Haiyan Wu

Fault isolation based on residual evaluation and contribution analysis

Abstract A new fault isolation strategy for industrial processes is presented based on residual evaluation and contribution plot analysis. Based on the space projection, the residual evaluation and contribution plot are unified into a framework. First, parity space and subspace identification methods are used to generate residuals for fault detection. Then, the optimal residuals are utilized to obtain a process fault isolation scheme. A new contribution index is calculated according to the average value of the current and previous residuals. The smearing effect can be eliminated, and the fault evolution can be acquired based on this index. This would be helpful for engineers to find the fault roots and then eliminate them. A numerical model and the Tennessee Eastman process are considered to assess the isolation performance of the proposed approach. The results demonstrate that the smearing effect is improved and the primary faulty variable can be located accurately when several different faults occur simultaneously. A superior isolation performance is obtained compared with the PCA-based isolation method. The reasonability of the isolation results is analyzed using fault propagation.

PLS-based process analysis for glycosylation reaction

In this work, a novel methodology of evaluating the glycan distribution has been proposed. A multivariate statistical regression method of partial least squares (PLS) is used to establish the relationship between the manipulated variables and the response variables. All the fitting relations are contained in the glycosylation process gain matrix K obtained by PLS. According to the singular value decomposition of K, the importance degree of input enzymes to the desired glycan state is clearly illustrated. After that, we analyze the controllability of the glycan distribution affected by appropriate variables of enzymes and nucleotide sugar donor concentrations. Compared with the results of the method of standard analysis of variance (ANOVA), similar conclusions can be achieved: We can change appropriate manipulated variable concentrations to direct the glycan distribution. For some obvious different results produced by the two methods, we make a detailed analysis and interpretation. First, we will discuss how to implement the design of experiments (DOE), which is based on the glycosylation reaction network model, to generate the output data of glycosylation process; Then the PLS model is established utilizing the glycosylation data to obtain the gain matrix K, and to get the singular value decomposition of the K; Finally, we analyze the controllability of desired glycan state utilizing the singular values σi of the high-dimensional matrix K. The results may provide a foundation for the controlling glycosylation on-line in the future.

Unified Architecture of Active Fault Detection and Partial Active Fault-Tolerant Control for Incipient Faults

Incipient faults are difficult to be detected due to the intrinsic fault tolerance of traditional controller, but it should be eliminated as soon as possible before it deteriorates with time into something more serious. As a consequence of an intrinsic inability to assess whether a fault occurs based on output residual, the existing detection methods are failure for incipient fault. So the aim of active fault detection (AFD) is to make the system be unstable when incipient fault has occurred, which drives rapid fault detection. The fault-tolerant control (FTC) is designed to maintain the system stable and eliminate the fault impact without shutting the process down even if faults occur. In this paper, the Youla–Jabr–Bongiorno–Kucera (YJBK) parameter is employed to build the AFD and the FTC based on the relationship analysis between the fault and the dual YJBK parameter. A new structure of the tolerant controller parameter for FTC is designed, named as partial active FTC (PAFTC). PAFTC is dependent upon the fault detection information but not the fault size considering the parameter fault with unknown size and known form. A unified operation architecture for AFD and PAFTC with different YJBK parameters for incipient faults is proposed. Some illustrative examples are given to indicate the effectiveness of the proposed unified operation architecture.

An improved kernel exponential discriminant analysis for fault identification of batch process

An improved batch process fault identification approach with kernel exponential discriminant analysis (KEDA) is proposed, in which performance index based on difference degree is given to identify fault classification. This method takes the advantages of both the kernel technology and the exponential discriminant analysis technique. The proposed KEDA method shows powerful ability in dealing with nonlinear, small sample size data and it has a noticeable improvement in classification performance. During the real applications to fault identification, both the normal data model and the fault data model for known faults are established according to the historical data. Then online measurement data is fed into these models to identify the current operation status, i.e., is the system in normal or fault condition, what type of fault occurs, or does new fault appear? Finally, the proposed method is applied to a typical penicillin fermentation process and the simulation results show the effectiveness of the proposed KEDA algorithm and the good performance in fault classification.

SOM-based visualization monitoring and fault diagnosis for chemical process

Data-based fault diagnosis technology applied in chemical industry process has attracted great attention, in which the effective methods for visualizing the process variation are still challenging. The self-organizing map (SOM) is an unsupervised learning algorithm of neural network, which is presented to solve the visualization monitoring and fault diagnosis problem. The high-dimensional input space is mapped to the two-dimensional output space through training the large sample data sets of SOM method. Meanwhile, the fault data sets can be automatically clustering by SOM so that the faulty category information will be obtained. Distinguish between other methods, SOM can preserve the topological structure and the density distribution of original data, so the visualization of on-line process monitoring and fault diagnosis can be effectively realized. Then, the Iris data benchmark is used to test the clustering results of SOM algorithm. Finally, a case study of the Tennessee Eastman (TE) process is employed to illustrate the fault diagnosis and monitoring performance of the SOM-based visualization monitoring method.

Kinetic model reconstruction in silico of NSD metabolism network

Basing on the kinetic model analysis of cell metabolism network can visualize the movement of intracellular metabolism, and provide the link between cell culture and manufacture of biological products. The kinetic model reconstruction in silico of nucleotide sugar donor (NSD) metabolic network has been put forward by Philip et al., which involved with 34 metabolites mass balances and 60 metabolic reaction rates. In this paper, the NSD metabolic network kinetic model is partly modified basing on the Philips model. 26 significant kinetic parameters are estimated again by the genetic algorithm (GA) in this model. Compared to the traditional methods, the better simulation results are obtained in new kinetic parameters for metabolite concentrations. It offers an optimizable platform for the quality control of biological products.