Metaheuristic

Algorithm

Residual

Engineering

Residual stress

Statistics

Metallurgy

Materials science

Subroutine

Design of experiments

Operating system

Particle swarm optimization

Computer science

Mechanical engineering

Response surface methodology

Mathematics

Artificial intelligence

Composite material

Artificial neural network

Machine learning

Journal of Materials Research and Technology

Residual stresses (RS) induced in machined components have substantial impact on the quality and lifetime of the final products. There are several cutting parameters and conditions that affect the generation of RS, so understanding the relationship between the RS generation and those parameters to minimize the induced tensile RS is a crucial issue. This paper presents a study on the utilization of artificial intelligence-based methods to model the RS generation during dry turning of DT4E pure iron. The experiments were designed based on central composite design method. The effects of the cutting parameters such as cutting speed, feed and cutting depth on the generated RSes in both circumferential and radial directions are investigated. Two hybrid artificial neural network (ANN) models are used to predict the process responses after training them using the experimental results. The prediction accuracy of the two models are enhanced via integration with two different metaheuristic optimization algorithms, namely particle swarm optimization (PSO) and flower pollination algorithm (FPA). These optimization algorithms are used as subroutine algorithms to determine the optimal parameters of the ANN model. The predicted results by the proposed models were compared with the experimental results as well as those obtained by standalone ANN. The accuracy of all models was evaluated using different statistical measures. The ANN-FPA had the best prediction accuracy followed by ANN-PSO. The coefficient of determination of ANN-FPA has high values of 0.996 and 0.997 for radial RS and circumferential RS, while they were 0.971 and 0.992 for ANN-PSO and 0.649 and 0.815 for standalone ANN.

Prediction of residual stresses in turning of pure iron using artificial intelligence-based methods