Multi-objective optimization of yield rate and tact time of ACF in the TFT-LCD short-bar attach process using the fuzzy dominant directed graph method

作     者：Kuo, J-L Hsieh, C-H 

作者机构：Natl Kaohsiung First Univ Sci & Technol Dept Mech & Automat Engn Kaohsiung 811 Taiwan 

出 版 物：《PROCEEDINGS OF THE INSTITUTION OF MECHANICAL ENGINEERS PART B-JOURNAL OF ENGINEERING MANUFACTURE》 (机械工程师学会会报；B辑：工程制造杂志)

年 卷 期：2011年第225卷第B12期

页      面：2257-2275页

核心收录：

学科分类：08[工学] 0802[工学-机械工程] 

主　　题：ACF attach process TFT-LCD manufacturing process yield rate tact time fuzzy Taguchi method multiple performance characteristics index (MPCI) fuzzy dominant directed graph (fuzzy DDG) multi-objective optimization 

摘      要：This paper proposes a fuzzy-DDG method to solve the multi-objective optimization problem in regard to the anisotropic conductive film (ACF) attach process for thin-film transistor liquid crystal display (TFT-LCD). As such, this paper has been motivated by a desire to decrease the usage of ACF materials in the ACF attach process. Herein, the experimental Taguchi method will be used to find the optimal solution for this problem. This paper will study the optimization problem of yield rate and tact time with respect to control factors of TFT-LCD. In using the Taguchi method, five test cases are discussed: the first involves the optimal yield rate problem;the second, the optimal tact time problem;and the third, solving multiple problems by simultaneously considering yield rate and tact time. Additionally, the conventional multiple performance characteristics index (MPCI) method is used, although it exhibits a less than optimal fuzzy inference structure, and fuzzification as well as defuzzification processes are required;in the fourth and the fifth cases, a new fuzzy DDG method is used to determine the multiple objective problems. Notably, the conventional fuzzy inference structure is not required in this algorithm and only graph-based matrix operations are required. Testing results show that the fourth and the fifth cases are more convenient in regard to arithmetical operations, compared to the third case. The fourth case illustrates the problem with the vertex order included only;only crisp potential values are derived. The fifth case illustrates the problem with the vertex S/N ratio values included and fuzzy potential values derived. Experimental results are provided to verify the validity of the proposed fuzzy DDG method.