基于ELM-MPLS的质量相关故障检测方法研究

Research on the Quality-related Fault Detection Method Based on ELM-MPLS

在工业生产过程中,很多产品的质量难以使用传感器来直接测量。因此,出现了偏最小二乘和潜在结构全投影等过程监测方法,但它们的质量相关故障检测效果并不理想。为此,提出了一种基于极限学习机(ELM)与改进的偏最小二乘(MPLS)的质量相关故障检测方法。该方法首先采用ELM进行质量预测,然后根据预测结果进行过程监测,最后采用田纳西伊斯士曼(TE)过程验证其质量相关故障检测的有效性。实验结果表明,ELM-MPLS不仅可以实现质量预测还提高了质量相关故障检测率,具有良好的质量相关故障检测效果。

In the process of industrial production,the quality of many products is difficult to be measured directly by sensors.Therefore,process monitoring methods such as partial least squares and total projection to potential structure are developed,but their quality-related fault detection effects are not ideal.To this end,a quality-related fault detection method based on extreme learning machine(ELM)and modified partial least squares(MPLS)is proposed.In this method,ELM is used for quality prediction firstly.Then,process monitoring is carried out according to the prediction results.Finally,TE process is used to verify the effectiveness of quality-related fault detection.The experimental results show that ELM-MPLS can not only achieve quality prediction but also improve the detection rate of quality-related faults,and has a good quality-related fault detection effect.