Optimization of machining parameters using artificial Intelligence techniques

Abstract

In the industrial scenario, there is a requirement for robust tools to estimate the parameters of machining processes precisely. This can be achieved by employing optimization techniques in conjunction with state of the art prediction models to predict the best set of parameters to perform the operation. In this work, the turning operation on titanium alloy is optimized for improved surface roughness while mitigating the effects on the material removal rate by employing non-traditional optimization algorithm. The optimization algorithm demands for a prediction model that establishes an empirical relationship between the parameters in the machining operation and its response variables. The prediction models are employed on turning of Ti-6-Al 4-V Titanium alloy dataset and were contrasted by prediction metrics to determine the best prediction technique. The Artificial Neural Network (ANN) outperformed other prediction models by achieving a mean absolute percentage error of 1.08% as it could learn the non-linear and interaction effects in the data. The algorithms have been implemented in the functional programming paradigm in python (general purpose programming language). The accuracy of prediction model has been improved in contrast to the existing ANN models in the literature by fine tuning the ANN to determine the best architecture and hyper parameters. The ANN was then integrated with the genetic algorithm, a non-traditional meta-heuristic optimization algorithm to arrive at the best set of combination that minimizes the surface roughness while maximizing the material removal rate. A bias factor was introduced to control the optimization between the contradictory objectives. The optimized solutions offered the best combination of operating parameters to achieve the best performance in the given operation.