A Research of Using Data-Mining Techniques in Manufacture Process Failure Mode--A Case Study of Switching Power Supply Manufacturers

Autor: CHUN-HUI LU, 呂俊輝
Rok vydání: 2012
Druh dokumentu: 學位論文 ; thesis
Popis: 100
Due to global climate warm up, the awareness of environmental energy-saving is rising. Designer has to focus on energy-saving criteria as design the power supply equipment. Moreover, the electronic products change into light, thin, and small, result in lifting up the quality control of material and production process. The raw material used in power supply product is complicated, caused many failure roots. Under the influence of various factor interaction, increasing the complexity relatively. It is difficult to find out the process root cause as any quality variation. This study collected SPS(Switching Power Supply)repair history records, including characteristics of product, process, and failure modes to find out the reason of failure. We used the Self Organizing Map to categorize the failure types, and Decision Tree algorithm to obtain the failure rules. We then inducted the Cause and Effect Analysis Charts for affected product’s material supply failure, manufacturing failure, and research/design failure. Finally, we proposed the solution of failure problems. The results showed that power supply equipment failure is not entirely caused by manufacturing process, on the contrary material and research/design are the major causes. We took the failure phenomenon that higher than risk value of 200 as an example, the main failure causes are blow-up, OCP no good, low power factors, poor effectiveness and high-pressure protection defect. Through the failure handling mode, the hit rate of failure mode is 71%, it proved that the failure handling mode is really could be a reference of failure prediction as design or trial production on power supply product. It may also assist the online repairer quickly get the solution and prevent the problem rise again.
Databáze: Networked Digital Library of Theses & Dissertations