| Summary: | 碩士 === 朝陽科技大學 === 工業工程與管理系碩士班 === 92 === Since recently developed electronic products are extremely thin and light, the Surface Barrier Layer (SBL) chips are key elements of traditional disk capacitors which are widely used in many small electronic components. Because surface defects exist on the SBL chips, they usually result in function failures of storage and release power. Therefore, it is very important to inspect the surface defects on the SBL.
The research proposes a surface defect classifier based on Adaptive Neuro Fuzzy Inference System (ANFIS) for SBL chips. Computer vision techniques are used to extract image features from SBL surfaces. The feature items in this research are extrated from color, gray, and binary images. This research proposes using analysis of variance to select better features and Duncan’s multiple range test method to establish fuzzy rules and fuzzy membership functions. A fuzzy inference system based on the fuzzy rules and fuzzy membership functions is developed to classify the SBL defects. The proposed defect classifier can be developed without using experts’ experience rules.
Experimental results show correct classification rate for normal chips is 100%, crater defects is 100%, reaction defects is 91.9% and ricochet defects is 100%. In order to exemplify the efficiency of the classification system, the IRIS data are classified by the developed system. The results show the best correct classification is 98.67%, the worse correct classification is 90.67%, and average of the correct classification is 94.67%.
|
|---|
