Research On The Simulation Of The Wire Electrical Discharge Machining Process Based On The Genetic Neural Network

Wire Electrical Discharge Machining (WEDM) process is changeable and difficult to control quantitatively. Many factors influence the process and they are related and contradictive. It is a key problem that under setting machining parameters reasonably reaching the processing effect. Depend on the experience of technical staff singly, it is difficult to get the good surface roughness and machining accuracy, and achieve fast processing speed. So it's necessary to establish a reasonable and effective WEDM processing model. It can realize the forecasting of the processing effect in the whole scope. This article focuses on the application of a computational intelligence approach which integrated with genetic algorithm and neural network, and proposes to modeling and optimization of WEDM based on the genetic neural network.The neural network imitating the structure and function of human brain cells is very suitable to solve the nonlinear problem. Especially back propagation neural network (BP neural network for short) can approach random function with random accuracy. The advantage of BP algorithm is of good accuracy to optimize parameters. But at the same time it has some mainly disadvantages such as running into apices in part, slow convergent rate and oscillation effect. Genetic Algorithm (GA) based on natural selection theory and genetics has good global search property. Combining the neural network with the genetic algorithm is to be a mix algorithm-genetic neural network (GA-BP). By analyzing the law of WEDM processing, this article selects the main factors affecting process: the thickness of the work-piece, pulse width, pulse interval, pulse height, as the input vectors. And select the processing speed, surface roughness as output vectors to establish WEDM processing model. By the model forecasting the processing speed and surface roughness under some processing requirement, find out the processing law. By adoption of the second rotation design, do experiment and compare the experimental data with the predicting results. The results show that the established model reaches high fitting and forecasting precision. At the same time, use genetic algorithm to optimize processing parameters. From the verification of experiment, the optimized results are validity and practicability.