Abstract：

Demand-controlled ventilation can effectively reduce the overall oil mist concentration in the breathing zone of the work cell where machine tools are located in mechanical processing plants primarily engaged in cutting operations. Moreover, the ventilation rate is closely related to the oil mist concentration distribution in the breathing zone. However, the concentration of oil mist is often difficult to measure stably over a long period. Therefore, this paper aims to predict the distribution of oil mist in the breathing zone using the method of machine learning combined with dimensional reduction techniques. Firstly, a simplified model of a unit in a mechanical processing workshop is created through numerical simulation. This analysis identifies the factors influencing the distribution of oil mist in the breathing zone. It is found that the distribution of oil mist in large spaces with multiple emission sources is mainly affected by the exhaust air volume, the strength and the location of the emission sources. Through simulation of different working conditions, it is found that the distribution of oil mist changes with the exhaust air volume, and the local concentration variations are weakly correlated with the overall changes. Then, based on the simulation results, a database is established to train various predictive models. Considering R2, the GBDT (gradient boosting decision tree) method is chosen for prediction. This approach ensures a prediction accuracy of over 80%, laying the foundation for subsequent optimization of ventilation control methods.

Keywords:mechanical processing workshop; oil mist particle; numerical simulation; breathing zone; influencing factor; machine learning; concentration prediction