Establishment and correction of the model for smoke diffusion in minimum quantity lubrication cutting

The large amount of smoke generated during minimum quantity lubrication (MQL) processing not only pollutes the ambient air but also directly endangers the health of operators. Establishing a smoke diffusion model is crucial for achieving precise control of MQL smoke. Currently, accurate smoke diffusion models in this field are lacking. In this study, a smoke diffusion model under MQL was established to predict the mass concentration of PM₁₀. The cutting speed, depth of cut, feed rate, and nozzle injection rate were integrated into the model using an extreme learning machine (ELM) to improve the accuracy of predicting the spatial distribution of smoke particles. A nonlinear equation reflecting the variation in concentration over time was solved using a backpropagation (BP) neural network. Finally, a spatiotemporal prediction model for smoke concentration during MQL turning was established. Comparing the predicted values of oil mist concentration in the test set with the true values through validation experiments, the results show that the absolute error of the prediction model at the measurement point tends to decrease with the increase of time, and the prediction accuracies are all above 90%. The maximum and minimum errors between the predicted and true values at different times are 9.77% (at the 0th second) and 4.11% (at the 6th second), respectively, which are less than 10%. Thus, the establishment of a highly accurate MQL cutting oil mist diffusion prediction model was realized.