De-based SVM Classifier And Its Application On Flexible Integrated Circuit Substrate Manufacturing Process Anomaly Detection

Flexible Integrated Circuit Substrate(FICS),as one of the most important electronic interconnection technology,has been widely used in various electronic products because of its superior performance.However,as a high-precision product,the manufacturing process of FICS is pretty complex and precise,thus imposing a high demand for its manufacturing process.With FICS tending to be increasingly precise and in higher integration level,the density and precision of the circuit become higher and higher,thus making the process control more important to ensure the quality of the FICS.Aiming at the manufacturing process of FICS,the thesis studies the support vector machine(SVM)classification method and its application on the abnormal pattern recognition of FICS manufacturing process,which is of great importance in timely detecting abnormal states and ensuring the quality of FICS.The main research contents are as follows.1.The thesis analyses the control chart pattern of manufacturing process of FICS,then summarizes its abnormal pattern types,anomaly detection and evaluation methods.2.The thesis proposes a novel multi-strategy differential evolution algorithm,which consists of a hybrid parameter setting strategy based on the fuzzy logic inference and a population size adaptation strategy.The algorithm improves the performance of differential evolution while avoids the vicious circle by the adaptive parameter setting mechanism.3.The proposed novel differential evolution algorithm is used to optimize the kernel parameter and penalty factor of SVM.Experiments on the standard data sets show that the proposed SVM model with Gaussian kernel obviously improves its classification accuracy and generalization ability.4.The thesis analyses the characteristics of the technology in the manufacturing process of high density FICS,then summarizes the possible types and sources of defects and anomalies,and stimulation data are generated accordingly.The proposed SVM which based on multistrategy differential evolution is employed to recognize the abnormal pattern of manufacturing process of high density FICS,so as to rapidly response to the anomaly in the manufacturing process.