of the Charge Amplifier. (3.3) and (3.6) 39 .The maximal force.3. so we have ܳ ൌ ܸ ܥ ܭ ܴ ݊݅ ܥേ 10ܸ ՜ ܨൌ ? ܰ ܨൌ where ܸ : output voltage ܸ ܥ ܭ ܴ ܨ ௫ ܳ௫ (3. 2008] Because of the high open loop gain. The relation of force and voltage can be obtained using the simple ratio comparison [Bishop. the range capacitor can be found .Fig. ܳ௫ ൌ 5000 ݊݅ ܥേ 10ܸ ՜ ܨ ௫ ൌ 6000ܰ ܳ ൌ ܸ ܥ ݊݅ ܥേ 8ܸ ՜ ܨൌ ? ܰ െܳ ܥ (3.1) is reduced to the simplified type by ignoring the capacitance terms.3) (3.5 Simplified charge amplifier model [Bishop. we have ܸ ൌ From Appendix 4.5) Comparing (3.4) The gain of charge amplifier ܭ in different voltage range is different. F୫ୟ୶ is 6000N after the unit conversion. this leads to the relation of force and voltage. Eq.2) (3.5).

1.1 to 3.8).1.2.Since parameters in Appendix 4.1 Numbers of data points read at different camshaft speeds Camshaft Speed(rpm) 1368 2165 Data Points one cycle 632 399 3.2.5 Comparison Group There are 4 groups of data taken in the experiments.68ܸ ሺݐሻሺNewtonሻ (3. Number of Data points in one cycle at different speeds is expressed as Eq. Different camshaft speeds were generated using different groups of pulley (Appendix 4.7) Experimental data ܨሺݐሻ at different speeds are presented in Appendix 3. this leads to Eq.7) ܨ ௫ : the maximal force in standard full voltage ܳ௫ : the maximal charge in standard full voltage ܥ : range capacitor ܭ : the gain of charge ampliϐier ܴ : the ratio of the full voltage in the acquisiton and the standard full voltage (3.2.4 to 4.8). They are presented in Table. ܿ݁ݏ ሻ min Data points ൌ ሺsample rateሻ ൈ camshaft speedሺrpmሻ 60ሺ ܨሺݐሻ ൌ 103. The faster the acquisition the greater the accuracy. (3.3. The experimental data at 1368 rpm and 2165 rpm were included in Appendix 3.6.6. The maximal speed in this duty motor (Appendix 4.1.2. Data points at 1368 rpm are 632 in one cycle. and those at 2165 rpm are 399.5 to 3. They are presented in the Table 3.2). The speed relation between motor and camshaft is given by r1 ൈ w1 ൌ r2 ൈ w2 40 and . Many important factors can affect the experimental accuracy.2.1 to 3. One of them is the sample rate in the acquisition.8) Table 3. based on different speeds and different installed spring direction (Appendix 3.4. The maximal sample rate in this DATAQ Acquisition is 14400 Hertz. The experimental data at 2372 rpm cited from Lin’s paper [2006] was also included in Appendix 3. (3.2.2) is 3450 rpm.

w2 is the camshaft speed. r1 is the pulley radius on the motor.where w1 is the motor speed.2 The comparison groups in the experiment Group 1 2 3 4 Camshaft Speed(rpm) 1368 1368 2165 2165 Spring Installed Direction Normal Opposite Normal Opposite w2 ൌ w1 ൈ r1 r2 41 . and r2 is the pulley radius on the camshaft Table 3.

4. Fig. (2. and represents nonlinear curve.024m .35).0386m 0. DRAM= 1GB.1. It is assumed that the bottom end of the spring rests in a static reference plane.0485(m) 0. 1368. for the boundary condition at we have yሺ݈.2 are also the basic parameters in this simulation besides those in Table 4.1 and 4. and the top end of the spring follows the cam profile. Table 4.1 Introduction This chapter provides a simple computational tool based on a finite difference solution of the equations. derived using an internal virtual force term and Hamilton Principle for analyzing the dynamic modeling of helical valve springs subjected to dynamic loading. One variable-pitch coil spring was used as the simulation model at three different speeds.1).2.86 GHz). A computer program (Appendix 1) was written in MATLAB. and it took approximately 5 minutes to obtain a converged solution on a laptop (INTEL Core Solo processor T1350. Fig4.1 is the set of physical parameters used in this simulation.1 is the first partial derivative of displacement with respect to the arc length. In the same valve spring model at different speeds.2 shows the cam profile.0399(m) 0. the damping coefficient is not always the same.0255m 0.4.Chapter 4 – Simulation 4.2. CPU= 1. Table 4. Another key in this simulation is to determine the damping coefficient. Eq. tሻ ൌ hሺtሻ ൌ spring installed length െ cam proϐile Fig. It involves many factors (referring to SECTION.004m 0. 2165 and 2372 rpm.1 The related parameters of the helical spring in this simulation Terms D(coil diameter) d(wire diameter) Original Free Height Preloaded Height(without transducer) Preloaded Height(with transducer) Solid height 42 Value 0. Hence.

000015(sec) 0.4.s/m) 70.7462(N. 2165.3731(N.s/m) 30.01(m) Fig.Displacement parameters( with transducer) Mass Density 7860 Shear Modulus Camshaft speed (rpm) Damping coefficient(1368rpm) Damping coefficient(2165rpm) Damping coefficient(2372rpm) Time increment k Spatial increment h(Node) 77.1 The pitch angle of helical springs along arc length which is along the helix on the initial condition without the preload 43 .9Gpa 1368. 2372 60.8705(N.s/m) 0.

20). the first derivative of displacement (referring to Eq. Fig. Fig. 21.4.0704 (sec) at 1368 rpm. Fig. Force and Frequency 4.2.2. 44 . so the chosen node in this simulation should be as near to lower nodes as possible. please refer to Experiment and Verification chapter (Chapter 4).2 Displacement.Fig. 1987. The transducer is installed at the seating end of the valve spring.4 is the dynamic displacements at node 8. the moving boundary section). The displacements on upper nodes should always be larger than those on lower nodes (referring to SECTION.3.2 The cam profile 4.4.3. but it is for the pitch angle. For experimental detail.19) and (2. In addition. and 34.1 Displacement This section shows the dynamic displacements.4. The solution looks reasonable and is verified with experimental data.5 has similar phenomenon as Fig. 1989 and 2006].1. they agreed well with the results in related technical papers [Lin.4.4. This simulation and optimization program in Optimization chapter (Chapter 5) always uses node 5 to run the computation. (2.4.3 shows a snapshot of the dynamic solution at t = 0 and t = 0.

3 The spring length varies with preload. and 34 varies at 2372 rpm 45 .4 The spring length on node 8.4. 21.Fig.4. and compression at 1368 rpm Fig.

4.7.15. 4. A few related books or papers suggested it should be at least 15 or even more. numerical solutions were computed at three different cam shaft speeds: 1368.10. Resonance can increase the individual coil deflection and stress levels well above the amounts predicted by static loading conditions. in addition. solution also included spring force data and the power spectrum with different installed directions. 4.2. It is observed in these figures that there are spikes in the spring 46 .Fig. 4. 4.14 and 4.11. These spring force figures are shown in Fig. Resonance can also cause spring surge. lowest natural spring frequency should be at least 13 times that of the base frequency of the cyclic loading. which can result in considerably lower loads than calculated at the maximum spring deflection when the spring force is needed most for the return of valve mass.2 Force and Frequency Resonance occurs in a spring when one of the component frequencies of the cyclic loading is near one of the natural spring frequencies. 2165 and 2372 rpm.6.5 The change in pitch angle on node 5 4.4. Numerical solutions were not limited to just displacement at each node. Then physical experiments were performed at these speeds for comparison with numerical solutions. To avoid resonance. For the accuracy of dynamic equations.

In Fig. Rothbart.4. 1963. (2. the resonance of the valve spring with the 19th harmonic number was pronounced.8. In order to obtain the natural spring frequency.1). 4.12 and 4.12. This process is shown in Fig.23. the phase can be used to determine the excited harmonic number. resonance of the valve spring with the 12th harmonic number was pronounced. As a result. 1993]. the resonance of the valve spring with the 11th harmonic number was pronounced. its phase always follows the phase of the cam profile. The output.8 and 4. as mentioned in SECTION.4. The power spectral curve of the valve spring force revealed the excitation of valve spring harmonics. the input in FFT is spring force data. 4. so the fundamental natural frequency can also be obtained as 47 . The same method is used to determine the excited harmonic number in other phase figures as well.19 shows the phase comparison between the simulation at 1368 rpm and the cam profile. These figures agreed qualitatively with those in some papers or textbooks [Wahl.17. 2004. Before resonance of the valve spring is pronounced. Here the comparison of numerical solutions with experimental data is based on Eq. Lin. 4. the moving boundary solution is very accurate.2) and (2. At 1368 camshaft rpm. 1983. indicating the occurrence of coil clash at the lower end of the spring.4. 4. In Fig. Fig.4. These power spectral plots are shown in Fig. 2004]. the amplitude on the harmonic number gradually decreases till the excited harmonic number forms a spike.4.9.2ሺHertzሻ ሻ 60ሺ min Furthermore. the fundamental natural frequency of the valve spring excited by the cam profile is the 19th harmonic.27).force amplitude. The phase is the displacement with respect to the excited force.2. Shigley. power spectrum was obtained using FFT (referring to SECTION. 1993.2.13.17. In addition.4. Pisano.16 and 4. so the fundamental natural frequency is the fundamental natural frequency ൌ 1368ሺ݉ݎሻ ܿ݁ݏൈ 19ሺ݄ݐሻ ൌ 433.9. It is obvious to find out the phase on the 19th harmonic number doesn’t follow that and shows chaos [Thomson.16 and 4. so the fundamental natural frequency could be obtained as the fundamental natural frequency ൌ 2165ሺ݉ݎሻ ܿ݁ݏൈ 12ሺ݄ݐሻ ൌ 433ሺHertzሻ 60 ቀminቁ In Fig.13.13. the harmonic content of the data was similar to those in 2165 and 2372 rpm.4. As the harmonic number increases.2.

the valve spring was installed in a normal (smaller pitch end is the stationary end) or opposite direction (smaller pitch end is the moving end) presenting an opposite pitch curve and the computer program simulates that very well.41).004݉.4.21ሺHertzሻ 2 ݉ ܽ ൌ 422.29ሺHertzሻ ଶ 2ߨ ܰ ܦ2ߩ where ܩൌ 79 ൈ 10ଽ ܲܽ.0255݉. The simulation shows a phenomenon that springs should have better performance in a normal installed direction than that in an opposite one due to the smaller amplitude. Besides the above method obtaining the natural frequency.3 is the excited amplitude comparison table.86ሺHertzሻ ቁ 60 ቀ min The above results indicated the fundamental natural frequency of this valve spring is around 433 to 434 (Hertz). The valve spring presents nonlinear variable pitch angle.42݉. In Optimization chapter. ݀ ൌ 0.39 to 2.18. ߩ ൌ 7860 the fundamental natural frequency ൌ 1 ݇ ඨ ൌ 423. ݈ ൌ 0. the amplitude of the power spectra on the excited harmonic number was used as the objective function for optimization by sequential quadratic programming (SQP). other methods. ܦൌ 0.2. In order to verify that. other parameters: Table 4. (2. more power spectra are simulated at different speeds.02ሺHertzሻ 2݈ In the experiment. Table 4. can be used for calculation as well. hence it is obvious to find these below calculated values presents a deviation. and ݉ଷ 48 . the fundamental natural frequency ൌ the fundamental natural frequency ൌ ݀ ܩ ඨ ൌ 423. and presented in Fig.the fundamental natural frequency ൌ 2372ሺ݉ݎሻ ܿ݁ݏൈ 11ሺ݄ݐሻ ൌ 434. Eq. The comparison table is presented in Table 4.1 ݇݃ .

4.6 A normal installed direction at 1368 rpm in the simulation Fig.4.7 An opposite installed direction at 1368 rpm 49 .Fig.

4.Fig.8 The power spectrum with a normal installation at 1368 rpm Fig.9 The power spectrum with an opposite installation at 1368rpm 50 .4.

10 A normal installed direction at 2165rpm Fig.4.Fig.4.11 An opposite installed direction at 2165rpm 51 .

4.13 The power spectrum with an opposite installation at 2165rpm 52 .Fig.12 The power spectrum with a normal installation at 2165rpm Fig.4.

14 A normal installed direction at 2372rpm Fig.15 An opposite installed direction at 2372rpm 53 .4.Fig.4.

Fig.4.16 The power spectrum with a normal installation at 2372rpm Fig.17 The power spectrum with an opposite installation at 2372rpm 54 .4.

Fig.19 The phase comparison between the simulation at 1368 rpm and the cam profile 55 .4.4.18 The power spectrum at different speeds Fig.

and shown in Fig.06 432.Table 4. Fig. FEA software ABAQUS.22 shows a maximal stress 3. Fig. was used to verify the resonant effects for the stress.4 Natural Frequency (Hertz) 432.4.31 ൈ to 5.2 The excited harmonics comparison table Camshaft speed (rpm) 1367 1443 1528 1624 1732 1856 1998 2165 2362 2598 2887 Excited Number 19 18 17 16 15 14 13 12 11 10 9 Table 4.23 also shows the maximal stress up . 56 10ଵଶ Pa on the first mode of excited resonance.21 shows a spring with the loading around 150 Newton downward and its stress distribution.4.4.59 6079.06 433 433. The maximal stress in the dynamic loading is 9.93 433.88 432.4.01 ൈ 10ଵଶ Pa on the second mode of excited harmonic number. The first and second mode related stress graphical chart were observed. and 4.05 4.22.9 432. Destructive effects under resonance are very obvious. Fig. According to Wahl’s papers [1963].03 433 433.64 4807.3 Resonant Effects for the Stress In this section.64 8340. the first mode is most easily to take place. over-torsion. Fig. not the valve spring in this experiment.3 The amplitude comparison table Group 1 2 3 4 5 6 Camshaft Speed(rpm) 1368 1368 2165 2165 2372 2372 Installed Direction Normal Opposite Normal Opposite Normal Opposite Amplitude on an excited number of harmonics (Power Spectrum) 4983.23 6924.4 ൈ 10଼ Pa.4.20 shows the standard valve spring model.67 7133.9 433 433.23.

21 The stress distribution under a normal static loading 57 .20 The three dimensional helical spring Fig.4.Fig.4.

4.23 The second mode of excited resonance 58 .4.Fig.22 The first mode of excited resonance Fig.

The combination of the QP subproblem and a constrained steepest descent method is called sequential quadratic programming (SQP). So it can be expressed as ܴ หݒ ห ݑ ୀଵ ୀଵ where ݒ ݅ ݏthe Lagrange multipliers for the equality constraint and free in 59 0 ൏ ߛ ൏ 1 ՜ ߚ ൌ ߛԡ݀ ԡଶ . Taking into account the power spectral distribution and a series of technological constraints.2 Algorithm For this optimization problem. and could also treat the inequality and equality constraints. and estimate the initial ݔ . 1984. variable coil diameter D(s). 2001. ߛ. ܴ . penalty parameter. among which an optimal spring was selected for an automotive application. A special objective function is implemented to optimize the power spectral density on the excited harmonic number. namely to replace the valve spring of a sporty vehicle.Chapter 5 . Venkataraman. and the constant. We describe the equations and steps as follows [Vanderplaats. This method is very powerful and useful. we use the sequential quadratic programming (SQP) to solve that. The excited harmonic is also verified by ABAQUS and shown in SECTION. Set iteration counter k=0. 5. Arora. 2002] 1. ܴ is the sum of all the Lagrange multipliers of the QP subproblem at the point ݔ . it will solve the quadratic programming (QP) subproblem at each iteration and then use the BFGS to estimate the Hessian of the Lagrangian.5. The design goal is to reduce the maximal power spectral density on the excited harmonic number.1 Goal A modified methodology for the optimum design of nonlinear helical springs is presented in this chapter. Several design variables that have an influence on the mechanical properties of the spring must be considered: variable pitch angle p(s). For this method.Optimization 5. many types of springs were devised. and variable wire diameter d(s).5.

ܴ . fmincon. we must find the maximum constraint violation. ߙ . Quasi-Newton. compute the cost and constraint functions and their gradients. and it will generate C୬ . ܸ 3. could handle the optimization problem with constraints. The command. Check the necessary condition of penalty parameter. The FFT command in MATLAB is fft(data input). More guessed initial points need to be used to search for the global optimal value. 5. Here. Update the iteration counter as k = If the criterion is satisfied here then we stop the steps. Power spectrum is defined by these two coefficients as: . and their gradients we got from step 2 to k + 1 and then return to step 2. Check the convergence criterion 2. ܸ ߝ1 and ԡ݀ ԡ ߳2. 4. Use the cost. or other search algorithms. fminunc. The default command. the commercial software MATLAB is used to solve this optimization problem. is based on the interior-reflective Newton method and uses the preconditioned conjugate gradients to search the direction. At ݔ .3 Demonstration For the accuracy and the computational speed. fmincon and fminunc. If the initial guessed solution was ideal the program may not find the global optimal value. Set ݔାଵ ൌ ݔ ߙ ݀ . ܽ݊݀ ݑ . Let the present penalty parameter ܴ ܴܽ ݏାଵ. 6. ݒ . ݀ . ݑ is the Lagrange multipliers for the inequality constraint. the QP subproblem may not find the solution. This method is gradient-based where the objective and constraints functions have the continuous first derivatives. and ݀ is the search direction Then we can use that equation to calculate the step size.4). 5. constraints functions. and then minimize the descent function to get the step size.2. sign. obtain the search direction. along the search direction. The search result most likely would be dependent to the chosen algorithm. a real part a୬ and a imaginary part b୬ (referring to 60 SECTION. The starting point would affect the performance of the algorithm based on these commands. It is composed of two parts. we could use the fmincon or fminunc functions that are based on the SQP. Other guess initial points need to be used to search the optimal value in MATLAB. In MATLAB. At some points.

݊ is time index in the grid. ଶ C୬ ൌ ඥaଶ ୬ b୬ మ 5. ሺ݊ െ 1ሻ݇ሻ݁ ି௪௧ ܰ 1 ൌ ݕହ. ݇ହ ሻ݁ ି௪௧ . ݇ is time step. a Fast Fourier Transform was performed on computed force data.38). ܰ ୀ ୀ ேିଵ ேିଵ (5. ݁ ି௪௧ . where ݊ is 0. ୀ ݕሺݏ.1. ݐሻ݁ ଵ The force transducer in this experiment was installed at the stationary end of the valve spring.4 Fast Fourier Transform in MATLAB In order to evaluate resonance of valve springs. and Substituting Eq. the spatial length ݏshould be close to zero at the stationary end.Spring performance can be evaluated.1). where ݄ is space step. (2. we have ି௪௧ ܥ ൌ ே ∑ேିଵ . (2. It generated a power spectral plot. it leads to the power spectral amplitude equation 1 ܥ ൌ ሺ݇ଵ ݕସ. (5. as a result. ܰ is total samples. ݉ is the harmonic number. The chosen node is expressed as follows.1) ݆ is space index in the grid. In order to approximate the experimental data. ܰ ୀ ேିଵ (5. For the amplitudes on these harmonics. the 5th node along the spring helix was chosen to process Fast Fourier Transform. ݆ ൌ 5 and ݐൌ ݊݇ (n time steps) This leads to the new expression 1 ܥ ൌ ݕሺ5݄.2 … N െ 1 ݏൌ 5݄.34) into Eq. referring to Eq.2) 61 .

5.0385 ݈ 0.000015 ݇ As a result. Variable Pitch Angle Variable Coil Diameter (Conical Spring) Variable Wire Diameter Variable Pitch Angle. hence we have where for ݉ 8 (the excited harmonic number always appears as m is larger than or equal to eight due to the limited camshaft speed). ݀ସ . ܦଵ . (5. ݀ଵ . ܦସ ሻ and the spring height constraint 0. ݇ସ ሻ ൌ ݂ ൫ሺݏሻ. the objective function is defined as which is subject to the maximum spring force constraint 120 ݃ሺݔሻ 230ሺܰሻ ൌ the power spectral amplitude on the excited harmonic number (5. Total samples in Eq.3 . 5. ݀ଶ . In addition.2) and constraints (Eqs. ସ .4) In engine design. 2. ଷ .0387. and ܦሺݏሻ. In optimization. 0. ݓሺcamshaft speedሻ is 2372 ሺrpmሻ.02529511 ሺsecሻ. ݀ଷ . This leads to total samples as ܰൌ ߬ 0. ݀ଵ . and Variable Wire Diameter In this Chapter. ܦସ ሻ τ ሺtime of one cycleሻ is where ݇ ሺtime stepሻ is 0. ܦଶ . 5. Four types of optimization were discussed in this chapter.000015 ሺsecሻ. ݀ሺݏሻ. ଶ . ଷ . ܦሺݏሻ൯ ൌ ݂ ሺଵ . maximal spring force was constrained in a reasonable range. dimensions of valve train components are always constrained in a certain range. 1. 3.2) is obtained as follows: ܰሺtotal sampleݏሻ ൌ ߬ ݇ ܥ ൌ ݂ ሺ݇ଵ . ݀ሺݏሻ. ܦଷ . the objective function (Eq. ସ .3) min ݂ ሺଵ . Variable Coil Diameter. ݀ସ . and ௪ ൌ 0.1) for the convenience of comparison. ܦଷ . ଶ .02529511 ൌ ൌ 1686. ݀ଶ . spring height was also constrained as approximate as original height (Appendix 4.4) are the 62 . ܦଵ . 4. ܦଶ . (5.The parameters ݇ଵ and ݇ହ included ሺݏሻ. ݀ଷ .

1 The flow chart of optimization program 5. Fig. For the initial trial point in SQP.4. ܦଶ ൌ 0.1. One end will have a higher rigidity (or spring rate) than the other end.5. For upper and lower bound of the design variables. ݀ଶ ൌ 0. ݀ଷ ൌ 0.1 Variable Pitch Angle A variable pitch valve spring has different levels of compression. In this case.0255ሺ݉ሻ ݂ ሺଵ . The optimization flow chart is shown in Fig. the diameter of the wire and the mean coil diameter must be kept as constants. so we have ݀ଵ ൌ 0.5.004ሺ݉ሻ and then the objective function is defined as ܦଵ ൌ 0. ܦଷ ൌ 0. each case is different and has been described at each section.same used in the four types of optimization. ଷ . we have 63 . ଶ . ݀ସ ൌ 0. ସ ሻ where ݉ 8 (the excited harmonic number always appears as m is larger than or equal to eight due to the limited camshaft speed). ܦସ ൌ 0. or it will look more compressed on one side than the other.

079511. ଷ ൌ 0. Fig.2187.4).6.1 ଷ 0. The original parameter is used as the initial trial point for the convenience of comparison.20). After the optimization computation. upper and lower bounds of ଵ . (2. limited to minimize the power spectral amplitude on the excited harmonic number. ଶ ൌ 0.2 shows the optimal variable pitch angle plot. It is obvious to find. ସ ൌ െ0.2 ଶ 0.145742.1 ଵ ൌ െ0. In comparison with the original value 6924. Fujimoto.5.63(N). ଷ ൌ 0.defined in Eq.5. The optimal results show that maximum spring force is 159.3 െ0. the optimal parameter is obtained as and Fig. According to some technical papers [Wahl.5.1 ସ 0. ଶ . ଶ ൌ 0. and their polynomial expression is െ0. ଵ ൌ െ0. it decreased. this optimal parameter was also verified in SECTION.00009. this optimum result can also be explained by the element natural frequency distribution. In comparison of the original value 182.3) and (5. ସ ൌ െ0. Besides constraint (5.1.3 shows the power density comparison plot between the optimal design and the original one. ܽ݊݀ ସ are constrained as follows.5. this original design was pretty good.0585.1716.5.1 െ0. Hence the improvement in this type is very small.11 (N). In addition. The optimal power density on the 11th harmonic is 6911. 64 . 1963.000198.4 ଵ 0 െ0. It is presented in SECTION. ଷ . the power spectral amplitude on the excited harmonic number can be reduced by variable pitch springs and the denser end coil. In the ABAQUS-MATLAB program.268699 . 2007].

5.2 The optimal variable pitch angle in optimization case one Fig.5.Fig.3 The power spectrum comparison of the original and the optimal designs (case one) 65 .

They offer the advantage of a reduced solid height compared to straight compression springs. ଶ ൌ 0.004 0. ݀ଶ ൌ 0.004 ܦଶ 0. variable pitch angle and the diameter of the wire must be kept as constants. we have Besides constraints (5. ܦଶ . (2. ܦସ ൌ 0. there are some features in conical springs. ଷ ൌ 0.5.000198.029 ܦଵ ൌ 0. ܦଷ . ݀ଷ ൌ 0. ସ ൌ െ0. ݂ ሺܦଵ . ܦସ ሻ. 2002].0585. ݀ଵ ൌ 0. ܦଶ ൌ 0. ܦଷ ൌ െ0. Each spring features a variable pitch to achieve the constant spring rate and coils which nest during deflection to provide a solid height approximately equal to two wire diameters.00009. After the computation.004731. The mean coil diameter along the spatial length is defined in Eq. According to Handbook of Spring Design [SMI.4.2 Variable Coil Diameter (Conical Spring) Conical compression springs are often specified where the large end is meant to work in a bore and the small end is meant to work over a rod.005 െ0.006 ܦଷ 0.22).3) and (5.02345. and we have the objective function as where ݉ 8 For the initial trial point in SQP.0003 ܦଵ 0. the optimal parameters are obtained as: ܦଵ ൌ െ0. 66 .0002 െ0. ܦଷ ൌ 0.2187. lower and upper bounds are listed as follows: െ0. Small solid height Variable spring rate Good stability Resonance and vibration is reduced Not easy installed in the valve train due to variable diameter In this case.004397.1716. especially when capable of telescoping. Conical springs are cone-shaped compression springs designed to provide a near constant spring rate and a solid height lower than a normal spring.022 ܦସ 0. so we have ଵ ൌ െ0. and ݀ସ ൌ 0.0255ሺ݉ሻ. and ܦସ ൌ 0.004ሺ݉ሻ.4). ܦଶ ൌ 0.

4 The optimal variable coil diameter in optimization case two 67 . The improvement is up to 51. Fig.22 percent. The power spectral amplitude on the 13th harmonic number is 3378.5 shows the power density comparison figure. it offered a lighter spring due to the smaller mean coil diameter. Fig. Maximum spring force 195.4 shows the variable coil diameter plot.5.Fig. For weight.5. The original value in the 11th harmonic number is 6924.52(N) in comparison of the original one is very close.5.

5.00009. ସ ൌ െ0. and we .3 Variable Wire Diameter Basically. They offer the advantage of remaining the same height compared to conical springs which have space limitations. and ܦସ ൌ 0. so we have ଵ ൌ െ0.Fig. especially when capable of flexible design. variable pitch angle and the mean coil diameter must be kept as constants as the original one. ܦଵ ൌ 0. The same spring height Flexible rate Good stability Resonance and vibration is reduced Shorter fatigue life due to larger stresses and smaller wire diameter Difficultly manufacturing technique In this case.0585. The features are listed as follows. ଷ ൌ 0. 68 The diameter of the wire along the spatial length is defined in Eq.5 The power spectrum comparison of the original and the optimal designs (case two) 5.0255ሺ݉ሻ. ଶ ൌ 0.25). this type of spring has the similar advantage as conical springs. ܦଷ ൌ 0.1716.2187. ܦଶ ൌ 0.4. (2.

we have Besides constraints (5. The improved percentage is up to 60. ݀ଵ ൌ 0.004 ݀ଶ 0.0035 ݀ସ 0.have the objective function as where ݉ 8 ݂ ሺ݀ଵ . െ0.001993 . lower and upper bounds are constrained as follows. ݀ଶ ൌ 0. It provided good performance in comparison with the original value of 6924. and ݀ସ ൌ 0. ݀ଶ . From the results. the optimal parameter is obtained as Fig. ݀ସ ൌ 0. ݀ସ ሻ For the initial trial point in SQP.005998.004 m.008 െ0.004 ݀ଷ 0.000111. ݀ଶ ൌ 0.5.25 percent. The optimal power density on the 13th harmonic number is 2752. ݀ଷ .002 െ0. 69 . ݀ଷ ൌ െ0. ݀ଷ ൌ 0.0048 ݀ଵ ൌ 0.4).0046.7 shows the power density comparison plot.3) and (5. Fig.0001 ݀ଵ 0.004ሺ݉ሻ After the computation.6 shows the variable wire diameter curve. the maximum spring force 166.0003 െ0. The original design value is a constant 0. The weight is slightly increased due to the larger wire diameter.11.5.78(N) in comparison with the original one is relatively close.

Fig.6 The optimal variable wire diameter in optimization case three Fig.7 The power spectrum comparison of the original and the optimal designs (case three) 70 .5.5.

1 ସ 0. and we have ଵ ൌ െ0.0002 െ0. the mean coil diameter.3 െ0. By a discussion of three previous optimal cases. ܦସ ሻ.1 Besides constraints (5.3) and (5.1716. Coil Diameter Theoretically.2187. ݀ଵ ൌ 0. all bounds are listed as follows: െ0. െ0. and ܦସ ൌ 0. ସ ൌ െ0. these values are chosen as the same as the original one. ܦଷ .006 ܦଷ 0.4).005 െ0. The predicted performance should not be below the optimal case three – variable wire diameter. Wire Diameter. Small solid height Variable spring rate Good stability Resonance and vibration is reduced Difficult to manufacture the variable pitch angle. and the diameter of the wire along the spatial length are defined in Eqs. 2. ଶ ൌ 0. ݀ଷ ൌ 0. ସ . Its features should be similar as variable wire diameter springs. ଷ ൌ 0.4 Variable Pitch Angle. 2. variable pitch angle.0035 ݀ସ 0.0255ሺ݉ሻ. ݀ଷ .0003 െ0.25.1 െ0. variable wire diameter and variable coil diameter – a total of twelve design variables to be optimized.004 ܦଶ 0. ݀ଶ ൌ 0. ଷ .0048 െ0.22 and 2. ܦଵ . ܦଶ ൌ 0. ݀ସ ൌ 0.00009. ଶ . ݀ଵ . the more design variables the better the performance.0001 ݀ଵ 0.008 െ0.2 ଶ 0.0003 ܦଵ 0.5. This case included all design variables. ܦଶ .1 ଷ 0. ܦଷ ൌ 0.004ሺ݉ሻ.20. and we have the objective function as whereas for ݉ 8 ݂ ሺଵ .002 . ݀ଶ . the mean coil diameter and the diameter of the wire more easily obtained better performance.004 71 െ0.4. ܦଵ ൌ 0. For the initial trial point in SQP.4 ଵ 0 െ0.0585.004 ݀ଷ 0. ݀ସ .004 ݀ଶ 0.

0.0014.8 percent. According to the design specification of valve springs [Wahl. 0.26. In general. -0.002.10. 0.2518.0002. and ܦସ ൌ 0. 0. Shigley. but this case shows that the top end is heavier. The improved efficiency is up to 68.0024. Fig.10 shows the variable wire diameter plot. decreased significantly. 0.0004. It is seen that this type of spring presents an extreme nonlinear spring force along the spatial length according to Fig.11 shows the power density comparison of the case and the original one. ଶ . ݀ଷ . ଶ .5.0317] The power density of the excited harmonic is 3129. ସ ൌ 0. ܦସ ሿ = [-0. ସ .0003.2. The spring in this optimization case four is very powerful with the lowest power density. ܦଷ ൌ 0.003628 .5.5. In reality. ݀ସ . All amplitudes on the harmonics are lower than those in the original design. and can be up to 10 percent.000363 . ܦଵ . 0. ܦଵ . Let the parameter [ଵ .9 on the 13th harmonic number. ݀ଶ . 0.149103 . ܦଵ .000231 .1640. ܦଶ ൌ 0. the performance springs usually have the heavier bottom end. 0.028783. ܦଶ .0003. 0. ܦସ ሿ = [-0. ܦଶ . That is a little bit unreasonable and contrary to those in optimization cases two and three. ଷ .000072 . 0.0017. ݀ଶ ൌ െ0. ݀ସ ൌ 0.5. ݀ଷ .279762. ݀ଷ . and only has a peak 2159. ܦଷ .9 and 5. -0. Fig. ସ .5. ݀ଵ .5. as its 72 .0042.8 shows the variable pitch angle plot and has the similar result as Fig. ݀ସ . -0.001972 .0033.0001. 2004]].4. ݀ଶ .After the computation. The maximum spring force of 129. Fig.0018 0. ݀ଶ .0052. ݀ଷ ൌ 0. However. 0.022 ܦସ 0. ݀ଵ .002625 .004683 ܦଵ ൌ െ0.0001. 0. 0.0259] The power density of the excited harmonic is 4944.0029. the result is shown as follows: 0. ܦସ ሿ േ 10%. the spring with the larger coil diameter and the heavier weight on the top end has the stability problem [Wahl. 0.1342.3077. In addition. ܦଷ .091418 .0823. ଶ . ଶ ൌ 0. 1963. ଷ . and the optimal parameters with the upper deviation are as follows: [ଵ . ଷ ൌ 0.029 Fig. ଷ .001535 ݀ଵ ൌ െ0. the optimal parameters with the lower deviation are as follows: [ଵ .5. the deviation is often happened in the manufacturing quality. ݀ଵ . ଵ ൌ െ0.9 shows the variable coil diameter plot which has a significantly different curve from Fig.16(N) in comparison of the original value. 0. ܦଶ .0040. ܦଷ . ݀ସ .1006.61. ସ . -0. 0. 1963]. 0.

9 The optimal variable coil diameter in optimization case four 73 . As a result. Fig. the power density on the excited harmonic increases over 30 percent. the manufacturing deviation would affect spring performance so much.5.8 The optimal variable pitch angle in optimization case four Fig.manufacturing deviation is around 10 percent.5.

Fig.5.10 The optimal variable wire diameter in optimization four Fig.5.11 The power spectrum comparison of the original and the optimal designs (case four) 74 .

then do the natural frequency computation in ABAQUS. Spring Seat Force (N) 182.0415 0.4.11 195. For the excited amplitude.5. Dia Vari.056 0.5. but also enlarge the fundamental natural frequency. into a finite element model.verification of optimal results using finite element method. it is seen that these optimal designs improved the spring performance. Wire. The verification flow chart for the obtained optimal parameters in different programs is shown in Fig.4.64 6911. The optimal parameters in SECTION.9 The excited order(th) 11 11 13 13 13 Original Vari. The fundamental natural frequency of the optimal case two. Coil. 75 . but its limitations is obvious.52 166. The role of MATLAB in this program is only to generate the geometric model by the obtained optimal parameters in SECTION.4 were used as an input in the analysis and implementation in ABAQUS-MATLAB program. three. and reduced the excited amplitude. Wire Dia Va. The optimal case two.5.1 shows the power density comparison of all optimal cases.12. Pitch Vari. The spring model was developed systematically from a solid model.0738 Max. Pitch.11 2159.1 Introduction and Program Structure This section involved the modeling and verification of the optimal valve spring with the focus on the non-linear behavior of high speed valve springs . Coil Dia 5.From these optimal results. three and four don’t only reduce the excited amplitude.78 129.68 3378 2752.0415 0.1 The results from different optimal designs term Weight (kg) 0. 60 ሺminሻ Table 5.16 The scale of the excited harmonics 6924.63 134.4. there is a large room for improvement for the optimal cases two. and four. The optimal case one can improve that as well. Table 5.1 to 5.5.0415 0. three and four is 2372 ሺrpmሻ sec ൈ 13 ሺthሻ ൌ 514 ሺHertzሻ.5 Verification in ABAQUS-MATLAB Program 5.

The verified result is shown in Fig. the valve springs are designed to have a high natural frequency of vibration.4).5.1 is 433 (Hertz) which is also verified in ABAQUS-MATLAB program.5. To lessen any tendency towards surging within motion speed range of the engine.13 which is has a fundamental natural frequency around 436 Hertz. It is seen that the result is the same as the result of optimization program and shows a limited improvement in a variable pitch angle type. From the results.5. and progressive rate springs (referring to SECTION. All development steps were continually checked with experiments and simulations in previous chapters.5.2. The primary concept used in this section is to check the obtained fundamental natural frequency in SECTION.and finally into a dynamic model.12 Flow chart for comparing results from different methods 5. mechanical spring dampers. Fig.5.4.2 Variable Pitch Angle This fundamental natural frequency of the optimal parameter in SECTION.4. A possible reason is that the original design greatly improved. the dynamic model was verified and good correlations were found. 76 .4. Various additional measures may be taken to minimize surge such as the use of double springs. especially at high engine speeds where valve train dynamics play an important role.

the two results show the same phenomenon that the top end needs the smaller coil diameter.5.5. Here we could calculate easily the engine speed limit by this result.14 which has the similar natural frequency around 510 Hertz as that in Fig.5.3 Variable Coil Diameter The optimal parameter in SECTION. so the calculated engine speed red line is round 4006 (rpm).5. The calculation is expressed as engine speed red line 1 min ሺrpmሻ ൈ ൬ ൰ ൈ 13ሺthሻ 2 60 sec The original fundamental natural frequency is around 433 (Hertz). The optimal parameter in 77 ൌ fundamental natural frequency.5.Fig. In fact.2 is taken into MATLAB-ABAQUS program to verify the result. It’s clear to show the dynamic equation derived in this thesis is very accurate. We select the maximum harmonic order that could affect the spring resonance is 13th order [SMI. 2002]. and can be used to improve spring performance. but the bottom end needs the larger coil diameter to form a stable and lighter design. The verified result is shown in Fig. This condition coincides to conical springs sold in the current spring market.5.13 The generated fundamental natural frequency by ABAQUS-MATLAB program (Optimal parameters of optimization case one) 5.4. .

The ABAQUS-MATLAB program at least can be used to make sure the optimal results – case one and two. As in the previous discussion in optimization program. it cannot handle springs involving variable wire diameter due to the involving generating solid model for variable cross-section along helix.14 The generated fundamental natural frequency by ABAQUS-MATLAB program (Optimal parameters of optimization case two) 5. The comparison of results obtained from optimal parameters is shown in Table 5.5.5.4 Limitation Although the ABAQUS-MATLAB program in this section provided good verification with optimal results in SECTION.5.conical springs design has a value around 514(Hertz). Comparatively speaking. hence the engine speed red line is elevated up to 4726 (rpm). and variable wire diameter shows the most powerful spring. generated by optimization program are correct and accurate. variable coil diameter. unfortunately. the combination of variable pitch angle. Fig.2.4. 78 . the simulation by the finite difference method and optimization by FFT provided faster and relative accurate result. It is improved around 700 (rpm) by this optimization design. Another problem in this program is that it is too slow due to the calculation of FEA mesh.

004731 ܦଷ ൌ െ0.145742 ଷ ൌ 0.079511 ସ ൌ െ0.5.4 514 X Optimization Program ABAQUS-MATLAB PROGRAM Optimization Program ABAQUS-MATLAB PROGRAM Variable Coil Diameter ܦଵ ൌ െ0.4.Table 5.3 79 .000198 Natural frequency(Hertz) 433 436 Program Optimization Program ABAQUS-MATLAB Program Optimization Program ABAQUS-MATLAB Program 514 X Combination Referring to SEC.004397 ܦସ ൌ 0.2 Comparison of fundamental natural frequencies in different optimization cases Optimization Design Variable Pitch Angle Optimal Parameter ଵ ൌ െ0.02345 514 510 Vari.000198 ܦଶ ൌ 0. Wire Dia Referring to SEC.4.5.268699 ଶ ൌ 0.

For improving the speed of numeric solution using MATLAB. variable wire diameter. which is a GM-ISUZU 1987 engine. the improvement is only 52 percent. Some other scenarios.1 Summary 1. variable wire diameter d(s). and had approximate solutions. Finite Difference method was used to solve the modified wave equation.8 percent. and variable coil diameter D(s). a faster computational algorithm in Crank-Nicolson method was applied. For evaluating spring performance. but also its natural frequency is up to 512 Hertz. In Optimization chapter. This thesis offered a modified wave equation by inserting an internal virtual force term into the original wave equation. This modified wave equation is applicable to helical springs with variable pitch angle p(s). While solving the dynamic equation. 4. these simulated results were used in optimization. With these special techniques. a moving boundary solution was programmed to take into account the coil closing at the top and bottom boundaries. It shortened the computational time. The lowest natural frequency in this tested engine. In this situation not only the maximal power density in 13th harmonic (which was exciting the lowest natural frequency) is reduced to around 2200. Because Crank-Nicolson’s method is unconditionally stable for solving partial differential equations. 2. The most thorough approach to optimize a helical spring is to assume the spring under consideration for optimization has variable pitch angle. The power spectrum density in the best situation was improved for as much as 68. and variable coil diameter. Because the numeric solution had been verified by physical experiments to be accurate enough. such as in variable coil diameter optimization.Chapter 6 . to take the coil closing or clashing into consideration. a Fast Fourier Transform (FFT) was used to evaluate the power spectral density. the result represented an approximate but sufficiently accurate simulation data to be compared with the experimental data. 3. 80 .Conclusion 6. is close to 433 Hertz. the maximum power density in higher order harmonics was defined as the objective function. it is used to solve this dynamic equation.

ABAQUS-MATLAB program was utilized to verify the obtained optimized design – optimization case one and two. 2009]. 6. For accuracy. variable wire diameter and variable coil diameter spring. In the future. The obtained fundamental natural frequency in this thesis is based on processing dynamic solution data using Fast Fourier Transform.2 Suggested Future Works 1. the fundamental natural frequency of helical springs was obtained by a different method analyzing the amount of coil closing. variable wire diameter. 3D solid model should be possibly based on ACIS description as input file for ABAQUS. each element between nodes should be sufficiently small. it is also very difficult to manufacture. It is desirable to use special manufacturing technique to make helical springs with variable wire diameter. For the latest technical papers [Liu. ABAQUS-MATLAB program was not able to do variable wire diameter in Optimization chapter. and compared also with actual experimental data. It is possible to modify that by using the ACIS 3D Model to make the 3D model with different wire diameter along the helix. and variable coil diameter performs very well dynamically.5. and its results agreed well with Finite Difference results. 3. Then the optimal solution can be compared with the finite difference solution. Although we know the type of spring with variable pitch angle. 81 . 2. as an alternative and independent solution. and compare numerical solution with experimental data based on real variable pitch angle. Due to the difficulty in using programs to generate variable diameter wire in a solid model. It should be interesting to compare the result for the same spring using these two different methods.

E. Camb. “Pitch and Curvature Corrections for Helical Springs. 1958 Arora. 1961 Crank. Yuyi and Johnston. Yuyi. Structural Dynamics and Materials Conference. Robert H. International Journal of Automotive Technology. Phil. Yuyi and Pisano. McGraw-Hill. 1987 Lin. D. ASME. 2009 Love. 1989 Kreyszig. 2007 Greenwood. 1944 Main. CCAMMS. et al. 1994 Lin. Li and Lin. SAE. and Goodier... Erwin. Static and dynamic analyses of helical spring”. Mitsubishi Motors Technical Papers. John and Phyllis Nicolson.. Spring Industry Technical Symposium. “Valve Jump Prediction Using Dynamic Simulation on Direct Acting Valve Train”. “Optimal Design of Resonance Suppression Helical Springs”. 1989 KISTLER. N. and Kim. “Stress Analysis and Optimal Cross-Section Design of Noncircular Spring Wire”. Proceedings of the 30th AIAA/ASME/ASCE/AHS/ASC Structures. Nicholas P.. 1988 82 . ASME. Yuyi. Acturators. ” A practical method for numerical evaluation of solutions of partial differential equations of the heat conduction type”. 1988 Jiang. Principles of Dynamics. “Non-linear and Linear. Akihiro. Jr. Yuyi and Pisano. Iain G. Albert P. 1991 Le. 1947 Fujimoto. 2001 Bishop.. Donald T. 1989 Liu. Spring Design and Application. “Dynamic Modeling and Simulation of Variable Pitch”. ASME. 2006 Kurisu.. Yuyi and Harby. A treatise on the mathematical theory of elasticity. Advanced Engineering Mathematics. 2006 Lin. W. “Dynamic Modeling and Experimental Verification of Helical Spring with Variable Pitch”. J. Proc. Jasbir S. Dover Publications.. Mechatronic Systems. et al. H. “General Dynamic Equations of Helical Sprsings With Static Solution and Experimental Verification”. Automotive Handbook. “A Study of Jump and Bounce in a Valve Train”. 1999 Lin. C. KISTLER Pizeo-Instrumentation. Introduction to Optimum Design. CRC Press. Soc. et al. ASME.Reference Anker. Scott. Cambridge University Press. “Three Dimensional Dynamics of Helical Spring for the automotive valve trains”. ”EFFECTS OF END COILS ON THE NATURAL FREQUENCY OF AUTOMOTIVE ENGINE VALVE SPRINGS”. A. McGraw-Hill. H.. Donald. Vibrations and Waves in Physics. 2007 Chironis . Toru.. 2008 Bosch.” ASME. et al. Sensors. J. Prentice Hall.. 1993 Lin. Wiley.

and Freudenstein. William T. Stephen. Cam Design Handbook. Dover Edition. McGraw-Hill. “Effects of Pitch Angle and Damping on Stresses in Helical Spring”. Theory of Elasticity.. P. 1951 Timoshenko. Highly Flexible Structures: Modeling.. 1963 Vanderplaats. 1983 Reddy. Appl. Calculus of Variations. P. 1974 83 . McGraw-Hill. P. Dover. Numerical Optimization Techniques For Engineering Design with Applications. An Introducation To The Finite Element Method... ASME. J.. Prentice Hall... 1890 Nunney.H.. G. 1993 Timoshenko.Mitchell. P.. 1984 Rothbart. Mechanical Springs. A. Cleveland Ohio. FEM for Springs.. F. W. Light and heavy vehicle technology. ASME. Stephen. Strength of Materials. Inc. M. 1990 Pisano. Harold A. 1995 Weinstock. McGraw-Hill. Technology & Engineering. and Tait. 2007 Paranjpe. Math. 1956 Timoshenko. Applied Optimization with MATLAB Programming. J. Mess. D. 1995 Thomson.. Inc. McGraw-Hill. Garret N. M. van Nostrand Company... Robert. McGraw Hill.. “An Experimental and Analytical Investigation of the Dynamic Response of a High-Speed Cam-Follower System Part 1: Experimental Investigation”. 1984 Venkataraman. 1883 Thomson. Oxford. JSSR. M. Theory of Vibration with Application.. Mech. 1935 Wahl. “The small deformation of curves and surfaces with the application with to the vibrations of a helix and a circular ring”.. 2002 Thomson. Lumped/Distributed parameter Dynamic Model”. F. J. W. Frank. A. et al. “Dynamic Analysis of a Valve Spring With a Coulomb-Friction Damper”. Wiley-Interscience.S.. 2003 SMI. 2007 Pai. Stephen. Malcolm James... N. R. Penton Publishing CO. et al. A. Computation. J. ASME. Applied Mathematics. 1983 Pisano. and Experimentation. Takeshi. and Freudenstein. Handbook of Spring Design. 1963 Weinberger. “An Experimental and Analytical Investigation of the Dynamic Response of a High-Speed Cam-Follower System Part 2: A Combined. “Numerical Partial Differential Equations: Finite Difference Methods”. McGraw-Hill Book Company. Theory of Elastic Stability. 2002 Wahl. American Institute of Aeronautics & Astronautics. 2004 Sawanobori. JSME. “Helical Compression and Tension springs”. Treatise on Natural Philosophy. A. 1986 Shimoseki. A first course in partial difference equations with complex variables and transform methods. Hans F.

%the shear modulus mu = 7. % compression ratio L = 0. % the diameter of every coil J = 2.(A) Main Program %the spring simulation program %the program could be divided into many parts %1: setup the basic parameters %2: setup the preload for 15% height %3:calculate the deflection %4:play the spring movie %5:calculate the spring seat force %6:calculate the power spectrum %7:output all data clear. %the wave speed F = 10. % G times J changes to E times I mg = mu*(pi*d. clc.^2+d. %frequency of surge waves a = sqrt((G*d.004.62e-11. %the radius of the coil D = 2*r. % the radius of the cross-section of wire r = 0. %the weight of the spring cc = 8. %poissoin ratio c = D/d. % the free height ys = 0.0386.^2)/(mu*(8*r.^2/4*L). % the solid height cr = 1 . %the damping coefficient fre_surge = d/(2*D*L)*sqrt(G/2/mu).%the material weight density d = 0.86e3.^2))). %the damping force per unit length of the wire per unit of velocity b = cc*L/mg.024. %the internal force in the spring causes the nonlinear deformation 84 . Program Code (Optimization program) 1. % the spring index yf = 0. fprintf('The Central Finite Difference Method\n').01275.ys/yf. %total length of the spring ws = 2400. %the rotational speed ei = (1+po)*G*J. G = 77e9.3. % polar moment of inertia po = 0.Appendix 1.42.

% the size of dt lamda = a*dt/dx. fprintf('The diameter of the cross wire: %9.000027.615/c.4f(m/s)\n'.4f(kg)\n'.4f \n'. %calculate the stability of the finite difference method fprintf('The amplification factor for the stability of the finite difference method: %6.^4*G/(8*D.4f(N/m)\n'.2f(m/s)\n'. end. fprintf('The damping coefficient: %6. fprintf('The spring density: %6. fprintf('The spring rate: %9. % the size of dx dt = 0.^2). 85 .01. if (lamda <= 1) fprintf('The simulation of this model is stable. fprintf('The weight: %9. dx = 0.ff = ei*(cos(5*pi/180)/r-cos(6*pi/180)/r).mg).lamda).\n'). %the length interval between every two points x = 0:dx:L.yf).4f(m)\n'. d).b).2f(rpm)\n'.ys).4f(m)\n'.^2. % k1 x L is the spring constant km = (4*c . fre_surge).ws). fprintf('The free height: %9.4f(m)\n'.mu). %the spring constant %output the data fprintf('The diameter of this coil spring: %9.\n'). %divide the length nx = length(x). fprintf('The frequency of surege waves: %7.%add on an force term to let the wave equation as nonlinear kc = pi*d.2f(hertz)\n'. fprintf('The wave speed: %9.4) + 0.a).k1).2f(kg/m^3)\n'.4f(m)\n'. else fprintf('The simulation of this model is not stable.1)/(4*c . %the time t1 = 0:dt:(1/w).D). % the correction factor k1 = kc/(pi*D*5). fprintf('The rotaional speed: %7. % it's used to calculate the cam profile for one cycle nt = length(t). %rev/min / 60 = rev/sec t = 0:dt:(1/w*4). fprintf('The solid height: %9. %the time interval between every two points w = ws/60.

ytempa = polyval(pp.3). n = 1.:) = ppo(1).1716 0.yd.%calculate the spatial curve of the spring when it is installed in the %opposed direction ppo = [-0.j+2)-2*yt(n+1.*x +ppo(4).15. pp = polyfit(x. %the curve of the spring installed on the normal direction ppn = c_pp_cal(L. %store the initial curve of the spring yt(:.00001:yf*preload_ratio].j+1)+ yt(n+1.:) = yt(1.^2. j = 1. end.00009].j+1) .nx).*x.yt(length(yd).*x. while(n <= length(yd)-2) while(j <= (nx-2)) temp = ((dt/dx*a).0) = f(x) yo = yt(1. yt(1.:).^2-4*(1+b*dt/2)*(dt.^2*ff+temp*yt(n.j)) + 2*yt(n+1. %boundary condition at x = 0 yt(:.j+1)).:).j+1) + b*dt/2*yt(n.*(yt(n+1.x).%preload is 10% yt = zeros(length(yd).ppo).0) = g(x) %calcaulate the pitch angle ypitch = 3*ppo(1).nx) ..1) = 0. %calculate the curve of the spring installed on the opposed direction %preload preload_ratio = 0. j = j +1. ta = (1 + b*dt/2)*yt(n. yt(n+2. yd = [0 0 0:0.j+1))).j+1) = real((ta-tb)/(2*(1+b*dt/2))). %initial condition for dy(x.*x+ppo(3).^2+2*ppo(2).^3+ppo(2).2187 0. tb = sqrt((-ta). yt(2.nx) = yt(1.. j = 1. %initial condition for y(x.:). end.yt(n.0585 -0..*x. %calculate the deflection fprintf('computating the spring deflection. 86 . n = n + 1.^2+ppo(3).j+1)+temp.\n').

nx).3)) -(a04 -2*a03 + a02) + b/dt*(dx/a).yr(:).2) + yt(n.*2.*2.dt.^2.7)) + 2*(dx/a).yt(n.2)) -(a03 -2*a02 + a01) + b/dt*(dx/a).^2*(a04 .2*yt(n+1.^2.yt(n.*(a02 .3)) + 2*(dx/a).5) + yt(n.yt(n. f05 = (dx/dt/a).:) = ynratio*x. clear t1.4)) + 2*(dx/a). by the finite difference method syms a01 a02 a03 a04 a05 a06 a07 a08 a09 a10 a11 a12 a13 a14 a15 a16 a17 a18 a19 a20 a21 a22 a23 a24 a25 a26 a27 a28 a29 a30 a31 a32 a33 a34 a35 a36 a37 a38 a39 a40 a41 while(n <= nt-2) f01 = (dx/dt/a). yt(1.^2*(a05 .6) + yt(n.^2*F*a04/L. yt = zeros(nt.w). yt(2. %initial condition for dy(x. ynratio = ytempa(nx)/L.2)) -(yt(n+1.2*yt(n+1.2*yt(n+1.7)) -(yt(n+1.3) + yt(n.8) .yt(n.*2.2*yt(n+1.0) = g(x) yt1 = yt.:) = yt(1.7) .6)) -(a07 -2*a06 + a05) + b/dt*(dx/a).*2. n = 1.^2*F*a05/L.*(a03 .5) + yt(n+1.2*yt(n+1. yr = cam_rise(t1. f02 = (dx/dt/a).6)) -(yt(n+1.3) .3)) -(yt(n+1.^2.2*yt(n+1.^2*F*a06/L. f04 = (dx/dt/a). clear ytempa.^2*F*a03/L.4) .*(a05 .6) .*(a04 .*(a01 .5) .^2*(a03 . yr = [yr 0 yr 0 yr 0 yr].2*yt(n+1.4)) -(yt(n+1.*2.7) + yt(n+1.6)) + 2*(dx/a).:).5)) + 2*(dx/a). yt(:.^2.4) + yt(n+1.4) + yt(n.ytempaL = ytempa(nx).2*yt(n+1.2*yt(n+1.2)) + 2*(dx/a).yt(n.5)) -(yt(n+1. %calculate every value on the grid.^2*F*a02/L.^2*F*a07/L.2*yt(n+1.^2.^2*(a06 . %length and time.2*yt(n+1.2) + yt(n+1.7) + yt(n. f06 = (dx/dt/a).3) + yt(n+1.1)) + b/dt*(dx/a).^2.2*yt(n+1.*(a06 . yr = yr(1:length(yr)-3). 87 .^2*(a01 .yt(n.6) + yt(n+1.1)) -(a02 -2*a01 + yt(3.4)) -(a05 -2*a04 + a03) + b/dt*(dx/a).^2*(a02 .5)) -(a06 -2*a05 + a04) + b/dt*(dx/a).nx) = yt(1. f03 = (dx/dt/a).nx) .*2.

11) .9)) + 2*(dx/a).9)) -(a10 -2*a09 + a08) + b/dt*(dx/a).^2*(a18 .*2.yt(n.14) + yt(n+1.^2*F*a19/L.^2*(a07 .*(a14 .yt(n.19))-(a20 -2*a19 + a18) + b/dt*(dx/a).12) + yt(n.14))-(a15 -2*a14 + a13) + b/dt*(dx/a). f12 = (dx/dt/a).7)) -(a08 -2*a07 + a06) + b/dt*(dx/a).13)) -(yt(n+1.^2.*(a15 .20)) + 88 .*(a12 . f15 = (dx/dt/a).2*yt(n+1. f08 = (dx/dt/a). f14 = (dx/dt/a).^2*F*a13/L.2*yt(n+1.^2.*2.^2*F*a14/L.^2*F*a09/L.^2.10) + yt(n+1. f17 = (dx/dt/a).*(a08 .^2.yt(n.12)) -(a13 -2*a12 + a11) + b/dt*(dx/a).*2.2*yt(n+1.2*yt(n+1.yt(n.11)) -(yt(n+1.2*yt(n+1. f13 = (dx/dt/a).^2*(a12 .11) + yt(n+1.16) .15) + yt(n+1.9) + yt(n+1.8)) -(a09 -2*a08 + a07) + b/dt*(dx/a).yt(n.^2*(a19 .21) .14)) + 2*(dx/a).14) .14)) -(yt(n+1.2*yt(n+1.*2.16)) + 2*(dx/a).10)) + 2*(dx/a).19)) -(yt(n+1.8)) -(yt(n+1.15)) -(yt(n+1.yt(n.12) .f07 = (dx/dt/a).^2*(a13 .*2.8)) + 2*(dx/a).*(a17 .14) + yt(n.2*yt(n+1. f10 = (dx/dt/a).^2*F*a18/L.18))-(a19 -2*a18 + a17) + b/dt*(dx/a).yt(n.2*yt(n+1.17)) + 2*(dx/a).8) + yt(n.^2*(a11 .9) .9)) -(yt(n+1.17) + yt(n.^2.*(a19 .*2.^2.2*yt(n+1.18) .12)) -(yt(n+1.^2*(a09 .2*yt(n+1.18) + yt(n.^2.*(a16 .8) + yt(n+1.15) + yt(n.^2*F*a16/L.2*yt(n+1.13)) + 2*(dx/a).yt(n.20) + yt(n.^2.19) + yt(n.2*yt(n+1.yt(n.10)) -(yt(n+1.^2*F*a15/L.^2.^2*(a15 .16)) -(yt(n+1.17)) -(a18 -2*a17 + a16) + b/dt*(dx/a).20)) -(yt(n+1.^2*(a16 .*(a07 .2*yt(n+1.18)) + 2*(dx/a).*2.^2.2*yt(n+1. f18 = (dx/dt/a).10)) -(a11 -2*a10 + a09) + b/dt*(dx/a). f09 = (dx/dt/a).13) .13) + yt(n+1.20) .17) .2*yt(n+1.19) .18) + yt(n+1.*(a18 .^2.16) + yt(n+1.11)) + 2*(dx/a).18)) -(yt(n+1.12) + yt(n+1.15) .yt(n. f19 = (dx/dt/a).2*yt(n+1.2*yt(n+1.*2.*(a10 .15)) -(a16 -2*a15 + a14) + b/dt*(dx/a).12)) + 2*(dx/a).^2*F*a12/L.*2.^2*(a17 .2*yt(n+1.2*yt(n+1.19)) + 2*(dx/a).10) + yt(n.^2.^2*F*a17/L.2*yt(n+1.*(a09 .15)) + 2*(dx/a).yt(n.16)) -(a17 -2*a16 + a15) + b/dt*(dx/a).19) + yt(n+1.*2.2*yt(n+1.17)) -(yt(n+1.2*yt(n+1.16) + yt(n.13) + yt(n.^2*F*a10/L.11))-(a12 -2*a11 + a10) + b/dt*(dx/a).*2.^2*(a14 .*2.9) + yt(n.^2.^2*F*a11/L.^2*(a08 . f11 = (dx/dt/a).2*yt(n+1.yt(n.^2*F*a08/L.*(a13 .yt(n.2*yt(n+1.*2.*(a11 .11) + yt(n.17) + yt(n+1. f16 = (dx/dt/a).2*yt(n+1.10) .13)) -(a14 -2*a13 + a12) + b/dt*(dx/a).2*yt(n+1.^2*(a10 .20) + yt(n+1.

2*yt(n+1.29)) + 2*(dx/a).29)) -(a30 -2*a29 + a28) + b/dt*(dx/a).*2.23) + yt(n.^2.^2*(a26 .*2.yt(n.29) + yt(n+1.26) + yt(n+1. f24 = (dx/dt/a).25) + yt(n+1.30)) -(a31 -2*a30 + a29) + b/dt*(dx/a).^2*(a30 .yt(n.^2*F*a25/L.2*yt(n+1.25))-(a26 -2*a25 + a24) + b/dt*(dx/a).yt(n.^2*F*a31/L.26)) -(yt(n+1.2*yt(n+1.23))-(a24 -2*a23 + a22) + b/dt*(dx/a).24) .32) + yt(n.22) .*(a25 .^2.2*(dx/a).2*yt(n+1.21)) + 2*(dx/a).yt(n.33) .^2.27)) -(yt(n+1.22) + yt(n.31) + yt(n+1. f25 = (dx/dt/a).^2*(a23 .*(a26 .*2.2*yt(n+1.2*yt(n+1.*2.^2.32)) + 2*(dx/a).27)) + 2*(dx/a).2*yt(n+1.yt(n. f28 = (dx/dt/a).2*yt(n+1.2*yt(n+1.^2.^2.23) .2*yt(n+1.24) + yt(n.2*yt(n+1.*(a29 .33) 89 .^2*F*a32/L. f30 = (dx/dt/a).24) + yt(n+1.2*yt(n+1.yt(n.28) + yt(n+1. f26 = (dx/dt/a).2*yt(n+1.29) .25) + yt(n.2*yt(n+1.22) + yt(n+1.29)) -(yt(n+1.27)). f27 = (dx/dt/a).yt(n.^2.*2.2*yt(n+1.*2. f32 = (dx/dt/a).^2.28)) -(yt(n+1. f21 = (dx/dt/a).2*yt(n+1.23)) -(yt(n+1.^2.28)) + 2*(dx/a).*(a27 .2*yt(n+1.^2*F*a21/L.*2.*(a21 .*(a28 .*(a31 .*2.31)) -(a32 -2*a31 + a30) + b/dt*(dx/a).*(a24 .^2.26)) -(a27 -2*a26 + a25) + b/dt*(dx/a).33) + yt(n. f22 = (dx/dt/a).*2.*2.^2*F*a24/L.*(a32 .32)) -(yt(n+1.*(a20 .32) . f29 = (dx/dt/a).33)) -(yt(n+1.2*yt(n+1.23) + yt(n+1.26)) + 2*(dx/a).^2*(a28 .^2*F*a29/L.^2.2*yt(n+1.2*yt(n+1.28)) -(a29 -2*a28 + a27) + b/dt*(dx/a).34) .20))-(a21 -2*a20 + a19) + b/dt*(dx/a).31)) + 2*(dx/a).yt(n.27) + yt(n.^2*(a31 .21) + yt(n+1.30)) -(yt(n+1.^2*F*a23/L.25)) + 2*(dx/a).26) + yt(n.31)) -(yt(n+1.24))-(a25 -2*a24 + a23) + b/dt*(dx/a).^2*F*a22/L.^2.27) + yt(n+1.21)) -(yt(n+1. f23 = (dx/dt/a).21))-(a22 -2*a21 + a20) + b/dt*(dx/a).*(a23 .^2*(a29 .29) + yt(n.2*yt(n+1.^2*F*a28/L.2*yt(n+1.^2*(a24 .^2*F*a20/L.28) .24)) + 2*(dx/a).^2.yt(n.2*yt(n+1.21) + yt(n.31) + yt(n.^2*F*a30/L. f20 = (dx/dt/a).^2*F*a27/L.30) .*(a22 .28) + yt(n.30) + yt(n+1.(a28 -2*a27 + a26) + b/dt*(dx/a).23)) + 2*(dx/a).*2.yt(n.*2.22))-(a23 -2*a22 + a21) + b/dt*(dx/a).^2*(a25 .25)) -(yt(n+1.^2*(a22 .27) .25) .^2*(a27 .^2*(a21 .^2*F*a26/L.22)) + 2*(dx/a).30) + yt(n.^2*(a20 .32) + yt(n+1.2*yt(n+1.22)) -(yt(n+1.yt(n.yt(n.2*yt(n+1.2*yt(n+1.30)) + 2*(dx/a). f31 = (dx/dt/a).24)) -(yt(n+1.26) .*(a30 .31) .*2.

f41).38)) -(a39 -2*a38 + a37) + b/dt*(dx/a).40)) -(yt(n+1.^2*(a36 .*2.*(a34 .*(a41 .f17.f29.f21.2*yt(n+1.f10.f22.*(a33 .2*yt(n+1.34) + yt(n+1.f25.*2.37)) -(a38 -2*a37 + a36) + b/dt*(dx/a).40) + yt(n.^2.33)) -(a34 -2*a33 + a32) + b/dt*(dx/a). f33 = (dx/dt/a).yt(n.38)) -(yt(n+1.2*yt(n+1.2) = double(vpa(result.yt(n.f40 .5) = double(vpa(result.38) + yt(n+1.39) .yt(n.f27.39) + yt(n+1.38)) + 2*(dx/a).f13.41) .^2.41) + yt(n.f06.39) + yt(n.f26.^2*(a32 .a03)).yt(n.f31.f02.^2*F*a34/L.2*yt(n+1.36)) -(a37 -2*a36 + a35) + b/dt*(dx/a).37)) + 2*(dx/a).yt(n.36) + yt(n+1.^2*F*a40/L.*2. yt(n+2.42) + yt(n. f36 = (dx/dt/a).38) + yt(n.34)) -(a35 -2*a34 + a33) + b/dt*(dx/a).^2.f28.^2*F*a35/L.36)) + 2*(dx/a).35)) -(a36 -2*a35 + a34) + b/dt*(dx/a).34)) + 2*(dx/a).37) .yt(n.35)) -(yt(n+1.40)) -(a41 -2*a40 + a39) + b/dt*(dx/a).^2*F*a36/L.33)) + 2*(dx/a).41)) -(yt(n+2.37) + yt(n+1.2*yt(n+1.^2.2*yt(n+1.f03.^2.39)) -(a40 -2*a39 + a38) + b/dt*(dx/a).a01)). yt(n+2.^2*(a35 .34) + yt(n.f19.42) + yt(n+1.39)) + 2*(dx/a).f35.2*yt(n+1.f15.*(a38 .32)) -(a33 -2*a32 + a31) + b/dt*(dx/a).^2*(a38 .yt(n.f37.42)) -(yt(n+1. f40 = (dx/dt/a).^2.*(a35 .2*yt(n+1.42)) + 2*(dx/a).37)) -(yt(n+1.f30.6) = double(vpa(result. yt(n+2.2*yt(n+1.4) = double(vpa(result.nx) -2*a41 + a40) + b/dt*(dx/a).f34.2*yt(n+1.39)) -(yt(n+1.36) + yt(n.f16.41)) + 2*(dx/a).^2*F*a39/L.3) = double(vpa(result.37) + yt(n.*2.35)) + 2*(dx/a).f05.2*yt(n+1.^2*F*a38/L.40) + yt(n+1.^2*(a34 .^2*(a33 .f18. f41 = (dx/dt/a).f39.*(a39 .yt(n.*2.*2.^2.f14.^2.*(a37 .2*yt(n+1.*(a36 .f04.+ yt(n+1.a02)).41)) -(yt(n+1. yt(n+2.a04)).f23.^2*F.^2*F*a37/L.40) .f12.2*yt(n+1.*2.^2*F*a33/L.34)) -(yt(n+1.f38.2*yt(n+1. f34 = (dx/dt/a). f37 = (dx/dt/a).^2*(a39 .2*yt(n+1.yt(n.2*yt(n+1.^2*(a37 .f07.^2*(a40 .f09. 90 .40)) + 2*(dx/a).f08.f36.2*yt(n+1.nx) .f33.42) .*2. f35 = (dx/dt/a).a05)).2*yt(n+1.^2*F*a41/L.f11.*(a40 . yt(n+2.*2.35) + yt(n.f24. result = solve(f01.38) . f38 = (dx/dt/a). f39 = (dx/dt/a).35) + yt(n+1.^2*(a41 .41) + yt(n+1.36) .f 20.f32.yt(n.35) .36)) -(yt(n+1.^2.

a36)).9) = double(vpa(result.a33)). yt(n+2.a32)). yt(n+2.19) = double(vpa(result.26) = double(vpa(result.20) = double(vpa(result.a41)).8) = double(vpa(result.:) + (pp(1).a20)).a24)).a13)).a07)). yt(n+2.10) = double(vpa(result. yt(n+2. yt(n+2. yt(n+2.28) = double(vpa(result.a38)).^2+pp(3).a25)).25) = double(vpa(result.*x.a22)).31) = double(vpa(result.a21)). yt(n+2.a40)). yt1(n+2. yt(n+2.a12)). yt(n+2.30) = double(vpa(result.a35)).7) = double(vpa(result.a29)).32) = double(vpa(result.a39)).24) = double(vpa(result. yt(n+2.a11)).14) = double(vpa(result. yt(n+2. yt(n+2.a17)). yt(n+2.a14)).41) = double(vpa(result.21) = double(vpa(result.16) = double(vpa(result. yt(n+2. yt(n+2.a15)). yt(n+2.a18)). yt(n+2. yt(n+2. yt(n+2. yt(n+2.a08)). yt(n+2.35) = double(vpa(result.23) = double(vpa(result. yt(n+2.12) = double(vpa(result.yt(n+2.42) = double(vpa(result.37) = double(vpa(result.39) = double(vpa(result.a31)).13) = double(vpa(result. yt(n+2. yt(n+2.*x.36) = double(vpa(result.a09)).a26)).a28)).:) = yt(n+2. 91 . yt(n+2.18) = double(vpa(result.29) = double(vpa(result.27) = double(vpa(result.a27)).17) = double(vpa(result.^3+pp(2).a30)).*x)*(1-yr(n+2)/(ytempaL .33) = double(vpa(result. yt(n+2.*x+pp(4) ynratio.40) = double(vpa(result.a16)).22) = double(vpa(result.a34)). yt(n+2.a19)).a23)). yt(n+2. yt(n+2.38) = double(vpa(result.34) = double(vpa(result. yt(n+2.a06)).11) = double(vpa(result.a10)).15) = double(vpa(result. yt(n+2. yt(n+2. yt(n+2. yt(n+2.a37)).ys)). yt(n+2.

va = 0.nn))/dt. fa(i) = mg/L*x(nn)*va(i)/dt.yt1(i.nt)..nn))/dt)/2.n = n + 1.r.nn))/dt. i = i + 1. %calculate the change condition of the pitch angle on the seat while(i <= nt) pps(i) = ((yt1(i.nn) .%the impulse force -M*dv/dt while(i <= nt) if(i==1) va(i) = (yt1(i+1. fa(i) = mg/L*x(nn)*va(i)/dt. end.yt1(i-1. end. fprintf('Finished computating the spring deflection.yt1(i. %used node i = 1.%the velocity of the spring fall down fa = 0. %calculate the velocity i = 1.yt1.nn-1))/(x(nn) . pps = 0. end. fa(i) = mg/L*x(nn)*va(i)/dt..\n').nn) .yt1(i-1. end.x(nn-1))).nn))/dt + (yt1(i. else if(i < (nt-1)) va(i) = ((yt1(i+1. else va(i) = (yt1(i. g1 = 0.nn) .nn) .L. nn = 6.nn) .. %play the spring animation play_movie(x.yt1(i. i = i + 1. %calculate the spring seat force i = 1. end. 92 .

g4(i) = ei*sin(pps(i))/ri*g3(i). %calculate the spectrum of the spring sp = cal_spec(Fe2(926*2+1:926*3)). fprintf('The Maximum Spring Seat Force: %7.^2/r). end. Fe2(i) = Fe(i) + h4(i).1401). Fe(3:nt) = Fe(3:nt) . % The modifed spring Force by the energy terms. i = 1. torsion and bending terms %F=GJ/r*cos(p)*delta(torsion) .M*dv/dt while(i <= nt) ri = r*cos(pps(i))/cos(pps(1)). %read the experimental data t1 = linspace(0.g2 = 0. %the new change radius when the spring is compressed Fe = 0.082. i = i + 1.^2/dy .0575.0828. i = i + 1.cos(pps(1)). Fe(i) = -(g2(i) .^2.0. ri = 0.EI/r*sin(p)*delta(curvature) + %1/2*EI*(delta(curvature)). g3(i) = (cos(pps(i)).0583. %cut the beginning error Fe = Fe + fa.1401). end. %The Spring Force by Wahl's equation F = GJ/r*cos(p)*delta(torsion) EI/r*sin(p)*delta(curvature) .sin(pps(1))*cos(pps(1))/r). %setup the modified force h4 = 0. %t1 = linspace(0.2f(N)\n'. g4 = 0.^2/ri .g4(i)). g1(i) = (sin(pps(i))*cos(pps(i))/ri . 93 . g3 = 0.max(Fe2)).^2 while(i <= nt) h4(i) = 1/2*ei*((pps(i)-pps(1))/(x(nn)-x(nn-1))).M*dv/dt Fe2 = 0. %the second bending force 1/2*EI*(d^2y/ds^2).0.Fe(3). %read the experimental data ytc = read_expt_op(). g2(i) = G*J*cos(pps(i))/ri*g1(i).

yt1(1525.05]). ylabel('Spring Seat Force(N)').ytc.t.3). fprintf('The scale of the spectra at this harmonic is %7. %output the data subplot(3.:)). fprintf('The fundamental natural frequency is %7. legend('node 8'.spmax).ypitch).'time is 0. legend('modified eq'.085 0 0.pps).2.2. xlabel('spatial length(m)').2).spi] = max(sp(8:20)).0575 0.x.x.Fe2. fprintf('The %7. title('variable pitch angle by the CTCS'). ylabel('displacement(m)').2.a/2/L).6). subplot(3. 94 .8).2). axis([0.t1. xlabel('time(s)'). xlabel('spatial length(m)').055 0.'time is 0 with preload'.4).t.yt1(:.[spmax.2.'experimental data').:). plot(t. ylabel('displacement(m)').0152 with preload'. plot(x.1). title('Power spectrum'). title('Pitch angle at time is 0'). subplot(3.5). xlabel('Number of Harmonics'). axis([0.'.2. subplot(3. plot(sp(1:20)). ylabel('pitch angle(rad)'). subplot(3. plot(x.2f th harmonic would be excited\n'. xlabel('Time(s)').21). legend('no preload'.'node 21'.':').'node 34'). subplot(3.8+spi-1).2f(hertz)\n'.yt1(:.yo. plot(t.'-.075 -50 200]).2.34)).yt1(:.yt1(3.2f\n'. plot(t. ylabel('Power of Harmonics').

. nx = length(x)... y1 = pp0(1)...^3+ppo(2). title('the change of the pitch angle'). 95 .. xlabel('time(s)').001:L. while(i <= imax) y3(i) = y2(imax ...08]). nc = 123... %get the curve coefficients 1.40625 + 1.. y2 = ymax .nc1:nc2) = 0.*x. cp(2.06 0.*x. end. cp = 0:dg:360....(B)Calculate the Curve of the Opposite Installation Direction %calcuate the curve of the spring installed on the opposed direction function ppn = c_pp_cal(L....085 0... ppn = polyfit(x.. 758 847 945 1050 1169 1292 1427 1571 1728 1884. i = 1.deg.40625.75/1. i = i + 1. nc1 = 298.y3..(C) Read the Cam Profile %read cam profile %insert the values into the cam profile to get that function yr = cam_rise(t....ppo) x = 0:0. dg = 1.(nc+1):(nc1-1)) = [5 9 13 17 17 22 26 34 43 56.. 259 284 314 348 386 432 483 542 606 678. ylabel('the pitch angle').displacement].y1..3).125/dg + 1.. 72 89 106 123 144 161 178 195 216 238..1:nc) = 0.i + 1). 1.. ymax = max(y1). cp(2. cp(2...dt.^2+ppo(3).055 0. imax = length(x).w) %[.*x +ppo(4).axis([0. nc2 = 360/dg + 1..

1)) yr(1) = cp(1.01389.j) .cp(2..\n').. 5229 5326 5411 5478 5533 5576 5601 5610 5605 5584. 3734 3899 4069 4238 4399 4560 4716 4860 4996 5114.j)) j = j + 1... if(i>1) yr(i) = (theta(i) . while(theta(i) > cp(1. theta = 2*pi*w.nt) fprintf('Computing the spring animation...(D) Play Spring Animation %play the spring animation(the same coil diameter) function play_movie(x...j-1))*(cp(2.r.. i = 1. yr = 0... j = 1. end..*t/pi*180. end. 96 .j-1).. %calcualte the cam values while(i <= i_max) if(theta(1) == cp(1...%the play times kk = 0. j = 1..cp(1. i = i + 1..yt1.j) ..L.. %the approximate length for one turn coil xx = 0:0. 1618 1457 1309 1169 1042 923 818 720 635 559.j-1))/(cp(1. 1.. end.1).001:L. i=1. %the initial play index ni_max = 2. 161 149 132 119 106 94 81 68 56 39 26 13 5]/1000000.. 492 432 381 339 301 267 242 216 195 178... 3849 3586 3336 3095 2858 2621 2397 2189 1986 1795. end..cp(1.j-1)) + cp(2. ni = 1.2045 2215 2388 2557 2731 2909 3078 3247 3413 3578. 5546 5483 5394 5284 5124 4983 4788 4577 4344 4103. i_max = length(t).

ztt1 = ppt(1).(F) Calculate the Power Spectrum %calcualte the spectrum of spring %input the spring seat force for one cycle function sp = cal_spec(fe) a = fft(fe).L) %p = [p1 p2 p3 p4]. end.yt1(i.%get the imaginary part sp = sqrt(b.ztt1). fprintf('Finished the animation playing\n').^2+ppt(3).(G)The Main Optimization File %spring optimization %function x = cal_ops(d. B=-A.02 -0.^3+ppt(2). axis([-0./kk)./kk). if(i == nt & ni <=ni_max) i = 1. while(i<=nt) ppt = polyfit(x.:).ytt1. %get the power spectrum 1. b = real(a).%get the real part c = imag(a).*xx+ppt(4). plot3(xtt1. %the coefficients of the height profile 97 .04]).1(sec)'). title('Playing time is 0 ~ 0.*c).01 0. pause(0.02 0.(E) Read the Experimental Data %Experimental data of the spring function C = read_expt() A=[experimental data ].xtt1 = r. 1.*b + c.01).*xx. %the x coordinate of the spring ytt1 = r.3). i = i + 10. %the y coordiante of the spring fprintf('Playing the movie\n'). end.D.02 0.02 -0. 1. C=(B+210).*sin(xx. ni = ni + 1.*xx.*cos(xx.

210. 1. %the total helix length c = ['the main program to call the maximum spring force' . x0 = [-0.4 -0.@confun.1 -0.[]. ceq = [].x0.[].[]. %the coefficients of the wire diameter %c = [c1 c2 c3 c4].(H) Objective Function %objective function %[p1 p2 p3 p4 w1 w2 w3 w4 c1 c2 c3 c4] function obj = objfun(x) %point the path to the main program obj = 'simulation main program'.1 0.007 -0.1 0.options). -'the main program to call the maximum spring force' + 120. 1.004 0 0 0 0.0048 0. %lower bounds ub = [0 0.2817 0.0002 0.00009 0 0 0 0. fval] = fmincon(@objfun.(I) Constraint Function %constraint function %[p1 p2 p3 p4 w1 w2 w3 w4 c1 c2 c3 c4] function [c ceq] = confun(x) xf = 0.(J) clear. Natural Frequency Distribution % original one 98 .003 0.ub.%w = [w1 w2 w3 w4].2 -0.[]. %[] lb = [-0.0585 -0.0003 -0. 1.1 -0.'off').029].lb.004 0.0002 -0.2187 0.].0255].003 0. %tranfer the optimal value to the constraint to get the constraint values [c ceq] = confun(x). %the coefficient of the spactial length x0 = [-0.007 0. %x is the optimal value.004 0 0 0 0.1716 0.0255].022].004 -0.0002 0.42.006 0.005 0.1716 0.0585 -0.00009 0 0 0 0. clc. fval is the optimal value of the objective function [x. %the coefficients of the coil diameter %x0 = [p w c].0035 -0.3 0. %upper bounds %set up the large or medium scale options = optimset('LargeScale'.

dSL(i)= s2t(i+1) .00001. s2t(i+1) = s2 . s12(i) = (s1 + s2t(i+1))/2. %mass density % original one 99 . n = 50.ds.1716 0.091418 0. while (i <= n) while (ddy >= 0.s1.s1.*(s2 .^2 .079511 -0. dy = 0.268699 0.149103 0. % variable pitch angle -> frequency distribution equation ps = p0. s1 = s2t(i+1).145742 0. ds = 0.*(s2.^3 . % type 4 p4 = [-0.0585 -0. i = 1.(ps(1). ddy = 1.42 0].00009].0386/n. ddy = 1.0001) ddy = dy .86e3. %optimization types % type 1 p1 = [-0.000198]. i = i + 1. end. G = 77e9.p0 = [-0.001535].^3) + ps(2).^2) + ps(3).*(s2.s1)).2187 0. s2 = s2 + ds. % constant pitch type pf = [ 0 0 0. s2 = 0.0386/0.279762 0. s1 = 0.s1. %shear modulus mu = 7. s2t = 0. end.

% frequnecy distribution equation fre = sqrt((G. ds) format long. D4 = [-0.000072 -0. ylabel('Frequency Distribution at each element').004.fre. %type 3 D3 = [-0.^2 + d(3).002625 0.^3 + d(2).*s12 + d(4)).*s12.^3 + d(2).0255].004]. bs.*s12 + D(4)).003628 0./(mu.000111 0.02345]./dSL. %type 4 p4 = [-0.000363 0. Program Code (ABAQUS-MATLAB program) 2.^3 + D(2). %optimization types %type 2 d2 = [0. d4 = [-0.*((D(1).279762 0.*s12. %variable wire diameter -> frequnecy distribution equation d = d0.2187 0.%wire diameter 100 .r-').*(d(1).*(2.149103 0.*s12.004683]. d = 0.(A) Main File function Fre = opt_spring(as.005998 -0.*s12.004397 0. %the parameter of the variable pitch angle py = [-0. cs.00009].^2)+(d(1). plot(s12.004731 -0.^2 + D(3).*s12.001535].001972 0.*s12 + d(4). xlabel('spatial length(m)'). %the parameter of the variable coil diameter(optimization input) D = [as bs cs ds].000231 0.028783].001993 0. title('Frequency Distribution') 2.^2 + d(3).'. D0 = [0 0 0 0.000198 0.^2)))).0585 -0.^2). %variable coil diameter -> frequnecy distribution equation D = D0.1716 0.d0 = [0 0 0 0.*s12.0046].091418 0.

y1 = DD/2*sin(t(i)). s = 0. f(2) = dl .^2+2*py(2).*0.^2 + py(3).0001:10*2*pi.*ss + D(4).0386.*ss. ss = s. tx(k) = DD/2*cos(t(j-1)).*ss + py(4). x2 = DD/2*cos(t(j)).*0.^2 + D(3).^3 + D(2).^2 + (y2-y1).^3 + D(2).*ss.024. tz(1) = 0.42.^2).*ss+py(3)). x(1) = (D(1). if(abs(f(2)) >= abs(f(1))) syt = 2. k = k + 1. while syt == 1.dsl. y(1:length(t)) = 0.%preloaded height ys = 0. i = j. y2 = DD/2*sin(t(j)). ty(k) = DD/2*sin(t(j-1)). %compute the coordinate of the spring with the variable pitch angle %and variable coil diameter while ss <= L+s DD = D(1).yf = 0. x1 = DD/2*cos(t(i)). tx(1) = x(1). k = 2.002. dl = sqrt((x2-x1). syt = 1. tz(k) = py(1).*0 + D(4))/2. i = 1. 101 .*ss.*ss.^2 + D(3). ss = ss + s. j = 1. ty(1) = 0. dsl = s*cos(3*py(1). f(1) = 1.%total spatial length t = 0:0. f(1) = 1.%solid height L = 0.^3 + py(2). x(1:length(t)) = 0.*ss.

end fclose(fidr). while 1 tline = fgetl(fidr). end fclose(fidr). end. break.'\n'). fprintf(fidw. if ~ischar(tline).tline(1:80)).'rt'). %write global section fidr = fopen('g2. j = j + 1. x = tx. end. end. 102 end end . fprintf(fidw. if ~ischar(tline). is = 1.else f(1) = f(2). max_xy = length(x). y = ty.txt'. %write a IGES wire file fidw = fopen('part1. z = tz. ig = 4.'rt').txt'.'wt'). clear tx ty tz. %write start section fidr = fopen('g1. fprintf(fidw. while 1 tline = fgetl(fidr).tline(1:80)). break. fprintf(fidw. clear x y.igs'. syt = 1.'\n').

end. j = 1.']. %write the parameter section while(count <= (max_xy-1)) st1 = ['110.:)=[' 110' numstr(j.5f') '.'D')]. 103 0 0 10000 0 0 0 1 0 Line end . count = count + 1.5f') '.' num2str(y(count).'%6.[de(2. fidr = fopen('temp_de.'rt').' ') ' 000000001' numstr(count.' num2str(x(count+1).' num2str(z(count).'%6. if ~ischar(tline). id = 2*(max_xy-1). count = count + 1. end fclose(fidr). st1 = [st1 blanks(64-length(st1)) numstr(j.0.5f') '.'\n'). %write DE section while(count <= 2*(max_xy-1)) de(1.' num2str(z(count+1).' num2str(y(count+1).'%6. de(2.txt'.:) '\n']).' ') numstr(count. count = count + 1. fprintf(fid_de.:) '\n']).tline(1:80)).5f') '.'P')].[st1 '\n']). fprintf(fid_de. break.fid_de = fopen('temp_de.' num2str(x(count).'%6. j = 1. fprintf(fidw. count = 1.[de(1.'%6.' '0. j = j + 1. count = 1.'%6. fprintf(fidw. fclose(fid_de).5f') '.'wt').txt'. fprintf(fidw.5f') '.'D')].:)=[' 110 0 0' numstr(count. while 1 tline = fgetl(fidr).

end. fprintf(fidw. end.'D') numstr(ip. fprintf(fidw. fprintf(fidw. fidr = fopen('template_p2.j = j + 2. while 1 tline = fgetl(fidr). while 1 tline = fgetl(fidr).'S') numstr(ig.igs' '"' '\n']). if ~ischar(tline).py'.[temp_t blanks(72-length(temp_t)) numstr(1. break. 'rt').txt'. ip = max_xy-1.tline). fidw = fopen('opt_spring. i = i + 1.'P')].tline). fclose(fidw). end 104 end end .'G') numstr(id. fidr = fopen('template_p1. break. imax = length(workdic).'\n'). while(i<=imax) if(workdic(i) == '\') workdic(i) = '/'.'wt'). end fclose(fidr).'T')]). fprintf(fidw. 'rt'). %write the terminate section temp_t = [numstr(is. fprintf(fidw. i = 1. fprintf(fidw.'\n').['igsFile =' '"' workdic '/part1. %check the current working directory workdic = pwd.txt'. if ~ischar(tline). end.

end if(i == 9) if(tline(27:31) == 'Value') Vmax = str2num(tline(35:46)).['mdb. fidr = fopen('spring_report. end.'\n').txt'.7f %8. fprintf('The Finite Element Analysis in ABAQUS is done. end fclose(fidr).py'). while i <= i_max tline = fgetl(fidr). bs.models[modelName].7f %8. fprintf(fidw. end. ds).tline).7f %8. 'rt'). Fre = str2num(tline(55:62)).as. break.\n'). 105 end .CircularProfile(name=''CircularProfile''. fprintf('The parameter: %8.fclose(fidr). fidr = fopen('template_p3. fprintf(fidw. Vmax = 0.rpt'. if ~ischar(tline).. i_max = 9. while 1 tline = fgetl(fidr). 'rt'). fclose(fidw). i = i + 1. %run the Abaqqus fprintf('The Finite Element Analysis in ABAQUS is processing\n'). break. [istatus.7f\n'. if ~ischar(tline). fprintf(fidw.result] = dos('ABAQUS cae noGUI=opt_spring. Fre = 0. %read the field report in ABAQUS i = 1. r=' num2str(d/2) ')' '\n']).. cs.

%lower bound lb = [-0. %upper bound ub = [0 0.end fclose(fidr).Fre). fprintf('06. fprintf('Running the wire frame mode\n').2009\n').0585 -0.Display = 'iter'.[]. %set up the optimization option options = psoptimset.4f\n'. %x is the optimal value.@confun1.1].options). %outoput the result fprintf('The fundamental natural frequency is %6.(B) Optimization Main File %optimal main program clear.x(1).ub.x(2).1].20. options.x(2). fval).2f\n'.00009].15f\n'. 2.[].3 0.[].Vmax).15f %16.1 0. fprintf('in University of Missouri at Columbia\n'). fval is the optimal value of the objective function fprintf('Valve Spring Optimization\n').(C) Objective Function %objective function %minimize the maximal stress function obj = objfun1(x) %point the path to the main program obj = -opt_spring(x(1).x(4)).x(3).2187 0.x(4)). %tranfer the optimal value to the constraint to get the constraint values [c ceq] = confun1(x). x0 = [-0.x0.1 -0. [x fval] = patternsearch(@objfun1.x(3). fprintf('This program was designed by Yu-Cheng Su\n'). fprintf('the optimal fundamental natural frequency: %6. clc.4 -0.(D) Constraint Function 106 . fprintf('The value at this frequency is %7.15f %16.1716 0.2 -0. 2. 2.[].lb. fprintf('the best parameter: %16.5e\n'.

%constraint function function [c ceq] = confun1(x) xi = 0. 107 . ceq = []. c = [].

3. Experimental Data

3.1 Installation Direction

Appendix 3.1.1 Spring normal installation

Appendix 3.1.2 Spring opposite installation

108

3.2 The Force Figures at 1368, 2165, and 2372 (rpm)

3.2.1

Normal Installation at 1368rpm

Appendix 3.2.1 Force with a normal installation at 1368rpm

3.2.2

Opposite Installation at 1368rpm

**Appendix 3.2.2 Force with an opposite installation at 1368rpm
**

109

3.2.3

Normal Installation at 2165rpm

Appendix 3.2.3 Force with a normal installation at 2165rpm

3.2.4

Opposite Installation at 2165rpm

Appendix 3.2.4 Force with an opposite installation at 2165rpm

110

6 Opposite Installation at 2372rpm Appendix 3.5 Normal Installation at 2372rpm Appendix 3.6 Force with an opposite installation at 2372rpm 111 .5 Force with an opposite installation at 2372rpm 3.2.2.2.2.3.

Lbs@RPM: 87@3400 Compression Ratio: 9.4.9mm YEAR: 1987 4.05~0.96mm for journal diameter 0.932~25.5kg@39.1 GM-ISUZU Engine DISPLACEMENT: 90inchଷ or 1. Equipment in the Experiment 4.1mm for Clearance Valve Spring: Free Length: 48.6:1 Bore: 77mm Stroke: 79mm Camshaft: 25.50mm Valve Closed Length: 21.2 Dayton Compressor Duty Motor 112 .5 liters HP@RPM: 70@5400 TORQUE Ft.

05% accuracy 5 digits LCD display Appendix 4.3 Photo Sensor Tachometer 4.4 KISTLER Force Transducer Model: 9031A 4.0.3 Cen-Tech Photo Sensor Tachometer RPM range: 2.4.5 KISTLER Charge Amplifier Model: 5004 113 .5~99999 Digital photo Stores last minimum and maximum Auto-zero adjustment +/.

5 The charge Amplifier The data of the charge Amplifier: 114 .Appendix 4.

6 DATAQ Acquisition Data: 115 .6 DATAQ Acquisition Model: DI-158UP Appendix 4.4.

63”. HY-T PLUS.7 V-Belt Manufacturer: Goodyear Type: V-Belt.8 The pulleys with different size 4.8 Timing Pulley Type: Mean Radius: 2. B56(5L590) V-Belt.2 cm Mean Radius: 5. HY-T PLUS.9 KISTLER Cable Model: 1361A(X) Function: to connect the transducer and the charge amplifier Wide Temperature range:-195 to 240Ԩ Capacitance: 30pF/ft Specify length to 30 meters “A” versions available in standard lengths of 1 through 5 meters 116 . 59”.1 cm Mean Radius: 3.5 cm Mean Radius: 6 cm X1 X1 X1 X1 Appendix 4. B60(5L630) X1 X1 4.4.

49 – to be used to acquire the experimental data MATLAB R2007b – to run optimization program one and two with the derived dynamic equations in this thesis or FEA software.8. Software WINDAQ Ver. ABAQUS CATIA V5R17 – plot the spring 3D structure ABAQUS Ver.1 – use Finite Element Method to verify the optimal results based on the derived dynamic equation and the solutions in Finite Difference Method 117 .5. 2. 6.

- -28sici-291099-0887-28199901-2915-3A1-3C9-3A-3Aaid-cnm219-3E3.0.co-3B2-y
- DPS PPT2
- designofgasholde00pill.pdf
- Pde
- Calculation of Voided Slabs Rigidities
- 31295000275171.pdf
- Civil-1
- AEPSPAN PS162_BoxRib
- 4 Effect of Axial Load & Frames
- 15 Flexure Section Modulus
- 14 about the subject
- 17 - Yield Line Theory for Slabs
- mech
- 110010
- DST BK - PENNY & DIANA ppt.pptx
- Stability_Group 2 (Obias,Ramos).pptx
- A Basic Course in Partial Differential Equations - Qing Han
- Mlep Sem05 Revised
- Detroit Medical Center - Biomech2
- Timoshenko Beam - 1.pdf
- Strength of Materials (S.O.M.) Model Question Paper (Q.P.) Solution
- 200-Foot Simple Span Bridge Girder Design Using NU2000 V1
- 56544
- International Codes V8i
- Ycmou Ast Pqp t35 s05 200802 Revised
- Chen HF Pure Shakedown and Limit Analysis of 90 Degree Pipe Bends Under Internal Pressure Cyclic in Plane Bending and Cyclic Thermal Loading May 2011
- The Design Procedures for Prestressed Reinforced Concrete
- Solve Non-Linear Parabolic Partial Differential Equation by Spline Collocation Method
- R09222104-AEROSPACEVEHICLESTRUCTURES-I
- Witz 1996

- 2
- tajhizat-ravankari(1)
- Generating an Advanced Bill of Material.pdf
- Generating an Advanced Bill of Material.pdf
- Knowledgeware_2.pdf
- HEIDENHAIN TNC 426 Training - Basis
- 329_203-26.pdf
- Geometry - Do These Equations Create a Helix Wrapped Into a Torus_ - Mathematics Stack Exchange
- Advanced.pdf
- لیست دروس کارشناسی ارشد رشته ساخت و تولید دانشگاه تبریز
- O-RING
- joining aluminium.pdf
- Joining Aluminium
- 1
- حداقل حقوق 96 مشخص شد ؛ 930 هزار تومان + جزئیات
- Sweeps in SW
- Angry Bird Sharpner
- حداقل حقوق 96 مشخص شد ؛ 930 هزار تومان + جزئیات.pdf
- Understanding Injection Mold Design
- ME-0604-PS-001

Read Free for 30 Days

Cancel anytime.

Close Dialog## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

Loading