Anda di halaman 1dari 7

# International Journal of Non-Linear Mechanics 46 (2011) 834840

## International Journal of Non-Linear Mechanics

journal homepage: www.elsevier.com/locate/nlm

## Non-linear analysis of structures using two-point method

Hamed Saffari n, Iman Mansouri
Department of Civil Engineering, Shahid Bahonar University of Kerman, P.O. Box 76169133, Kerman, Iran

a r t i c l e i n f o
Article history: Received 25 May 2010 Received in revised form 2 January 2011 Accepted 20 March 2011 Available online 30 March 2011 Keywords: Non-linear analysis Space trusses NewtonRaphson method Two-point method

a b s t r a c t
Non-linear algebraic equations must be solved by an iterative method, the non-linear equations being linearized by evaluating the non-linear terms with the known solution from the preceding iteration. The NewtonRaphson method, which is based on the Taylor series expansion and uses the tangent stiffness matrix, has been extensively used to solve non-linear problems. In this paper, a new Newton Raphson algorithm is developed for analyses involving non-linear behavior. Our method, here named as a two-point method, is constructed as a predictorcorrector one, most frequently taking Newtons method in the rst iteration. It should be noted that our concern in this research ignores the problem of passing limit points. The presented method incorporates the known information at each stage of the loading process to determine the subsequent unknown variables. Compared with the classic Newton Raphson algorithm, it offers a strategy that can be deployed to reduce both the number of the iterations and the computing time involved in non-linear analysis of structures. & 2011 Elsevier Ltd. All rights reserved.

1. Introduction The various types of space structures differ in their behavior under load, and so a different method of analysis must be used for each type. Space truss systems are one of the most popular forms of space frames. Nowadays, they provide an easy and economic method of roong large areas. The lighter and more efcient these structural systems become, the more likely they are to have nonlinear stability problems. Space trusses may suddenly collapse if their failure modes are not realized. These structures may be designed to maintain their integrity even after an initial inelastic failure: a progressive failure that redistributes loads, normally causing other members to fail until the total structure becomes unstable . For accurate modeling of a progressive failure, the analysis method that can create a load vs. displacement response during instability is required. Methods that have been used involved creating nite element models that update their stiffness according to non-linear geometric response . The analysis of geometrically non-linear structural problems has been a subject of interest for over three decades. The solution of a non-linear problem reduces to that of tracing a non-linear loaddisplacement path by solving a system of non-linear algebraic or differential equations. An abundance of procedures exists for attacking the non-linear equilibrium equations. These include

Corresponding author. Tel.: 98 9131411509; fax: 98 341 3220054. E-mail addresses: hsaffari@mail.uk.ac.ir (H. Saffari), im_pce@yahoo.com (I. Mansouri). 0020-7462/\$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.ijnonlinmec.2011.03.008

the NewtonRaphson method, the modied NewtonRaphson procedure, the perturbation method , the self-correcting incremental procedure [2,3], the incremental stiffness procedure , the initial value approach  and many more. The non-linear behavior of trusses has been investigated by other researchers using various analytical techniques. Saffari et al.  used a modied normal ow algorithm to study non-linear behavior in truss structures, Tabatabaei et al.  applied the NewtonRaphson iterative algorithm along the ow path normal in non-linear static analysis. Papadrakakis and Gantes  presented some procedures to truncate the NewtonRaphson method. Pina et al.  presented a formal solution of quasi-static for solving non-linear problems. Greco et al.  proposed a new geometric non-linear formulation for space truss analysis that uses nodal positions rather than nodal displacements. Tabatabaei and Saffari  studied large strain analysis of planar frames using normal ow algorithm. Thai and Kim  presented the large-deection inelastic analysis of space truss structures including both geometric and material non-linearities. Papadrakakis  employed the dynamic relaxation scheme for the secondorder and large deection analysis of trusses. Bellini  suggested a new mathematical model to study snap-buckling problem. Kassimali and Parsi-Feraidoonian  studied the non-linear behavior of prestressed cable trusses. Inelastic postbuckling analysis of truss structures by the dynamic relaxation method investigated by Ramesh and Krishnamoorthy . Of all numerical computation, systems of non-linear equations are perhaps the most difcult to solve. There are several approaches for nding solutions. For a non-linear problem, the solution of the non-linear set of equations normally takes up most

## H. Saffari, I. Mansouri / International Journal of Non-Linear Mechanics 46 (2011) 834840

835

of the computing time. Because the NewtonRaphson method has the quadratic convergence characteristics, it is commonly used for solving non-linear equations. At each iterative step of the NewtonRaphson method, a set of linear algebraic equations must be solved. Many iterative methods have been offered for improving NewtonRaphson approaches for solving non-linear equations [16,17]. However, many depend on the second or higher derivatives from the computing process which restricts their practical application because of inherently intensive and extensive nature of the computation involved In recent years, there has been some progress in devising iterative methods designed to improve application of the NewtonRaphson method while at the same time, not requiring the computation of second derivatives (Hessian matrix) for solving the non-linear equations [18,19]. They can be classied as either belonging to the one-step or the two-step class . Two-step methods have been suggested by combining the NewtonRaphson method with another one-step method . This approach, constructed in conformity with predictor corrector methods, and so taking most frequently Newtons method in the rst iteration, is used in this paper. Multipoint iterative solvers belong among such powerful methods for solving non-linear equations that they overcome the theoretical limits of one-point methods concerning the convergence order and computational efciency . Although the NewtonRaphson is powerful in providing the response of a structure at a specied set of loads, divergence occurs when the solution point is close to the limit point. As a result, taking this as the failure load of a structure may not be correct. However, these problems can be overcome using a form of displacement control method [5,11]. This study ignores this inability of the NewtonRaphson method, because the main object is to develop the load control method for a wide range of problems which have non-linear behavior, but without passing limit points and to do so by using a multipoint iterative solver that improves efciency (by reducing the number of iterations and time taken) in non-linear analysis.

solution of non-linear problems may be regarded as variations of this technique. A summary of the NewtonRaphson approach is presented in this section. A detailed discussion is provided in references [24,25]. The rst iteration in this process is similar to the linear incremental method except at the end of the rst iteration, when member forces are calculated and transformed into the global coordinates. Then, the degree imbalance in forces between the internal nodal forces and the applied external load are determined. Subsequent iterations are continued until a predened convergence criterion is satised. 3.2. Two-point method In this subsection, we review the mathematics of the twopoint method concept and in the next section it will be employed in form of applications to solving structural engineering problems. In many areas of the engineering, there arises the classical problem of nding the roots of a given non-linear function f: i.e. that of nding the solutions of the equation f (x)0. In two-point method two formulas are used for nding the approximate root for non-linear equations. One formula is appropriately called the predictor in a sense that it gives prediction about the root using the initial estimate. The second formula is called the corrector, since it gives the correct approximation using the value predicted by the rst point. Two-point methods are sometimes also called as predictorcorrector methods . Recently, Petkovic and Petkovic  proposed and analyzed the following two-point iteration scheme for solving non-linear equations: 8 < yn xn ff0xnn x n 0,1, . . . 3 : xn 1 yn f0yn f yn To reduce the number of evaluations and thus increase the computational efciency, Chuns approach  is used to approximate f0 (y) by f 0 yn % f 0 xn gtn 4

2. Geometrical non-linear analysis of structures Elementary information can be found in . 2.1. System equilibrium equations The equations of equilibrium of the system can be written as ff xg fPg 1

in which tn f yn f xn 5

in which {f} is the resultant of the nodal internal loads and {P} represents external nodal loads. The member force deformation relationships denote that {f} is a highly non-linear function of {x}.The differential form of Eq. (1) is tfDxg fDPg 2

assuming that the function g and its derivatives g0 and g00 are continuous in the neighborhood of 0. Now the two-point scheme (3) has the form 8 < yn xn ff0xn xn n 0,1, . . . 6 : xn 1 yn f0yn gtn f xn

where {Dx} and {DP} represent increments of displacement and load, respectively, and [t] is the system tangent stiffness matrix. A detailed discussion to calculate [t] is provided in Ref. .

3.2.1. Efciency evaluation By means of the efciency index, iterative methods can be compared easily. If w dened as the number of functional evaluations required per iteration step, and p is the order of convergence of an iteration method, then the efciency index e is given by 

3. Non-linear solution algorithm 3.1. NewtonRaphson method for the non-linear problems As mentioned, the NewtonRaphson method is one of the most powerful and widely used techniques for the solution of non-linear problems. Most iterative algorithms adapted in the

e p1=w

For example, the classical Newtons method xn 1 xn f xn =f 0 xn , is quadratically convergent, and needs to evaluate f and f0 . Therefore, the efciency index of Newtons method is 2(1/2) 1.4142. The efciency of two-point method can be evaluated as follows.

836

## H. Saffari, I. Mansouri / International Journal of Non-Linear Mechanics 46 (2011) 834840

Let a be a simple root of f, using Taylor series expansion of f around xn a and taking into account f(a) 0, gives  8 f xn f 0 a en c2 e2 c3 e3 c4 e4 Oe5 n n n n where en xn a, ck f k a=k!f 0 a, Oeln
1 X il

Functions g6g9 are applied in the proposed method. Also, the following function :   1 2 p 1 g10 t 18 t 1 14t satises the conditions of g(0) 1 and g0 (0) 2 in a limit case:

ci ei n

limg10 t 1,
t-0

## 0 lim g10 t 2 t-0

19

Using Eq. (8) the derivation form of f is f xn f 0 a 1 2c2 en 3c3 e2 4c4 e3 Oe4 n n n
0

10

Using Eq. (6) and applying the Taylor series to1/f0 (xn) one can reach
f xn 2 3 rn yn a en 0 c2 e2 2c3 c2 e2 7c2 c3 4c2 3c4 e4 Oe5 n n n n f xn

and is applied for new approach. Among these choices, the results (Tables 29) show that choice of a2, b0, c1, q 1: gt t 3 2t 1 1t 2 20

## is more efcient than the other forms obtained by trial.

Table 2 Comparison of number of iterations for star truss.

12

## Newtons method 106

Two-point method 71

## Decrease percentage 33.02

If function g be described by its Taylors polynomial of the second order at the point t 0 gtn g0 g 0 0tn g 00 0 2 t , 2 n tn f yn f xn 13

Using Eqs. (6) and (10)(13), the error equation (en 1 xn 1 a) can be obtained as follows:
2 f yn c2 4g0g 0 02 2c3 g0 1 e3 en 1 xn 1 a rn gtn 0 n f xn c2 c3 14g04g 0 07 3c4 g0 1
3 c2 413g0 7g 0 0g 00 0=2e4 Oe5 n n

Table 3 Comparison of the CPU time for star truss. Function g1 g2 g3 g4 g5 g6 g7 g8 g9 g10 Proposed g Decrease percentage 17.71 18.41 19.26 17.21 20.13 17.93 19.33 21.41 20.62 17.93 24.85

14

If function g has the conditions g0 1 and g 0 0 2 then the error equation is simplied   ! g 00 0 2 c2 c3 e4 Oe5 en 1 c2 5 n n 2 15

16
Table 4 Comparison of number of iterations for geodesic dome. Newtons method 99 Two-point method 77 Decrease percentage 22.22

This means that if g has above conditions, the two-step method has fourth-order convergence. As a consequence, its efciency index is equal to 4(1/2) 2 which is better than that of classic Newtons method 1.4142. 3.2.2. Choice of function g The function g has been searched in a general rational form 1 at bt2 ct3 gt 1 a2t qt 2 17

Table 5 Comparison of the CPU time for geodesic dome. Function g1 g2 g3 g4 g5 g6 g7 g8 g9 g10 Proposed g Decrease percentage 9.05 4.88 7.16 7.34 6.05 5.83 6.33 8.01 9.17 7.62 11.06

which satises conditions given in Eq. (15) and its parameters (a, b, c, q) have been selected using trial and error approach. In this way many different functions can be generated (Table 1). In this paper functions g1g5 have been chosen from .
Table 1 Various forms of function g. a g1(t) g2(t) g3(t) g4(t) g5(t) g6(t) g7(t) g8(t) g9(t) 0 2 0 0 2-c (cAR) 1 4 1 1 b 0 0 1 0 1 3 1 2 0 c 0 0 0 0 0 1 1 1 1 q 0 0 0 aAR 0 1 1 1 1

Table 6 Comparison of number of iterations for circular dome truss. Newtons method 104 Two-pint method 83 Decrease percentage 20.19

## H. Saffari, I. Mansouri / International Journal of Non-Linear Mechanics 46 (2011) 834840

837

Table 7 Comparison of the CPU time for circular dome truss. Function g1 g2 g3 g4 g5 g6 g7 g8 g9 g10 Proposed g Decrease percentage 7.18 7.94 8.11 7.78 7.66 5.84 6.12 7.02 6.54 6.01 8.85

B is transformation matrix that indicates the relationship between the nodal forces of the member in global and local coordinates systems. Step 7: Calculate the unbalanced load fDQ i 1 g fPi 1 gfF i 1 g 24

Step 8: Check for equilibrium. Using the latest available values of all geometric quantities, determine [T] for each member. Assemble the system tangent stiffness matrix, [t]. Step 9: Applying the unbalanced nodal forces as a load increment, determine the correction vector, {Dxi 1}, from the incremental relationship fDxi 1 g t1 fDQ i 1 g Step 10: Calculate {y}: 25

Table 8 Comparison of number of iterations for Schewdelers dome truss. Newtons method 90 Two-point method 61 Decrease percentage 32.22

## fyi 1 g fxi 1 g fDxi 1 g

26

update global nodal displacements. Step 11: Determine the new member forces: {fyi 1} and construct the vector of internal global forces, n o n o Fyi 1 B fyi 1 Step 12: Determine the unbalanced load: fDQyi 1 g fPi 1 gfFyi 1 g Step 13: Calculate the two parameters t and g: ti 1 fDQ i 1 g fDQyi 1 g
T fDQ i 1 g fDQ i 1 g T

Table 9 Comparison of the CPU time for Schewdelers dome truss. function g1 g2 g3 g4 g5 g6 g7 g8 g9 g10 Proposed g Decrease percentage 14.09 14.28 13.71 14.32 13.71 12.77 11.28 14.03 13.24 10.65 15.71

27

28

29

where the superscript T represents transpose; select function gi 1 based on Section 3.2. If the correction displacement vector is not small enough, determine the new node locations: fxi 2 g fyi 1 g gi 1 t1 fDQyi 1 g 30

3.3. Implementation of two-point method in computational mechanics A program implementing the two-point algorithm has been written in MATLAB and representative results are provided. Material non-linearities are not presently included in the algorithm. The algorithm is structured as follows: Step 1: Initialization of variables and parameters. Step 2: Input rst incremental load, {DP}, boundary conditions, connectivity, material properties and structural geometry. Step 3: Determine for each element, the tangent stiffness matrix, [T], in global coordinates based on the latest information concerning element geometry and forces. Assemble the system tangent stiffness matrix [t]. Step 4: Solve for the incremental global nodal displacements: fDPi 1 g t fDxi 1 g, i Z0 21

repeat Steps 613 until {Dx} is sufciently small according to a prescribed criterion. Step 14: Apply next load increment, {DP}, and return to Step 2. The convergence criterion, based on displacement, used herein is given by s P Dxi 2 re Pi 2 i Dxi

31

Step 5: Update global nodal displacements and determine the new geometric conguration from the equation: fxi 1 g fxi g fDxi 1 g Step 6: Compute the element force vector: {f construct the vector of internal global forces, n o n o F i 1 B f i 1
i1

22 }. Now,

23

838

## H. Saffari, I. Mansouri / International Journal of Non-Linear Mechanics 46 (2011) 834840

where e is the error tolerance. The graphical representation of this process is shown in Fig. 1. 4. Numerical examples Four numerical examples were solved in a microcomputer environment so that the efciency of the proposed procedure used together with the NewtonRaphson method in non-linear behavior of space trusses could be compared. The computer program was developed based on the procedure described in this paper. All examples have been solved with a 32 bit Pentium 2.00 GHz processor (2 CPUs). For the solution of non-linear equations, a new iterative method is adopted and the iterative process will stop when the convergence criteria are satised. All the numerical examples presented here use a tolerance of 10 3. The proposed method can be used for various structural systems, but in this paper numerical examples are applied for space trusses.

18
Newton's method new method

16 14 12
P (kN)

10 8 6 4 2 0

4.1. Star truss The rst example is shown in Fig. 2 with the two axes of the Cartesian co-ordinate system and the normal z-axis indicating the out-of-plane dimensions of the nodes. The dome is subjected to a single central concentrated load. The structure has 24 members and 13 nodes with a total of 21 degrees-of-freedom. The following properties are used for each member of the truss : the elastic modulus of the members E6895 kN/cm2 and the cross-sectional area A 6.452 cm2. Increment of load DP0.4448 kN and P 15.568 kN. This 3D truss was analyzed using the method developed in this paper. Fig. 3 shows the loaddisplacement curve obtained by applying the method developed during the present study. The results show that the number of iterations and the computing time used in deploying the classic NewtonRaphson approach is more than that used by applying our method. The computational results presented in Tables 2 and 3 show that in all of the cases the two-point scheme converges more rapidly than Newtons method and requires the less number of iterations, so that they can compete with Newtons method.

10
Displacement (cm)

15

20

## Fig. 4. Geodesic dome truss, dimensions are given in cm.

4.2. Shallow geodesic dome The geometric dimensions of the geodesic dome truss shown in Fig. 4 are taken from Ramesh and Krishnamoorthy . This truss has 156 members and 61 nodes with pin supports at the outer nodes and one vertical load P3.8 kN, DP 0.1 kN at the center. The elevation of the truss is dened by the following equation: x2 y2 z 7:22 60:84 32

## Fig. 2. Star dome truss, dimensions are given in cm.

All members have identical cross sections, i.e., with E6895 kN/cm2, A6.5 cm2. The loaddisplacement curve for this structure is shown in Fig. 5. The comparisons between the results of applying the two methods (NewtonRaphson and twopoint) are listed in Tables 4 and 5. The test results show that for all the functions tested, the two-point method introduced in the present work has equal or better performance as compared to the other methods of the same order.

## H. Saffari, I. Mansouri / International Journal of Non-Linear Mechanics 46 (2011) 834840

839

Newton's method

new method

900 800

Newton's method

new method

700 600

P (kN)

P (kN)
2 1 0 0 0.2 0.4 0.6 0.8 1 1.2 1.4

## 500 400 300 200 100 0 0 1 2 3 4 5 6

Displacement (cm)
Fig. 5. Loaddisplacement curve for geodesic dome.

Displacement (cm)
Fig. 7. Loaddisplacement curve for circular dome truss.

## Fig. 8. Schewdelers dome truss, dimensions are given in cm.

4.3. Circular dome truss The circular dome truss taken from  is shown in Fig. 6. This structure is subjected to a vertical load P 820 kN at the apex and has 168 elements with 73 nodes with a total of 147 degrees-offreedom. The out-of-plane motion has been constrained with pin supports added to each end of the truss. The cross-sectional area A is equal to 50.431 cm2 for all the members. The elastic modulus of the members E is 2.04 104 kN/cm2. Increment of load DP is 20 kN. Fig. 7 illustrates the numerical responses obtained from the proposed formulation for the geometric non-linear analysis. The computational results in Tables 6 and 7 show that two-point technique requires less CPU time than NewtonRaphson method.

Therefore, they are of practical interest and can compete with NewtonRaphson method and work well than it. 4.4. Schewdelers dome truss This truss, shown in Fig. 8 with 264 elements and 97 nodes with pin supports at the outer nodes, gives a possibility of comparison with results in the literature . The axial stiffness for all members is EA640 103 kN. The external loading was equipment loading which consists of P30 kN at the crown node and DP 1 kN. Fig. 9 shows the variation of vertical displacement with the load P. To compare the performance of the two-point method, the results of analyses are summarized in Tables 8 and 9. The computations are carried out with a high precision. It can be

840

## H. Saffari, I. Mansouri / International Journal of Non-Linear Mechanics 46 (2011) 834840

35

Newton's method

new method

comments and suggestions which resulted in considerable improvement in the quality of this paper.

30

References
25
 G.E. Blandford, Progressive failure analysis of inelastic space truss structures, Comput. Struct. 58 (1996) 981990.  W.E. Haisler, J.A. Stricklin, F.J. Stebbins, Development and evaluation of solution procedures for geometrically nonlinear structural analysis, AIAA J. 10 (1972) 264272.  D.A. Massett, J.A. Stricklin, Self-correcting incremental approach in nonlinear structural mechanics, AIAA J 9 (1971) 24642466.  J.A. Stricklin, W.E. Haisler, W.A. Von Riesemann, Self-correcting initial value formulations in nonlinear structural mechanic, AIAA J. 9 (1971) 20662067.  H. Saffari, M.J. Fadaee, R. Tabatabaei, Nonlinear analysis of space trusses using modied normal ow algorithm, J. Struct. Eng. ASCE 134 (2008) 9981005.  R. Tabatabaei, H. Saffari, M.J. Fadaee, Application of normal ow algorithm in modal adaptive pushover analysis, J. Constr. Steel Res. 65 (2009) 8996.  M. Papadrakakis, C.J. Gantes, Truncated Newton methods for nonlinear nite element analysis, Comput. Struct. 30 (1988) 705714.  J. Pina, A. Costa, J. Appleton, Formal solution of quasi-static problems, Int. J. Non-Linear Mech. 45 (2010) 525534.  M. Greco, F.A.R. Gesualdo, W.S. Venturini, H.B. Coda, Nonlinear positional formulation for space truss analysis, Finite Elem. Anal. Des. 42 (2006) 10791086.  R. Tabatabaei, H. Saffari, Large strain analysis of two-dimensional frames by the normal ow algorithm, Struct. Eng. Mech. 36 (2010) 529544.  H. Thai, S. Kim, Large deection inelastic analysis of space trusses using generalized displacement control method, J. Constr. Steel Res. 65 (2009) 19871994.  M. Papadrakakis, Post-buckling analysis of spatial structures by vector iteration methods, Comput. Struct. 14 (1981) 393402.  P.X. Bellini, The concept of snap-buckling illustrated by a simple model, Int, J. Non-Linear Mech. 7 (1972) 643650.  A. Kassimali, H. Parsi-Feraidoonian, Nonlinear behavior of prestressed cable trusses, J. Constr. Steel Res. 7 (1987) 435450.  G. Ramesh, C.S. Krishnamoorthy, Inelastic post-buckling analysis of truss structures by dynamic relaxation method, Int. J. Numer. Methods Eng. 37 (1994) 36333657.  C. Chun, A note on some iterative schemes for nonlinear equations, Appl. Math. Comput. 188 (2007) 271274.  M.A. Noor, K.I. Noor, Fifth-order iterative methods for solving nonlinear equations, Appl. Math. Comput. 188 (2007) 406410.  C. Chun, Iterative methods improving Newtons method by the decomposition method, Comput. Math. Appl. 50 (2005) 15591568.  J. Kou, Y. Li, X. Wang, A modication of Newton method with third-order convergence, Appl. Math. Comput. 181 (2006) 11061111.  M.A. Noor, F. Ahmad, S. Javeed, Two-step iterative methods for nonlinear equations, Appl. Math. Comput. 181 (2006) 10681075.  M.A. Noor, Some iterative methods free from second derivatives for nonlinear equations, Appl. Math. Comput. 192 (2007) 101106.  M.S. Petkovic, L.D. Petkovic, Families of optimal multipoint methods for solving nonlinear equations: a survey, J. Appl. Anal. Discrete Math. 4 (2010) 122.  A. Kassimali, E. Bidhendi, Stability of trusses under dynamic loads, Comput. Struct. 29 (1988) 381392.  S.L. Chan, P.P.T. Chui, Non-linear Static and Cyclic Analysis of Steel Frames with Semi-rigid Connections, rst ed, Elsevier Science Ltd., The Netherlands, 2000.  M.A. Criseld, Non-linear Finite Element Analysis of Solids and Structures, vol. 1, John Wiley, England, 1991.  M.S. Petkovic, L.D. Petkovic, A one parameter square root family of two-step root-nders, J. Appl. Math. Comput. 188 (2007) 339344.

P (kN)

20

15

10

0 0 2 4 6 8 10

Displacement (cm)
Fig. 9. Loaddisplacement curve for Schewdelers dome truss.

easily seen that less number of iterations and computing time are used by two-point technique due to its higher order.

5. Conclusions The purpose of this paper was to develop a more efcient implementation for deployment of load control methods. Only geometric non-linearity is considered in this study. The nonlinear equations are solved using the two-point method. A numerical procedure was extended and implemented into a computer program. Several well-known algorithms and a new method, was proposed, and results compared with those obtained by deploying the classic NewtonRaphson method. The numerical examples represented a good compromise between reducing the computing time, reducing the number of the iterations and obtaining results sufcient accuracy. Deployment of our twopoint method can improve work efciency for a wide range of problems which have non-linear behavior but for which passing limit points is not an issue.

Acknowledgements The authors would like to express their deep gratitude to Dr. Jim Peterson, Prof. Miodrag Petkovic, and reviewers for their