A study on deep drawn cups and the selection of optimal settings deploying ANN training and architectural parameters using the Taguchi ARAS approach

Abstract

In recent years, the use of artificial neural networks (ANN) for modeling and optimizing metal forming processes has gained popularity. Numerous benefits are provided by ANNs, which can only be attained by constructing a high-performance ANN model. However, selecting an ANN’s appropriate training and architectural parameters remains challenging. Typically, these parameters are chosen using a trial-and-error approach in which many ANN models are produced and evaluated. This article describes the use of the Taguchi method for optimizing an ANN model trained on architectural parameters by the Levenberg–Marquardt algorithm and the additive ratio assessment method for selecting the optimized parameter. The method’s operation is demonstrated using the Altair Inspire form, the Forming Limit Diagram, and the Maximum Thinning Rate. The ANN model was trained using the orthogonal array, and its prediction performance was compared to the experimental validation. Analysis of variance assisted in determining whether the model was significant. The analysis of means helped us in deciding the appropriate ANN level for each parameter. Taguchi orthogonal array, ANN optimized for additive ratio assessment method. This model was developed and was found to be the most accurate. Through learning and testing, it was demonstrated that a systematic method for selecting the optimal ANN training and architecture parameters saves time and money by avoiding trial and error. The Taguchi-based additive ratio assessment method algorithm can then be used to determine the ideal processing parameters.