Study On The Optimizing Of Wire Electrical Discharge Machining Process Parameters In Engineering Ceramics And The Prediction Of Surface Roughness

Based on the realizing of national inside and outside engineering ceramics material of WEDM (Wire Electrical Discharge Machining), taking electric Al₂O₃/TiC as example, analyzing the theory of WEDM and the theory of surface roughness,surface remnant stress,surface decaying layer of engineering ceramics of WEDM,it studied on the effect of impulse width,impulse separation,work voltage,the number of power transistors on the engineering ceramics of WEDM.studying faintness system and artifical neural networks, focusing on some disadvantages in neural networks algorithm, such as low convergence rate,easily falling into local minimum point and weak global search capability,the author used a new learning presented algorithm that used the faintness and Modular arithmetic to train neural networks, taguchi method and a standard analysis of variance (ANOVA) to study on theory of WEDM and the experiment. During the experiment, it introduced in S/N in order to deal with data and eliminate disturb.The prediction model of surface roughness in engineering ceramics of WEDM based on faintness Modular neural networks was proposed in detail.An L16 experimental matrix design based on Taguchi method was conducted to process Al₂O₃/TiC ceramics of electrical discharge machining.Based on a standard analysis of variance (ANOVA),it has been found that the relative significance of each factor on surface roughness was arranged in decreasing order of work voltage(F=37．16%), the number of power transistors (F=27．7%), impulse separation(F=19．0%), impulse width (F=16．14%),while on process rate was arranged in decreasing order of impulse separation(F=44．42%),the number of power transistors (F=25．45%),work voltage(F=20．71%),impulse width(F=9．42%).Under the optimum factors levels,it had least surface roughness (1．215μm)and most process rate(24．78 mm²/min),the prediction value in good agreement with experimental result.And it also analysed the remnant surface stress and the decaying layer after ceramics is machined with electrical charges.The result indicates that the remnant surface stress of engineering ceramic after electrical discharge machining is tension stress,and the decaying layer of the surface averaging 50～100μm becomes loose and weakened.The remnant surface stress and the decaying layer of the surface will increase with the incresase of pulse energy (impulse width,the number of power transistors,work voltage)whereas decrease with the increase of interval between two pulses. The result shows that the electro-discharge machining has a damaging effect on the surface of ceramics,giving low flexural strength and low Weibull modulus by Weibull statistical method.Taking engineering ceramics of WEDM for L16 orthogon-al array and experimental results as the stylebook,the experimental and simulating results showed that the improved faintness Modular neural networks can not only effectively overcome the problems of easily falling into local minimum point,but also the model can obtain higher accuracy of prediction,error 7%,certainly having practical value.