A Machine Learning Approach for Automated Cost Estimation of Plastic Injection Molding Parts

Abstract

Market competition leads to shorter cycle times for new or updated products. Therefore, flexibility in reacting to market changes, product development, and all related processes must be accelerated. In this regard, accurate cost estimation in the early stages of product development is critical for assessing the economic viability of a product. However, cost estimation requires data and expertise from several departments. Machine learning approaches could improve the accuracy and reduce the time needed for cost estimation. To investigate the eligibility of machine learning based cost estimation, a case study was conducted on an industrial company that produces plastic molding parts as key components of its products. The study involved training various supervised machine learning algorithms on a dataset of plastic injection molding parts using three different cost calculation methods. The three methods differed in the extent to which they considered the different process steps involved in the production of the parts. Different tree-based machine learning regression models and neural network models were trained to identify the most suitable approach for cost estimation in the given context. The results showed that tree-based machine learning algorithms outperformed neural networks and that individually predicting manufacturing parameters for cost calculation of each manufacturing process step leads to the most accurate cost estimation. This paper demonstrates how machine learning can support cost estimation in the early stages of the product lifecycle, reducing development times and improving cost estimation accuracy.