论文部分内容阅读
基于故障可检测性条件,提出了概率主元个数选择方法,根据能将所有故障检测出的条件,确定出主元的个数.但是在有些实际工业过程中,其故障形式往往不可尽知,从而无法得出主元的个数,给监控带来了困难.为了能够有效地检测出故障,进一步提出一种多概率主元分析(PPCA)模型的检测方法,首先选择不同的主元个数,建立PPCA模型,然后联合这些PPCA模型进行检测,如果有一个主元模型的指标值超出控制限,则认为过程出现故障,从而实现故障检测.
Based on the condition of fault detectability, a method of selecting the number of principal components of probability is proposed, and the number of principal components is determined according to the conditions that can detect all the faults. However, in some practical industrial processes, the failure modes are often inexhaustible , So that the number of the principal components can not be obtained, which brings difficulties to the monitoring.In order to detect the faults effectively, a new method of detecting the PPCA model is proposed. First, different principal components PPCA model is established, and then these PPCA models are combined to detect. If there is a principal component model whose index value is beyond the control limit, the process is considered to be faulty so as to achieve fault detection.