Research on grinding force prediction of spiral bevel gear based on generalized regression neural network and undeformed grinding chips

This is the authors' abstract. We don't add key points for this paper.

This study is rooted in a generalized regression neural network (GRNN) and undeformed grinding chips to predict the grinding force of spiral bevel gears. Firstly, establish a grinding force model for an individual abrasive particle, and integrate it with the dynamic number of abrasive particles in the grinding wheel to obtain a spiral bevel gear grinding force model based on undeformed grinding chips. Secondly, using the duplex spread blade design technique, the spiral conical gear blank geometric dimensions, machine tool adjustment parameters, and tool specifications are calculated, and the simulation of milling and grinding is carried out in three-dimensional solid modeling software. The undeformed chips and related parameters are obtained through Boolean intersection. Due to the complexity of calculation and simulation in the development procedure of spiral bevel gears, the GRNN is established to predict the undeformed grinding chip parameters of non-simulated gears. Finally, each group's comprehensive coefficient of cutting and friction forces were calculated through grinding experiments at different generating motion speeds. Predict these coefficients through experimental data. The results showed that the maximum standard deviation between the expected grinding force and the experimental value was 3.3260, which verifies the method's effectiveness. The GRNN algorithm can help select process parameters more reasonably, avoid tooth surface burns, and improve machining efficiency.