基于机器学习的切丝后含水率预测及控制方法

Prediction and control method of moisture content after cutting based on machine learning

选取云烟(A)牌号制丝生产过程稳态数据样本,采用递归特征消除法分析模型的影响变量。基于车间温湿度SARIMAX预测模型,利用蒙特卡洛仿真、神经网络算法和XGBoost算法建立切丝后含水率控制模型,通过预测值与实际值对比的方法进行模型检验。结果表明,在工艺标准值±0.15%的误差范围内,切丝后含水率准确率由62.57%提升至86.49%;切丝后含水率的过程能力指数达标率由91.44%提升至97.30%。该方法实现了前后工序参数协同和精准控制,有效保证了制丝过程中切丝后含水率的稳定性。

In order to improve the stability of the moisture content after cutting in the production process of silk making,the matching degree of the indexes between the moisture content at the outlet of loose moisture return,the moisture content at the outlet of moistening leaf feeding and the moisture content after cutting was guaranteed.The steady state data samples of"Yunyan(A)"brand silk production process was selected and the influence variables of the model were analyzed by Recursive Feature Elimination Method(RFE).Based on the temperature and humidity prediction model of the workshop,Monte Carlo simulation,Neural Network algorithm and XGBoost algorithm were used to establish the moisture content control model after cutting.The model was tested by comparing the predicted value with the actual value.Within the error range of±0.15%of the process standard value,the accuracy of moisture content after cutting increased from 62.57%to 86.49%.CPK compliance rate increased from 91.44%to 97.30%.The prediction and control method of the moisture content after cutting based on machine learning can realize the coordination and accurate control of the process parameters before and after cutting,and effectively ensure the stability of the moisture content after cutting in the process of silk making.