Sie sind auf Seite 1von 9

Viscoplastic Materials

Univ.-Prof. Dr.-Ing. habil. Josef BETTEN RWTH Aachen University Mathematical Models in Materials Science and Continuum Mechanics Augustinerbach 4-20 D-52056 A a c h e n , Germany < betten@mmw.rwth-aachen.de >

Abstract This worksheet is concerned with the linear and nonlinear theory of viscoplasticity. For example, in the following the rate-dependent deformation of a Fe-0.05 weight percent carbon steel at temperatures in excess of a homologous temperature of 0.5 has been discussed in comparison with suitable experiments. In order to describe the tensile tests at elevated temperatures, a viscoplastic constitutive equation with three parameters has been proposed. These parameters can be determined by using the LEVENBERG-MARQUARDT algorithm. Keywords: linear and nonlinear viscoplasticity; tensile tests at elevated temperature; viscoplastic constitutive equation; parameter identification; L-one and L-two error norms

Linear Theory of Viscoplasticity A linear viscoplastic model, called BINGHAM body (1922), consists of a linear viscous dashpot (NEWTON) and a solid friction element (MISES) connected in parallel, i.e., the HOOKE element of the KELVIN model has been replaced by the MISES element acting as a rigid body, if the shear stress is smaller than a critical value k, and sliding at the constant friction k with a shear rate larger than zero, if the shear stress is larger than k. Thus, for simple shear the viscoplastic material law is given by > restart: > 2*eta*d[12]=piecewise(F<0,0, F>=0,F*sigma[12]); 2 d12 = { 0 F 12 F<0 0F

> where eta is a material constant ( plastic viscosity) analogous to the shear viscosity of a fluid. Stress and stress rate are expressed by sigma and d, respectively. The function F is defined as > F:=1-k/abs(sigma[12]); F := 1 k 12

> The friction constant k is interpreted as the yield stress in pure shear of the material considered. In generalizing the BINGHAM model to multiaxial states of stress, we arrive at the following tensorial constitutive equation (BETTEN, 2008): > restart: > 2*eta*d[ij]=F*s[ij]; 2 dij = F sij > The tensors on the left and on the right hand side are the rate-of-deformation tensor and the stress deviator, respectively. The yield function F is defined as > F:=1-k/sqrt(J[2]); F := 1 > with the quadratic deviator invariant defined as: > J[2]:=s[ij]*s[ji]/2; J2 := 1 s s 2 ij ji k J2

> where the summation convention has been used introduced by EINSTEIN (1916). Inverting the above constitutive equation with respect to the stress deviator, we obtain the relation > s[ij]:=(2*eta+k/sqrt(K[2]))*d[ij]; sij := 2 + k dij K2

> consisting of two parts: a viscous part (parameter eta) and a solid part (parameter k) and with the quadratic invariant of the rate-of-deformation tensor: > K[2]:=d[ij]*d[ji]/2; K2 := 1 d d 2 ij ji

> The stress deviator s is a traceless tensor, trace(s) = 0, and defined as: > s[ij]:=sigma[ij]-sigma[kk]*delta[ij]/3= sigma[ij]-J[1]*delta[ij]/3; sij := ij 1 1 kk ij = ij J1 ij 3 3

> where J[1] is the linear stress invariant. From the constitutive equation we read that trace(s) = 0 implies trace(d) = 0. Thus, the constitutive equation is valid for incompressible materials. Compressible behavior has been investigated, for instance, by BETTEN (1982) in more detail.

Nonlinear Theory of Viscoplasticity

In the nonlinear case the isotropic constitutive equation may be expanded by an additional term, > restart: > 2*eta*d[ij]=F*(s[ij]+kappa*t[ij]); 2 dij = F ( sij + tij ) > which is quadratic in stresses and characterizes the second-order-effect regulated by the parameter kappa. The tensor t in the second term is traceless, trace(t) = t[jj] = 0, and defined as > t[ij]:=Diff(J[3],sigma[ij])=s[ik]*s[kj]-2*J[2]*delta[ij]/3; tij := 2 J3 = sik skj J2 ij ij 3

> where J[2] is the quadratic deviator invariant, while J[3] is the cubic deviator invariant, which can be expressed as > J[3]:=s[ij]*s[jk]*s[ki]/3; J3 := > The yield function is generalized according to > restart: > F:=1-k/f(J[2],J[3]); F := 1 k f( J2, J3 ) 1 s s s 3 ij jk ki

> as a function of both deviator invariants. For example, the function f may be assumed in the form > restart: > f(J[2],J[3]):=sqrt(J[2]+alpha*J[3]/Y); f( J2, J3 ) := J2 + J3

Y > where Y is the yield stress in uniaxial tension. The admissible range of the parameter alpha is: > alpha:=[-3, 3/2]; 3 := -3, 2 > resulting from the convexity of the surface f = constant as has been pointed out in more detail by BETTEN (1979). The proposed nonlinear constitutive equation is compatible with the minimum polynomial representation (BETTEN, 1985) > d[ij]:=f[ij](s[pq])= phi[0]*delta[ij]+phi[1]*s[ij]+phi[2]*s[ij]^2; dij := fij( spq ) = 0 ij + 1 sij + 2 sij > > s[ij]^2=s[ik]*s[kj]; # deviator power two
2

sij = sik skj > Hence, comparing this tensorial polynom with the proposed constitutive equation, we arrive at the following identities: > restart: > [phi[0], phi[1], phi[2]]= [-2*J[2]*phi[2]/3, F/2/eta, kappa*phi[1]]; F 2 [ 0, 1, 2 ] = J2 2, , 1 2 3 > [phi[0], phi[1], phi[2]]= (F/2/eta)*[-2*J[2]*kappa/3, 1, kappa]; 2 F J2 , 1, 1 3 [ 0, 1, 2 ] = 2 > Because of the assumed incompressibility the function phi[0] can be expressed by phi[2]. The parameter kappa in phi[2] regulates the second-order-effect as mentioned before. The nonlinear constitutive equation, valid for isotropy, has been modified to anisotropy and to compressible behavior by BETTEN (1982).

Viscoplastic Behavior of Metals For example, in the following the rate-dependent deformation of an Fe-0.05 weight percent carbon steel at temperatures in access of a homologous temperature of 1/2 should be discussed in comparison with suitable experiments. In order to describe the tensile tests at elevated temperatures, we propose the following viscoplastic constitutive equation > restart: macro(eps=epsilon,sig=sigma): > sig(eps[p]):=Y+b*(1-exp(-c*eps[p])); sig(eps[p]):=a-b*exp(-c*eps[p]); ( p ) := Y + b ( 1 e ( p ) := a b e
( c ) p

( c ) p

> The first representation consists of two parts, a friction element ( yield stress Y) and a KELVIN element, in contrast to the BINGHAM body, where sigma and epsilon[p] are the stress and the plastic strain in a uniaxial tensile test. The second one with a - b = Y may be more convenient for numerical evaluations and is therefore utilized in the following. The three parameters [a, b, c] have been determined from results of continuous isothermal tension tests at a number of constant strain rates, performed by P.J. WRAY of the U.S. Steel Research Laboratory and published by ANAND (1982). The strain-rate and temperature ranges spanned by these tests were 1.4*10^(-4)/s to 2.3*10^(-2)/s and 1173 to 1573 K, respectively. In this temperature range the steel has a fcc crystal structure called austenite.

The three parameters [a, b, c] can be determined by using the LEVENBERG-MARQUARDT algorithm in connection with the following experimental data for three plastic strain rates: > Diff(eps[p],t)={0.00014, 0.0028, 0.023}/s; { 0.00014, 0.0028, 0.023 } p = t s > which are identified in the following as > [0.00014, 0.0028, 0.023]=[1, 2, 3]; [ 0.00014, 0.0028, 0.023 ] = [ 1, 2, 3 ] The corresponding experimental data and the constitutive equation are given as follows: > data[1]:=[0.0075,10.234],[0.015,11.72], [0.0225,12.94],[0.03,13.95],[0.0375,14.78],[0.045,15.46], [0.0525,16.02],[0.06,16.48],[0.0675,16.86]; data1 := [ 0.0075, 10.234 ], [ 0.015, 11.72 ], [ 0.0225, 12.94 ], [ 0.03, 13.95 ], [ 0.0375, 14.78 ], [ 0.045, 15.46 ], [ 0.0525, 16.02 ], [ 0.06, 16.48 ], [ 0.0675, 16.86 ] > > [a[1], b[1], c[1]]=[18.62, 10.20, 26.01]; # determined by using the LEVENBERG-MARQUARDT algorithm [ a1, b1, c1 ] = [ 18.62, 10.20, 26.01 ] > sig[1](eps[p]):=18.62-10.20*exp(-26.01*eps[p]); 1( p ) := 18.62 10.20 e
( 26.01 ) p

> > data[2]:=[0.0075,14.65],[0.015,17.00], [0.0225,18.97],[0.03,20.63],[0.0375,22.02],[0.045,23.20], [0.0525,24.19],[0.06,25.02],[0.0675,25.72]; data2 := [ 0.0075, 14.65 ], [ 0.015, 17.00 ], [ 0.0225, 18.97 ], [ 0.03, 20.63 ], [ 0.0375, 22.02 ], [ 0.045, 23.20 ], [ 0.0525, 24.19 ], [ 0.06, 25.02 ], [ 0.0675, 25.72 ] > > [a[2], b[2], c[2]]=[29.43, 17.56, 23.03]; # determined by utilizing the LEVENBERG-MARQUARDT algorithm [ a2, b2, c2 ] = [ 29.43, 17.56, 23.03 ] > sig[2](eps[p]):=29.43-17.56*exp(-23.03*eps[p]); 2( p ) := 29.43 17.56 e
( 23.03 ) p

> > data[3]:=[0.0075,17.28],[0.015,21.35], [0.0225,24.70],[0.03,27.44],[0.0375,29.70],[0.045,31.54], [0.0525,33.05],[0.06,34.29],[0.0675,35.31]; data3 := [ 0.0075, 17.28 ], [ 0.015, 21.35 ], [ 0.0225, 24.70 ], [ 0.03, 27.44 ], [ 0.0375, 29.70 ], [ 0.045, 31.54 ], [ 0.0525, 33.05 ], [ 0.06, 34.29 ], [ 0.0675, 35.31 ] > > [a[3], b[3], c[3]]=[39.95, 27.65, 26.44]; # determined by using the LEVENBERG-MAQUARDT algorithm

[ a3, b3, c3 ] = [ 39.95, 27.65, 26.44 ] > sig[3](eps[p]):=39.95-27.65*exp(-26.44*eps[p]); 3( p ) := 39.95 27.65 e


( 26.44 ) p

> The three stress-plastic strain curves are represented in the following Figure. > alias(th=thickness,co=color): > p[1]:=plot({seq(sig[i](eps[p]),i=1..3)},eps[p]=0..0.08, axes=boxed,xtickmarks=4,th=3,co=black, title="stress-plastic strain curves"): > p[2]:=plot({seq([data[i]],i=1..3)},eps[p]=0..0.08,0..40, style=point,symbol=cross,symbolsize=40,th=3,co=black): > p[3]:=plots[textplot]([0.02,35,`T = 1373 K`],co=black): > plots[display](seq(p[k],k=1..3));

This Figure illustrates the good agreement between theory and experimental data. The three curves correspond to plastic strain rates 0.00014/s, 0.0028/s, and 0.023/s. The dimension of the stresses on the ordinate is MPa. The yield stresses, Y = a - b, are given at epsilon[p] = 0: > [Y[0.00014],Y[0.0028],Y[0.023]]=[8.42,11.87,12.3]*MPa; [ Y0.00014, Y0.0028, Y0.023 ] = [ 8.42, 11.87, 12.3 ] MPa > L-one and L-two Error Norms In the following let us calculate the distance between stress-plastic strain curves and the experimental data expressing by the corresponding L-one or L-two error norm. These vector norms are defined as follows: > V:=vector([seq(v[i],i=1..5)]);

V := [ v1, v2, v3, v4, v5 ] > Norm(V,1)=norm(V,1); Norm( V, 1 ) = v1 + v2 + v3 + v4 + v5 > L[1]:=Norm(V,1)/number_of_points=norm(V,1)/5; Norm( V, 1 ) 1 1 1 1 1 = v1 + v2 + v3 + v4 + v5 number_of_points 5 5 5 5 5 > Norm(V,2)=norm(V,2); L1 := v1 + v2 + v3 + v4 + v5 Norm( V, 2 ) = > L[2]:=Norm(V,2)/sqrt(number_of_points)=norm(V,2)/sqrt(5); L2 := Norm( V, 2 ) number_of_points = 1 5 v1 + v2 + v3 + v4 + v5
2 2 2 2 2 2 2 2 2 2

> Where v[3], for instance, is the difference between the y-coordinates of the third experimental point and the corresponding curve point. In the following a distance vector V is generated by using a do-loop. Corresponding to the three parameters [0.00014, 0.0028, 0.023] = [1, 2, 3] we arrive at the following results: 1) plastic strain rate = 0.00014/s: > for i from 1 to 9 do x[i]:=evalf(subs(eps[p]=data[1][i][1], (sig[1](eps[p])-data[1][i][2])),5) od: > X:=vector([seq(x[i],i=1..9)]); X := [ -0.0063, -0.0050, -0.0012, -0.0044, -0.0059, -0.0044, -0.0036, -0.0021, -0.0025 ] > L[1][strain_rate=0.00014]:=Norm(X,1)/9=evalf(norm(X,1)/9,5); 1 Norm( X, 1 ) = 0.0039333 9 strain_rate = 0.00014 > L[2][strain_rate=0.00014]:=Norm(X,2)/3=evalf(norm(X,2)/3,5); L1 := L2 > 2) plastic strain rate = 0.0028/s: > for i from 1 to 9 do y[i]:=evalf(subs(eps[p]=data[2][i][1], (sig[2](eps[p])-data[2][i][2])),5) od: > Y:=vector([seq(y[i],i=1..9)]); Y := [ 0.006, -0.001, 0.001, 0.0003, 0.0062, 0.0009, -0.0011, 0.0002, -0.0004 ] > L[1][strain_rate=0.0028]:=Norm(Y,1)/9=evalf(norm(Y,1)/9,5); 1 Norm( Y, 1 ) = 0.0019000 9 strain_rate = 0.0028 > L[2][strain_rate=0.0028]:=Norm(Y,2)/3=evalf(norm(Y,2)/3,5); L1 :=
strain_rate = 0.00014

:=

1 Norm( X, 2 ) = 0.0042593 3

L2 >

strain_rate = 0.0028

:=

1 Norm( Y, 2 ) = 0.0029580 3

3) plastic strain rate = 0.023/s: > for i from 1 to 9 do z[i]:=evalf(subs(eps[p]=data[3][i][1], (sig[3](eps[p])-data[3][i][2])),5) od: > Z:=vector([seq(z[i],i=1..9)]); Z := [ -0.006, 0.003, -0.002, 0.001, -0.009, -0.0033, -0.0001, 0.0012, -0.0011 ] > L[1][strain_rate=0.023]:=Norm(Z,1)/9=evalf(norm(Z,1)/9,5); 1 Norm( Z, 1 ) = 0.0029667 9 strain_rate = 0.023 > L[2][strain_rate=0.023]:=Norm(Z,2)/3=evalf(norm(Z,2)/3,5); L1 := L2 > Rsum In this worsheet it has been illustrated that Maplesoft furnishes powerful tools in fitting viscoplastic constitutive equations to experimental data. The very good approximation of theory to experiments has been expressed by L-one and L-two error norms.
strain_rate = 0.023

:=

1 Norm( Z, 2 ) = 0.0040077 3

References ANAND, L. (1982). Constitutive Equations for the Rate-Dependent Deformation of Metals at Elevated Temperatures, Tranactions of the ASME 104: 12-17. BETTEN, J. (1979). ber die Konvexitt von Fliekrpern isotroper und anisotroper Stoffe, Acta Mechanica 32: 233-247. BETTEN, J. (1982). Pressure-Dependent Yield Behavior of Isotropic and Anisotropic Materials, in: P. VERMEER and H. LUGER (eds.), Deformation and Failure of Granular Meterials, A.A. Balkema, Rotterdam, pp. 81-89. Presented at the IUTAM Symposium in Delft, Sept. 1982. BETTEN, J. (1985). The Classical Plastic Potential Theory in Comparison with the Tensor Function Theory, Eng. Fracture Mech. 21: 641-652. Presented at the Int. Symposium Plasticity Today in Udine, June 1983. BETTEN, J. (2008). Creep Mechanics, Third Edition, Springer-Verlag, Berlin / Heidelberg / NY. BINGHAM, E.C. (1922). Fluidity and Plasticity, McGraw-Hill Book Comp., New York.

EINSTEIN, A. (1916). Die Grundlagen der allgemeinen Relativittstheorie, Annalen der Physik 4, Folge 49: 769-822. >

Das könnte Ihnen auch gefallen