The Reaearch On Ion Beam Polishing Process Parameters Algorithm Based On Neural Network

Ion beam polishing technology belongs to the ultra-precision machining technology. The key variables of the technology are dwell time and removal function. So the thesis carried out research on theses two factors.Firstly, We researched on the principle of the ion beam polishing. Through the software simulation, we analyzed the relationship between the sputtering yield and the incident yield, and the relationship between the sputtering yield and the incident energy. Then we draw the conclusion that the sputtering yield would go up and come down with the increasing of the incident angle, and we got the maximal sputtering yield at74°.The sputtering yield would increase following with the increasing of the incident energy.Secondly, we got removal function model. And we known the key factors of the model is sputtering yield. Due to the sputtering yield has close relationship with angle, so we set the BP neural network to calculate the angle according to the sputtering yield.Thirdly, the CEH model adopted the system of linear equations to solve dwell time. We found that the method could eliminate the edge effect completely. However, Due to this method existed ill-posedness, therefore, TSVD regularization method was adopted to compensate for the defects. We got the "L" shape processing curve. By anglicizing the curve, we known that the parameter k on the turning points was the best value.But the shortcoming of TSVD algorithm was the low solving speed for large matrices. In order to overcome this problem, we used iterative algorithm to solve the problem for large matrix. But this method would appear the edge effect. therefore, we used the iterative method of smoothing the edge data and the "L" shape processing curve. Through the curve, we got the conclusion that the parameter y on the turning points was the best value.