论文部分内容阅读
针对实际过程中存在的各种不确定性因素,提出了基于主成分分析和贝叶斯AR模型相结合的概率故障预测算法.该算法首先对过程数据进行主成分分析,得到T~2和SPE统计量,并对其进行Box-Cox变换.对变换后的两个序列分别建立贝叶斯AR模型,得到一系列下一时刻预测值.将所得预测值经Box-Cox反变换还原为原分布下的统计量预测值,并由核密度估计及相应控制限计算统计量超限概率,实现故障预测.最后,仿真结果表明了该方法的有效性.
Aiming at the various uncertain factors existing in the actual process, a probabilistic fault prediction algorithm based on principal component analysis and Bayesian AR model is proposed.The principal component analysis of the process data is carried out firstly, and then T ~ 2 and SPE And carry on Box-Cox transformation to them.We establish Bayesian AR model respectively for the two transformed sequences to obtain a series of predictions at the next moment.All the predicted values are restored to the original distributions by Box-Cox inverse transformation And the predictive value of the statistic is calculated, and the overrun probability of the statistic is calculated from the kernel density estimation and the corresponding control limit to realize the fault prediction.Finally, the simulation results show the effectiveness of the method.