Velocity Analysis of the Inverted Slider-Crank Mechanism using Artificial Neural Network Approach

In this study, velocity analysis of the inverted slider-crank mechanism using a supervised learning algorithm is introduced. Although velocity analysis of a slider-crank mechanism is simple and very well-established, there has been not been an application of artificial neural networks to such an analysis, in particular, to an inverted slider-crank mechanism. In this study, the velocity analysis of the inverted slider-crank mechanism is first solved analytically, with a total of 50 datasets determined. The Levenberg- Marquardt (LM) backpropagation supervised learning algorithm is then employed. The LM supervised learning algorithm is chosen because of its specific characteristics: its speed, incorporation of the Gauss-Newton training algorithm alongside the steepest descent method, and its capacity to guarantee a stable convergence of training error. There are two main objectives of this study: (i) to understand how to use an artificial neural network approach to solve the velocity analysis problem for linkages, specifically an inverted slider-crank mechanism, and (ii) to demonstrate an application of artificial neural networks to mechanism analysis. Results show that the LM algorithm is a suitable supervised learning algorithm for the artificial neural network solution approach to mechanisms velocity analysis.