基于滑窗B样条偏最小二乘的浮选过程质量指标软测量

摘要

In the industrial field,due to the complexity of the process of flotation,concentrate grade is difficult to be measured online.To solve this problem,a nonlinear modeling method of moving window B spline partial least square（BS-PLS） is proposed for soft sensor of the concentrate grade.In first instance,the features of the bubble are extracted from the foam image by the digital image processing technique.Then the data is filtered via wavelet transform.Finally,the BS-PLS is adopted to establish a regression model about bubble features and concentrate grade,and a dual online modification strategy is used to update the parameters with moving windows and compensate the output with the offset of last output of the model.The experimental results with industrial data demonstrate that the proposed method could effectively predict the concentrate grade of flotation process.%在工业现场,由于浮选过程的复杂性,精矿品位很难在线检测。针对这一问题,提出一种基于滑窗B样条偏最小二乘（B-Spline Partial Least Squares,BS-PLS）方法对其进行软测量。该方法首先应用数字图像处理技术,从实时获取的泡沫图像中提取泡沫特征;然后对获得的特征数据进行小波滤波预处理,再利用BS-PLS建立泡沫特征关于精矿品位的回归模型;最后采用滑窗滚动更新参数和偏差补偿输出两种策略配合对模型进行在线校正,实现精矿品位实时软测量。工业数据的仿真结果表明,该方法能有效软测量浮选过程的精矿品位。