HYBRID MODELLING OF ELECTRO STREAM DRILLING PROCESS

Abstract

Since the intelligent optimization techniques themselves have certain limitations in modelling complex manufacturing processes, such as neural network, genetic algorithm and simulated annealing etc, their applications in hybrid form have emerged as a new area of research. In artificial intelligence, the hybrid system refers to the combine of any two intelligent techniques e.g. artificial neural network and genetic algorithm. Recently the application of hybrid approach in modeling and optimization the process have paved the way for the wide scale use of nontraditional machining processes in different industries such as aviation, space, automobile, and electronics and computers. Due to the high complexity of the nontraditional machining processes such as electro-stream drilling (ESD), it has been hard to establish models that accurately predict and correlate the process variables and performance characteristics. In the proposed hybrid approach to model ESD process, the capability of neural network to model and predict ill- structured data is exploited together with the power of genetic algorithm for optimization. The proposed hybrid approach involves (a) artificial neural network to predict the objective parameters of the process, (b) desirability function to reduce the multi response problem into single response i.e. the network output in case of multi-response problem get transformed into single response which in turn acts as the fitness function to genetic algorithm, and (c) genetic algorithm to optimize. The developed methodology can be highly beneficial to manufacturing industries interested in maintaining close control over the process.