Description
The dominating mass-manufacturing process of today is plastic injection molding. This production process uses economies of scale because parts are produced in seconds at marginal cost of plastics. However, upfront investment costs for the tooling of molds are the basis for deciding if a mold is tooled and hence if a part is viable for mass-production. If tooling costs are too high, a product may not viable for production. If tooling costs are estimated too low by the tool shop, contract implications may arise. Because injection molds differ in their complexity, price estimations for the tooling of molds are an ongoing quest. There are various methods for estimating the costs of injection molds such as rule based, analytical or data driven approaches. The advantage of data driven approaches is the ability of adjusting to historical production data as well as readjusting while training on new batches of recent data. The focus of our research was to support the quotation process of tool shops. To this end, we studied a data driven machine learning approach. The goal of this research is to develop a method with humanlike quotation accuracy, achieve standardization, factor in historic quotation data and shorten quotation process times. The machine learning approach developed is based on geometry data of parts and additional meta-information. Within this research, a system was developed to interact with live production systems of an electronic part producing tool shop. The method developed was trained and validated on production data in a case study. To enhance the quotation process, the method developed was embedded into a server-based application with a web user interface and interfaces to live production systems for the automation of processes.