Two-Position Synthesis of the Four-Bar Planar Linkage Mechanisms Using Artificial Neural Networks

Abstract

The four-bar linkage mechanism is a fundamental and widely recognized mechanism with diverse applications, including various vehicle components, rehabilitation robotics, and rotary and reciprocating engines. Traditional textbooks introduce graphical and analytical solutions for the kinematic synthesis problem of the four-bar mechanism in different positions, while research articles explore its applications in various fields. Recently, artificial neural network (ANN) methods have gained popularity across different research domains. Researchers have proposed different solution approaches using ANN algorithms for the inverse and forward kinematic analysis problems of these mechanisms. However, the specific use of ANN algorithms for solving the two-position kinematic synthesis problem of the four-bar planar linkage mechanism has not been explored yet. This study aims to address this gap by introducing a solution for the two-position kinematic synthesis problem of the four-bar planar linkage mechanism using an artificial neural network algorithm. The Levenberg-Marquardt backpropagation neural network algorithm is chosen due to its speed, combination of Gauss-Newton training algorithm and steepest descend method, and ability to provide stable convergence of the training error. The neural network algorithm is trained, validated, and tested using a total of 50 randomly split data sets. Additionally, an additional test is conducted using all 50 data sets to evaluate the performance of the trained neural network algorithm. The study presents and discusses the results of the artificial neural network algorithm solution.