A missing variable approach for decentralized statistical process monitoring

A latent feature oriented dictionary learning method for closed-loop process monitoring

Industrial cyber-physical system (ICPS), by its powerful computing, communication, precise control and remote operation functions, has become the mainstream of modern industrial process. The observed variables of the closed-loop process in ICPS are subject to the degradation of equipment and other factors, resulting in exhibiting a stationary/nonstationary mixture feature and dynamic feature. Moreover, due to the frequent change of working conditions in the closed-loop process, the traditional open-loop process monitoring method always triggers false alarms, which will impose a negative impact on the safety and trustworthiness of ICPS. Therefore, for the closed-loop process in ICPS, a latent feature oriented dictionary learning (LFDL) method is proposed, which realizes the precise separation of latent features of raw data through three stages. First, closed-loop process variables are separated into stationary and nonstationary variables to mine the local information spatially. Then, from the temporal viewpoint, the static and dynamic features were extracted for stationary and nonstationary variables on the basis of the slow feature analysis method and cointegration analysis for local monitoring. Finally, the global monitoring results are obtained by utilizing the dictionary learning method to fuse respectively the local monitoring results of the static and dynamic features. Since the proposed method has taken the feature of the close-loop process from temporal and spatial viewpoints simultaneously, it can distinguish the normal change of operating conditions and actual faults accurately. Extensive experiments including the three-phase flow, the Tennessee Eastman process and an industrial roasting process are conducted to demonstrate the feasibility and effectiveness of the proposed method.

Fault detection for chemical processes based on non-stationarity sensitive cointegration analysis

Due to the time-varying operation conditions, chemical processes are characterized by non-stationary characteristics, which makes it a great challenge for conventional process monitoring methods to capture the non-stationary variations In the non-stationary processes, the abnormality would cause the stationary variables to be non-stationary. In this article, a non-stationarity sensitive cointegration analysis monitoring method is proposed to explore potential non-stationary variations. First, the essential non-stationary variables are distinguished using Augmented Dickey-Fuller test to eliminate the influence of essential non-stationary under normal conditions. Then by further analyzing the faulty data, the variables which are sensitive to the non-stationary variations are selected. On this basis, cointegration analysis models are established for both the essential non-stationary variables and non-stationarity sensitive variables to explore long-term dynamic equilibrium relationship, respectively. With the selection of non-stationarity sensitive variables, the potential faulty information is emphasized in the process monitoring model, which makes the model capable to handle the non-stationary variations. Finally, the monitoring results are combined through Bayesian inference criterion. The proposed method is applied on the Tennessee Eastman process and a vinyl acetate monomer plant model, and the feasibility and performance are demonstrated.

Variable-weighted FDA combined with t-SNE and multiple extreme learning machines for visual industrial process monitoring

The visualization of an operating state of industrial processes allows operators to identify and diagnose faults intuitively and quickly. The identification and diagnosis of faults are important for ensuring industrial production safety. A method that combines variable-weighted Fisher discriminant analysis (VWFDA), t-distributed stochastic neighbor embedding (t-SNE), and multiple extreme learning machines (ELMs) is proposed for visual process monitoring. First, the VWFDA weighs variables on the basis of their contribution to the fault, thereby amplifying the fault information. The VWFDA is used to extract feature vectors from industrial data, and normal state and various fault states can be separated from each other in the space formed by these feature vectors. Second, t-SNE is used to visualize these feature vectors. Third, given that t-SNE lacks a transformation matrix during dimension reduction, one ELM is used for each class data of t-SNE to obtain the mapping relation from its input data to its mapping points. Finally, the VWFDA and multiple trained ELMs are combined for online process monitoring. The performance of the proposed approach is compared with that of FDA-t-SNE and other methods on the basis of the Tennessee Eastman process, thereby confirming that the proposed approach is advantageous for visual industrial process monitoring.

Recursive correlated representation learning for adaptive monitoring of slowly varying processes

Correlated representation learning has found wide usage in process monitoring. However, slow and normal changes frequently occur in practical production processes, which may lead to model mismatch and degrade monitoring performance. Therefore, updating the monitoring model online and involving recently processed data information are important. This study proposes a recursive correlated representation learning (RCRL) incorporating an approach for online model updating for adaptive monitoring of slowly varying processes. First, an initial canonical correlation analysis-based monitoring model is established using historical process data. Second, an online model updating criterion is developed, and updating procedures are provided to reflect online data information and update monitoring model in a timely manner. Then, monitoring statistics are established and decision making logic is established to identify process status. The fitness of the monitoring scheme is increased because the online process information is considered to update the model. The proposed RCRL-based monitoring scheme is applied on a numerical example and a lab-scale distillation process. The effectiveness and superiority of the RCRL approach are verified.

Nonlinear robust fault diagnosis of power plant gas turbine using Monte Carlo-based adaptive threshold approach

This paper addresses the robust fault diagnosis of power plant gas turbine as an uncertain nonlinear system using a new adaptive threshold method. In order to determine the bounds of the adaptive threshold and to identify neural network thresholds modelling, an approach based on Monte Carlo simulation is employed. To evaluate the performance of the proposed fault detection method, a fault sensitivity analysis is provided. In addition, the neural network-based estimators are considered to estimate the magnitude of faults according to the values of residuals. The proposed fault diagnosis system is evaluated during different scenarios. The obtained results indicate the high sensitivity, accuracy, and robustness of the proposed method for fault detection and isolation in the nonlinear uncertain systems, even in dealing with small faults.