Yingwei Zhang

Modeling and Monitoring of Dynamic Processes

In this paper, a new online monitoring approach is proposed for handling the dynamic problem in industrial batch processes. Compared to conventional methods, its contributions are as follows: (1) multimodes are separated correctly since the cross-mode correlations are considered and the common information is extracted; (2) the expensive computing load is avoided since only the specific information is calculated when a mode is monitored online; and (3) after that, two different subspaces are separated, and the common and specific subspace models are built and analyzed, respectively. The monitoring is carried out in the subspace. The corresponding confidence regions are constructed according to their respective models.

Fault Detection of Non-Gaussian Processes Based on Model Migration

In this paper, a new modeling approach is proposed for common and specific feature extraction. The original space of a mode can be separated into two different parts, namely, the common and specific ones. There are both non-Gaussian similarity and dissimilarity in the underlying correlations of different modes. After two different non-Gaussian blocks are separated, one can obtain the common and specific blocks, respectively. They play different roles in industrial batch processes, which are referred to as repetitive and complementary effects, respectively. Then, the common block and specific block are analyzed. A new multiblock monitoring method is proposed and the monitoring process is carried out in each block. The proposed method is applied to process monitoring of a continuous annealing process. Application results indicate that the proposed approach effectively captures the non-Gaussian relations to build the process model and improves the detection ability.

Fault Isolation of Nonlinear Processes Based on Fault Directions and Features

In this brief, a new fault isolation method is proposed. The disadvantages of the conventional contribution plot method are: 1) the fault cannot be identified accurately due to process control and recycle loops in process flowsheets. To overcome this disadvantage, the fault-relevant-independent components (ICs) are extracted in this brief, which clearly represent different fault feature and 2) the conventional fault identification method is not available for the nonlinear process. In this brief, the nonlinear fault direction information is extracted. Then, the fault isolation method in the nonlinear process is proposed, where the historical fault information is used to build the model. The proposed method is applied to a simple nonlinear process and the electro-fused magnesia process. Compared with IC analysis (ICA) method and kernel ICA method, the results clearly show the effectiveness of the proposed method.

Fault Detection for Time-Varying Processes

In this brief, a new manifold learning method is proposed. Then, a process monitoring approach is proposed for handling the multimode monitoring problem in the electro-fused magnesia furnace based on the proposed manifold learning method. In the conventional methods, only partial common information is shared by different modes, i.e., the common eigenvectors. Compared with the conventional methods, the contributions are a new method of extracting the common subspace of different modes is proposed based on the manifold learning. The common subspace extracted by the proposed manifold learning method is shared by all different modes, and after those two different subspaces are separated, the common and specific subspace models are built and analyzed, respectively. The monitoring is carried out in the manifold subspaces.

FUZZY DIRECT ADAPTIVE SLIDING MODE DECENTRALIZED CONTROL FOR INTERCONNECTED SYSTEMS

Abstract Fuzzy direct adaptive sliding mode controller is designed to realize the decentralized control for a class of nonlinear interconnected large-scale systems with unknown functions. It uses the methods of fuzzy adaptive control, fuzzy logic approximation and fuzzy sliding mode control to approximate the ideal controller. Through the adaptive process of a parameter, the affection of interconnected terms on the subsystems is counteracted. The fuzzy sliding mode control is introduced to attenuate the fuzzy approximation errors. Simultaneity, the close-loop system is stable in Lyapunov sense and the tracking error converges to a neighbourhood of zero.

An Optimized Kernel Principal Component Analysis Algorithm for Fault Detection

Abstract An optimized kernel principal component analysis based on chaotic particle swarm optimization method (CPSO-KPCA) is put forward in this paper. The improved method adequately makes use of the characteristics of normal data and fault data to optimize the parameters of the mixture kernel function through chaotic particle swarm optimization so that the optimal nonlinear feature can be discovered and nonlinear fault can be detected accurately. Based on monitoring statistics charts in the feature space, this method is applied to fault detection in rolling process which is a nonlinear process. Practical application shows that the presented method has higher accuracy in fault detection.

Sensor Fault Compensation for Nonlinear Systems Using Fuzzy Adaptive Sliding Control

An active sensor fault compensation control law is developed for a class of nonlinear systems to guarantee the closed-loop stability in the presence of a fault, based on a fuzzy logic system and sliding mode. Through the adaptive process of the parameters, the dynamics caused by the fault is counteracted. The fuzzy sliding mode control is introduced to attenuate the fuzzy approximation error. Simultaneously, the closed-loop system is stable in Lyapunov sense and the tracking error converges to a neighbourhood of zero. The example of the proposed design indicates that the fault compensation control law is effective for a nonlinear system.

Fault Detection for MIMO Networked Control System

Abstract In this paper, we present a reduced-order memoryless state observer for networked control system with long delays in the control vectors. Then we use z-transform to deal with the long delays. By computing the residual between the output of the practical system and the output of the observer, we detect the fault of the networked system. An illustrative example shows the effectiveness of the approach.

Complex Process Monitoring Using KUCA With Application to Treatment of Waste Liquor

In this paper, a new nonlinear process modeling method called kernel uncorrelated component analysis (KUCA) is proposed. Several advantages of KUCA are summarized as follows. First, KUCA is superior to traditional data-driven modeling methods such as principal component analysis, independent component analysis, and so on, for it has no constraints on data distribution and data structure. In other word, it can be applied to modeling non-Gaussian and nonlinear process. Second, the proposed KUCA allows us to extract uncorrelated components (UCs) from observations directly without considering the second-order or higher order statistic, which is of low computational load. In addition, KUCA is applied to a numerical example to evaluate its performance of source signal separation. Finally, on the basis of KUCA, the corresponding processes monitoring scheme is developed, and the hot galvanizing pickling waste liquor treatment process is taken to evaluate the monitoring performance of the proposed methods. Experiment results demonstrate the feasibility of the proposed monitoring scheme.

Process Monitoring of Multimode Processes Using Kernel Independent Component Analysis

Abstract In the paper, a new process monitoring approach is proposed for handling the multimode problem in the industrial processes. The original space can be separated into two different parts, which are the common part and the specific part. There are both similarity and dissimilarity in the underlying correlations of different modes, which play different roles in the industrial processes. Because the industrial processes have the non-Gaussian and nonlinear characteristics, modified kernel independent component analysis is used to monitor the multimode processes in this paper. The global multimode basis vector and the multimode sub-basis vector are obtained based on the modified KICA. Then, the common part and specific part in one mode are respectively analyzed. The proposed method is applied to monitor the continuous annealing process. The proposed approach effectively captures the non-Gaussian and nonlinear relationship in different modes in the industrial processes.