Data-driven cost estimation for additive manufacturing in cybermanufacturing

Cybermanufacturing is a new paradigm that both manufacturing software and hardware tools are seamlessly integrated by enabling information infrastructure and are accessed as services in cyberspace. This paradigm encourages tool sharing and reuse thus can reduce cost and time in product realization. In this research, a new cost estimation framework is developed based on big data analytics tools so that the manufacturing cost associated with a new job can be estimated based on the similar ones in the past. Manufacturers can use this cost analytics service in their job bidding process, which is currently ad hoc and subjective in industry practice. The new framework is implemented and demonstrated for additive manufacturing, where the similarities of 3D geometry of parts and printing processes are established by identifying relevant features. Machine learning algorithms for dynamic clustering, LASSO and elastic net regressions are applied to feature vectors to predict the cost based on historical data.