A deep belief network model for predicting the material removal rate in the polishing process

The material removal rate (MRR) is an essential indicator for regulating process parameters and evaluating machining effects in the polishing process. However, as is known to all, it is difficult to accurately describe the MRR in theoretical approach due to the fact that it is affected by multi-parameter coupling such as pressure, velocity, abrasives, and the complex machining environment makes it challenging to monitor the MRR. This paper proposed a novel model based on the combination of the genetic algorithm and deep belief network (DBN) in order to predict MRR in the polishing process. First, the random forest algorithm was applied to select the parameter variables having a significant influence on the MRR, and these parameter variables were arranged as the input variables of the DBN model. Second, the hyperparameters of network were optimized by using the genetic algorithm with more powerful global search ability. Finally, the network model was trained and tested with the aid of the dataset provided by the PHM2016 Data Challenge, which resulted in an root mean square error value of 2.5487 and an R 2 value of 0.9937 for the complete dataset. Meanwhile, the prediction error was reduced by more than 2.2% compared with predictive models available in the literature. The results show that the model investigated in this study more accurately predicts the MRR in the polishing process.