SLM parameter optimization based on improved deep belief network

In response to the problem of low quality and strength of parts, large surface roughness, and difficulty in twisting between parts in current aviation additive manufacturing, the selective laser melting (SLM) process was analyzed. By examining the main parameters of this process, an improved deep belief network (DBN) was employed to optimize these parameters and enhance the quality of the SLM manufacturing process. First, the main influencing factors in the 3D printing process of SLM were analyzed, such as laser angle, laser scanning speed, laser power, powder layer preheating temperature, powder thickness, and laser radius. Based on their characteristics, a parameter optimization process with optimal adaptability was designed. To address the practical constraints associated with SLM processing variables, a model was established with the objective function of fitting energy density to obtain manufacturing parameters. An adaptive DBN algorithm was employed to analyze the influence of various parameters on the processing outcomes. Finally, printing experiments were conducted based on the parameter settings of the design model for the turbine blades of aircraft engines, thereby validating the advanced nature and effectiveness of the proposed algorithm in SLM manufacturing applications.