TRIZ Problem-solving Model for Multiple-to-Multiple Parameter Contradiction in Case-based Reasoning

Abstract

The engineering parameter and contraction matrix (CM) summarized by Altshuller based on the patents of traditional industries in 1950s can hardly be applied in today』s industry due to the following two problems: 1) the basic physical and chemical principles of the contemporary science and technology industries are totally different from those of the traditional industries; 2) problems faced by the industries are not necessarily one-to-one parameter contradiction correspondence. Based on this, this paper uses the chemical mechanical polishing (CMP) processing in the semiconductor industry as an example to establish industrial parameters, and employed the case-based reasoning (CBR) method to establish the multiple-to-multiple parameter corresponding case database, in order to obtain the correspondence of the inventive principles (IPs) of the contradiction combinations. This paper first reviews the patent summaries and establishes the multiple-to-multiple parameter correspondence patent case database. Through the operational mode of CBR, similarity coefficient is used to compare the similarity of the problems. Similar problems have similar corresponding IPs solutions. The weighted integration of solutions to highly similar problem cases can identify the available inventive solutions. The correctly solved cases after validation can be added into the case database to make it have learning and growing characteristics. The contributions of this study include: 1) demonstrating the low applicability of the classical matrix in multiple-to-multiple parameter contradiction problems; 2) constructing the prototype case database of multiple-to-multiple parameter contradiction of CMP processing problems; 3) establishing multiple-to-multiple parameter contradiction mathematical solutions, improving the disadvantages of mathematical tools mainly based on qualitative description, logical reasoning and lack of accuracy as well as quantitative analysis, and providing solution sequencing; 4) providing highly similar cases to problems to be solved as reference to new problems; 5) being able to replace the classical matrix to resolve one-to-one parameter contradiction.