Product Design Cost Estimation for Make-to-Order Industry: A Machine Learning Approach

Abstract

This research addresses the need for accurate design cost estimation in the Make-To-Order (MTO) industry. The complexity of product customization is key to differentiation. While many studies focus on manufacturing cost estimation, few explore design cost estimation. To improve the accuracy of design cost estimation, this research proposes a new cost driver based on design features available in Computer-Aided Design (CAD) data. The design feature is analyzed to the actual industry cost using machine learning methods, including Artificial Neural Networks (ANNs) and Support Vector Regression (SVR). The cost drivers identified as significant consisted of twenty-six 3D CAD features and four 2D CAD features. The results showed that the ANN models outperformed the SVR models in correctly estimating product design costs, as evidenced by the high R2 values in the training and testing phases. The proposed method allows early identification of cost drivers, a significant advantage at the order initiation stage when detailed design features are often ambiguous. The novelty of this research is the use of 3D CAD technology for cost estimation, which quantifies costs based on product design complexity, providing valuable insights into the impact of design adjustments on costs early in the design process.