Você está na página 1de 33

Residual Stresses and Strains in Cross-linked Polyethylene Power Cable Insulation

Lorant Olasz

Doctoral thesis no. 63 TRITA, HFL-0412 KTH Engineering Sciences SE-100 44 Stockholm, Sweden

TRITA HFL-0412 ISSN 1104-6813 ISRN KTH/HFL/R06/11SE

Akademisk avhandling som med tillst nd av Kungliga Tekniska H gskolan i Stockholm a o framl gges till offentlig granskning f r avl ggande av teknisk doktorsexamen tisdagen a o a den 13 juni kl. 10.15 i sal F3, Kungliga Tekniska H gskolan, Lindstedtsv gen 28, Stocko a holm.

Preface
When I started as a Ph.D. student at the Department of Solid Mechanics I was determined that to work with computer modeling without having to perform experiments, having done way too many experiments in the past. Luckily, during the course of the project I came to realize how important it is to base a model on experiments and to keep a sound balance between experiments and theory in building a model. I would like to thank my advisor prof. Peter Gudmundson for keeping me on track and for his continuous support and advice. Financial support for the project was provided by Fax nLaboratoriet, ABB e Power Technologies AB and Borealis AB. I am also grateful to the various people involved in the project for being available for questions and discussion. I specially want to thank Mr. Tommy Johansson, ABB Power Technologies AB, for his dedication and for always being available. I owe special thanks to Mr. Hans Oberg for his ideas and help with the lab equipment and experiments and to the staff in the machine shop for the quick and precise machining of various parts, but also for contributing with good ideas to solve experimental problems. To my colleagues I want to say thanks for interesting discussions and for making the workplace enjoyable.

Lorant Olasz April 2006, Stockholm

ii

List of appended papers


This dissertation consists of a summary and four appended papers:

Paper A:
Olasz, L. and Gudmundson, P. Viscoelastic model of cross-linked polyethylene including effects of temperature and crystallinity, Mechanics of Time-Dependent Materials, (Online First), 2006

Paper B:
Olasz, L. and Gudmundson, P. Prediction of residual stresses in cross-linked polyethylene cable insulation, Polymer Engineering and Science, 45(8):11321139, 2005

Paper C:
Olasz, L. and Gudmundson, P. A non-linear viscoelastic constitutive model for XLPE and its application in residual stress prediction for high voltage cable insulation, Report 410, Royal Institute of Technology, Department of Solid Mechanics, Stockholm, Sweden, 2006

Paper D:
Olasz, L. and Gudmundson, P. Residual strains in XLPE high voltage cable insulation, Report 411, Royal Institute of Technology, Department of Solid Mechanics, Stockholm, Sweden, 2006

iii

iv

Abstract
The subject of this thesis is modeling of the manufacturing process of high voltage power cables with the aim of predicting residual stresses and strains in the cable insulation. The studied material is cross-linked polyethylene (XLPE), which at room temperature is a semi-crystalline viscoelastic solid. Above the crystallization temperature the material exhibits a rubber type behavior due to the crosslinks. An extensive set of uniaxial tensile relaxation tests were used for the mechanical characterization of XLPE. These experiments were complemented by pressure volume temperature experiments as well as density and crystallinity measurements. Based on the experiments, initially a linear and later a non-linear viscoelastic power law model was formulated, incorporating temperature and crystallinity dependence. The non-linear model is based on the Schapery formulation. Evaluations of the model were performed with additional uniaxial experiments. These included comparisons between predicted stress responses and measured values during relaxation tests with transient temperature histories, during two step relaxation experiments and during uniaxial tests with constant strain rate loading. The initial modeling work focused on the prediction of residual stresses which develop during the cooling stage of the manufacturing process. As the constitutive model incorporates temperature and crystallinity dependence, the mechanical problem is coupled to the heat transfer and crystallization problems. Calculations were performed for a vertical manufacturing line. The effects of a viscoelastic material model are illustrated by a comparison to a stress state predicted by a thermo-elastic material model. A nal study concerns the modeling of the residual strains in the insulation. It was found that strains originating at cross-linking of the molecules play a signicant role for buildup of residual strains. Calculations are performed for the same vertical process line as before. Good agreement was found between predicted and experimentally obtained residual strains. Based on the residual strain state an estimate is made for the upper limit of shrink-back of the cable insulation.

vi

Contents
1 Introduction 2 Cable manufacturing process 3 Structure and properties of PE and XLPE 4 Thermal residual stresses 5 Constitutive modeling 5.1 Linear viscoelastic models . . . . . . . . . . . . . . . . . . . . . 5.2 Non-linear viscoelastic models . . . . . . . . . . . . . . . . . . . 1 1 2 4 5 5 7

6 Experimental characterization 8 6.1 Physical properties . . . . . . . . . . . . . . . . . . . . . . . . . 8 6.2 Mechanical testing . . . . . . . . . . . . . . . . . . . . . . . . . 8 6.2.1 Pressure Volume Temperature (pVT) measurements . . . . 9 6.2.2 Stress relaxation experiments . . . . . . . . . . . . . . . 11 6.2.3 Additional uniaxial tests . . . . . . . . . . . . . . . . . . 13 6.3 Constitutive model for XLPE . . . . . . . . . . . . . . . . . . . . 13 7 Process model 8 Discussion and outlook 9 Summary of appended papers References 14 15 18 20

vii

viii

1 Introduction
Cross-linked polyethylene (XLPE) is a widely used material for power cable insulation, where it has largely replaced oil impregnated paper. Although both the material and the process have been known and used for several years, there is a need in the industry for a deeper understanding of the manufacturing process and its inuence on the nal product properties. Mechanical, thermal and chemical modeling of the manufacturing process of a cable can be used as a tool for prediction and optimization of the nal product. The present work focuses on the mechanical modeling of insulation during the manufacturing process, namely on nding a suitable constitutive model for XLPE and utilizing it in prediction of thermal residual stresses and strains in the cable insulation.

2 Cable manufacturing process


By the process of cross-linking, polyethylene (PE) is changed from a thermoplastic into a single, interwoven molecule. Improved heat resistance is the most important advantage of XLPE over low density polyethylene (LDPE). In a power cable, the conductor temperature increases with the current transmitted through the cable, thus more power can be sent through an XLPE cable than a non-crosslinked cable of the same size. The most widely used method for cross-linking polyethylene in the cable industry is the continuous vulcanization (CV) method with peroxide additives. High temperature triggers the decomposition of peroxide into free radicals. The free radicals react with polyethylene molecules, extracting hydrogen atoms, thus forming a radical on the polymer chain which reacts with another polymer chain radical to form a cross-link. A typical manufacturing line for high voltage power cable utilizing crosslinked polyethylene (XLPE) is usually of two types: a vertical continuous vulcanization (VCV) line or a catenary continuous vulcanization (CCV) line (Boysen, 1970; Huotari, 1992; Huotari and Sistola, 1997). These differ in the physical location of the cable that is transported through the process. In a VCV line the cable is positioned vertically, whereas in a CCV line the cable is fed through the system with a slight downward slope. The important stages of cable manufacturing can be seen in the schematic of Figure 1. In a rst step of the continuous process, the polymer is extruded onto a metal conductor, after which heating of the cable takes place. During this stage, cross-linking of the molecules is achieved at temperatures above the melting point of the polymer. The last stage of the process is cooling of the cable to room temperature, achieved by circulating water or gas around the cable. Starting at the outer ra1

Conductor Die Extruder

Direction of cable travel through the process

Curing Heat activated cross! linking of PE molecules. Cable insulation Cooling Solidification and partial crystallization of the insulation

Figure 1. Schematic of power cable manufacturing process with its most important parts. dius and continuing inwards, partial crystallization of the insulation takes place. Cooling of polymer products is often a problematic manufacturing stage as it introduces residual stresses into the material. In particular, for high voltage cable insulation the knowledge of such stresses is desirable as it inuences both mechanical and electrical properties of the cable. It is widely known that electrical breakdown strength is inuenced by the mechanical stress. But electrical breakdown strength is also inuenced by the presence of voids or other defects, the formation of which is related to the buildup of residual stresses during manufacturing. During installation or repair of a cable, residual stresses may cause problems at cable joints or terminations, since the insulation tends to shrink back in the axial direction on the conductor. This phenomenon is commonly referred to as shrink-back in the industry and there is a maximum allowed level for it regulated by an international standard IEC (2004).

3 Structure and properties of PE and XLPE


The result of cross-linking in PE, a single large interwoven molecule called gel, makes up for most part of a cross-linked sample. XLPE used for high voltage cable insulation is usually made starting with LDPE and its microstructure and properties are similar to those of polyethylene in general. Polyethylene is a partially crystalline solid, whose properties are strongly in2

Lamellae Interlamellar links

Figure 2. A schematic representation of the structure of a spherulite. uenced by the relative amounts of crystalline and amorphous phases present. Density is strongly related to crystallinity, with higher density polyethylenes having higher degree of crystallinity. Other important characteristics inuencing the physical properties are basic molecular properties such as chain branching, average molecular weight and molecular weight distribution. The basic structure of PE is the chain (CH2CH2)n , which has no branches on the backbone. In commercial PE, n may be between about 400 and above 50000 and chain branches are usually present on the chain backbones. The smallest crystalline units, called lamella, are planar in shape and consist of chains that are perpendicular to the plane and fold back and forth, every 5-15 nm, although sometimes less frequently. Lamellae are interconnected by a few chains which pass from one lamella, through a small amorphous region, to another lamella. As it can be seen in gure 2, the lamellae are ordered into much larger spherically shaped units, called spherulites. The lamellae extend radially in the spherulite, with the molecule chains oriented in a direction that is perpendicular to the radius of the spherulite. Spherulites consist of both crystalline and amorphous parts and are interconnected by amorphous regions. Spherulites are thought to develop from an initial central lamellae, which subsequently branches to form the spherulite. It is also possible that spherulites grow as much that they impinge into each other, thus obtaining the shape of polyhedra. Density of a pure PE crystal has been calculated to 1.00 g/cm3 , whereas that of an amorphous phase is believed to be about 0.855 g/cm3 . LDPE contains short chain branches on the chain backbone. These short chain branches disrupt chain folding and thus reduce the degree of crystallinity to 403

60% from about 70-90% of unbranched PE. The crystalline melting point T m is also reduced to 105-115C from 135C of unbranched PE. Melting usually occurs over a wide temperature range, beginning at 70-80 C for a polymer of density 0.92 g/cm3 . The density is 0.912-0.935 g/cm3 compared to the higher 0.9600.970 g/cm3 of unbranched PE. A wide interval in degree of cross-linking and gel content is possible to achieve in cured XLPE samples depending on the process and base material. Gel contents close to 100% are possible to achieve, but for LDPE, which is the common base material for high voltage cables, crosslinked with dicumyl peroxide the gel content is more likely to be between 70% to 80%. According to several authors (Nilsson et al., 1998; Phillips and Kao, 1986; Gohil and Phillips, 1986a,b) the fully developed spherulite crystal structure is not preserved in cross-linked polyethylene. Instead XLPE was found to have what is described as a sheaf like microstructure. These sheaves look like undeveloped spherulites and are considerably smaller in size than spherulites in PE. It seems like the crystal development in XLPE stops at various stages before reaching the fully developed spherulites of PE. It is also reported that the microstructure changes with the degree of cross-linking. Gohil and Phillips (1986b) reported that high cross-link density results in parallel compact lamellar bundles and with decreasing cross-link density the lamellae have a tendency to splay apart and form sheaf-like structures. The nucleation density of these lamellar bundles was found to increase with cross-link density.

4 Thermal residual stresses


The problem of thermal residual stresses is found in a wide variety of polymer products where rapid inhomogeneous cooling is part of the manufacturing process. For power cables, cooling starts at the outer layers of the insulation where the temperature drops almost instantaneously from well above the melting point to room temperature. As a result the outer layers stiffen, but the inner layers still occupy a larger volume introducing stresses into the insulation. As cooling proceeds radially into the insulation, solidication produces further thermal stresses. Usually thermal stresses partially persist in the nished product since most polymers are viscoelastic. Furthermore, the stress development is also inuenced by the semi-crystalline microstructure of certain polymers like PE and XLPE. In the past, the development of thermal stresses has been treated mostly as a thermo-elastic problem and it has been recognized that the predicted stress state can be viewed as a limiting case to the actual stress state in polymers which show viscoelastic behavior (Struik, 1990; Djurner et al., 1980). Crystallization during cooling changes the density of the material and the con4

stitutive relationship. As a result of the rapid inhomogeneous cooling, the varying sub-cooling of the polymer melt at solidication inuences the degree of crystallinity and the the shape and size of the crystals. For the case of poly(etheretherketone) (PEEK) it is shown by Eduljee et al. (1994) that as crystalline regions grow during cooling, the elastic modulus of the polymer changes which inuences the development of the residual stresses. For this polymer it is also found that at high cooling rates crystallization is suppressed and thus the process becomes similar to amorphous polymers.

5 Constitutive modeling
5.1 Linear viscoelastic models
At innitesimal strain, linear viscoelasticity can be used for modeling of the mechanical behavior. For the uniaxial case, the normal stress ! is related to the normal strain " by the relaxation modulus, E(t), according to

! (t) =

t 0

E(t # )

d" d# . d#

(1)

Early on it has been recognized that characterization of the relaxation modulus can for many materials be simplied if isothermal relaxation curves at different temperatures can be brought together to form a master curve by shifting them parallel to the logarithmic time axis, (Morland and Lee, 1960; Muki and Sternberg, 1961; Lee and Rogers, 1965). This is called time-temperature shifting and the class of materials that exhibit this behavior is denoted as thermorheologically simple materials (TSM). Equation (1) changes now according to

! (t) =

t 0

E $ (t) $ (# )

d" d# , d#

(2)

where the reduced time $ is dened by

$ (t) =

t 0

a T (# ) d# .

(3)

In equation (3) the shift factor a depends on the instantaneous temperature. Materials in the TSM category are amorphous polymers like e.g. polymethylmethacrylate (Muki and Sternberg, 1961; Lee and Rogers, 1965). In general it can be said that polymeric materials whose behavior can not be described by equation (2) fall into the category of thermorheologically complex materials (TCM). For a class of TCM materials, master curves can be formed by applying both horizontal and vertical shifts to the relaxation curves (Harper 5

and Weitsman, 1985; Uesaka et al., 1989; Popelar et al., 1990). In this case the constitutive equation can be written as

! (t) = A T (t), % (t)

t 0

E $ (t) $ (# )

d" d# , d#

(4)

where A T (t), % (t) is the vertical shift factor which depends on temperature and other material parameters represented by % (t). Equation (3) also changes to

$ (t) =

t 0

a T (# ), % (# ) d# ,

(5)

in order to incorporate dependence of the horizontal shift factor a on other parameters. For paper, the need of vertical shifting has been attributed to the inuence of moisture content, which enters as a parameter into the shift function, (Uesaka et al., 1989). For semi-crystalline polymers like polyethylene it is believed that changes in crystallinity with temperature cause the need for both horizontal and vertical shifting of relaxation curves (Tobolsky and McLoughlin, 1955; Tobolsky, 1960; Popelar et al., 1990). As described by several textbooks on the subject (Fl gge, 1967; Tschoegl, u 1989), there are many ways to mathematically represent viscoelastic behavior. A common, traditional way is the use of spring and dashpot models, like the generalized Maxwell or Kelvin models. For any of these models, the constitutive equation can take the form of a differential equation consisting of terms which are derivatives of stress or strain (Fl gge, 1967). The resulting expression for the u relaxation modulus, a series of exponential terms, is also known as a Prony series. An example where this has been applied to describe relaxation of polyethylene can be found in a paper by Popelar et al. (1990). A drawback of these models can be the large number of parameters involved. For some materials the relaxation modulus appears to be proportional to time raised to some fractional power, at least in a certain nite time interval. Such observations are the basis of power law and fractional derivate models. According to the simplest form of power law model, the relaxation modulus can be expressed by t m , 0 < m < 1, (6) E(t) = E0 t0 where E0 and m are material parameters, whereas t0 can be chosen arbitrarily. According to fractional derivative models, integer order derivatives of the generalized Maxwell or Kelvin models are replaced by fractional order derivatives (Bagley, 1989). Mainly applied in modeling of dynamic problems (Adolfsson and Enelund, 2003), these models have a theoretical basis developed from molecular theory (Bagley and Torvik, 1983; McKenna and Gaylord, 1988). 6

5.2 Non-linear viscoelastic models


Whereas some polymers show a large linear region others, like most types of PE and XLPE, exhibit non-linear behavior from a very low strain. For this reason a non-linear constitutive equation is necessary for accurate models. Non-linear viscoelastic constitutive models, just as the linear viscoelastic models, are based on the idea that stress depends on the entire deformation history, or, alternatively, that strain depends on the entire stress history. The different models can be categorized under three major types: the multiple integral, the single integral and the differential equation formulation. A constitutive model, proposed by Green and Rivlin (1957), utilizes a multiple integral formulation. Detailed accounts of non-linear viscoelasticity based on multiple integral models are given by e.g. Lockett (1972) and Haddad (1995). Although theoretically general enough to describe most kinds of viscoelastic behavior, these types of mathematical models are inconvenient in practice due to the difculties associated with the experimental determination of the large numbers of necessary parameters. Less general, but still applicable to a wide range of polymers, single integral formulations of the constitutive behavior are more accessible in terms of experimental characterization. One such model is due, in its original form, to Schapery (1969). The transient part of the constitutive equation incorporates strain dependence into the linear viscoelastic equations according to

! (t) = h1 (" )
where

t 0 t 0

E $ (t) $ (# )

d (" h2 (" )) d# , d#

(7)

$ (t) =

a (T (# ), q(# ); " (# )) d#

(8)

denes the reduced time and h1 , h2 are material properties that depend on strain. A strain dependence may also be found for the shift factor, a, which generally depends on temperature T . Non-linear viscoelastic behavior found in some epoxy adhesives is facilitated by incorporating dependence of the shift factor on free volume (Popelar and Liechti, 1997). Other authors (Knauss and Emri, 1987; Losi and Knauss, 1992) introduced other state variables, here represented by q, to account for different effects. E.g. the effects of physical aging on the mechanical behavior of certain semi-crystalline polymers are quantied in this way by the free volume (Losi and Knauss, 1992). Numerous papers are available on the application of Schapery type models both in nite element formulations and experimental studies. A selection shows that among the studied materials are both amorphous polymers (Lai and Bakker, 1996; Haj-Ali and Muliana, 2004) and semi-crystalline polymers (Rand et al., 7

1996; Beijer and Spoormaker, 2002), as well as paper composites with polymer resin (Nordin and Varna, 2005). A somewhat different approach to the modeling of non-linear behavior is taken by Xia et al. (2003) and Nitta and Suzuki (1998) who introduce non-linearity into the relaxation kernel, by considering non-linear springs in generalized Kelvin or Maxwell models. In one case the non-linearity or anharmonicity of the springs is related to the Gr neisen constant which is a material parameter describing the u anharmonicity of molecular interaction.

6 Experimental characterization
Characterizing the time dependent constitutive behavior of polymers is a complex task. Depending on the type of material several techniques can be used, however the choice between methods is not trivial. It can depend on the type of polymer but also on the type of behavior that is of interest e.g. short- or longtime behavior, dynamic properties, etc. Care has to be taken to conduct the experiments in a controlled environment as most polymers have mechanical properties which change signicantly with temperature or even moisture. Material properties such as density or crystallinity may affect the mechanical behavior and need to be measured before the mechanical tests. For the same reason the specimens may need to be conditioned prior to experiments.

6.1 Physical properties


In the present study experiments were performed on different XLPE, as it was expected to nd a dependence on some state parameters, as degree of crystallinity or gel content, of the constitutive law. These different types of XLPE were characterized with respect to gel content, density and fraction of crystalline material. Details of these tests are presented in Paper A. The results are reproduced here in Table 1. Degree of crystallinity was determined from both density measurements with a density column and differential scanning calorimetry (DSC).

6.2 Mechanical testing


For the mechanical characterization of XLPE a number of different tests were performed of which the majority are reported in Paper A. Additional testing pertinent to the characterization of the non-linear behavior is presented in Paper C.

XLPE type Gel content /wt.% Density /(g/cm3) Crystallinity (from density) /wt.% Crystallinity (DSC) /wt.%

A 0.15 0.9205 49.09 52.37

B 54.5 0.9186 47.73 50.99

C 66.2 0.9172 46.76 49.68

D 75.15 0.9155 45.60 47.63

E 80.5 0.9155 45.55 47.19

F 83.1 0.9152 45.38 46.93

Table 1. Gel content, density and crystallinity for the types of XLPE utilized in the present study.
3.5 3 2.5

K / GPa

2 1.5 1 0.5 0 0 50 100 150 T / C 200

XLPE Type A XLPE Type B XLPE Type C XLPE Type D XLPE Type E XLPE Type F

250

300

Figure 3. Bulk modulus versus temperature for various types of XLPE (cf. Table 1). 6.2.1 Pressure Volume Temperature (pVT) measurements

Bulk behavior of a material can be obtained in a so called pVT apparatus in which a small sample of a material is surrounded by mercury and enclosed in a piezometer cell. The cell is inside a pressure vessel where hydrostatic pressure can be applied to it. At any constant temperature and pressure the change, relative to a reference state, in specic volume is recorded after equilibrium is reached. These measurements yielded both the bulk modulus and thermal expansion coefcient of XLPE, shown here in Figures 3 and 4, respectively.

103 1.4 1.2 1

& / C

0.8 0.6 0.4 0.2 0 0 100 200 300

T / C

Figure 4. Linear thermal expansion coefcient versus temperature for XLPE at ambient pressure calculated from PVT measurements.

Figure 5. Photograph showing the experimental setup for the stress relaxation experiments.

10

6.2.2

Stress relaxation experiments

During stress relaxation experiments the strain is kept constant on a specimen while the relaxation of the carried load is recorded. These tests are useful in determining the shear relaxation modulus of a polymer. However, there are some limitations to what it is possible to determine with such experiments. For example, the instantaneous modulus is not accessible by any experiments. It is also not possible in stress relaxation experiments to get the functions G(t) for the entire time domain 0 t < ' but only for a nite domain called the experimental window of the measuring technique. There is a short-time limit, determined by the eigenfrequency of the initial vibrations of the system or the performance of the test machine, usually between 0.01 and 1 s. The long-time limit is determined by the stability of the equipment and thermostating system, the thermal stability of the specimen and not to forget the patience of the experimenter. In favorable cases the time domain can be as wide as over eight magnitudes on the time scale, or 0.1 to 107 s. There are two possibilities for the long time behavior of polymers, which can be observed when experiments can be conducted over sufciently long times. One possibility is viscoelastic behavior without ow, this can be found in cross-linked polymeric systems, such as XLPE, and the other possibility is viscoelastic behavior with ow, as seen in uncross-linked polymeric systems. In the present work stress relaxation tests on XLPE were carried out for the various combinations of temperature, strain and crystallinity shown in Table 2. As it can be seen in Figure 5 the equipment consisted of a servo-hydraulic tensile test machine and a simple enclosure with heating elements. The temperature range for the experiments was chosen to cover the interval where residual stresses develop during cooling. Similarly, the time frame was chosen to cover the time a power cable would normally spend in the cooling stage of the manufacturing process. Earlier measurements of residual strains in cable insulation by Paulsson and Patel (2001), indicated residual strains below 10%. The strain levels for the current experiments were chosen to cover this strain range as good as possible. Variation of the crystallinity was achieved by selecting different grades of XLPE for the experiments. As it can be seen in the example of Figure 6 relaxation curves consisting of two linear regions (when plotted in a log-log diagram) were found for XLPE. For the constitutive modeling, a formulation based on equation (6) was adopted. Analysis of the experimental results with respect to temperature and crystallinity dependence can be found in Paper A, whereas strain dependence was explored in Paper C.

11

Temperature 23 C 40 C 60 C 78 C Strain 1 %3 %5 %7 %1 %3 %5 %7 %1 %3 %5 %7 %1 %3 %5 %7 % XLPE Type A: 0.4 wt.% peroxide B: 0.8 wt.% peroxide C: 1.2 wt.% peroxide D: 1.6 wt.% peroxide E: 2.0 wt.% peroxide F: 2.4 wt.% peroxide

Table 2. Relaxation experiments. Each marker corresponds to at least two tests at the appropriate conditions.

10

Increasing strain

1% strain 3% strain 5% strain 7% strain

G(t) / MPa

p1 p

10

10

!2

10

t /s

10

10

Figure 6. Relaxation curves for XLPE Type F (cf. Table 1) at 40 C.

12

6.2.3

Additional uniaxial tests

To be able to conrm the constitutive model based on the stress relaxation tests described in Section 6.2.2 and to further characterize the non-linear material behavior, additional tests were performed according to the following: Stress relaxation tests with a transient temperature prole. These tests were performed to conrm the validity of the proposed temperature and crystallinity dependent constitutive model and can be found in Paper A. Two step stress relaxation experiments. During these experiments the strain was increased to various levels in a second step during the relaxation tests. These tests were used for developing the non-linear material model and are presented in Paper C. Another type of test performed to check the validity of the non-linear constitutive model was the uniaxial loading of a specimen with constant strain rate. These tests were also included in Paper C.

6.3 Constitutive model for XLPE


Based on the insight into the mechanical behavior of XLPE gained from experiments, a constitutive model was formulated. Initially the inuence of temperature and crystallinity on the mechanical behavior was explored in Paper A. This was later extended to a non-linear model in Paper C. The non-linear viscoelastic model is summarized below. A useful observation was that both above and below its crystallization temperature XLPE can be considered as incompressible. The shear relaxation modulus Gr (t) at a chosen reference temperature and reference crystallinity was expressed as t p1 t p2 , c21 Gr (t) = max c11 , (9) t0 t0 where the parameters c11 ,p1 ,c21 and p2 were all experimentally determined. By calculating shift factors A((T, (% ) and a((T, (% ) for each material state dened by deviations (T and (% from the reference temperature and reference crystallinity, respectively, the shear relaxation modulus can be determined at all states. The equations for these shift factors can be found in both Paper A and Paper C. The non-linear viscoelastic constitutive equation was formulated in a manner similar to the Schapery model. The stress deviator si j was expressed as si j = 2A((T, (% )h("e f f )
t 0

Gr ($ (t) $ (# ))

dei j d# , d#

(10)

13

where ei j denotes the strain deviator and the reduced time is dened as

$ (t) =

t 0

a ((T, (% ) d# .

(11)

The material property h depends on the von Mises effective strain according to

"e f f =

2 ei j ei j . 3

(12)

As it can be seen in equation (10), strain dependence is separated from the temperature and crystallinity dependence. For the strain dependent function h an exponential strain dependence was found, h(" ) = ce) " , (13)

where the parameters c and ) were evaluated from experiments, see Paper C. The u so determined value of parameter ) agreed with the value of the Gr neisen parameter for various polyethylenes, used in certain non-linear viscoelastic models (Nitta and Suzuki, 1998; Wada et al., 1969; Rodriguez and Filisko, 1987) It was also observed that the above viscoelastic material model can be used below approximately 70 C. Above this temperatures it was found that an almost immediate relaxation of stresses results in the material.

7 Process model
Utilizing the material model presented in the thesis, a mechanical model is formulated for the XLPE cable insulation as it passes through the manufacturing process. In Paper B and C the focus is on the part of the cooling stage where the residual stresses develop in the insulation. Whereas Paper D targets the development of residual strains and incorporates stages of the process starting from the cross-linking of the polymer. Considering the incompressibility of XLPE and also the circular symmetry of the cable in the vertical manufacturing line, the problem can be greatly simplied, which has been exploited in the modeling work. Strains of non-mechanical origin, from e.g. thermal expansion and crystallization, are thus sufcient in determining the strain components in the insulation. With the knowledge of strain, the stress state can be determined from the constitutive equations and the equilibrium equations. Details of the model are presented in Paper B, where the linear viscoelastic model is applied in the calculations. In Paper C the same framework is used, but the material model is changed 14

2 1 0

2 1 0

!i j / MPa

-1 -2 -3 -4 -5 0

!i j / MPa !r !% !z

-1 -2 -3 -4 -5 0

!r !% !z

0.25 0.75 1 0.5 (r ri ) / (ro ri ) (a) Prediction by the linear viscoelastic material model.

0.25 0.75 1 0.5 (r ri ) / (ro ri ) (b) Prediction by the non-linear viscoelastic material model.

Figure 7. Predicted residual stress in the insulation. to the non-linear viscoelastic model. Predictions of the residual stress in a cable insulation by both models are presented in Figure 7. An estimation of the residual strains in the insulation can be made if a simulation of unloading from the predicted residual stress state is made. This analysis can be found in Paper B. The obtained strain eld proved to be much lower then experimentally observed residual strains in the insulation. This motivated the modeling work presented in Paper D, where it is recognized that the residual strains originate at the time of cross-linking in the insulation. The strain is locked at the time of cross-linking, which makes this strain state a reference state to which the material returns to after sufcient relaxation. Results predicted this way showed excellent agreement with experimental measurements as it can be seen in Figure 8.

8 Discussion and outlook


As set out in the Introduction the goal of this thesis was to provide a modeling tool which can be utilized in optimization of the manufacturing process with regards to the residual stress and strain of the cable insulation. This has been largely achieved during the course of the work. The experimental investigation of XLPE that was initially performed provides a rm foundation for the modeling work. The constitutive equations incorporate crystallinity and temperature dependent parameters. This has the benet that the model can be coupled to heat transfer and crystallization models for the same process. However, this also requires accurate models to be available for these latter problems. The experimental and modeling work together led to important ndings, namely that different mechanisms are important in the stress buildup and strain buildup 15

"iR j

0.25

0.5 0.75 (r ri ) / (ro ri )

Figure 8. Experimental (Lethagen, 2006) and predicted residual strain in the insulation. Circular, square and triangular symbols denote experimenR R R tally determined " r , "AB, and "z , respectively. Filled symbols refer to measurements on the convex side and open symbols refer to the concave side. Model predictions are represented by curves in the gure.

16

in the insulation. Stress development takes place mainly during cooling and is governed by the viscoelastic laws formulated in this thesis. Strain development has to be modeled starting from a reference state at the moment of cross-linking and onward through the manufacturing process. Calculations were performed based on a heat transfer model which provided realistic temperature data. But a correspondingly accurate model was not available for the crystallization problem. As a result the degree of crystallinity of XLPE undergoing cooling had to be estimated based on experimental data. A realistic model for the crystallization of XLPE would certainly contribute to the overall accuracy of the results. Stress and strain predictions are inuenced by the crystallization process in yet another way, as the volume strain is made up by a temperature and a crystallinity dependent part. Currently these two parts are not treated separately, instead experimental results are used for an estimation of the total thermo-chemical expansion. The current machanical model allows however for an easy incorporation of this separation of strains if a crystallization model will be available in the future. The process model was formulated with not only the accuracy of the predictions but also with the speed and ease of execution in mind. It is implemented in scripts which require small computer resources and can be run without expert knowledge. Areas that were not investigated in this work include defect formation in the insulation during manufacturing and predictions of the cable geometry. These issues could be important for the electrical properties of a cable, and could be implemented in the future by using the constitutive model presented in this thesis. Another possibility for future studies is in the estimation of shrink-back. A more accurate model can be formulated by considering shear stresses that develop between the insulation and the cable. This could be modeled by e.g. a shear-lag model.

17

9 Summary of appended papers


Paper A: Viscoelastic model of cross-linked polyethylene including effects of temperature and crystallinity
Characterization of the mechanical behavior of cross-linked polyethylene (XLPE) commonly used in high voltage cable insulation was performed by an extensive set of isothermal uniaxial tensile relaxation tests. The relaxation experiments were complemented by pressure volume temperature experiments as well as density and crystallinity measurements. Based on the experimental results, a viscoelastic power law model with four parameters was formulated, incorporating temperature and crystallinity dependence. It was found that a master curve can be developed by both horizontal and vertical shifting of the relaxation curves. The model was evaluated by making comparisons of the predicted stress responses with the measured responses in relaxation tests with transient temperature histories.

Paper B: Prediction of residual stresses in cross-linked polyethylene cable insulation


Modeling of the manufacturing process of high voltage power cables is performed with the aim of predicting residual stresses in the cable insulation. A viscoelastic power law type constitutive model is used which incorporates temperature and crystallinity dependence of the parameters. Thus the mechanical problem is coupled to the heat transfer and crystallization problems allowing for incorporation of different models. In order to illustrate the effects of relaxation, a comparison to a thermo-elastic material model is made.

Paper C: A non-linear viscoelastic constitutive model for XLPE and its application in residual stress prediction for high voltage cable insulation
A non-linear constitutive model, based on the Schapery non-linear viscoelastic model, was developed from stress relaxation experiments on cross-linked polyethylene (XLPE). Experiments with varying load histories were used for validation of the constitutive relations. The main application area for XLPE is as insulation for high voltage power cables, where residual stresses may inuence the electrical properties of the material. Therefore modeling of the manufacturing process was also performed with the aim of predicting residual stresses in the cable insulation.

18

Paper D: Residual strains in XLPE high voltage cable insulation


The residual strains in XLPE high voltage cable insulation are modeled in this study. It is found that strains originating at cross-linking of the molecules play a signicant role for buildup of residual strains. Detailed calculations are made for the case of the vertical continuous process line. Also, an estimate is made for the shrink-back of the cable insulation. The results are compared to experimental measurements of residual strain in cable insulation.

19

References
Adolfsson, K. and Enelund, M. (2003). Fractional derivative viscoelasticity at large deformations. Nonlinear Dynamics, 33(3):301321. Bagley, R. L. (1989). Power law and fractional calculus model of viscoelasticity. AIAA Journal, 27(10):14121417. Bagley, R. L. and Torvik, P. J. (1983). A theoretical basis for the application of fractional calculus to viscoelasticity. Journal of Rheology, 27(3):201210. Beijer, J. G. J. and Spoormaker, J. L. (2002). Solution strategies for FEM analysis with nonlinear viscoelastic polymers. Computers and Structures, 80(1415):12131229. Boysen, R. L. (1970). An analysis of the continuous vulcanizing process for polyethylenes. In IEEE Summer Power Meeeting and EHV Conference, pages 926933. Djurner, K., Kub t, J., and Rigdahl, M. (1980). Inre sp nningar och orientering a a i formsprutgods Del 1: Grundbegrepp. Technical Report 17, Plast- och Gummitekniska Institutet. Eduljee, R. F., J. W. Gillespie, J., and McCullough, R. L. (1994). Residual stress development in neat poly(etheretherketone). Polymer Engineering and Science, 34(6):500506. Fl gge, W. (1967). Viscoelasticity. Blaisdell. u Gohil, R. M. and Phillips, P. J. (1986a). Crystallinity in chemically crosslinked low density polyethylenes: 3. Morphology of the XLPE-2 system. Polymer, 27:16871695. Gohil, R. M. and Phillips, P. J. (1986b). Crystallinity in chemically crosslinked low density polyethylenes: 4. Inuence of crosslink density on morphology. Polymer, 27:16961704. Green, A. E. and Rivlin, R. S. (1957). The mechanics of non-linear materials with memory: Part I. Archives for Rational Mechanics and Analysis, 1:121. Haddad, Y. M. (1995). Viscoelasticity of engineering materials. Chapman & Hall. Haj-Ali, R. M. and Muliana, A. H. (2004). Numerical nite element formulation of the Schapery non-linear viscoelastic material model. International Journal for Numerical Methods in Engineering, 59:2545. 20

Harper, B. D. and Weitsman, Y. (1985). Characterization method for a class of thermorheologically complex materials. Journal of Rheology, 29(1):4966. Huotari, P. (1992). Simulation based optimal control of some industrial cable insulation processes. PhD thesis, Tampere University of Technology. Huotari, P. and Sistola, M. (1997). Production of XLPE insulated high and extra high voltage cable cores on catenary CV lines. Wire Industry, 64(762):366368. IEC (2004). Power cables with extruded insulation and their accessories for rated voltages above 30 kV (Um = 36 kV) up to 150 kV (Um = 170 kV) - Test methods and requirements. IEC, 3rd edition. Knauss, W. G. and Emri, I. J. (1987). Volume change and the nonlinearly thermo-viscoelastic constitution of polymers. Polymer Engineering and Science, 27(1):86100. Lai, J. and Bakker, A. (1996). 3-D schapery representation for non-linear viscoelasticity and nite element implementation. Computational Mechanics, 18:182191. Lee, E. H. and Rogers, T. G. (1965). On the generation of residual stresses in thermoviscoelastic bodies. Journal of Applied Mechanics, 32:874880. Lethagen, J.-E. (2006). Residual strain in XLPE Insulation in a 1200 mm2 Cable Measured on Sliced Material. Technical report, ABB Corporate Research, V ster s, Sweden. Condential. a a Lockett, F. J. (1972). Nonlinear viscoelastic solids. Academic Press. Losi, G. U. and Knauss, W. G. (1992). Free volume theory and nonlinear thermoviscoelasticity. Polymer Engineering and Science, 32(8):542557. McKenna, G. B. and Gaylord, R. J. (1988). Relaxation of crosslinked networks: theoretical models and apparent power law behaviour. Polymer, 29(11):2027 2032. Morland, L. W. and Lee, E. H. (1960). Stress analysis for linear viscoelastic materials with temperature variation. Transactions of the Society of Rheology, 4:233263. Muki, R. and Sternberg, E. (1961). On transient thermal stresses in viscoelastic materials with temperature dependent properties. Journal of Applied Mechanics, 28:193207. 21

Nilsson, U. H., Dammert, R. C., Campus, A., Snec, A., and Jakosuo-Jansson, H. (1998). Morphology of polyethylene for power cable insulation: Effects of antioxidant and crosslinking. In 1998 IEEE International Conference on Conduction and Breakdown in Solid Dielectrics, pages 365367. IEEE. Nitta, K. and Suzuki, K. (1998). A constitutive equation for nonlinear stress-strain curves of crystalline polymers. Journal of Material Science, 33(4):10151021. Nordin, L.-O. and Varna, J. (2005). Nonlinear viscoelastic behavior of paper ber composites. Composites Science and Technology, 65:16091625. Paulsson, G. and Patel, P. (2001). Measurements of residual strains and stresses in an XLPE insulated cable. Technical report, ABB Corporate Research, V ster s, a a Sweden. Condential. Phillips, P. J. and Kao, Y. H. (1986). Crystallinity in chemically crosslinked low density polyethylenes: 2. Crystallization kinetics. Polymer, 27:16791686. Popelar, C. F. and Liechti, K. M. (1997). Multiaxial nonlinear viscoelastic characterization and modeling of a structural adhesive. Journal of Engineering Materials and Technology, 119:205210. Popelar, C. F., Popelar, C. H., and Kenner, V. H. (1990). Viscoelastic material characterization and modeling for polyethylene. Polymer Engineering and Science, 30(10):577586. Rand, J. L., Henderson, J. K., and Grant, D. A. (1996). Nonlinear behavior of linear low-density polyethylene. Polymer Engineering and Science, 36(8):1058 1064. Rodriguez, E. L. and Filisko, F. E. (1987). Thermal effects in high density polyethylene and low density polyethylene at high hydrostatic pressures. Journal of Material Science, 22(6):19341940. Schapery, R. A. (1969). On the characterization of nonlinear viscoelastic materials. Polymer Engineering and Science, 9(4):295310. Struik, L. C. E. (1990). Internal stresses, dimensional instabilities and molecular orientations in plastics. John Wiley & Sons. Tobolsky, A. V. (1960). Properties and Structure of Polymers. Wiley. Tobolsky, A. V. and McLoughlin, J. R. (1955). Viscoelastic properties of crystalline polymers: Polytriourochloroethylene. Journal of Physical Chemistry, 59(9):989990. 22

Tschoegl, N. W. (1989). The Phenomenological Theory of Linear Viscoelastic Behavior An Introduction. Springer-Verlag. Uesaka, T., Kodaka, I., Okushima, S., and Fukuchi, R. (1989). History-dependent dimensional stability of paper. Rheologica Acta, 28(3):238245. Wada, Y., Itani, A., Nishi, T., and Nagai, S. (1969). Gr neisen constant and theru mal properties of crystalline and glassy polymers. Journal of Polymer Science: Part A-2: Polymer Physics, 7:201208. Xia, Z., Hu, Y., and Ellyin, F. (2003). Deformation behavior of an epoxy resin subject to multiaxial loadings. part II: constitutive modeling and predictions. Polymer Engineering and Science, 43(3):734748.

23