Research And Development On Expert System Of Process Database For WEDM

Reciprocating WEDM is developed by our country, which has good development in the technology level, machine structure and process performances. But compared with international low-speed WEDM, there still exists difference in the stability, reliability, automation degree, processing accuracy and others. With the development of processing quality required by moderm products, the requirement of machining tools such as reciprocating WEDM will continuously improve. Therefore, in order to adapt to the requirement, we need to make intensive study in the processing precision, reliability, stability and automation of reciprocating WEDM. This thesis will discuss the performance parameter optimization and the process model construction of reciprocating WEDM.Because of the mechanism of reciprocating WEDM is very complex and there is a highly nonlinear relationship between performance parameter and performance goal, in order to analyze it, an orthogonal array experiment is designed, with choosing performance voltage, amplifiers tube number, pluse width, pulse interval as the performance parameters, and choosing process time, surface roughness as the performance goal. At the same time, orthogonal level analysis method will be used to analyze the experimental results. In order to reduce the influence of random factors to performance parameters, signal-to-noise ratio method is used to analyze the experimental results of orthogonal array experiment, and the influence of performance parameters to signle performance objective can be obtain, which is prepared for the multiple performance objectives. Thereafter, based on the analysis of single performance objectives, the grey relation analysis is used to make multiple objectives optimization, and then the optimization performance parameters combination with multiple performances objectives is gained. The experimental results show that whether orthogonal level analysis method and signal-to-noise ratio analysis method for single process objective optimization or grey relation analysis method for multiple objectives optimization, which can improve the precision, reliability and stability of reciprocating WEDM simultaneously.Owning to the highly nonlinear relationship between performance parameters and performance goal, in order to build a process model, this thesis adopts the orthogonal experiment results as the learning sample, and builds a predictive model with multiple performance objectives of reciprocating WEDM by using the RBF neural network. The simulation results show that the process model can well solve the nonlinear relationship between performance parameters and performance goal.Finally, based on the process model developed by RBF neural network, with the characteristic of portability, expansibility, robustness of COM component technology, an expert system of process database for reciprocating WEDM using VC++ & MATLAB mixed programming is developed. The comparative experiments show that the system can well predict the performance goal through the given performance parameters and predict the performance parameters through the given performance goal, the system has certain application value in the industrial production.