Multiobjective Optimization of Green Sand Mould System Using DE and GSA

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Multiobjective Optimization of Green Sand Mould System using DE and GSA

T.Ganesan1
1

P.Vasant2
2

I.Elamvazuthi2,
3

Ku Zilati Ku Shaari4,
1

Department of Chemical Engineering, Universiti Technologi Petronas, Malaysia, email: tim.ganesan@gmail.com

Department of Fundamental & Applied Sciences, Universiti Technologi Petronas, Malaysia.

Department of Electrical & Electronics Engineering, Universiti Technologi Petronas, Malaysia.

Department of Chemical Engineering, Universiti Technologi Petronas, Malaysia.

AbstractMost optimization cases in recent times present themselves in a multi-objective (MO) setting. Hence, it is vital for the decision maker (DM) to have in hand multiple solutions prior to selecting the best solution. In this work, the weighted sum scalarization approach is used in conjunction with two meta-heuristic algorithms; differential evolution (DE) and gravitational search algorithm (GSA). These methods are then used to generate the approximate Pareto frontier to the green sand mould system problem. Some comparative studies were then carried out with the algorithms in this work and that from the previous work. Examinations on the performance and the quality of the solutions obtained by these algorithms are shown here. Keywords-multi-objective (MO); industrial optimization; green sand mould system; weighted sum approach;differential evolution (DE); gravitational search algorithm (GSA); Hypervolume Indicator (HVI) ; approximate Pareto frontier.

frontier approximations produced by the algorithms were carried out using the Hypervolume Indicator (HVI) metric [11], [12]. Comparison studies were then conducted on the individual best solutions as well as the frontier approximations obtained in this work against those obtained in Surekha et al [8]. This paper is organized as follows. In section II of this paper, the computational techniques are presented. The HVI metric is discussed and this is followed by description on the green mould sand MO problem statement in section III. Section IV discusses computational results and finally, the concluding remarks are given in section V. II. COMPUTATIONAL TECHNIQUES A. Differential Evolution (DE) DE is a class of evolutionary meta-heuristic algorithms first introduced in 1995 by Storn and Price [9]. This core idea of this technique is the incorporation of perturbative methods into evolutionary algorithms. DE starts by the initialization of a population of at least four individuals denoted as P. These individuals are realcoded vectors with some size N. The initial population of individual vectors (the first generation denoted gen = 1) are randomly generated in appropriate search ranges. One principal parent denoted xpi and three auxiliary parents denoted xai is randomly selected from the population, P. In DE, every individual in the population, P would become a principle parent, xpi at one generation or the other and thus have a chance in mating with the auxiliary parents, xai. The three auxiliary parents then engage in differential mutation to generate a mutated vector, Vi. Algorithm 1: Differential Evolution (DE) Step 1: Initialize individual size N, P, CR and F Step 2: Randomly initialize the population vectors, xGi. Step 3: Randomly select one principal parents, xpi Step 4: Randomly select three auxilary parents, xai Step 5: Perform differential mutation & generate mutated vector, Vi Step 6: Recombine Vi with xpi to generate child trial vector, xchildi Step 7: Perform knock-out competition for next generation survival selection

I. INTRODUCTION In recent times, many concerns have been raised regarding emerging technologies in engineering optimization which present themselves in a multiobjective (MO) setting [1], [2]. Some well known scalarization techniques include the Weighted Sum method [3], [4], Goal Programming [5] and NormalBoundary Intersection method (NBI) [6]. In MO optimization problems, determining the most efficient solution set can be a very daunting process. Many varieties of concepts (such as; diversity and convergence) have been proposed in the last years. These ideas were then used as indicators to evaluate the solution set produced by the optimization algorithm [7]. Such evaluations were used to benchmark the algorithms performance. This work aims to produce a set of solutions that dominantly approximates the Pareto frontier in the objective space of the green sand mould system. This problem was presented and attempted by using of genetic algorithm (GA) and particle swarm optimization (PSO) techniques (using weighted sum scalarization method) in Surekha et al [8]. In this work, the green mould sand system problem [8] was tackled using Differential Evolution (DE) [9] and Gravitational Search Algorithm (GSA) [10] in conjunction with the weighted sum approach to generate a series of solutions that dominantly approximate the Pareto frontier. The dominance ranking among the

978-1-4673-5118-8 c 2012 IEEE

255

Step 8: If the fitness criterion is satisfied and t= Tmax , halt and print solutions else proceed to step 3 B. Gravitational Search Algorithm (GSA) The GSA algorithm is a meta-heuristic algorithm first developed in 2009 by E.Rashedi et al [10]. This technique was inspired by the law of gravity and the idea of interaction of masses. This algorithm uses the Newtonian gravitational laws where the search agents are the associated masses. Thus, the gravitational forces influence the motion of these masses, where lighter masses gravitate towards the heavier masses (which signify good solutions) during these interactions. The gravitational force hence acts as the communication mechanism for the masses. The position of the masses correlates to the solution space in the search domain while the masses characterize the fitness space. As the iterations increase, and gravitational interactions occur, it is expected that the masses would conglomerate at its fittest position and provide an optimal solution to the problem.

objective functions and the range of the decision variables are shown as follows:
f 1 = 17.2527 - 1.7384A - 2.7463B + 32.3203C + 6.575D + 0.014A 2 + 0.0945B 2 - 7.7857C 2 1.2079D + 0.0468AB - 0.1215AC - 0.0451AD + 0.5516BC + 0.6378BD + 2.689CD)
2

(1)

f 2 = 1192.51 - 15.98A - 35.66B + 9.51C - 105.66D + 0.07A 2 + 0.45B 2 4.13C 2 + 4.22D 2 + 0.11AB + 0.2AC + 0.52AD + 1.19BC + 1.99BD - 3.1CD
(2)

f3 = 38.2843- 0.0494A+ 2.4746B+ 7.8434C+ 7.774D+ 0.001A2 - 0.00389B2 1.6988C2 - 0.6556D2 - 0.0015AB- 0.0151AC0.0006AD- 0.075BC - 0.1938BD+ 0.65CD
f 4 = 1.02616 + 0.01316A - 0.00052B 0.06845C + 0.0083D - 0.00008A 2 + 0.0009B 2 + 0.0239C 2 - 0.00107D 2 0.00004AB - 0.00018AC + 0.00029AD - 0.00302BC - 0.00019BD - 0.00186CD
52 A 94 8 B 12 1 .5 C 3 3 D 5

(3)

Algorithm 2: Gravitational Search Algorithm (GSA) Step 1: Initialize no of particles, mi and initial positions, xi(0) Step 2: Initialize algorithm parameters G(0) , . Step 3: Compute gravitational & inertial masses based on the fitness map Step 4: Compute the gravitational constant, G(t) Step 5: Compute distance between agents, Rij(t) Step 6: Compute total force, Fid(t) and the acceleration aid(t) of each agent. Step 7: Compute new velocity vi(t) and position xi(t) for each agent Step 8: If the fitness criterion is satisfied and t= Tmax , halt and print solutions else proceed to step 3 III. APPLICATION DATA In the green sand mould system, the response parameters of the mould heavily influence the quality of the final product. In Surekha et al [8], these parameters are selected as the objective functions. The responses parameters are; green compression strength (f1), permeability (f2), hardness (f3) and bulk density (f4). These objectives on the other hand are influenced by on the process variables which are; the grain fineness number (A), percentage of clay content (B), percentage of water content (C) and number of strokes (D). The

(4)

(5)

The MO optimization problem statement for the green mould system problem is shown as follows:

Max ( f1 , f 2 , f 3 , f 4 ) 52 A 94 8 B 12 1 .5 C 3 3 D5

subject to

(6)

The algorithms used in this work were programmed using the C++ programming language on a personal

256

2012 4th International Conference of Soft Computing and Pattern Recognition (SoCPaR)

Figure 1: Pareto frontiers of the objectives obtained by the GSA method

Figure 2: Pareto frontiers of the objectives obtained by the DE method

2012 4th International Conference of Soft Computing and Pattern Recognition (SoCPaR)

257

computer (PC) with an Intel dual core processor running at 2 GHz. IV. RESULTS AND ANALYSIS In this work, the solution sets which are the approximations of the Pareto frontier were obtained using the DE and the GSA methods. The quality of these solutions was measured using the HVI. For the approximation of the Pareto frontier, 21 solution for various weights were obtained for both the algorithms. The approximate Pareto frontiers obtained using the GSA and DE algorithms are shown in Figures 1 and 2 respectively. It can be observed using the HVI that the best solution obtained by DE algorithm dominates the best solution produced by the GSA algorithm by 21.19%. Besides, the number of function evaluations taken to obtain the best solution by the DE method is lesser than the GSA method. The comparison of the best candidate solutions obtained by the DE and PSO methods in this work as well as the PSO method in Surekha et al [8] is shown in Table 1. The HVI computed for the entire frontier of each solution set produced by an algorithm gives the true measure of dominance when compared with another algorithm. In this work, the HVI for the entire frontier was computed for each of the algorithm. The HVI for the entire frontier for the solution sets produced by the PSO [8], DE as well as GSA and the associated execution time is shown in Table 2.
TABLE I: THE COMPARISON OF THE BEST SOLUTIONS
OBTAINED BY THE ALGORITHMS

the algorithm to produce the entire approximate Pareto frontier, the DE method takes the longest time followed by the GSA and the PSO [8] method respectively. Although the DE algorithm produces the most dominant frontier, it sacrifices computational time as compared to the GSA and PSO [8] methods. In this work, the DE and GSA algorithms performed stable computations during the program executions. All Pareto-efficient solutions produced by the algorithms developed in this work were feasible and no constraints were compromised. Although, the DE algorithm performs well relative to other algorithms used in this work, it can be clearly seen that the execution time is much higher than the one obtained by Surekha et al [8] using the PSO method. Since DE is an evolutionary-type algorithm, the diversification of the search space is high and thus resulting in high computational time as compared with the GSA and PSO methods which are swarm-type algorithms. The GSA method can be said to be the second best optimizer as compared to the DE method. Besides, in comparison with the PSO algorithm the GSA method produces more superior results.
V. CONCLUSIONS

Description f1 Objective Function f2 f3 f4 x1 Decision Variable x2 x3 x4 HVI

PSO [8] 55.4112 107.8949 84.7936 1.5079 52.0001 11.9998 2.8452 4.9999 8876.715583

DE 50.2281 137.769 86.289 1.50073 53.0457 9.39392 2.99849 4.39392 71665.771

GSA 34.6562 205.619 81.1777 1.46948 52.0015 8.00543 2.49446 3.19889 59134.56

In this work, a new local maximum and a more dominant approximation of the Pareto frontier was achieved using the DE method.. When gauged with the HVI metric, the DE algorithm produced the most dominant approximate of the Pareto frontier as compared to the GSA and the PSO [8] methods. For future works, other meta-heuristic algorithms such as Genetic Programming (GP) [13], Analytical Programing (AP) [14], Hybrid Neuro-GP [15], MO evolutionary algorithm [16], [17] and Hybrid NeuroSwarm [18] should be applied to the green mould system. More large-scale MO problems should be studied using the DE and GSA algorithms for a better understanding of the mentioned algorithms performance and efficiency. ACKNOWLEDGEMENTS The authors acknowledge the financial support by Universiti Teknologi PETRONAS, Malaysia. REFERENCES
[1] [2] H. Eschenauer, J. Koski, and A. Osyczka, Multicriteria Design Optimization, Springer-Verlag, Berlin, 1990. R.B. Statnikov and J.B. Matusov, Multicriteria Optimization and Engineering, Chapman and Hall, New York, 1995. P.C. Fishburn, Additive Utilities with Incomplete Product Set: Applications to Priorities and Assignments, Operations Research Society of America (ORSA), Baltimore, MD, U.S.A, 1967. E. Triantaphyllou, Multi-Criteria Decision Making: A Comparative Study. Dordrecht, The Netherlands, Kluwer Academic Publishers (now Springer), 2000, pp 320. M. L. Luyben and C. A. Floudas, Analyzing the interaction of design and control. 1. A multiobjective

TABLE II: THE HVI OBTAINED BY THE ALGORITHMS AND THE


COMPUTATIONAL TIME FOR THE ENTIRE FRONTIER

PSO [8] HVI Computational time (secs) 130,108.48 0.013

DE 492, 5326.52 39.08

GSA 309, 5494.45 21.20

[3]

A new optima is achieved by the DE method (see Table 1) since it outperforms the GSA and the PSO [8] methods. Thus, it can be said that the DE method in this work outweighs the overall optimization capabilities of GSA and PSO [8] (see HVI value in Table 2). In terms of computational time taken for

[4]

[5]

258

2012 4th International Conference of Soft Computing and Pattern Recognition (SoCPaR)

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

[15]

[16]

[17]

[18]

framework and application to binary distillation synthesis, Computers and Chemical Engineering, 18 (1). 1994, pp 933-969. I. Das and J.E. Dennis, Normal-boundary intersection: A new method for generating the Pareto surface in nonlinear multicriteria optimization problems, SIAM Journal of Optimization 8(3) ,1998, pp 631657. C. Grosan, M. Oltean., D.Dumitrescu, Performance Metrics for Multiobjective Optimization Evolutionary Algorithms, In Proceedings of Conference on Applied and Industrial Mathematics (CAIM), Oradea, 2003. B. Surekha, Lalith K. Kaushik, A. K. Panduy, A.P. R. Vundavilli & M. B. Parappagoudar, Multi-objective optimization of green sand mould system using evolutionary algorithms, International Journal of Advance Manufacturing Technoloqy, , 2011, pp 1-9. R. Storn and K. V. Price, Differential evolution a simple and efficient adaptive scheme for global optimization over continuous spaces, ICSI, Technical Report TR-95-012, 1995. E.Rashedi, H.Nezamabadi-pour, S.Saryazdi, GSA: A Gravitational Search Algorithm, Information Sciences ,Vol. 179, 2009, pp 22322248. E. Zitzler and L. Thiele. Multiobjective Optimization Using Evolutionary Algorithms - A Comparative Case Study. In Conference on Parallel Problem Solving from Nature (PPSN V), 1998, pp 292301. E. Zitzler and L. Thiele. Multiobjective evolutionary algorithms: a comparative case study and the strength Pareto approach. IEEE Transactions on Evolutionary Computation, 3(4), 1999, pp 257271. J.R. Koza, Genetic Programming: On the Programming of Computers by means of Natural Selection, MIT Press, USA, 1992. I. Zelinka, Analytic programming by Means of SOMA Algorithm. Mendel 02, In: Proc. 8th, International Conference on Soft Computing Mendel02, Brno, Czech Republic, 2002, pp 93-101. T. Ganesan, P. Vasant and I. Elamvazuthi, Optimization Of Nonlinear Geological Structure Mapping Using Hybrid Neuro-Genetic Techniques, Mathematical and Computer Modelling, 54 (11-12) ,2011, pp 2913 2922. B.Y. Qu and P.N. Suganthan, Multi-objective evolutionary algorithms based on the summationof normalized objectives and diversified selection, Information Sciences, 180 , 2010, pp 31703181. Ke Li, S. Kwong, J. Cao, M. Li, J. Zheng and R. Shen, Achieving Balance Between Proximity And Diversity In Multi-Objective Evolutionary Algorithm, Information Sciences, 182, 2011, pp 220242. I. Elamvazuthi, T. Ganesan, and P. Vasant, A comparative study of HNN and Hybrid HNN-PSO techniques in the optimization of distributed generation (DG) power systems, International Conference on Advanced Computer Science and Information System (ICACSIS), 2011, pp 195 200.

2012 4th International Conference of Soft Computing and Pattern Recognition (SoCPaR)

259

You might also like