半监督多源域适应集成的球磨机负荷参数软测量

Soft measurement for a ball mill load parameters based on integration of semi-supervised multi-source domain adaptation

针对球磨机工况改变后,历史数据与待测数据分布差异导致的模型失配问题以及待测工况样本少的问题。研究了基于半监督域适应的球磨机负荷参数软测量方法。该方法考虑输出标签对特征变换矩阵的影响,首先集成约束条件寻找特征变换矩阵,将历史数据和待测数据投影到公共子空间;然后根据投影后的历史数据及少量有标签的待测数据建立回归模型,获得无标签待测数据负荷参数;考虑到不同工况历史数据有信息互补的特点,建立基于半监督多源域适应集成的软测量模型,进一步提高软测量模型的准确性。基于实验室球磨机多工况实验的数据测试,表明该方法能够有效提高球磨机负荷参数的预测精度。

Aiming at the model mismatch problem caused by distribution difference between historical data and data to be measured after changes of a ball mill’s working conditions and the problem of less samples of working conditions to be measured,a soft measurement method for ball mill load parameters based on semi-supervised domain adaptation was studied here.Considering effects of output label on the characteristic transform matrix,firstly constraint conditions were integrated to search the characteristic transform matrix,and historical data and data to be measured were projected into the common subspace.Then,a regression model was established according to the projected historical data and less labeled data to be measured to obtain load parameters of unlabeled data to be measured.Considering different working conditions’historical data having information complementation feature,a soft measurement model based on integration of semi-supervised multi-source domain adaptation was built to further improve the correctness of soft measurement model.The measured data of multi-working condition tests of ball mills in laboratory showed that the proposed method can effectively improve the prediction accuracy of ball mill load parameters.