ANN Application for Predicting and Investigating Dynamic Response of a Slider-Crank Mechanism with an Elastic Element
Abstract
Incorporating elastic elements into planar mechanisms offers significant advantages in terms of energy efficiency, impact load reduction, and the elimination of dynamic reaction forces during operation. However, the mathematical models describing the dynamics of these mechanisms typically involve complex numerical computational processes that require considerable time when directly applied to response analysis or design optimization tasks. In this study, a surrogate model based on Artificial Neural Networks (ANN) is developed to predict the dynamic response of a slider-crank mechanism with an attached elastic element. The ANN model is trained using data from prior publications and validated through standard error metrics. The results demonstrate that the proposed model achieves a prediction error of less than 6% and reduces computational time by a factor of more than 30 compared to the direct mathematical model. Comparative benchmarking against Support Vector Regression (SVR) and Polynomial Regression further validates ANN as the only surrogate model achieving consistent accuracy across all four output variables for this strongly nonlinear, multi-output dynamics problem. This approach opens a sustainable and promising avenue for addressing multibody dynamics problems in the future.