论文部分内容阅读
MSPCA方法在生产过程监控方面有着广泛应用。本文在研究该方法的基础之上,提出了一些改进,在其进行小波分解后即对其小波系数进行阈值处理,使小波消噪与MSPCA方法合为一体,并运用统计控制图中的平方预测误差(SPE)图方法检测引起过程变化或故障的过程变量。在保证其MSPCA算法复杂度不变的前提下,能够消除数据的噪声污染,使故障诊断的误报大为减少。经检验,该算法确实可行,相对于小波消噪与MSPCA方法分别进行,效率提高了大约13%～17%。
The MSPCA method is widely used in the production process monitoring. Based on the study of this method, some improvements are proposed. After wavelet decomposition, the wavelet coefficients are thresholded, the wavelet denoising is combined with MSPCA method, and the square prediction in statistical control chart The error (SPE) plot method detects process variables that cause process changes or failures. Under the premise of ensuring the complexity of MSPCA algorithm, it can eliminate the noise pollution of data and greatly reduce the false alarm of fault diagnosis. After testing, the algorithm is feasible, compared with the MSPCA wavelet denoising methods, respectively, efficiency increased by about 13% to 17%.