An Improved Mixture of Probabilistic PCA for Nonlinear Data-Driven Process Monitoring

Abstract

An improved mixture of probabilistic principal component analysis (PPCA) has been introduced for nonlinear data-driven process monitoring in this paper. To realize this purpose, the technique of a mixture of probabilistic principal component analyzers is utilized to establish the model of the underlying nonlinear process with local PPCA models, where a novel composite monitoring statistic is proposed based on the integration of two monitoring statistics in modified PPCA-based fault detection approach. Besides, the weighted mean of the monitoring statistics aforementioned is utilized as a metrics to detect potential abnormalities. The virtues of the proposed algorithm are discussed in comparison with several unsupervised algorithms. Finally, Tennessee Eastman process and an autosuspension model are employed to demonstrate the effectiveness of the proposed scheme further.

Fault Detection Method for Suspension Systems of Maglev Train Based on Optimized Random Matrix Theory

“…In sum, F selected by equation (9) and n obtained by equation (12) can meet the requirements well as can be seen in Fig.14, and the computation is very small too. Therefore, it is feasible to obtain n by equation (12) in this case.…”

“…In this paper, to verify the proposed method, about 2.68 × 10 4 samples from the sub-health stage (include the overload fault) are selected and used to calculate the auto-correlation from MSU, then the corresponding auto-correlation lengths are gained. Therefore, F is 4943 by equation 9, and n is 28 by equation (12). This paper will analyze the M SR and discuss the parameters (F and n) below.…”

“…Although 4000 selected from 1000 to 13000 (By default, n is obtained by equation (12)) is preferable, it is not a quick method. If this method is used to seek F when there are many variables in the system and n is not obtained by equation (12), it will require a large computation. Moreover, F obtained by equation (9) can also meet the requirements well as can be seen in Fig.13, and the computation is very small too.…”

“…Since the product of m and F is fixed, only can M λ be maximized when (n × b) ≈ (F/n), that is, a viable value of n is taken by equation (12). Finally, the maximum value of M λ is acquired.…”

AI-based modeling and data-driven evaluation for smart manufacturing processes

“…However, PCA-based approaches project features to another space based on a linear combination of original features. Therefore they cannot be interpreted in the original feature space [29]. Moreover, most of the PCA-related work has considered linear PCA, which is not efficient in exploring nonlinear patterns.…”