| Summary: | 碩士 === 輔仁大學 === 統計資訊學系應用統計碩士在職專班 === 103 === High frequency is the trend for electrical products. The appropriate control of characteristic impedance is the key step to the success of FPC process.The issue of shorten the lead time for setting characteristic impedance and speed up producing time are extremely important for FPC process. Typically,FPC companies face many difficulties when performing simulations of controlling characteristic impedance, because it is not suitable for the higher price or professional electrical program needed. How to efficiently predict and accurately evaluate the characteristic impedance, it is the issue and cannot be ignored. This paper used test data as a benchmark to predict the characteristic impedance. The purpose of this study is to predict the characteristic impedance by using important explanatory variables. The proposed approaches include Multiple Regression(MR),Artificial Neural Network(ANN),Multivariate Adaptive Regression Splines(MARS), MRANNand MARS-ANN. The experimental results show that the ANN approach has the best accurate performance among those six methods. In addition, the MR-ANN and MARS-ANN are suggested to be used for the purpose of convenience.
|
|---|
