An information value based approach to design procedure capture

Capturing engineering knowledge and managing it effectively is important for enterprises to stay competitive in today's global market. In our research, we focus on specific engineering knowledge, called design procedures, and attempt to develop effective ways to capture this operational knowledge from the design events monitored during design process. We proposed a novel method called Information Value based Mining for Sequential Pattern, or VMSP for short. VMSP does not require any predefined design-task-specific rules for classification of design events. The basic idea of VMSP is to treat any list of monitored design events as a sequential pattern and search for most frequent and informative sequential patterns. VMSP automatically generates candidate templates of sequential patterns, and then identify valuable patterns as design operations based on the interestingness we set. The interestingness is a synthesized index of the two evaluation criteria: One is intrinsic value representing how many design events constitute the template; and the other is extrinsic value indicating how often the sequence appears in the design process. We have evaluated VMSP through two case studies using real CAD operation data recorded through gear design and automotive door design processes, respectively. A case study using synthetic data was also carried out to test the scalability of the proposed algorithm.