Design and implementation of Improved Differential Evolution Algorithm for Engineering Optimization Problems

Abstract

Nature Inspired Algorithms (NIA) are used to solve the complex optimization problems. If the user is not very cautious about the exact solution of the problem then NIA may be a good choice to provide the solutions in a given time frame. The NIA are population based meta-heuristics and generally categorizes into Evolutionary Strategies and Swarm Intelligence based strategies. In this synopsis, Diand#64256;erential Evolution (DE) algorithm, which is an evolutionary algorithm, is selected as a key research area. The algorithm is selected due to its eand#64259;ciency to solve the complex optimization problems and its simplicity to implement. Researchers are gradually working to improving the solution search capacity of the DE algorithm. The DE algorithm has been applied to several real world engineering optimization problems and proved an eand#64259;cient strategy in the and#64257;eld of evolutionary optimization algorithms. But like other evolutionary algorithms DE also inherits some drawbacks. In DE, the variation in solutions during the solution search process is controlled by two signiand#64257;cant control parameters, namely scale factor (F) and crossover probability (CR). These parameters play important role for balancing the exploration and exploitation capabilities in the solution search region. Therefore, and#64257;ne tuning of these parameters are very necessary to obtain the global optima. Researchers are continuously working to and#64257;nd a dynamic and#64257;ne tuning strategy for these parameters but yet to not succussed. Further, DE sometimes stagnated to a point which is neither local optima or global optima. The DE also not able to balance the diversity and convergence abilities of the solutions during the solution search process. Therefore, there are signiand#64257;cant possibilities to improve the eand#64259;ciency of the DE search process and develop eand#64259;cient variants of DE. Further, the developed variants are to be tested by applying them to solve the real world engineering optimization problems. This synopsis contains a brief review of the DE algorithm as well as to improve the eand#64259;cie