论文部分内容阅读
多牌号产品生产过程经常涉及到牌号切换,而切换后新牌号生产过程的变量关系可能随之发生变化,故采用单一的故障检测和诊断方法,无法对多牌号产品连续生产过程出现的异常做出有效的判断。这就需要及时准确地识别出新牌号,并对每个牌号有相应的故障检测和诊断模型。为此,本文引入人工神经网络(ANN),将其用于牌号识别,提出了牌号识别和主成分分析(PCA)相结合的方法,即利用历史数据建立各个牌号的BP神经网络(BPNN)模型和PCA模型,在线数据经过BPNN识别确认牌号类型后,调用对应牌号的PCA模型进行故障检测和诊断。结果表明,BPNN不仅可以准确识别牌号,识别率较规格界限法更高,而且可以对牌号过渡过程进行判断。另外,与不进行牌号识别仅采用单一牌号正常样本或者所有牌号正常样本混合建立的PCA模型相比较,采用牌号识别后进行故障检测时的精度更高,证明了该方法的有效性。
Multi-brand production process often involves the brand switching, and the relationship between the variable relationship of the new brand production process after switching may change, so the use of a single fault detection and diagnosis methods, can not be more than the continuous production of multi-brand products to make abnormal Effective judgment. This requires timely and accurate identification of new grades, and each grade has a corresponding fault detection and diagnosis model. Therefore, this paper introduces Artificial Neural Network (ANN), which is used for brand identification, and puts forward a method of combining brand identification and principal component analysis (PCA), that is, using BPNN model And PCA model. After online data is identified by BPNN, the PCA model is called for fault detection and diagnosis. The results show that BPNN can not only accurately identify the grade, the recognition rate is higher than the specification limit method, but also can judge the grade transition process. In addition, compared with the PCA model, which is constructed by using only a single normal sample or a normal sample of all the grades without brand identification, the accuracy of the fault detection after using the brand recognition is higher, which proves the effectiveness of the method.