An Intelligent SMT Soldering Diagnosis System Using NeuroFuzzy Approach

• Main Authors: Tsung-Nan Tsai, 蔡聰男
• Other Authors: Shih-Yang Liu Ph.D.
• Format: Others
• Language: zh-TW
• Published: 1999
• Online Access: http://ndltd.ncl.edu.tw/handle/03577514891674171335

碩士 === 義守大學 === 管理科學研究所 === 87 === SMT(Surface Mount Technology) is slowly replacing the transitional wave-soldering assembly method and becoming increasingly significant in the modern electronics industry. SMT offers faster electrical signal processing capability, higher density, lighter weight and better electronic performance in products. Because the SMT process interconnects with several critical steps and field researchers examine these steps individually and then determine some complicated formulas, this process leads to the following shortcomings. (1)Difficulty in capturing the overall process knowledge; (2)Ignorance of the interactions among process factors; (3)Special tools are required to measure the parameters derived from the given formulas; (4)Difficulty in process deplyment without the advanced training. Moreover, both SMT process knowledge and expertise are difficult to translate into documentary procedures and rules of thumb to assist SMT engineers in dealing with malfunction processes and soldering defects. Overcoming the problems discussed above is the driving force behind the development of intelligent SMT soldering diagnosis system. The system helps engineers to troubleshoot the malfunction process, improve solderability, and complement traditional SPC by providing optimal process parameters and process simulation facility. The system consists of four ingredients:(1) NeuroFuzzy model(FAMs) for the extraction of fuzzy production rules; (2) A complete set of peculiar SMT experimental data; (3) exclusive SMT domain knowledge and expertise; (4)SPC data from mass production lines. In addition to the ingredients examined above, the fuzzyTECH software and MS Visual Basic are the development tools for building a system with the following features. (1) Provide optimal SMT process parameters; (2)Guide the users through the Graphical User Interface in SMT process diagnosis; (3)Train users by example simulations and (4)Perform online NeuroFuzzy training from extra user-defined rules. This intelligent SMT soldering system provides roughly 85% accuracy in SMT process diagnosis as proved by the empirical evaluations at VeriFone Taiwan Ltd. Also, it is amendable to some degree of accommodation to varied SMT machine models.