V.

Desikan Raju Sethuraman1

Department of Mechanical Engineering, Indian Institute of Technology Madras, Chennai600036, India

Analysis of Material Nonlinear Problems Using Pseudo-Elastic Finite Element Method

A method based on linear elastic nite element analysis is presented for stress eld determination of elasto-plastic problems. Henckys total deformation theory is used to dene effective material parameters, which are treated as spatial eld variables and considered to be functions of nal state of equilibrium stress and material properties. These effective material parameters are obtained in an iterative manner using straincontrolled projection method, arc-length method, and Neuber rule applied on experimental uniaxial tension test curve. Three problems of von Mises material are considered to illustrate the application of the proposed method: a thick-walled cylinder subjected to internal pressure characterized by general work-hardening behavior, a V-notch specimen subjected to remote tensile load having elastic-perfectly plastic behavior, and a rotating disk with material having elastic linear work-hardening behavior. Obtained results for all the cases are compared with standard nonlinear nite element results and are found to be in good agreement. S0094-99300000104-9

Introduction
In classical design of structural components, the norm dened on the stress tensor is bounded within the elastic limit; but in practice it would be advantageous to consider the elasto-plastic behavior of structural materials. This may be due to the unexpected presence of notches, voids, microcracks, overloads, etc. In pressure vessel design to enhance the pressure-carrying capacity, residual stresses are developed by introducing initial partial plastic yielding, i.e., autofrettageing. Thus, the behavior of the materials in the plastic range is essential in order to understand the structural behavior completely and to have reliable estimate of life. However, the nonlinear stress-strain relationship and the loading path dependency in the plastic range makes the analysis tedious. Over the years, nite element method has been successfully employed in analyzing the material behaviors in elastic and elastoplastic range. Neuber 1 obtained elasto-plastic stress and strain at the stress concentration point using elastic solution, assuming that the elastic strain energy is conserved at the notch tip. Marcal and King 2 solved elasto-plastic problems based on stiffness method. They developed incremental stress-strain relationship using partial stiffness coefcient. Later, Yamada et al. 3, using Prantl-Reuss incremental stress-strain relation, obtained explicit elasto-plastic stress-strain constitutive matrix for von Mises materials. Their method is based on nding the varying load increment needed to bring the elastic material point, which is very close to the yield surface, onto the yield surface. The iterative process is repeated till the summation of the loading increments reaches the total applied load. Later, Owen and Hinton 4 provided nite element computer implementation of elasto-plastic problems based on incremental theory. Seshadri 5 developed GLOSS method based on two-linear elastic nite element analysis which is used to evaluate the approximate plastic strains at certain local regions. A lower-bound limit-load theorem approach for showing the effectiveness of reduction of local modulus of elasticity was given by Mackenzie et al. 6 and Shi et al. 7. Jahed et al. 8 developed an analytical method to solve pressure vessel problems in the
1 Corresponding author, e-mail: [email protected] Contributed by the Pressure Vessels and Piping Division for publication in the JOURNAL OF PRESSURE VESSEL TECHNOLOGY. Manuscript received by the PVP Division, January 9, 1999; revised manuscript received June 27, 2000. Associate Technical Editor: R. Seshadri.

elasto-plastic range. In their method, the cylindrical pressure vessel was considered to be assembly of nite number of strips, and closed-form elastic solution was used within each strip. The material properties are considered to be varying along the radial direction, but constant within the strips. The application of this method is restricted to problems for which the nite strip has closed-form solutions. Recently, Babu and Iyer 9 developed a robust method using relaxation method, which is based on GLOSS method of analysis. Here, an attempt was made to satisfy force equilibrium in the plastic range. This paper presents a pseudo-elastic nite element method for the determination of inelastic solutions. Linear elastic nite element analysis is carried out for solving elasto-plastic problems using material parameters as eld variables. The method is based on Henckys total deformation theory and is applicable for proportional loading. Three algorithms, namely projection method, arc-length method, and Neuber rule, are used to calculate the material parameters, and their behavior with respect to computational efciency in terms of number of iterations needed for convergence are established. The effectiveness of the method is demonstrated using the following problems: a pressure vessel subjected to internal pressure for general hardening behavior, a stress concentration problem of V-notch type subjected to tensile load having elastic-perfectly plastic behavior, and a body force problem of rotating disk having linearly work-hardening behavior.

Effective Material Parameters Determination

The principle of virtual work 10 for a static equilibrium body under elastic or elasto-plastic stress eld, for every kinematically admissible virtual displacement eld, is given as

i ji jd

B m u md

T k u k dS (1)

The strain-stress relationship can be taken in the form i j f i j (2) where the total strain tensor i j is the sum of conservative elastic and nonconservative plastic part i j iej ipj (3)

Journal of Pressure Vessel Technology

Copyright 2000 by ASME

NOVEMBER 2000, Vol. 122 457

Downloaded 16 Jan 2008 to 203.199.213.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

The elastic strain tensor is related to the stress tensor and is given by Hookes law for isotropic material as iej

eff E eff

1 i j kk i j E E

1 o E 2 o o

n1

(17)

(4)

The plastic strain tensor is related to the deviatoric part of stress tensor and is given by Henckys total deformation relation ipj S i j where Si ji j 1 3 kk i j (6) (5)

is the deviatoric stress tensor, and is a scalar valued function, given by

p 3 eq 2 eq

The effective material parameters are functions of the nal state of stress elds. Hence, E eff and eff can be thought of eld variables describing the material properties at each point, since the nal state of stress at every point is unique. Clearly, for constant values of E eff and eff , Eq. 10 describes the elastic behavior of the body. It is known that every state of stress at point in a equivalent sense follows experimental uniaxial tensile curve. From that point of view and observing Eqs. 13,14,16, the reciprocal of E eff is nothing but secant modulus dened on the experimental uniaxial curve. Hence, the effective modulus can be obtained from the experimental material curve, the evaluation is described in the next section.

(7)

Finite Element Implementation

The virtual work statement of Eq. 1 is solved using nite element method. The discrete form of the equation is obtained by discreting the given domain into nite domains using constant strain triangular element CST, which has the constant stressstrain eld representation. Since the stress eld is constant over the element domain, the effective material parameter is also constant over the element, thereby leading to the determination of nite number of material parameters over the entire domain. The element stiffens matrix for the CST element is given by 11
K ele

the equivalent plastic strain and the equivalent stress are dened as
p eq

2 p p 3 ij ij

and

eq

3 S S 2 ij ij

(8)

Substituting Eqs. 4 8 in Eq. 3, we get i j

1 1 i j kk i j E E 3

(9)

All the variables inside the parentheses in Eq. 9 are involved with the material properties, nal equivalent total plastic strain, and equivalent stress. Equation 9 can be written as i j

B T D eff B d

(18)

1 eff eff i j E eff E eff kk i j

(10)

where eff and E eff are the effective Poisson ratio and effective Young modulus, respectively, and are treated as material parameters in the present study. Comparing Eqs. 4 and 10, we get the effective values of the material parameters as 1 1 2 E eff E 3 (11)

eff E eff

E 3

(12)

Using Eq. 7 for elastic-perfectly plastic material, E eff is obtained as 1 1 p E eff E o (13)

where B matrix is the conventional strain displacement matrix, and D eff matrix is obtained from constitutive Eq. 10. The determination of effective material parameters needed to calculate D eff , using projection method, is as follows. First, a linear elastic nite element analysis is carried out. Suppose, for a particular element, the equivalent stress is calculated from linear nite element analysis and point A, as shown in Fig. 1, is obtained. This point has crossed the yield stress. Keeping the strain value the same, i.e., strain controlled, and projecting point A on the experimental uniaxial curve, the effective value of Youngs modulus for the next iteration is obtained. Substituting this effective value in Eq. 12, the effective Poisson ratio is obtained. These effective values are obtained for all the elements which have yielded. With this new set of effective material parameters, the next linear elastic nite element analysis is performed. This

Similarly, for linearly work-hardening material having tangent modulus E T , the E eff is obtained as

o o 1 E eff E ET

(14)

Substituting Eqs. 13 and 14 in Eq. 12, the effective Poisson ratios for elastic-perfectly plastic and linearly hardening materials are obtained. For Ramberg-Osgood model o o o

(15)

effective Youngs modulus and effective Poissons ratio are obtained as 1 1 o E eff E o

n1

1 o

(16)

Fig. 1 Projection method of E eff determination

458 Vol. 122, NOVEMBER 2000

Transactions of the ASME

Downloaded 16 Jan 2008 to 203.199.213.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 2 Arc length method of E eff determination Fig. 4 Comparison of stress variation along the thickness direction for cylinder with internal pressure

Fig. 3 Neubers method of E eff determination

iterative procedure is repeated and elastic analysis with currently evaluated E eff and eff is performed until all the effective material parameters converge and equivalent stress falls on the experimental uniaxial stress-strain curve. In arc-length method, the yielded point A, as shown in Fig. 2, is projected onto the experimental uniaxial curve, considering OA as the arc radius and O as the center. The intersecting point B on the experimental uniaxial curve is used to obtain the new effective modulus for the next iteration. Substituting this effective modulus in Eq. 12, the new effective Poisson ratio is obtained and is used for the next iteration. In Neuber method, the yielded point A, as shown in Fig. 3, is projected onto the uniaxial curve at point B. While projecting the total strain, energy is assumed to be conserved; i.e., the area ODAEO is the same as area OCBFO. Thus, a corresponding point on the uniaxial material curve which has the same strain energy is found. With this point, the effective Young modulus is obtained and is substituted in Eq. 12 to get the effective Poisson ratio for the next iteration.

stress components are nondimensionalized with respect to initial yield stress. The ratio of the internal pressure to uniaxial yield stress is varied from 1.0 to 2.15. The cylinder has the geometric dimension of r 2 / r 1 5. The material having Youngs modulus 2.0* 1005 N/mm2 , Poissons ratio 0.3, and yield stress 2.0* 1002 N/mm2 is considered for the analysis. The material is of von Mises type and assumed to be characterized by nonlinear behavior obeying Ramberg-Osgood model with 3/7 and n 5. Figure 4 shows the variations of radial and hoop stresses along the thickness direction of the cylinder for plane stress condition for pressure ratio1.0. The results are being compared with variable property approach method 8 and NISA 12. It can be seen that the results are identical with variable property approach method. When compared with NISA, it is observed that the stresses are in good agreement, except for slight deviation in hoop stress near elasto-plastic boundary. Figure 5 gives the stress variations for different interanal pressure considering the problem as plane strain condition. Here, convergence is assumed to be achieved when the difference between the equivalent stress and current ow stress during the present iteration is less than 1 percent for all the elements. Figure 6 shows how a pseudo-elastic

Results and Discussions

Three case studies are considered to illustrate the effectiveness of the pseudo-elastic nite element analysis. First, a cylindrical vessel under plane stress and plane strain conditions subjected to internal pressure which is sufcient enough to cause yielding is considered. Due to symmetry, only one-quarter of the problem is discretized using uniform constant strain triangular element having 627 nodes and 1152 elements. The applied pressure and all the Journal of Pressure Vessel Technology

Fig. 5 Stress variation for Ramberg-Osgood model under plane strain condition

NOVEMBER 2000, Vol. 122 459

Downloaded 16 Jan 2008 to 203.199.213.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 8 Geometry of V-notch

Fig. 6 Convergence path for a particular pseudo-elastic point

point converges and nally falls on the experimental uniaxial curve with successive linear elastic nite element analysis. Figure 7 shows the state of stress for all the material points after convergence. It can be seen that in the present method, all the material points fall exactly on the uniaxial material curve. Second, a stress concentration problem of V-notch type is analyzed. Figure 8 gives the geometrical and loading details of the V-notched tension specimen. The material is assumed to be elastic-perfectly plastic with Youngs modulus 1.962* 105 N/mm2 , Poissons ratio 0.3, and yield stress 294.3 N/mm2. Due to symmetry in geometry and loading, only onequarter of the problem is considered. The geometry is discretized with 733 elements and 406 nodes. For comparison purpose, the problem considered is very similar to the problem solved by Yamada 3. Figure 9 shows the distribution normal stresses along section AA. The results are compared with Yamadas method and observed as are in close agreement. For the comparison of computational efciency, Table 1 shows the number of iterations taken by the present method with three iterative schemes, Yama-

das method, and NISA. It can be observed from the Table 1, the Neuber iterative scheme based on the present method converges fast for all the considered loading conditions. In NISA, full Newton-Raphson and modied Newton-Raphson techniques are used. Comparing the convergence rate with these techniques, it can be seen from Table 1 that Neuber rule performs better than modied Newton-Raphson, but full Newton-Raphson converges very fast for all the cases considered. Finally, a body force problem of rotating disk is considered. The diameter of the disk is 40.0 mm, and the disk is assumed to be

Fig. 9 Stress variation for V-notch problem Table 1 Number of iterations taken for convergence for stress concentrationV-notch problem

Fig. 7 Converged state of stress for material points

460 Vol. 122, NOVEMBER 2000

Transactions of the ASME

Downloaded 16 Jan 2008 to 203.199.213.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm

Fig. 10 Stress variation for rotating disk

n hardening exponent in Ramberg-Osgood model P r pressure ratio, normalized pressure with respect to initial yield stress r 1 internal radius of cylinder r 2 external radius of cylinder S traction prescribed surface Ti surface traction vector component ui displacement vector component yield offset in Ramberg-Osgood model experimental uniaxial stress o yield strength hoop/tangential stress r radial stress i j stress tensor experimental uniaxial total strain i j strain tensor Poissons ratio domain/volume variational symbol i j kronecker delta scalar valued function B strain displacement matrix K stiffness matrix Superscripts e p ele i, j,k,m elastic plastic element free and dummy indices used for tensor components

rotating with constant angular speed of 500 rad/s. The disk is made of linearly hardening material with Youngs modulus 2.9* 105 N/mm2 , Poissons ratio 0.3, tangent modulus 1.45* 105 N/mm2 , and yield stress 200.0 N/mm2. Due to symmetry, the geometry is modeled with axisymmetric solid elements. There are 200 elements and 126 nodes in the nite element discretization of the geometry. The obtained radial and tangential stresses are shown in Fig. 10 and are compared with NISA 12 results, and found to be in good agreement.

Subscripts eq equivalent eff effective T tangent

Conclusions
Material nonlinear problem has been solved with the proposed pseudo-elastic linear nite element method. The proposed method can be easily adapted to the existing linear elastic nite element code with suitable updating of elastic material properties with effective material parameters which are readily obtained from one-dimensional uniaxial tension curve of a given material. Von Mises material characterized by elastic perfectly plastic, linearly work hardening and general hardening obeying Ramberg-Osgood behavior is considered for the problem of pressure vessel subjected to internal pressure, a V-notch specimen subjected to remote tension, and a rotating disk subjected to body force. Stress elds obtained for all the cases are found to be in good agreement with the available nonlinear nite element results. The rate of convergence for the three algorithms, viz., projection method, arclength method, and Neuber rule which are used in the present method, were studied, and it was found that the Neuber rule has better convergence rate and also behaves better when compared with Yamadas algorithm. In NISA, full Newton-Raphson and modied Newton-Raphson techniques are used. Comparing the convergence rate with these techniques, Neuber rule performs better than modied Newton-Raphson, but full Newton-Raphson converges very fast for all the cases considered.

References
1 Neuber, H., 1961, Theory of Stress Concentration for Shear Strained Prismatical Bodies With Arbitrary Nonlinear Stress-Strain Law, ASME J. Appl. Mech., 28, pp. 554550. 2 Marcal, P. V., and King, I. P., 1967, Elastic Plastic Analysis of TwoDimensional Stress Systems by the F.E.M., Int. J. Mech. Sci., 9, pp. 143 155. 3 Yamada, Y., Yoshimura, N., and Sakurai, T., 1968, Plastic Stress-Strain Matrix and Its Application for the Solution of Elastic-Plastic Problems by the Finite Element Method, Int. J. Mech. Sci., 10, pp. 343354. 4 Owen, D. R. J., and Hinton, E., 1980, Finite Elements in Plasticity: Theory and Practice, Pineridge Press Limited. 5 Seshadri, R., 1991, The Generalized Local Stress-Strain GLOSS AnalysisTheory and Applications, ASME J. Pressure Vessel Technol., 113, pp. 219227. 6 Mackenzie, D., and Boyle, J. T., 1993, A Method of Estimating Limit Loads by Iterative Elastic AnalysisI: Simple Examples, Int. J. Pressure Vessels Piping, 53, p. 77. 7 Shi, J., Mackenzie, D., and Boyle, J. T., 1993, A Method of Estimating Limit Loads by Iterative Elastic AnalysisIII: Torispherical Heads Under Internal Pressure, Int. J. Pressure Vessels Piping, 53, p. 121. 8 Jahed, H., Sethuraman, R., and Dubey, R. N., 1997, A Variable Material Property Approach for Solving Elasto-Plastic Problems, Int. J. Pressure Vessels Piping, 71, pp. 285291. 9 Babu, S., and Iyer, P. K., 1998, Inelastic Analysis of Components Using a Modulus Adjustment Scheme, ASME J. Pressure Vessel Technol., 120, pp. 15. 10 Malvern, L. E., 1969, Introduction to the Mechanics of a Conitnuous Medium, Prentice-Hall, Engelwood Cliffs, NJ. 11 Zeinkiewicz, O. C., and Taylor, R. L., 1989, The Finite Element Method, Vol. 1, McGraw-Hill International Editions. 12 NISA II, 1998, Version 7.0 Users Manual, Engineering Mechanics Research Corporation, Michigan.

Nomenclature
B i body force component E Youngs modulus

Journal of Pressure Vessel Technology

NOVEMBER 2000, Vol. 122 461

Downloaded 16 Jan 2008 to 203.199.213.66. Redistribution subject to ASME license or copyright; see http://www.asme.org/terms/Terms_Use.cfm