Machine Learning Algorithm for Identifying Longer Straight-Line Crank Rocker Coupler Curves

Abstract

The paper addresses the longstanding need to transform uniform rotational motion into a significantly long approximate straight-line path at the application point. This requirement has inspired the search for a robust methodology to predict the possibility of tracing such an approximate straight-line path, with a length exceeding twice the crank length. Initially, traditional sources of such information, such as the Atlas, are briefly described. However, the limitations of these sources, including the lower yield of desired coupler curves and the considerable effort required to segregate them, are emphasized. To overcome these challenges, a solution is sought from the domain of Machine Learning to streamline the process, bypassing the extensive graphic analysis typically required for the selected crank-rocker. Recognizing the need for a substantial dataset to develop an efficient predictive algorithm, this paper demonstrates the development and utilization of the Naïve Bayes Classifier Algorithm as the initial step to identify the right mix of data points for future use. The development and successful validation of the Naïve Bayes Classifier algorithm for identifying the desired type of longer straight-line coupler curves are illustrated.