Abstract
Inconel 718 is a hard-to-machine alloy with very poor machinability and low thermal conductivity. Machining of such alloy is a critical manufacturing issue that should be carefully controlled to obtain machined components with acceptable accuracy and surface integrity. In this paper, hybrid machine learning (ML) models are developed to predict the induced residual stresses (RSes) during turning of Inconel 718 alloy. The developed models are composed of a traditional artificial neural network (ANN) incorporated with bio-inspired optimizers, namely pigeon optimization algorithm (POA) and particle swarm optimization (PSO). These optimizers are used to fine-tune the ANN parameters to enhance its prediction accuracy. The models were trained using measured RSes at different cutting conditions. The effects of the cutting conditions, such as cutting speed, cutting depth, and feed rate on the induced RSes are also investigated. The predicted RSes obtained by the developed models were compared with the measured ones as well as with those predicted by traditional ANN. The prediction accuracy of the models was statistically evaluated using seven statistical measures. The ANN–POA and ANN–PSO outperformed the traditional ANN. The coefficient of determination of ANN–POA, ANN–PSO, and ANN was 0.991, 0.938, and 0.585, respectively, while root mean square error was 11.870, 31.487, and 119.437, respectively.