Abstract: To address the dynamic monitoring of small shifts in automated process quality,a dynamic monitoring method for process quality based on Empirical Mode Decomposition—Improved Chimp Optimization Algorithm—Bidirectional Gated Recurrent Unit (EMD-ICHOA-BiGRU) was proposed.The EMD-ICHOA-BiGRU model was constructed to predict quality characteristics.The Empirical Mode Decomposition (EMD) algorithm decomposed the original data into Intrinsic Mode Function (IMF) components at different scales.Each IMF was input into a Bidirectional Gated Recurrent Unit (BiGRU) for prediction,with hyperparameters optimized by the Improved Chimp Optimization Algorithm (ICHOA).The predicted components were superimposed to obtain the predicted quality characteristics.An Exponentially Weighted Moving-Average (EWMA) chart was constructed by combining the original data and predicted values.The chart was continuously updated through a moving window to achieve dynamic monitoring of small shifts in process quality.The method was validated using the quality data set from an automatic screw-tightening process for a rotating cover plate.The prediction accuracy of EMD-ICHOA-BiGRU model was significantly better than other models.The EWMA control chart based on EMD-ICHOA-BiGRU could monitor the small deviation of process quality,providing an effective method for dynamic monitoring of automated process quality.

Key words: process quality, small shifts, quality prediction, exponentially weighted moving-average chart, dynamic monitoring

摘要： 针对自动化工序质量微小偏移动态监测问题,提出基于EMD-ICHOA-BiGRU的工序质量动态监测方法。构建EMD-ICHOA-BiGRU模型进行质量特性预测,利用经验模态分解(EMD)算法分解原始数据,得到不同尺度的本征模函数分量(IMF),将IMF输入双向门控循环单元(BiGRU)进行预测,采用改进的黑猩猩优化算法(ICHOA)优化其超参数,将预测分量叠加得到质量特性预测值;结合原始数据和预测值构建EWMA控制图,通过移动窗口不断更新,实现工序质量微小偏移动态监测。利用旋变盖板螺栓自动拧紧工序质量数据集进行实例验证,结果表明,EMD-ICHOA-BiGRU模型的预测精度显著优于其他模型,基于EMD-ICHOA-BiGRU的EWMA控制图能够监测工序质量微小偏移,为自动化工序质量动态监测提供一种有效方法。

关键词: 工序质量, 微小偏移, 质量预测, 指数加权移动平均控制图, 动态监测