Multi-objective optimization of process parameters for laser metal deposition of NiTi shape memory alloy based on neural network and genetic algorithm

Due to the poor machinability and weldability of nickel-titanium (NiTi) shape memory alloy, application of NiTi alloy components prepared by smelting technology has been limited in aerospace and other fields. Laser metal deposition (LMD) technology opens a new way for the fabrication of NiTi alloy components. However, deposition quality and deposition rate are significantly influenced by the employed LMD process parameters. In this work, a small-sample prediction and optimization model based on BP-GA neural network for LMD process parameters selection was developed to improve the deposition quality and deposition rate of NiTi alloy components. The initial step involved the design of a central composite experiments consisted of thirty small-sample single-track experiments and building of the prediction model for deposition quality and deposition rate, wherein the inputs were consisted of the process parameters such as the laser power, scanning speed, and powder feeding rate. The responses, on the other hand, included the microhardness, roughness, and deposition rate. Prediction of the single-track cladding results by regression models of the response surface methodology, the ML models of back propagation neural network and random forest algorithms were comparatively analyzed and the prediction model was established. Then, based on the prediction model, non-dominated sorting genetic algorithm-II algorithm was applied to optimize the three input process parameters with the multi-objective of maximizing microhardness and deposition rate, and minimizing roughness. And the optimal combination of process parameters was obtained as a laser power of 1292.14 W, a scanning speed of 8.79 mm/s and a powder feeding rate of 16.78 g/min. Finally, single-track deposition experiments under the optimal combination of process parameters were carried out. The results showed that microhardness of 267.01 (5.23% improvement) and roughness of 7.86 µm (20.04% improvement) was achieved while maintaining a high deposition rate of 36.45 mm³/s. Objective of this study was to improve the deposition quality of NiTi alloy components with high deposition rate.