These slides were presented at the NAFEMS World Congress 2025, held in Salzburg, Austria from May 19–22, 2025.

Abstract

Designing sheet metal components is a challenging and specialized task that demands a deep understanding of engineering principles and extensive industrial experience. Traditionally, this process has heavily relied on heuristic knowledge and practical expertise acquired over many years. While this approach has been effective, it is inherently time-consuming and prone to human error, limiting the efficiency and accuracy of the design process. With the development of artificial intelligence (AI), a significant transformation is underway in the design and optimization of sheet metal components. AI methods, renowned for their exceptional performance, are now being integrated into the design process to streamline operations and improve outcomes. These methods aim to simplify the inherently complex design tasks, reduce reliance on manual expertise, and significantly shorten the time required to develop and refine designs. This shift not only enhances the overall efficiency of the design process but also improves accuracy and innovation. Among the most promising AI methods in this context are Artificial Neural Network (ANN), particularly Multi-Layer Perceptron (MLP). MLP is especially effective in addressing engineering design challenges and minimizing errors in experimental data. It is well-suited for optimizing design parameters and making predictions based on datasets, which would be too time-consuming with traditional simulation methods. MLP can significantly reduce the time spent on simulations by learning from existing data and providing faster, more accurate predictions. The objective of our research is to develop an integrated methodology that combines forming simulation with MLP to approximate design parameter functions and evaluate design performance, ultimately enabling the identification of optimal designs. In this methodology, forming simulations are initially employed to generate training data for the MLP. The well-trained MLP is then used to predict the performance of different designs. This methodology not only accelerates the design process but also provides a reliable means of exploring design variations and assessing their effectiveness. To ensure the reliability of the developed MLP, its performance is compared with other machine learning and ANN methods. The results clearly demonstrate that the proposed methodology is highly effective, excelling not only in predicting and evaluating designs but also in estimating various design variations. This integrated approach offers a robust and efficient solution for optimizing sheet metal component design, setting a benchmark for future advancements in the field.