t ANALYSIS

OF A PLANAR SPIRAL DISPLACER SPRING

FOR USE IN FREE-PISTON STIRLING ENGINES,

A Thesis Presented to The Faculty of the College of Engineering and Technology Ohio University

In Partial Fulfillment

of the Requirements for the Degree

Master of Science

by Roger Stage!
-r-

November, 1991

ACKNOWLEDGMENTS

First and foremost I wish to express my sincere thanks to my graduate advisor, Mohammad Dehghani, Ph.D., P.E., without whose support and advice this thesis would not have been possible. Secondly, I would like to thank Mr. Bhavin Mehta for his patience in answering my innumerable questions and for his unflagging efforts to keep the computing resources necessary for this study up and running.
I would also like to thank Sunpower Incorporated for the use

of their lab equipment and facilities.

Special thanks goes to Mr.

Jarleth McEntee of Sunpower for his insightful views on how to approach this design analysis problem. Two other individuals of

Sunpower certainly deserving mention are Mr. Albert Schubert, the designer and developer of the planar spiral spring testing rig used in the lab and Mr. Andy Weisgerber for his photography work which appears in the Appendix. Finally, I wish to express my gratitude to 0 . E. Adams, Ph.D., P.E., and Joseph A. Recktenwald, Ph.D., for serving as members on my graduate committee. Their comments and advice were constructive

and much appreciated.

TABLE OF CONTENTS
Page List of Figures Chapter 1: Rationale Introduction Alternate Spring Designs Planar Spiral Springs Purpose of Study Closure Chapter 2: Review of Literature vi
1 1

1 2 2
5

7
7 7

Definition and Purpose of Mechanical Springs Spring Design Force-Deflection Curves and Spring Rate Stress and Fatigue Energy Storage Planar Spiral Springs Comparisons with Conical Springs Background of Spring Design Theory Problems in Using Conical Spring Design Theory with Planar Spiral Springs Hypotheses Closure Chapter 3: Methods

8
9

15

15
16

21

23 25 25 27 27 28 28

Selection of Analysis Method Equipment and Software Finite Element Model Analysis Approach

33

Closure Chapter 4: Results

36
38 38

Experimental Results Finite Element Solutions for Deflection Finite Element Solutions for Stress Closure Chapter 5: Discussion Experimental and Finite Element Force-Deflection Results Finite Element Stress Results Closure Chapter 6: Conclusions and Recommendations

38

53

64
83

83
86

91 93 93

Conclusions Recommendations References Appendix

LIST OF FIGURES

Figure 1a 1b 2 3 4 Face of a Planar Spiral Spring Rotated View of a Planar Spiral Spring
o-N Diagram

Page 3

3
11 12 13 17 18 19 19 20 29

Goodman Diagram Modified Goodman Diagram Deformed Planar Spiral Spring Conical Spring Conical Spring Section View Planar Spiral Spring Section View Force-Deflection Curve for Conical Springs Planar Spiral Spring Testing Rig Transition Mesh Deformed Transition Mesh

5
6

7a

7b
8 9 10
11

32
32

12 to 16

Force-Deflection Curves for Laboratory Results 3 9 - 4 1 42 42 44 44 45-47 48-49 50-5 1 52 54

1 7 a Finite Element Mesh 1 7 b Finite Element Mesh Closeup 1 8 a Finite Element Mesh 1 8 b Finite Element Mesh Closeup

19 to 23
24 to 26 27 to 30 31 32

Force-Deflection Curves Convergence Studies for Deflection Force-Deflection Curves

Deflection vs. Thickness Deflection vs. Width of Material Cut

vii 33 to 35 Convergence Studies for Maximum Von Mises Stress 36 to 44 45 Force vs. Maximum Von Mises Stress 55-56 58-62 63

Maximum Von Mises Stress vs. Thickness

46
47

Maximum Von Mises Stress vs. Width of Material Cut 6 5 Von Mises Stress Contours
66

48 to 5 1 52 to 63 64

Position of Primary and Secondary Stress Peaks 6 7 - 7 0 Von Mises Stress Contours 71-82
98

Planar Spiral Spring Testing Rig

Table Compendium of Finite Element Results

CHAPTER 1: Rationale

Introduction Certain designs of Stirling engines incorporate a mechanical spring to help provide the restoring force necessary to maintain cyclic operation of the reciprocating elements (26). The spring may

act in combination with a gas spring to tune the displacer so that it will resonate at a frequency near the engine's design operating frequency (4). In many instances the displacer spring is a helical type of spring attached to the displacer rod. However, there are applications

of Stirling technology where a premium is put on compactness and a helical spring will not work. This study is an analysis of a spring

design intended for use in Stirling engines where helical types of displacer springs cannot be used. Alternate Spring Designs An example of a Stirling engine where a helical displacer spring is replaced by an alternate spring design is the Thermo-mechanical Generator (TMG) described by Cooke-Yarborough (3). The TMG is a special type of free-piston Stirling engine where the piston is actually a vibrating diaphragm (25). Instead of a helical spring,

Cooke-Yarborough described a displacer spring for a TMG made from a flat disk of stainless steel which consisted of four circular arcs (3, 4,

15).

Further development lead to a design consisting of a single ring Though these designs were

with four lugs as attachment points (4).

successful in TMG's, there are problems in using either of them in other Stirling applications. First, the four arc springs are difficult to

manufacture (4) and

secondly, the spring with lugs has a length of

stroke that is somewhat limited. Planar Spiral Springs There are applications of other types of free-piston Stirling engines that also impose space limitations so strict that a helical spring cannot be used on the displacer. However, these applications

call not only for a compact spring but one that has a stroke length or operating length greater than that of a TMG. Figure 1 shows the face

and profile of a design idea intended to meet these criteria. The coils of the spring are the result of removing two spirals of material from a round flat disk of spring steel. In this particular

design example, spirals the shape of involutes have been removed by cutting or stamping. involutes. The two resulting coils are also in the shape of

A spring of this type could have more than two coils

(resulting from an equal number of material cuts) and the coils could be in a shape other than involutes. The scope of this study, however,

is limited to the basic design of Figure 1. As it appears that a common name for such a spring does not exist, the author has chosen to use the descriptive term of planar spiral spring for this study. Purpose of Study Recognizing that any design must be analyzed to determine whether performance will comply with specification (22), the goal of this study is to determine the relationships between the geometric parameters of a planar spiral spring and the resulting performance characteristics. With regard to Stirling engine applications there are

three design criteria for the planar spiral spring: the spring must

material cut width

5.22 mm radius

>-\

--_,

Figure la: Face of a Planar Spiral Spring

Figure lb:

Rotated View of a Planar Spiral Spring

4 have a predictable, preferably constant, spring rate, the spring must have infinite fatigue life and the spring must be easily manufacturable. The spring rate is important in Stirling engine applications because of its effect on the resonant frequency of the system (26). The natural frequency of a simple spring mass system is a function of the spring rate (26). As springs are essential for providing the

restoring force necessary to maintain cyclic operation in free-piston Stirling engines (26), the spring rate of a planar spiral spring must be known if the spring is to be used to tune the displacer's natural frequency to the engine's operating frequency. In order for a spring to meet the second design criteria of infinite fatigue life the maximum working stress must be kept below the endurance limit of the material (21). The endurance limit of the

material is the stress level below which fatigue failure is not expected to occur regardless of the number of loading cycles (14). Consequently, the maximum working stress in a planar spiral spring must be known if comparison is to be made with the material endurance limit to determine the expected operating life of the spring. As fatigue cracks generally originate at the surface of a part, the surface condition has a significant influence on the fatigue life of a spring (25). A quality surface finish will improve a spring's Typically the highest stresses in a part occur at

resistance to fatigue. the surface (5).

Consequently, having a good surface finish at the

location of the maximum working stress will help ensure that a planar spiral spring has infinite life. Conversely, knowing where the

maximum stress occurs is useful to the extent that efforts can be made to ensure a good surface finish at least at that location. Economics is reason enough to look for an easily manufacturable design. Appearance suggests that planar spiral In the case of sizeable

springs could meet this last design criteria.

production runs the design lends itself to a stamping process thus obviating the need for any winding or grinding steps.

If determined

necessary, finishing processes such as heat treatment, shot peening or polishing could still be considered. By evaluating the spring rate, maximum working stress and the location of the maximum working stress, it can be determined whether the design criteria for a compact displacer spring in Stirling engines can be met by planar spiral springs. These baseline analysis

results may then be used by the spring designer as a general guide in selecting the geometry of a planar spiral spring such that the desired spring rate and working stress levels are achieved. Closure Some types of Stirling engines use a mechanical spring to help tune the displacer so that it resonates at a frequency at or near the engine's operating frequency. compactness, Some of these engines require

thus alternative designs to the typical helical displacer Two designs of flat springs for TMG's have

spring must be used.

been described but they do not fulfill the needs for compact springs in all Stirling engine applications. A new design of displacer spring, called a planar spiral spring, has been proposed to help fill this need. To be used, the spring must

have a predictable spring rate, an infinite fatigue life and be easily

6
manufacturable. An analysis is necessary in order to determine whether these criteria can be met by planar spiral springs.

CHAPTER 2: Review of Literature

Definition and Purpose of Mechanical Springs A mechanical spring may be defined as an elastic body whose primary function is to deflect or distort under load and which returns to its original shape when the load is removed ( 2 5 ) . Gross

( 1 1 ) gives a definition in a narrower sense in that any device which

is specially made for converting mechanical work into potential energy and re-converting it into mechanical work, by virtue of the elastic deformation of material, is a mechanical spring. It is the

ability to convert and store energy that makes mechanical springs useful and suitable for exerting force, providing flexibility and

reducing shock in machine design applications. In free-piston Stirling engines springs are essential to provide the restoring force necessary to maintain cyclic operation of the reciprocating components (26). In many designs of Stirling engines Some designs,

the restoring force is provided by a gas spring.

however, require that a mechanical spring be used to tune the displacer so that it will resonate at a frequency at or near the operating frequency of the engine (4). S ~ r i n r rDesign Gross notes that the whole purpose of doing spring design is to find the most suitable spring to fulfill a specific purpose ( 1 1 ) . work is centered on finding the three main aspects of spring performance: the force-deflection curve, the maximum stress intensity and the energy storage capacity. This

8
Force-Deflection Curves and Spring Rate The deflection as a function of applied force is an essential characteristic of any mechanical spring due to the fact that the forcedeflection curve is needed to obtain the stiffness. When coupled to

the displacer in a Stirling engine a spring will alter the natural frequency of the displacer by an amount which is a function of the spring's rate. Treated as a simple spring-mass system the natural

frequency of the displacer is determined by

where

fn

is the natural frequency, k is the spring rate and m is the

mass of the spring-mass system (25). Provided that working stresses do not exceed the elastic limit, most springs will have a linear force-deflection characteristic. That

is the deflection of the spring will be proportional to the load applied to it; doubling the load will double the deflection (25). This relation

holds for springs designed for torque or moment loads as well providing angular deflection is used in place of linear deflection (22). There are also a number of types of springs whose forcedeflection curves are not linear. These springs may exhibit nonlinear

stiffening or nonlinear softening or some other characteristic whose curve shape is not a straight line. In general, the relationship between force, F, and deflection, 6, can be given as (22):

F = F(6)
The spring rate is defined as (22):
k (6) = lim (AF /A6) = dF/d6

(2)

(3)

where 6 is measured in the direction of, and at the point of application of F. For springs where the force-deflection curves are

linear, k is a constant, known as the spring constant and equation (2) becomes (22):
k = F/6
(4)

Stress and Fatigue In addition to the required force-deflection characteristics of a spring from which the spring rate is obtained, the maximum allowable stress will also influence the design of a spring. This is

complicated by the fact that the maximum allowable stress depends on the material strength and the type of loading of the spring. For

convenience, the types of loading can be divided into static loading and cyclic loading. In static loading a spring is subject to load just In cyclic loading, the number of load cycles

once or only a few times.

the spring must be able to withstand may be only a few hundred or may be millions of cycles as is the case with Stirling engines. Consequently, springs loaded in this way are sub~ectto fatigue. In general, fatigue loading may be in the form of fully reversed stresses or loading may be either tension or compression. However,

regardless of the type of dynamic loads applied, a mechanical part's ability to withstand fatigue loading is dependent upon the properties of the material and the level of working stresses. The fatigue strength of a material is the level of stress above which fatigue failure can be expected to occur for a specified number of loading cycles (14). The characteristic behavior of a material

under uniaxial fatigue loading is typically shown using a stress-

10 number of cycles diagram (o-N diagram). Figure 2 is an example of such a diagram. For a material subjected to low-cycle fatigue, the fatiguestrength may be at or near the ultimate-strength of the material. very high-cycle loadings, the o-N curve forms a knee and the fatigue-strength levels off at a level known as the endurance limit (0,) or fatigue limit. The endurance limit is interpreted as the For

maximum stress which can be applied repeatedly for any number of cycles without causing failure (14). Springs are rarely subjected to both positive and negative stresses. Typically springs are subjected to fluctuating stresses In other words, the stresses, though The life of a

which never pass through zero.

cyclic, are either always compressive or always tensile.

spring, that is to say, the number of load cycles which it can withstand before failure, depends on the mean of the fluctuating stress but even more particularly on the amplitude of the fluctuation. At a given mean stress, life increases as the stress amplitude decreases (1 1). The fatigue strength can also be determined using the values for mean stress and stress amplitude along with either a Goodman diagram or a Modified Goodman diagram (21, 22, 23, 25). Figure 3

shows an example of a Goodman diagram and Figure 4 shows an example of a Modified Goodman diagram. The use of a Modified Goodman diagram is especially useful when the cyclic loading causes combined stress instead of uniaxial stress. When combined stresses, due to some combination of torsion,

OU

- ultimate strength

oe

endurance limit

10'

lo2

lo3

l o 4 lo5

lo6

lo7

Number of Cycles (Log Scale) Figure 2: o-N Diagram

o - ultimate stress , 4 - yield stress o, - endurance limit o, - alternating stress or - stress range om - mean stress

Figure 3:

Goodman Diagram

0 Mean Stress

o m

out

cry o, o, omoyc-

yield stress endurance limit alternating stress mean stress compressive yield stress out- ultimate tensile stress - tensile yield stress O Y ~

Figure 4:

Modified Goodman Diagram

14 bending and shear, from the fluctuating stresses a part is subjected to, the distortion-energy failure theory is used along with a Modified Goodman diagram to find fatigue strength. This is done by first

determining the mean Von Mises stress and the alternating Von Mises stress amplitude and then applying them to a Modified Goodman diagram. The justification for using this approach is that all

available experimental evidence shows it to be conservative (20, 21). Application of this approach to planar spiral springs is actually simpler than usual because, contrary to most spring applications, the spring is intended to be subjected to fully and equally reversed loading cycles. This results in a mean stress of zero, thus only the

alternating stress amplitude need be considered. To ensure that a spring will survive the cycles of variable loading expected during its service life, the spring must be designed such that the working stresses never exceed the endurance limit of the material. In order to do this a spring designer needs an

understanding of how working stresses are affected by geometric design parameters. As springs are primarily subject to some combination of shear forces, bending moments or torsional moments, the maximum stress will occur somewhere on the surface. This is because the maximum

stress in any part caused by any one of these forces occurs at the surface (5). Consequently, fatigue cracks, which initiate at points of The

high stress, tend to start on the surface of a spring (14)

tendency for a fatigue crack to start is increased if there are stress risers on the surface such as scratches, notches, scale or corrosion. spring's resistance to fatigue will be improved if the surface is of A

15 high quality (11, 14, 25), especially in the vicinity of the maximum working stress. Consequently, determination of the maximum stress

location is essential. Energy Storage All mechanical springs store energy when they are deformed. Springs whose purpose is to bring moving masses to a stop, such as buffer springs, are designed with energy storage capacity in mind. The energy stored in a spring is simply
U=

I,'

F(z)6dz
This equation reduces to

where z is the integration variable.

when the spring rate is constant (13) Planar S ~ i r a lSprings In Figure 1, the shape of the cut out spiral is that of an involute of a circle. The result of using two involute shaped cuts, 180' apart,

is that the coils are also involutes having constant width as they progress out from the center. The material cuts in planar spiral springs do not need to be in the shape of involutes necessarily. Cuts of various spiral shapes

would result in planar spiral springs but of different geometries. Springs of several coils can be made by introduction of nonintersecting material cuts. Two cuts, resulting in two coils, is the

minimum, where a coil is defined as the arm or spline of material that connects the outer ring to the center.

16 Comparisons with Conical Springs Figure 5 shows a planar spiral spring with an applied load. The

outer edge of the spring is fixed in position and the load, applied at the center causes the spring to deflect out of its plane. In this

distorted state the two coils appear to be similar to two conical springs acting in parallel. The conical spring, shown in Figure 6, is a type of helical spring where the mean diameter of the coils is not constant. When the conical spring is designed so that each coil nests wholly within an adjacent coil the spring can be compressed to a solid height of one coil. Comparatively, planar spiral springs have a

height or thickness of one coil in the unstressed state because all of the coils lie in a single plane. When loaded and deformed, a planar Hence, planar spiral

spiral spring takes on more of a conical shape.

springs can be thought of as being more like "conical extension springs" rather than as compression springs as conical springs are typically used. Conical springs are usually wound with wire having a circular cross-section. Conical springs made with material having a

rectangular cross-section with the long side of the cross-section parallel to the spring axis are called volute springs (1, 25). When the

long side of the cross-section is normal to the spring axis the orientation is similar to that of the planar spiral spring, see Figure 7. Regardless of the shape of the wire cross-section, conical springs have a force-deflection curve that is not fully linear. example curve is shown in Figure 8. As a conical spring is compressed it initially exhibits a constant spring rate. But as An

Figure 6: Conical Spring

: - pi g +S rn
I

Axis

+tb

Figure 7a: Conical Spring Section View

Spring Axis
I
I I I

rn m

F A \ ?

I I
I I I

=\ F A

Fs3

txsk

tsk+Y5Y

Figure 7b:

Planar Spiral Spring Section View

\ "bottoming out"
of active coils.
0

Deflection

Figure 8: Force-Deflection Curve for Conical Springs

21
compression is continued, a transition point is reached where the active coils begin to "bottom out" and the spring rate begins to increase. In contrast, as planar spiral springs are loaded in tension

all active coils remain active so the "bottoming out" phenomena is not possible. Another advantage of planar spiral springs is that they are potentially inexpensive to make in large numbers as it may be possible to stamp them out of sheet steel as opposed to the wire winding process used to make most types of springs. Also, as conical

springs have increased lateral stability compared to cylindrical helical springs (13), planar spiral springs may be quite stable too. Finally, by simulating two or more conical springs in parallel with the coils evenly spaced, planar spiral springs may exhibit less load eccentricity than do many sorts of helical springs. Planar spiral springs are not without certain disadvantages. For example, the coil orientation with the long side of the crosssection normal to the spring axis results in an unfavorable stress distribution. This is especially significant for infinite life

requirements for conical springs with similar coil cross-section orientations so it may also be a concern for planar spiral springs (25). Another disadvantage of planar spiral springs is the fact that very little fatigue data is available on any springs with rectangular coil cross-sections (25) and to the best of the author's knowledge, none has been published for planar spiral springs. Background of S ~ r i n gDesign Theorv A number of authoritative publications on the design and analysis of various types of springs are in existence. These include

helical springs in tension or compression, torsion springs, leaf springs, flat springs, conical springs, spiral springs, and special designs such as Belleville springs. In the literature are

methodologies for designing many types of springs for a needed force-deflection response, for determining working stress levels and for determining energy storage in a given spring type (1, 11, 13, 23,

25).
Computational procedures for determining the necessary geometry of a cylindrical helical spring for a required spring rate and maximum working stress level have been developed and are reasonably accurate (11, 25). The equations for spring rate can be

applied to conical springs if the spring is treated as a series of cylindrical springs. The rate for each coil or fraction of a coil if the wire has a circular cross-section is computed using (13):
P - ~d~ k = - - -----8 SD~N,

where

P - load, N

6 - deflection, mm

G - shear modulus, MPa

d - wire diameter, mm

D - mean spring diameter, mm

N, - number of active coils

If the wire is rectangular the equation to use is (13):

where

b - width, mm t - thickness, mm
K2 -

correction factor based on the ratio blt

Note that rectangular is the cross-sectional shape after winding. A keystone cross-section wire will deform to a rectangular shape when wound (13, 25). The rate for the complete spring is found by combining these individual rates according to the series relationship (13):
k=

1 l/kl + l/k2 + . . . + l/k,

(9)

where kl, k2, . . . kn are found using equations (7) or (8). To calculate the maximum stress at a given load, the mean diameter of the largest active coil is used. The effect of coil

curvature must also be accounted for when computing stress because neglecting curvature in rectangular coil springs may result in errors of 15 percent or more (25). Problems in Using Conical S ~ r i n gDesign Theorv with Planar S ~ i r a l Springs The computational procedures for the design of conical springs can be tedious as they involve an iterative process to find the best geometry for a given application. Using these procedures in the Also, by

design of planar spiral springs would be no less tedious.

looking at the equations it is not immediately clear what the relationships are between geometric parameters such as thickness and coil width and response characteristics such as spring rate and stress distribution. It should be noted that inspite of the similarities between conical springs and planar spiral springs there are significant differences. The equations have been developed for conical springs This allows the ends of the springs

loaded between two flat plates. to rotate freely.

Planar spiral springs, on the other hand, have their

coiled ends fixed at the outer ring and symmetry at the center controls rotations there. Furthermore, as conical springs are usually wound using round wire the maximum stress is found at the inside of the largest coil. planar spiral spring has a coil whose cross-section is that of a flat rectangle. That is, the long sides of the rectangle are perpendicular

to the spring axis.

A spring coiled in this way may have the

maximum working stress occur at the inside of the coil or at some point along the sides (25). Knowing where the peak stresses will occur can be helpful when specifying manufacturing processes to ensure that a quality surface finish, which improves fatigue life, will result at that location. The usual design equations do not give the

location of the maximum stress for a rectangular cross-section coil and an exact analysis which also takes into account the curvature effects of the coil involves elasticity theory and is quite complicated
(25).

Hence, the design theory for conical springs may prove insightful as to how planar spiral springs will behave but it cannot meet all of ones needs. Hypotheses In order to understand the relationship between the geometric parameters of a planar spiral spring and its performance characteristics some form of analysis is necessary. The most easily

controllable geometric parameters of greatest interest are thickness and coil width (which is controlled by the width of the removed material). With the results of a careful analysis it can be determined

whether planar spiral springs can meet the design criteria for use as displacer springs in Stirling engines. For this analysis two

hypotheses are put forth: (1) planar spiral springs are a viable design idea for application as displacer springs in free-piston Stirling engines, and (2) the working envelope for a specific type of planar spiral spring can be established with reasonable confidence without an extensive test program. Closure Mechanical springs can be defined by their ability to store and release energy. The process of spring design centers on finding the

three main aspects of spring performance: the force-deflection curve, the maximum stress intensity and the energy storage capacity.
A spring's rate, which is important in applying vibration theory

to spring-mass systems, is obtained from the force-deflection curve. The maximum stress intensity along with the number of loading cycles will determine the fatigue life of a spring. of a spring will also influence its fatigue life. The surface finish

The energy storage

26
capacity of a spring can be calculated using the force-deflection curve and the maximum deflection information. In appearance, planar spiral springs resemble conical springs in parallel. Inspite of the similarities there are enough differences such

that using the design equations for conical springs on planar spiral springs may not be judicious. can be quite tedious to use. In order to get an understanding of the relationships between the geometric parameters of a planar spiral spring and its performance characteristics an analysis of the design is necessary. Chapter 3 will discuss the method used to make such an analysis. These equations, should they apply,

CHAPTER 3: Methods

Selection of Analysis Method As discussed in Chapter 2, even though planar spiral springs share similarities with conical springs, it is not clear that the spring design equations for conical springs can be used to design a planar spiral spring. Regardless of the applicability of these equations to

planar spiral springs, they tend to get quite tedious, especially as the number of approximating arcs to the actual coil shape increases. Also, by looking at the equations it is not clear what the relationships are between geometric parameters and spring performance. A rigorous and thorough test program on a statistically significant number of planar spiral springs would generate data for checking whether planar spiral springs will meet the design criteria for displacer springs in Stirling engines. This data could also be used

to make graphs that would demonstrate relationships between design parameters and spring behavior. Such a test program,

however, would be time consuming and expensive as the commonly encountered large variation in results when testing for fatigue (5) would require a large number of tests. Also, the testing would

probably need to include tests such as photoelastic coating tests to determine maximum stress locations. Finite element analysis makes a more attractive, time efficient initial approach to this design analysis problem. Using finite element

models, a number of design parameters can be evaluated much more quickly than can be done through testing alone. Finite element

analysis does not supercede the testing program, it still should be

28 done, but the viability of planar spiral springs as a design concept can be largely substantiated using finite element analysis. Eauipment and Software The PATRAN software package by PDA Engineering (versions 2.4 and 2.5) was used to generate the finite element model for the spring and the finite element code ABAQUS by Hibbitt, Karlsson and Sorensen, Inc. (releases 4.8) was used for analysis. Results from

ABAQUS were output as data reports and were also post-processed using PATRAN. PATRAN was run on a Micro-VAX I1 host processor through an Intergraph Corporation Interpro 220 workstation. ABAQUS was run on a VAX 751 computer for small models and on a CRAY Y-MP8/864 for larger models. With the exception of the CRAY,

all of the software and equipment used are either owned or licensed by Ohio University. The Ohio Supercomputing Center's CRAY was

accessed remotely from the Ohio University campus. Data for generating force-deflection curves for actual springs was obtained using a testing rig designed and built by Sunpower Incorporated, Athens, Ohio. rig. The Appendix Figure 9 shows a schematic of the testing

contains a picture and a complete description of

the testing rig. Finite Element Model Selection of the proper element is the important first step in finite element modeling (8). Because the coil radii increase from the

center outward and because the ends are fixed, bending as well as torsion and shear forces are present in the coils. Shell theory based

elements are generally capable of modeling these combined forces

30 and shell elements are well suited for the geometry of planar spiral
springs in both the undeformed and deformed (i.e. unloaded and loaded) states. The material cuts in a planar spiral spring and the small center hole unavoidably introduce geometric discontinuities. Since any

discontinuity in a machine part alters the stress distribution in the vicinity of the discontinuity (22), stress concentrations were expected. Stress concentrations, having highly localized effects (22), If constant-

can cause steep stress gradients in these local areas.

stress elements such as four node quadrilaterals were used, an exceptionally fine mesh would be required for an accurate solution to such stress fields (8). Consequently, higher order quadratic elements

were used which gave linear instead of constant stress distributions across elements. In particular, the shear flexible S8R5 eight node

quadrilateral shell element in the ABAQUS element library was chosen. For these elements ABAQUS computes stresses at the upper

and lower surfaces of the shell (12). As the finite element models were refined, regions of low or constant stress were discretized using relatively large elements resulting in a coarse mesh. In regions having stress concentrations,

relatively small elements were used resulting in a fine mesh. Between these regions of coarse and fine meshes a transition in mesh refinement naturally had to exist. The transitions were accomplished by the use of multi-point constraints (MPC's). MPC's are a technique whereby the position of a The

node can be restrained to a line defined by two or more nodes. application of MPC's to the models used in this study are best

illustrated with an example. that form a transition.

31 In Figure 10 are shown three elements

Elements 2 and 3 are clearly topologically incompatible with element 1. An MPC can be used to ensure that nodes a and b, the mid-side nodes of elements 2 and 3, remain on the line or curve defined by nodes A, B, and C. Nodes a and b are allowed to "slide" along the ABC curve but they they must remain on that curve. constraint avoids what would be a compatibility problem at the boundary of elements 1, 2, and 3 as illustrated in Figure 11. A transition of this type without the use of MPC's could result in interelement gaps or overlaps which could affect the computed solutions (12, 18). The displacer rod in a Stirling engine is attached to the central hole of a planar spiral spring. The force transmitted by the displacer This

rod to a planar spiral spring was modeled as lumped forces equally distributed among the nodes on the edge of the center hole. This a d

hoc lumping results in mid-side nodes and corner nodes carrying equal loads which will result in less accurate solutions for coarse meshes. But with mesh refinement there will be convergence toward

the correct solution (2). A planar spiral spring is intended to operate with the outside edge of the spring fixed in place. This boundary condition was

modeled by constraining all of the element corner nodes on the outside edge of the spring in all six degrees of freedom. mid-side nodes on the outside edge were not constrained. The element The intent

of this approach was to give a slight amount of flexibility for deformation to the element edges forming the spring outside edge.

@
C
Figure 10: Transition Mesh

Figure 11:

Deformed Transition Mesh

33
It was felt that this would better represent the actual boundary conditions of the spring than would a fully-fixed outside edge. Young's modulus and Poisson's ratio for steel are, to a certain degree, dependent upon the alloy considered. As there does not

appear to be any reason that planar spiral springs must be limited to any particular alloy of steel, the widely used "mid-range" values of 200 GPa for Young's modulus and 0.30 for Poisson's ratio were used in this study. Analysis Approach Before any confidence can be placed in the solutions resulting from a finite element analysis, the validity of the finite element model must be proven (19). this. There are various techniques for doing

One is to compare the finite element solution with the analytical Perhaps the best

results available for a comparable problem.

validation technique is comparison of the finite element solution with results derived empirically (10). Another technique is to do a

convergence study, sometimes called a mesh refinement study, by comparing a number of trial solutions each with a different number of degrees of freedom in the model. If these trial solutions are

plotted, a converging curve can be made for a model that is sufficiently and properly refined (10, 19). two validation techniques were used. A set of planar spiral springs comprising five different geometries were manufactured by Sunpower, Incorporated of Athens, Ohio. In the laboratory facilities of Sunpower and using a In this study the latter

testing rig designed and built by Sunpower (see Appendix), each spring was subjected to a series of incremental loads and the

34
resulting deflections were recorded. This data was used to develop the force-deflection curve for each spring geometry in the set. PATRAN geometric models of these lab tested springs were developed and finite element models were generated using the geometric models. The finite element models were then translated load cases per

into ABAQUS input files. Using at least three different

model, a linear static analysis of each load case was run using ABAQUS. The computed deflections were then compared to the lab When the

results to verify the finite element model solutions.

computed results deviated from the lab results the finite element model was refined and the load cases were rerun. For three of the five experimental spring geometries, finite element model convergence studies for both deflection and maximum Von Mises stress were conducted. In refining the meshes

for these convergence studies, care was taken to ensure that all of the coarser meshes were included in the more refined meshes of each model. This along with using the same order of elements are

necessary in order for the sub-division process to converge to the exact solution (6, 10). Since the variation of stress within an element is of one order less than displacement, a model predicting an accurate displacement solution may indicate an inaccurate stress solution (8). Consequently,

a second test to ensure accurate stress solutions was used. For S8R5 eight node quadrilateral reduced integration shell elements, ABAQUS computes the stresses at the four Gaussian integration points. The stresses at the nodes of an element are then If a node

computed by extrapolating out from the integration points.

35 is shared by two elements then there will be two computed stresses

at that node, one from the extrapolation in the first element and another from the extrapolation in the second element. The computed

stresses at a node shared between two elements are, in general, not the same. The difference in the computed stresses for a node is the

inter-element stress jump at the given node. In developing finite element models in this study, a mesh was refined until the inter-element Von Mises stress jump at any shared node between elements was less than some A computed as:
A = omax - omin
oavg

A value of 4% was chosen for this A as small enough such that when the nodal stresses computed for each element sharing a given node were averaged, the resulting nodal stress average would be within 112 A or 2% of the exact solution. This would seem to be sufficiently accurate for design engineering purposes considering other variations inherent in making planar spiral springs such as the variation in material thickness, warpage or distortion due to heat treatment and imperfections in geometry from the cutting or stamping process. Also, the inherent scatter in fatigue test results

mandates a sizeable factor of safety if the goal is infinite life (22, 25). Thus, 2% error is not out of proportion in relation to other design considerations involved in designing and making planar spiral springs. When modeling the lab tested springs the actual measured material cut widths were used. On the other hand. the measured

thicknesses were compared to the thicknesses for various standard

36
gages of sheet steel (16) and the thickness of the closest matching gage steel was used. This was done because standard gage sheet

steel had been used to make the springs and the measured thicknesses were suspect to error due to machining burrs and warpage. With the modeling approach for planar spiral springs verified, a number of standard gage thicknesses and material cut widths were analyzed. Solutions for deflection, maximum Von Mises stress and

the distribution of Von Mises stress within a planar spiral spring were computed and recorded. Closure There are several possible methods that may be used to determine whether planar spiral springs will meet the design criteria for displacer springs in Stirling engines. Design equations for conical A rigorous

springs are not fully suitable for planar spiral springs.

testing program of a statistically significant number of planar spiral springs would be time consuming and expensive. Finite element

analysis is an attractive approach because a number of parameters can be tested in a timely fashion. Because of the anticipated forces in a planar spiral spring under load, shell elements were used in the finite element model. geometric discontinuities existed, steep stress gradients were also anticipated hence quadratic order elements were used which gave linear element stress distributions. Model validation is very important in order to justify confidence in any finite element solution. Both convergence studies As

and comparison with experimental data were used to validate the

37
models. After validating various models, analysis runs were made for several load cases for each model to determine the forcedeflection characteristics, maximum Von Mises stress levels and the Von Mises stress distributions in planar spiral springs. With a plan for model creation and validation in hand, the test work was scheduled and the geometric and finite element modeling began.

CHAPTER 4: Results

Ex~erimental Results Figures 12 through 16 show the force-deflection curves for the five spring geometries tested. Two springs were used to generate

Figure 12, six springs for Figure 13, one spring for Figure 14, ten springs for Figure 15 and three springs for Figure 16. Each of the five figures shows a straignt line fitted to the data. The computed slope of each line, which corresponds to the spring constant, k, of each spring is also shown. range from 1.90 N/mm to 5.95 N/mm. Finite Element Solutions for Deflection The meshes in the finite element models used to model the experimental springs varied in the number of elements and nodes depending upon the spring being modeled. As few as 1872 nodes These spring constants

and 518 elements to as many 5164 nodes and 1570 elements were used. In all of the finite element analysis runs, the node with the

maximum computed deflection was not one of the loaded nodes around the central hole but was at a node on the outer edge of where the spring coil connects to the central disk. The difference between

the maximum nodal displacement and the displacement at the center hole was always less than 1%. Since deflections in the lab were measured relative to the edge of the small center hole, finite element results for nodes at this same position were used in place of the maximum nodal deflections. Figures 17a and 17b show the finite element mesh for a model of a spring with a material cut width of 1.71 mm. Figure 17a shows

Force-Deflection Curve Laboratory Results

40

W(cut) = 1.71 mm W(coil) = 4.35 mm t = 1.062 mrn (1 9 gage) 0

4 6 Deflection, mm

10

Figure 12

Force-Def lection Curve Laboratory Results

k = 1.90 N/mm

W(CU!) = W(coil) = t = 0.836 10

4 Deflection,

mm

Figure 13

Force-Deflection Curve Laboratory Results

= 2.18 mm = 4.03 mrn

14 rnm (18 gage) 0

Deflection,

mm

Figure 14

Force-Def lection Curve Laboratory Results

k = 3.92 Nlmm

W(cut) = 2.20 mm W(coil) = 4.00 mrn t= 1.062 mm (19 gage)

Deflection,

mm

Figure 15

Force-Def lection Curve Laboratory Results

k = 2.75 NImm

W(cut) = 3.09 mm W(coil) = 3.16 rnm t = 1.062 mrn (19 gage)

10

4 6 Deflection, mm

Figure 16

17a: Finite Element Mesh

17b: Finite Element Mesh Closeup

43 a fully meshed spring and Figure 17b shows a close-up of the refined mesh area in the lower right of Figure 17a. Figures 18a and 18b

show the mesh for a spring with a material cut width of 3.09 mm for comparison. Figures 19 through 23 show the force-deflection curves of the five sets of lab tested springs with the results from the finite element analysis runs also plotted. The spring constant, determined

from the finite element results are shown along with the spring constant derived from the lab results. Figures 19, 20 and 23 show

that the finite element results predict a higher spring stiffness than that determined experimentally. For the spring of Figure 20 the

finite element stiffness is 15% greater than the lab determined stiffness. Figures 24, 25 and 26 show the results of convergence All three studies

studies for deflection for three different springs.

were done with a force of 5 Newtons and the results for lab tests at a force level of 5 Newtons are also shown. Figures 27 through 30 show the force-deflection curves for four spring geometries not tested in the lab. straight lines fitted to the data are shown. The computed slopes of

Just as in Figures 12

through 16, these line slopes also represent the spring constants. Figure 31 shows the relationship between spring thickness and deflection for four springs with different widths of material cut. All

four curves were generated using data from finite element analysis runs using a force of 5 Newtons.

18a: Finite Element Mesh

18b: Finite Element Mesh Closeup

Force-Deflection

Curve

k l a b = 4.19 kFEA=4.53

N/mm Nlmm

Experimental

FEA Results
0 2 4 6

W(cut) = 1.71 mm W(c0il) = 4.35 mm f = ( I 9 gage) Mesh SKID

10

Deflection,

mm

Figure 19

Force-Def lection

Curve

klab = kFEA=

1.90 Nlmm 2.1 9 Nlmm

Experimental FEA Results 0 2

4

W(cut) = 1.84 mm W(coil) = 4.32 mm t = 0.836 mm (21 gage) Mesh S121 10

Deflection,

mm

Figure 20

Force-Def lection

Curve

k l a b= 5.95 kFEA=5.81

Nlmm Nlmm

W(cut) = 2.18 mm Experimental

FEA Results
0

W(s~line)= 4.03 mm t = 1.214 mm (18 gage) Mesh SJl

6 mm

Deflection,

Figure 21

Force-Deflection

Curve

klab =

3.92 Nlmm

kFEA=3.95

Nlmm

W(cut) = 2.20 mm W(coi1) = 4.00 mm t 0 2 4 6 Deflection, mm 8

=: 1.062 mm (1 9 gage) Mesh SGl D

10

Figure 22

Force-Def iection

Curve

klab

= 2.75 Nlmm

kFEA =

2.89 Nlmm

FEA Results
0 2
4 6 Deflection,

W(cut) = 3.09 mm W(coil) = 3.16 mm t = 1.062 mm (19 gage) Mesh SH7D 12

10

mm

Figure 23

41

I I

Convergence Study for Deflection

FEA Results Lab Data scale

0

+.
0

W(cut) = 1.71 rnrn t = 1.062 rnrn (19 gage) Force = 5 N Meshes SKxD

1000 2000 Number of Nodes

3000

Figure 24

Convergence Study for Deflection

E
0 .0

4b

m-

1-

. c

a J -

n
0; 0

Q)

FEA Results Lab Data scale

I

W(cut) = 2.20 rnm t = 1.062 rnrn (19 gage) Force = 5N Meshes SGxD

1000

2000

3000

Number of Nodes

Figure 25

2.0 1.8

Convergence Study for Deflection

w

0 .C

1.6

W ( C U ~ ) 3.09 mm = t = 1.062 rnm (19 gage) Force = 5 N Meshes SHxD

0 Q
Q)

1.41.21.o 0

FEA Results
Lab Data
l ' l ' l ' l ' l '

1000

2000

3000

4000

5000

6000

Number of Nodes

Figure 26

Force-Def lection
40

Curve

W(cut) = 1.984 mm W(coil) = 4.29 mm t = 1.062 mm (19 gage) Mesh SB1 D 0

4 Deflection,

mm

Figure 27

Force-Def Iect ion

40

Curve

10

0 0 2 4
6

W(cut) = 2.381 mm W(coil) = 3.878 mm t = 1.062 mm (19 gage) Mesh SC8D 8

Deflection,

mm

Figure 28

Force-Def lection Curve FEA Results

40

W(cut) = 2.778 mm W(coil) = 3.46 mm t = 1.062 mm (19 gage) Mesh SD2D 0 2 4 6 Deflection, mm

10

Figure 29

W(cut) = 3.969 mm W(coil) = 2.14 mm t = 1.062 mm (19 gage) Mesh SF2D 10 Deflection, 20

mm

Figure 30

Deflection vs. Thickness

15

W(cut) = 3.09 mm

E
10
0 .0 C

W(cut) = 2.20 mm
0

W(cut) = 1.71 mm

al

al

Force = 5 N Meshes SKI ,SG1 ,SH7, S F2

0
0.6

0.8

1 .O Thickness, rnrn

1.2

1.4

Figure 31

53 Figure 32 shows the relationship between the width of material

cut and deflection.

The relationship is for a material thickness of

1.062 mm and a load of 5 Newtons. Finite Element Solutions for Stress Figures 33, 34 and 35 show the results of convergence studies for maximum Von Mises stress for three spring geometries. These

three geometries are the same as those used in the convergence study for deflection shown in Figures 24, 25 and 26. The

convergence studies for maximum Von Mises stress were also done with a force of 5 Newtons on each spring. Note that these and all

other computed stresses are at the surface of the shell elements. The Table shows a compendium of all of the analysis runs made using fully refined finite element meshes with a load level of 5 Newtons. The Table includes the number of nodes in each model, the

maximum average nodal Von Mises stress and the percent Von Mises stress jump at the node where the maximum was computed. Figures 36 through 44 show the relationship between applied force and the maximum Von Mises stress produced in a spring. Figures 36, 37, 39, 40 and 43 correspond to the spring geometries tested in the lab. Figures 38, 41, 42 and 44 are the results for

additional geometries. Figure 45 shows the relationship between spring thickness and the resulting maximum Von Mises stress for a force of 5 Newtons. The four widths of cut of material are the same in these figures as Figure 3 1.

Deflection vs. Width of Material Cut

3

t = 1.062 mm (1 9 gage) Force = 5 N

Meshes K,B,G,D,H,F

Width of Cut, mm

Figure 32

Convergence Study for Maximum Von Mises Stress

W(cut) = 1.71 mm t = 1.062 mm (19 gage) Force = 5N Meshes SKxD

0 1000 2000 3000

Number of Nodes

Figure 33

Convergence Study for Maximum Von Mises Stress

0

120 100-

5
V) V)

z
V) V)

80-

.-

5 s

>

40 20 60 01 ' 0
I
I

W(cut) = 2.20 mm t = 1.062 mm (19 gage) Force = 5 N Meshes SGxD

1000 2000 Number of Nodes 3000

Figure 34

Convergence Study for Maximum Von Mises Stress

n
m

120

100

W(cut) = 3.09 mm
t = 1.062 rnm (19 gage) Force = 5 N

Meshes SHxD

Number of Nodes

Figure 35

Model

Wcut

(mm)

thickness (mm)

nodes

(MPa) (Von Mises)

omax

%OjUmp

SKlAl 1872 1.71 0.607 0.24 294.5 1.71 1872 SKlBl 0.759 0.05 188.3 1.71 1872 SKlIl 0.836 0.19 155.6 1.71 1872 SKlCl 0.15 131.2 0.9 12 1.71 1872 SKlDl 0.4 97.65 1.062 SKlEl 1.214 - - - - - - - - - 1- - -1.71- - - . . . . . . . . . .1 . . .1.8 7 . . .1 . . . . . 75.68. . . . . .0.54 1 --..2 ..... .. SI211 - - - . - - - - - - - -0.836 - - - -4- - . - - - - -168.9- - - - - - - - - - - - - - - 1.84 484 0.06 ----------. SBlDl 1.984 ( 1.062 1898 99.10- - - - 0.5 S-J- ~ -E ~-= - - - - - - - . - - - 1.214 - - - - - - - - - - - - - - - -7 6 .-7-0- - - - - -6- - --3 --- - -- . -----2140 0.607 2.20 SGlAl 2.19 296.6 2140 2.20 0.759 2.60 191.9 SGlBl 2.20 2140 2.82 159.5 0.836 SGlIl 2140 2.20 0.912 2.89 135.2 SGlCl 2140 2.20 1.062 3.1 101.6 SGlDl 2 SGlEl - - - - - - - - - - - - 2.20 - - . . - - 1.214- - - - - - -1-4-0 - - - - - - 79.48 - - - - - 3.33 -------------101.1 2.381 - - - - -1.062- - - - - - - - - - - - - - - - - - - - - - - - - - - --------- ---------SD2D1 - - - - 2.778 - . , - -1.062- - - - - - - - - - - - - - - 107.7 - - - - - -0.28- --------------------1.12 5164 0.607 SH7A 1 3.09 337.9 5164 3.09 0.47 214.8 0.759 SH7B 1 3.09 5164 0.836 0.51 SH7I1 177.9 5164 3.09 0.47 150.0 0.912 SH7C1 5164 3.09 0.09 113.6 1.062 SH7D1 5164 3.09 SH7E - - - - - - - 1 - - - - - - - - - - . . - - -1.214- - - - - - - - - - - - - - - - -89.92- - - - - -0.11 --------0.74 4152 0.607 3.969 486.3 SF2A1 4152 3.969 321 .O 0.56 0.759 SF2B 1 4152 3.969 0.5 6 269.4 0.836 SF21 1 4152 3.969 0.52 2 3 0.2 0.912 SF2C1 4152 3.969 0.69 174.7 1.062 SF2D 1 4152 3.969 1.214 0.66 137.0 SF2E 1

-1

1-

:::; f

Table:

Compendium of Finite Element Results

Force vs. Max. Von Mises Stress

40

10 slope = 0.051 2 NIMPa

W(cut) = 1.7'1 mm t = 1.062 mm (19 gage) Mesh SKID 600

100

200

300

400

500

Max. Von Mises Stress, MPa

Figure 36

Force vs. Max. Von Mises Stress

20

slope = 0.0296 NIMPa 0 0 100 200 300 400 500 W(cut) = 1.84 mm t = 0.836 rnrn (21 gage) Mesh S121 600

Max. Von Mises Stress, MPa

Figure 37

Force vs. Max. Von Mises Stress

40

10 slope = 0.0505 N/MPa 0 0 100 200 300 400 500

W(cut) = 1.984 mrn t = 1.062 mrn (19 gage) Mesh SB1 D 600

Max. Von Mises Stress, MPa

Figure 38

Force vs. Max. Von Mises Stress

W(cut) = 2.18 mrn t = 1.214 mrn (18 gage) Mesh SJ1 E 0 100 2 00

300

400

500

Max. Von Mises Stress, MPa

Figure 39

Force vs. Max. Von Mises Stress

10 slope = 0.0492 N/MPa 0 0 200 400 600

W(cut) = 2.20 rnm t = 1.062 mm (1 9 gage) Mesh SG1D 800

Max. Von Mises Stress, MPa

Figure 40

Force vs. Max. Von Mises Stress

40

10 slope = 0.0495 N/MPa 0 0 200 400 600

W(cut) = 2.381 mm
t = 1.062 mm (19 gage) Mesh sc8D

800

Max. Von Mises Stress, MPa

Figure 41

Force vs. Max. Von Mises Stress

40

slope = 0.0464 NIMPa W(cut) = 2.778 rnrn t = 1.062 mrn (19 gage) Mesh SD2D

Max. Von Mises Stress, MPa

Figure 42

Force vs. Max. Von Mises Stress

40

slope = 0.0440 N/MPa

W(cut) = 3.09 rnm t = 1.062 rnrn (19 gage) Mesh SH7D 800

200 400 600 Max. Von Mises Stress, MPa

Figure 43

Force vs. Max. Von Mises Stress

40

slope = 0.0286 N/MPa

W(cut) = 3.969 mm t = 1.062 mm (19 gage) Mesh SF2D

1200

200

400

600

800

1000

Max. Von Mises Stress, MPa

Figure 44

Max. Von Mises Stress vs. Thickness

500

Force = 5 N Meshes SKI ,SG1 ,SH7, S F2

Thickness,

mm

Figure 45

64 Figure 46 shows the relationship between the width of material cut and maximum Von Mises stress. The results are for a spring with

a thickness of 1.062 mm and a load of 5 Newtons. Figure 47 shows the Von Mises stress contours for a spring with a material cut width of 1.71 mm, a thickness of 1.062 mm and a load of 5 Newtons. Figures 48 through 51 show closeup views of meshes in the area of stress concentration (see Figures 17 and 18) of four geometries. The material width cuts were 1.71 mm, 2.20 mm, 3.09 For a material thickness of 1.062

mm and 3.969 mm, respectively.

mm and a load of 5 Newtons the maximum Von Mises stress was computed for each and the node associated with the primary stress peak indicated. indicated. Figures 52 through 57 show the Von Mises stress contours for a spring with a material cut width of 1.71 mm and loaded with a 5 Newton force. Each figure is for a different thickness of spring. The secondary stress peak, when it occurred, is also

Figures 58 through 63 show the same sequence of material thicknesses but for a spring with a material cut width of 3.09 mm and a 5 Newton force. Closure Presented are the results from lab tests of five geometries of springs. Finite element model convergence study results are also

presented along with finite element analysis results for deflection and maximum Von Mises stress. The relationships between force,

material thickness and the material cut width and deflection and Von Mises stress are shown graphically and in contour plots.

Max. Von Mises Stress vs. Width of Material Cut

180

100 80 1

t = 1.062 mm (1 9 gage) Force = 5 N

Meshes K,B,G,D,H,F

2 3 Width of Cut, mm

Figure 46

CHAPTER 5: Discussion

Experimental and Finite Element Force-Deflection Results The spring constants of the experimental springs ranged from a minimum of 1.90 N/mm for a spring with a material cut width of 1.84 mm and a thickness of 0.836 mm, to a maximum of 5.92 N/mm for a spring with a material cut width of 2.18 mm and a thickness of 1.214 mm. The experimental results shown in Figures 12 through 16 This is

show a linear force-deflection curve for planar spiral springs.

reasonable because all coils remain active through the operating deflection distance. The force-deflection curves resulting from the finite element analysis runs are consistent in showing a linear deflection response to an applied force for all five geometries of springs tested. The

computed spring constants differed from the lab results by +8.1%, +15.3%, -2.3%, +0.8% and +5.1% as shown in Figures 19 through 23, respectively. The springs where the computed spring constant and

the experimental spring constant vary by -2.3% and +0.8% indicate a good correlation between the finite element model and the actual spring in terms of gross deflection. For the three spring geometries where the finite element spring constant deviates from the associated experimental spring constant by more than 5%, the finite element results are consistently higher. That is, the spring of the finite element model is stiffer than This is a characteristic of results using finite Because the

the actual spring.

element code which is based on displacement functions.

number of degrees of freedom, which is related to the number of

84 nodes, are fewer in a finite element model than actually exist, the finite element model is always stiffer than the actual part being modeled (6, 7). The deviations of 8.1% and 15.3%, though sizeable, do not necessarily indicate finite element models of marginal quality and/or of marginal use. The geometry of the finite element models, and in

particular the width of the material cuts, were measured by hand using a caliper. each spring. The measurements were only taken at one place on

Since all of the springs had been handled and many of

them tested or used to one degree or another they may have been slightly deformed. The result could have been that the measured

material cut widths do not reflect the true cut width of the actual spring. This would be important because the geometric model used

to make the finite element model was based on the measured dimension. true width, If the measured material cut width was smaller than the the resulting spring coil would be wider and hence a

stiffer spring model than the actual spring. Consequently, the two finite element models that deviate by 8.1% and 15.3% from the experimental results may actually be giving good results for the geometries modeled. modeled may, in fact, not match experimental springs. However, the geometries

the geometries of the actual

Because of this possibility it is necessary to

use convergence studies for deflection to help substantiate the validity of the finite element models and their results. The convergence studies shown in Figures 24, 25 and 26 were done using finite element models of experimental springs all having

the same material thickness, 1.062 mm.

85 In each of the three analysis

runs on each spring the applied load was the same, 5 Newtons. In all three convergence studies the finite element results show good stability. The percent change from the coarsest meshes to the This

most refined meshes were 5.0%, 3.7% and 0.0%, respectively.

stability indicates that the results will not improve appreciably regardless of how much more mesh refinement is done. Because of

this stability and because the spring constant results of three of the models were within 5% of the experimental results, it is concluded

that the general modeling technique used is viable for determining deflections in planar spiral springs. The force-deflection curves of Figures 27 through 30 are the results of finite element analysis runs for additional spring geometries. All four of these springs were modeled with a thickness

of 1.062 mm, only the material width cuts varied. The curves in Figure 31 show that there is a nonlinear relationship between deflection and spring thickness. As the

thickness increases the stiffness of the spring increases thus reducing deflection for a given load. This nonlinear relationship is reasonable

as the stiffness of any part in either bending or torsion is related in a non-linear way to geometry. That is, stiffness in bending is related

to the second moment of area or moment of inertia, I, which for a beam of rectangular cross-section is given as 1/12(bh3). For a rectangular cross-section in torsion the stiffness is given as where

pbt

(G/L)

p is a coefficient related to the ratio of width to thickness in a

nonlinear fashion.

86 The possible material cut widths range from zero, or no cut, to

a maximum of when the adjacent material cut paths just intersect. Figure 32 shows a nonlinear relationship between the width of the material cut and deflection for a given material thickness and load. As the width of cut increases, the width of the resulting spring coil decreases along with the stiffness of the spring. As with the change

in thickness, the change in material cut width which controls the resulting coil width, is related to stiffness in a nonlinear way and hence the resulting curve. For a given desired spring constant there may be several geometries of planar spiral spring that will result in the necessary force-deflection response. Figures 19 to 23 and 27 to 32 can be used

to identify the spring geometries which will satisfy the spring constant requirement. The best geometry for a given design

situation can then be identified by considering load induced stress which is discussed presently. Finite Element Stress Results No experimental tests were conducted to determine stresses in a planar spiral spring. In order to validate the finite element models

for stress, convergence studies were done for three different geometries. Because combined loading necessitated the use of Von

Mises stresses, the convergence studies were done using the maximum computed Von Mises stress. In Figures 33, 34 and 35 the stability of the finite element solutions is clear. Using a spring thickness of 1.062 mm and a force

of 5 Newtons in all three studies, refinement of the mesh did not appreciably change the computed maximum stress values. The

87
percent change from the medium refined mesh to the finest mesh were -0.6%, 1.5% and 0.3%, respectively. This indicates that the finite

element stress solutions would not change any appreciable amount regardless of the amount of further mesh refinement. The Table shows the results of another type of study used to help validate the finite element models for computed stresses. For

each analysis run made, the node with the highest computed Von Mises stress contributed by any single element was identified. Then,

the largest difference in computed Von Mises stress between any two elements sharing that node was determined. This difference,

divided by the average of the Von Mises stresses of the node contributed by each element sharing the node, gave the percent maximum inter-element stress jump at the node. these The values of

computed stress jumps are shown in column six of the Table. All of the inter-element stress jumps were less than 4%. In

fact, with exception to the SG1 series of models (width of cut of of 2.20 mm), all of the stress jumps were only 1.12% or less. Such small

stress jumps in the area of greatest interest, that is, in the area of greatest stress, indicates a well refined mesh. The node with the highest average Von Mises stress was also identified. Without exception, this node was the same node that had The maximum average

the highest stress from any single element.

Von Mises nodal stress for each model loaded with a 5 Newton force is shown in column 5. To put these stress values into perspective, an AISI 1095 heat treated steel, quenched in oil and tempered, has a yield strength of

813 MPa (21).

Since the Von Mises failure criteria without regard to

88 fatigue is when the maximum Von Mises stress exceeds Q I ~ ( ~ Y ) , where

OY

is the material yield strength (22, 24), failure is expected in

this steel when the maximum Von Mises stress exceeds 383 MPa. Column 5 of the Table shows that the computed maximum Von Mises stress for model SF2Al of 486.3 MPa is the only value in excess of the failure limit. As this steel has a rather high strength,

such a thin planar spiral spring with such a wide material cut probably cannot be made to work in Stirling applications under loadings of 5 Newtons or more. Especially since fatigue loading

reduces the allowable working stress in the spring. The other spring geometries may or may not be feasible for Stirling engines depending upon the factor of safety desired with regard to the endurance limit. A modified Goodman diagram can be

constructed for this material and, with a factor of safety in mind, the viable geometries for a 5 Newton load can be identified. For other

loads, Figures 36 to 46 can be used to get the approximate maximum Von Mises stress for various geometries. The relationships between the maximum Von Mises stress, force and geometry are shown graphically in Figures 36 to 44. each case the relationship is linear. This is reasonable as shear In

stress, bending stress and shear stress due to torsion are related to load in a linear fashion. That is, shear stress is determined by the Bending stress is determined by Shear stress due to torsion

equation z=F/A, which is linear in F,

o=Mc/I where M=Fx which is linear in F.

is determined by z=Tr/J where T=Fx which is also linear in F. The curves in Figure 45 show a nonlinear relationship between the maximum Von Mises stress and spring thickness. In general, the

89 wider the material cut, the steeper the slope of the resulting curve for a given thickness. This nonlinear relationship is reasonable as

energy, which increases nonlinearly with thickness because the spring rate is nonlinear with respect to thickness (see Figure 31), is stored in the form of stress. Note how similar the curves are for widths of cut of 1.71 mm and 2.20 mm. Inspite of a 29% increase in material cut width, the

maximum Von Mises stress changes by less than 1% for t=0.607 mm and by less than 5.0% for t=1.214 mm. Figure 46 shows a nonlinear relationship between the width of the material cut and maximum Von Mises stress for a given material thickness and load. As with the change in thickness, the energy,

which increases nonlinearly with the width of material cut because the spring rate is nonlinear with respect to the width of material cut (see Figure 3 2 ) , is stored in the form of stress. The Von Mises stress contours of the active coils in Figure 47 show that in general the stress at a point on the inside edge of a coil is higher than the point directly across from it on the outside edge of the coil. The figure also shows that the stresses increase as the coil This is probably due to the fact that as

spirals from the center out.

the coil spirals out from the center, the moment arm from the center to the coil continually increases. Hence, the torque, and the

accompanying shear stress due to torsion, increase. In Figures 48 to 51 can be seen what the effect of changing the material cut width has on the position of the maximum Von Mises stress. For a narrow width cut as shown in Figure 48, the maximum

Von Mises stress or stress peak is on the edge of the semicircular end

90 of the tool cut path. In Figure 49 for a wider cut, the highest or primary stress peak is also on the edge but a secondary peak which is 91% in magnitude of the primary peak has formed to the interior face of the coil. The same is true for Figure 50 but the secondary Figure 51 shows that

stress peak is 98% of the primary stress peak.

for a wide material cut the primary stress peak has shifted to the interior face of the coil and the secondary stress peak, which is 75% of the primary stress peak, is now at the cut edge.

A possible reason for this stress peak shift may have to do with
the degree of abruptness of the change in section. That is, the

narrow cut width of Figure 48 causes a sudden change in geometry which forms a severe stress concentration. As the material cut width

is increased the radius of the cut edge increases and the change in geometry becomes more gradual. This reduces the severity of the

stress concentration until the stress peak due to the combined stresses on the rectangular cross-section coil exceeds the stress peak due to stress concentration. Figures 52 through 57 and 58 through 63 show the effect of material thickness on the Von Mises stress distribution in general, and the position of the maximum Von Mises stress in particular. For

the narrow material cut width of Figures 52 through 57, as thickness is increased the maximum stress declines but the position of the stress peak remains unchanged; the stress peak stays on the edge of the cut. Though the magnitude of Von Mises stress associated with

each stress contour changes as the thickness changes, the overall pattern formed by the contours is virtually constant.

91
For the wider material cut shown in Figures 58 through 63, changes in material thickness have a much more pronounced effect. The magnitude of the Von Mises stress peak declines as thickness is increased but the position of the stress peak shifts as well. In Figure

58 the region of highest stress is on the interior face of the coil and a
secondary stress peak exists on the edge of the cut. Figure 59 shows

that with an increase in thickness the relative difference in magnitude between the primary and secondary stress regions decreases. Figures 60 and 61 show a continuation of this trend as In Figures 62 and 63 the position of

material thickness increases.

the stress peak has shifted to the edge of the cut and the secondary stress region is now on the interior face of the coil. This shift in position of the primary stress peak is important because it may influence the machining and/or the heat treatment process used to make the spring. That is, different processes may be

necessary in order to ensure a quality surface finish at the different primary stress areas. Closure Both the experimental planar spiral springs and the finite element models of such springs had linear force-deflection curves which means that the spring rate for these springs is a constant.
A

variety of thicknesses and material width cuts were analyzed and the results are presented here graphically. These graphs may be

used either directly or by interpolation to determine the approximate geometry necessary for a desired spring constant. The maximum Von Mises stresses induced in planar spiral springs are linear with respect to induced loads. The magnitude of

92 the stresses are dependent upon material thickness and the material
cut width. The results of finite element analysis runs for a variety of

thicknesses and cut widths are presented in both graphical and tabular form. The graphs or tabulated results may be used either

directly or by interpolation to determine stress levels induced in planar spiral springs of various geometries. Stress contour plots indicate the areas where stress peaks generally occur in planar spiral springs. Attention during

manufacturing to ensure a quality surface finish in these general areas will improve fatigue life.

CHAPTER 6:

Conclusions and Recommendations

Conclusions 1. The finite element modeling technique used in this study is

acceptable for determining deflections in planar spiral springs.

2.

The finite element modeling technique used in this study is

also acceptable for determining Von Mises stresses in planar spiral springs.

3.

Planar spiral springs are a viable design idea for application

as displacer springs in free-piston Stirling engines. 4. The response characteristics of various geometries of planar

spiral springs can be identified using the finite element analysis results presented.

Note that conclusions 3 and 4 satisfy the two original hypotheses of: (1) planar spiral springs are a viable design idea for applications in Stirling engines, and (2) the working envelope for a specific type of planar spiral spring can be established with reasonable confidence without an extensive test program. Recommendations This study does not complete the necessary analysis work to be done on planar spiral springs. This study was limited to planar spiral Other

springs having only two coils both in the shape of involutes. numbers and shapes of coils remain to be studied.

The effect of the Also,

length of the coil on response characteristics is still unknown.

for a given coil length, what effect do the radii of the starting and

94 stoping points of the material cut tool path have on spring rate and
stress? Changes in material properties will have an effect on results as well. Thus, reanalysis using the modulus of elasticity and poissons

ratio of other potential spring materials is in order. As some Stirling engines operate at frequencies of as much as

60 Hz., a modal analysis of planar spiral springs is called for in order

to understand their behavior under such rapid cyclic loading. Certainly more test work is needed as well. Test work to

further substantiate the results of this and further studies and to prove the durability and reliability of planar spiral springs.

1. Associated Spring Corporation (1949). Spring Design. Bristol, CT: Author.

Handbook of Mechanical

2. Cook, R. D. (1987). Theory and General Remarks. In H. Kardestuncer (Ed.), Finite Element Handbook New York: McGraw-Hill. (pp. 2.109-2.114).

3. Cooke-Yarborough, E. H. & Yeats, F. W. (1975). Efficient ThermoMechanical Generation of Electicity from the Heat of Radioisotopes. Proceedings of the 10th Intersociety Energy

Conversion Engineering Conference, (pp. 1003-10 11).

4. Cooke-Yarborough, E. H. (1985). Metal Spring for Tuning and

Positioning the Displacer of a Short-Stroke Stirling Engine. Proceedings of the 20th Intersociety Energv Conversion Engineering Conference, (pp. 3.212-3.2 17).

5. Dally, J. W., & Riley, W. F. (1991). Experimental Stress Analysis.

New York: McGraw-Hill. 6. Desai, C. S. (1972). Introduction to the Finite Element Method; a Numerical Method for Engineerin? Nostrand Reinhold. 7. Desai, C. S. (1979). Elementary Finite Element Method. Cliffs, NJ: Prentice-Hall. 8. Dietrich, D., & Levy, A. (1987). FEM Modeling and Preprocessing. In H. Kardestuncer (Ed.), Finite Element Handbook (pp. 4.1534.17 1). New York: McGraw-Hill. Englewood Analysis. New York: Van

96 9. Fuchs, H. 0 . (1961). Volute Spring Design Data. In N. P. Chironis (Ed.), Spring Design and Application (pp. 100-101). New York: McGraw-Hill. 10. Grandin, Jr. H. (1986). Fundamentals of the Finite Element Method. New York: Macmillan. 11. Gross, S. (1966). Calculation and Design of Metal S p r i n ~ s . London: Chapman and Hall. 12. Hibbitt, Karlsson, & Sorensen, Incorporated (1989). User's Manual. Providence, RI: Author. ABAOUS

13. Joerres, R. E. (1986). Springs. In J. E. Shigley & C. R. Mischke (Eds.), Standard Handbook of Machine Design (pp. 24.1-24.68). New York: McGraw-Hill. 14. Keyser, C. A. (1980). Materials Science in Engineering. Columbus, Ohio: Charles E. Merrill. 15. McBride, J. R. (1984). The Homach TMG: A New Stirling Power Source for Unattended Operation. Proceedings of the 19th Intersociety Energy Conversion Engineering: Conference, (pp. 1843-1848). 16. McGannon, H. E. (1987). In E. A. Avallone & T. Baumeister I11 (Eds.), Mark's Standard Handbook for Mechanical Engineers (pp. 6.12-6.46). New York: McGraw-Hill. 17. Meriam, J. L., & Kraige, L. G. (1986). Engineering Mechanics: Dynamics. New York: John Wiley & Sons. 18. PDA Engineering (1990). PATRAN Plus User's Manual. Vol. 11. Costa Mesa, CA: Author. 19. Rizzo, A. R. (1991, May). Estimating Errors in FE Analyses. Mechanical Engineering, (pp. 61-63).

97 20. Roark, R. J. & Young, W. C. (1982). Formulas for Stress and

Strain. New York: McGraw-Hill.

21. Shigley, J. E., (1977). Mechanical

McGraw-Hill.

engineer in^

Design. New York:

22. Shigley, J. E., & Mischke, C. R. (1989). Mechanical Engineering

Design. New York: McGraw-Hill. 23. Society of Automotive Engineers (1982). Design and Application of Helical and Spiral Springs - SAE HS 5795. Warrendale, PA: Author.

24. Ugural, A. C., & Fenster, S. K. (1975). Advanced Strength and

A~~lied Elasticity. New York: Elsevier.

25. Wahl, A. M. (1963). Mechanical S~rinrrs. New York: McGraw-Hill. 26. Walker, G., & Senft, J. R. (1985). Lecture Notes in Engineering:
Free Piston Stirling Engines. Berlin: Springer-Verlag.

APPENDIX

The planar spiral testing rig was designed and developed by Sunpower Incorporated. It included a Sunpower manufactured

variable inductor type fast linear displacement transducer (FLDT) to measure displacement. An OMEGA LCL-20-20 load cell with a full

active temperature compensated bridge was used to measure load. Two FLUKE 3 112 digit multimeters provided digital output. The

signal conditioner and filtered differential amplifier were designed and built by Sunpower. See Figure 64.

Figure 64:

Planar Spiral Spring Testing Rig