TAPABRATA RAY
Email Address
tsltray@nus.edu.sg
18 results
Publication Search Results
Now showing 1 - 10 of 18
Publication Leader identification and leader selection: Its effect on a swarm's performance for multi-objective design optimization problems(2004-09) Liew, K.M.; Tan, P.K.; Ray, T.; TEMASEK LABORATORIESIn this paper, three distinct swarm strategies for the optimization of engineering design problems with multiple objectives are presented. These strategies build upon the swarm algorithm of Ray et al by incorporating new processes which improve the performance of their predecessor. The constraint-matching strategy calls for the mating of solutions based on constraint satisfaction characteristics. Local search entails the thorough exploration of regions in parametric space where good solutions potentially reside. Migrating leaders prescribes the exchange of information between the best performing members of the swarm. As proof of their utility, the strategies were used to solve two well-studied optimal engineering design problems. Solutions obtained by the strategies are comparable with those of other optimization approaches documented in the literature.Publication A framework for optimization using approximate functions(2003) Won, K.S.; Ray, T.; Tai, K.; TEMASEK LABORATORIESPopulation-based, stochastic, zero-order optimization methods (e.g. genetic and evolutionary algorithms) are a popular choice in solving intractable, real-life optimization problems. These methods are particularly attractive as they are easy to use and do not require assumptions about functional and slope continuities unlike some of its gradient-based counterparts. Despite their advantages, these methods require the evaluation of numerous candidate solutions, which is often computationally expensive and practically prohibitive. We introduce a framework for optimization using approximate functions. The optimization algorithm is a population-based, stochastic, zero-order, elite-preserving algorithm that makes use of approximate function evaluations in lieu of actual function evaluations. The approximate function is constructed using a radial basis function (RBF) network and the network is periodically retrained after a few generations unlike other models which create and use the same approximate model repeatedly without retraining. A scheme for controlled elitism is incorporated within the optimization framework to ensure convergence in the actual function space. The computational accuracy and efficiency of the proposed optimization framework is assessed using a set of five mathematical test functions. The results clearly indicate that the optimization framework using approximations is able to arrive at reasonably accurate results using only a fraction of actual functions evaluations. © 2003 IEEE.Publication Optimal process design of sheet metal forming for minimum springback via an integrated neural network evolutionary algorithm(2004-02) Liew, K.M.; Tan, H.; Ray, T.; Tan, M.J.; TEMASEK LABORATORIESThe process of sheet metal forming is characterized by various process parameters. Accurate prediction of springback is essential for the design of tools used in sheet metal forming operations. In this paper, an evolutionary algorithm is presented that is capable of handling single/multiobjective, unconstrained and constrained formulations of optimal process design problems. To illustrate the use of the algorithm, a relatively simple springback minimization problem (hemispherical cupdrawing) is solved in this paper, and complete formulations of the algorithm are provided to deal with the constraints and multiple objectives. The algorithm is capable of generating multiple optimal solutions in a single run. The evolutionary algorithm is combined with the finite element method for springback computation, in order to arrive at the set of optimal process parameters. To reduce the computational time required by the evolutionary algorithm due to actual springback computations via the finite element method, a neural network model is developed and integrated within the evolutionary algorithm as an approximator. The results clearly show the viability of the use of the evolutionary algorithm and the use of approximators to derive optimal process parameters for metal forming operations.Publication A parallel hybrid optimization algorithm for robust airfoil design(2004) Ray, T.; Tsai, H.M.; TEMASEK LABORATORIESWe present a hybrid algorithm that consists of a population based, stochastic, zero-order optimization algorithm and a gradient based optimization algorithm for robust airfoil design optimization. The gradient based optimization algorithm used in the hybrid is to repair solutions (to satisfy the equality constraints) for the problem in contrast to other hybrids where it is usually used to improve the final solution obtained by the stochastic algorithm. Approaches involving equality constraints are known to pose difficulties to existing stochastic methods. The inequality constraints in the present stochastic algorithm are handled via the concept of non-dominance, instead of scaling and aggregating the constraint violations. To demonstrate the behavior of the current proposed hybrid algorithm, we present results of two single objective and two multi-objective airfoil design optimization problems. For comparison, a result of an airfoil design using aggregation is included to highlight some of the limitations of aggregation based formulations used in robust design. To improve upon the computational cost for such computationally expensive problems, we have implemented the algorithm to run on multiple processors based on master-slave architecture.Publication A Comparative Study of Evolutionary Algorithm and Swarm Algorithm for Airfoil Shape Optimization Problems(2003) Tan, C.M.; Ray, T.; Tsai, H.M.; TEMASEK LABORATORIESShape optimization of airfoils involves highly expensive, nonlinear objective(s) and constraint functions. Zero order, stochastic methods are often used to handle such problems. In this paper, we report the performance of two such methods; Evolutionary Algorithm (EA) and the Swarm Algorithm (SWARM) on five airfoil shape optimization problems. Both the EA and the SWARM algorithm used here are variants of their original form suited to handle multiple objectives and multiple constraints without aggregation. Their original form were meant to solve unconstrained single objective problems. The present studies indicate that the EA marginally outperforms SWARM for single objective problems in terms of the computational efficiency while for the multiple objective problems, the SWARM exhibits a better overall performance in locating the Pareto front. Both methods exhibit fast convergence capabilities and provide the designer the flexibility of cost function choices that is necessary to solve various forms of the shape optimization problems. © 2003 by T. Ray and H.M. Tsai.Publication Multilayer dielectric filter design using a multiobjective evolutionary algorithm(2005-11) Venkatarayalu, N.V.; Ray, T.; Gan, Y.-B.; TEMASEK LABORATORIESDesign of multilayer dielectric filters involve the identification of suitable dielectric material and appropriate thicknesses of the layers that best satisfies the desired frequency response for the application. Such problems, like any other practical design optimization problem require simultaneous consideration of multiple objectives and constraints. In this paper, we introduce a multiobjective evolutionary algorithm that is capable of handling unconstrained and constrained, single and multiobjective problems without any restriction on the number and nature of variables, constraints and objectives. The algorithm handles constraints and objectives separately using two fitness measures derived out of nondominance, unlike most of its counterparts which use a single fitness measure. Unlike most evolutionary algorithms where only the good parents participate in mating, our algorithm ensures that all solutions participate in mating, which is useful for exploring highly nonlinear search spaces. The diversity of the solutions is controlled by the partner selection scheme that prefers elites with distant neighbors as mating partners. The results of two multiobjective test problems, three multilayer dielectric filter designs (low-pass, bandpass and stopband) and one variable layer low-pass filter design are presented in this paper to highlight the benefits offered by our algorithm in terms of modeling flexibility, computational efficiency and its ability to arrive at competitive nondominated designs. A comparison of our results with those obtained using a single objective aggregated formulation for the stopband filter design is also presented. We have also compared the performance of our algorithm with non-dominated sorting genetic algorithm (NSGA-II) for the low-pass filter design where our results are better. © 2005 IEEE.Publication Aircraft configuration design using a multidisciplinary optimization approach(2004) Rao, C.S.; Ray, T.; Tsai, H.M.; TEMASEK LABORATORIESAt the conceptual phase of an aircraft design process, the aim is to determine the set of design features such as configuration arrangement, planform geometry, wing area, engine configuration and weight that meet various performance characteristics. Multidisciplinary design optimization (MDO) at this phase allows the designer to consider various options via simultaneous interactions of aerodynamics, propulsion, structures, stability and flight mechanics to arrive at better designs. Earlier attempts with MDO seek to optimize a design for a predetermined aircraft configuration or optimize a base design for different mission profiles. In this paper, we propose a MDO framework for conceptual aircraft design that considers different aircraft configurations simultaneously. The advantage of this approach is to allow the evolution of various aircraft configurations. The optimization algorithm used in the framework is based on computational intelligence which is a stochastic, zero-order, population based algorithm, especially suitable for multi-objective, constrained optimization problems involving computationally expensive functions. The work is aimed at providing insights in how different mission requirements dictate configuration choices. In the present work we study the evolution of different two-seater, propeller driven aircraft configurations for different mission requirements and we limit it to two configurations namely, conventional wing-tail and canard configurations.Publication Society and civilization: An optimization algorithm based on the simulation of social behavior(2003-08) Ray, T.; Liew, K.M.; TEMASEK LABORATORIESThe ability of an individual to mutually interact is a fundamental social behavior that is prevalent in all human and insect societies. Social interactions enable individuals to adapt and improve faster than biological evolution based on genetic inheritance alone. This is the driving concept behind the optimization algorithm introduced in this paper that makes use of the intra and intersociety interactions within a formal society and the civilization model to solve single objective constrained optimization problems. A society corresponds to a cluster of points in the parametric space while a civilization is a set of all such societies at any given point of time. Every society has its set of better performing individuals (henceforth, referred as leaders) that help others in the society to improve through an intrasociety information exchange. The intrasociety information exchange results in the migration of a point toward a better performing point in the cluster that is analogous to an intensified local search around a better performing point. Leaders of a society on the other hand improve only through an intersociety information exchange that results in the migration of a leader from a society to another that is headed by better performing leaders. This process of leader migration helps the better performing societies to expand and flourish that correspond to a search around globally promising regions in the parametric space. In order to study the performance of the proposed algorithm, four well-studied, single objective constrained engineering design optimization problems have been solved. The results indicate that the algorithm is capable of arriving at comparable solutions using significantly fewer function evaluations and stands out as a promising alternative to existing optimization methods for engineering design. Futhermore, the algorithm employs a novel nondominance scheme to handle constraints that eliminates the problem of scaling and aggregation that is common among penalty-function-based methods.Publication Constrained robust optimal design using a multiobjective evolutionary algorithm(2002) Ray, T.; TEMASEK LABORATORIESA major fraction of evolutionary optimization methods aims to find solutions that maximize performance. However, a solution that solely maximizes performance is of no practical use as it may be too sensitive to parametric variations (nonuniform material properties, inexact physical dimensions, uncertainties in loading and operating conditions, etc.). Furthermore, for design problems with constraints, a robust solution needs to be feasible and remain feasible under parametric variations. In this paper, a new evolutionary algorithm is proposed that is capable of handling constrained robust optimal design problems. A multiobjective formulation is introduced that considers an individuals' performance, the mean performance of its neighbors and the standard deviation of its neighbors' performance as three objectives for optimization. In order to handle feasibility, an innovative constraint-handling scheme based on the Pareto concept is introduced that considers an individual's self-feasibility and its neighborhood feasibility. Robust optimal solutions to two engineering design examples are reported in this paper. Results of simulations are also presented to illustrate the differences between an optimal solution and a robust optimal solution. © 2002 IEEE.Publication Golinski's speed reducer problem revisited(2003-03) Ray, T.; TEMASEK LABORATORIESThe article provides a comparison between the results reported by various sources to Golinski's speed reducer problem. A swarm algorithm has been used to solve the nonlinear constrained minimization problem and arrive at the best known feasible solution. A swarm size of 70 has been used and the results are after 70,000 function evaluations.