Application of machine learning to predict the product quality and geometry in circular laser grooving process

Abstract

Micro-grooves have various applications in different industries. Circular laser grooving is a new method for manufacturing micro-grooves on the circumference of cylindrical parts. In this process, the selection of appropriate parameters to reach experimentally the desired groove aspect ratio along with minimum dross formation requires tremendous efforts. Here, the role of a decision-making technique, similarly machine learning is highlighted to evaluate the significance of process inputs and provide the appropriate model for prediction concerning the input parameters. The experiments were conducted with the various inputs such as workpiece rotational speed, laser beam position, duty cycle, power and frequency. The outputs are the groove geometry in terms of width and depth of the circular groove as well as the groove quality considering the dross formation. Then, Random Forest (RF) technique was utilized to derive the most influential inputs on the outputs. The RF analysis revealed that the rotational speed, laser position and duty cycle are the most decisive process parameters in the groove geometry and groove quality. Also, the enhancement of the assist gas pressure does not improve the process outputs according to the results of RF analysis.