Sintered silicon carbide grinding surface roughness prediction based on deep learning and neural network

Surface roughness is vital to high-precision parts, and it directly affects performance and service time of parts. So, in the grinding process, roughness control and prediction of grinding surface is particularly important. However, it is difficult to accurately predict surface roughness due to interlaced influence of large factors in grinding process. This problem can be solved due to the unique working principle of neural network; thus, in this paper, for the task of ground surface roughness, an integration sparrow search algorithm-based deep belief net model (SSA-DBN) was established, based on data obtained from simulation and experiment. SSA refers to a type of swarm intelligence algorithm, and DBN refers to deep belief net and belongs to deep learning, respectively. The purpose of swarm intelligence algorithm is to optimize network weight, compared to DBN model and SSA-BP network model, that is, DBN without optimization and back-propagation neural network (BPNN) with optimization, for analyzing their performance separately. Ground surface roughness prediction based on SSA-DBN model effectiveness was proven through training. At the same time, compared to SSA-BP network model and DBN model, SSA-DBN model performance was the best.