Track: Data Analytics and Big Data

Abstract

During the design process of new products, it would be important to obtain the impact of design decisions in terms of their cost effects. In this way, a rough product cost calculation could be obtained in order to decide whether such a product is worth manufacturing or not. Often, companies have an upper limit for manufacturing costs back-calculated from the expected market price, which is to be adhered to in the product design. A large amount of these costs are caused by the used material costs. But how should material costs be estimated if new components that have not been used before are to be handled? Obtaining offers for only planned, potentially eligible components on the procurement market is certainly out of the question for reasons of expense and time. Especially since alternative components with the same or similar technical properties often exist and the designer has to choose one of them. In order to solve this problem, we investigated whether a good cost estimate or forecast can be realized with the help of intelligent methods. An innovative two-stage technique with Clustering and Neural Nets showed the best results. The basis for this approach is a large data collection of previously supplied components with their specifications and their procurement prices.