Surface roughness prediction model of GH4169 superalloy abrasive belt grinding based on multilayer perceptron(MLP)

Abstract

GH4169 nickel-based alloy materials are widely used in the design of aero-engine compressor terminal stators and rotor blades because of their high strength, oxidation resistance and gas corrosion characteristics in high-temperature environments. In addition, the advantages of abrasive belt grinding in the processing of nickel-based alloy blades, such as low temperature and strong vibration absorption, can effectively improve the surface integrity and be popularized and applied by aviation manufacturers. However, due to the random distribution of abrasive grains, grinding heat and machining chatter during the abrasive belt grinding process, there is a coupling and strong nonlinear model relationship between the grinding process parameters and the surface roughness after grinding. At present, the strong nonlinear model relationship of multi-dimensional parameters and their coupling relationship has not been resolved, so it is difficult to achieve accurate prediction of surface roughness.We propose a study on the multi-dimensional parameter model of abrasive belt grinding for the decision-making of the surface integrity characteristics of nickel-based alloy blades. Based on the MLP multi-layer perceptron, a prediction model of abrasive belt grinding surface roughness is proposed. According to the multi-layer perceptron model and the input parameter characteristics of the belt grinding system, the initial connection weight and paranoia, the backward propagation algorithm, and the neural network layer are optimized. Through the above algorithm, a multi-dimensional parameter surface integrity nonlinear model is obtained and combined with boundary conditions to achieve accurate prediction of the surface roughness of nickel-based alloy abrasive belt grinding.