Batch Process Fault Diagnosis Research Articles

Incipient Fault Diagnosis of Batch Process Based on Deep Time Series Feature Extraction

Intelligence modeling interpretable for incipient fault diagnosis of batch process represents a serious challenge. For the multivariate, nonlinear, and high-dimensionality characteristics of process data, existing fault diagnosis solutions are easily concealed by noise while neglect the low amplitude and noise interference of the incipient faults. In this paper, an intelligent incipient fault diagnosis model based on deep time series feature extraction network is proposed, which integrates denoising autoencoder (DAE), layer normalization and dropout layer added LSTM (LD-LSTM), and stacked autoencoder (SAE) to obtain the efficient features. DAE is applied to restore the data and eliminate noise interference. LD-LSTM is utilized to extract time series features, in which the layer normalization and the dropout layer are added between the layers of the LSTM to solve the problems of slow model convergence and over-fitting. SAE performs the feature extraction to obtain deep time series features and learns more efficient expressions. The softmax function is applied to achieve the incipient fault diagnosis based on the deep time series features. A case study on a fed-batch penicillin fermentation process suggests that the proposed method can obtain an accuracy rate of 98.97%, which has increased an average of 20% than the traditional shallow machine learning. The accuracy of fault recognition and comparative experiments demonstrate the effectiveness and feasibility of this model, and provide a solution for the application of intelligent fault diagnosis in batch industries.

Phase partition and identification based on kernel entropy component analysis and multi-class support vector machines-fireworks algorithm for multi-phase batch process fault diagnosis

For the characteristics of nonlinear and multi-phase in the batch process, a self-adaptive multi-phase batch process fault diagnosis method is proposed in this paper. Firstly, kernel entropy component analysis (KECA) method is used to achieve multi-phase partition adaptively, which makes the process data mapped into the high-dimensional feature space and then constructs the core entropy and the angular structure similarity. Then a multi-phase KECA failure monitoring model is developed by using the angular structure similarity as the statistic, which is based on the partitioned phases and the effective failure features by the KECA feature extraction method. A multi-phase batch process fault diagnosis method, which applies the multi-class support vector machines (MSVM) and fireworks algorithm (FWA), is proposed to recognize each sub-phase fault diagnosis automatically. The effectiveness and advantages of the proposed multi-phase fault diagnosis method are illustrated with a case study on a fed-batch penicillin fermentation process.

Phase Partition and Fault Diagnosis of Batch Process Based on KECA Angular Similarity

Fault monitoring of multiphase batch process is a difficult problem in multivariate statistical process monitoring. It needs to consider not only the process monitoring under stable mode, but also the transition mode with strong dynamic nonlinearity. Since the data has different correlations under different operating modes, it is necessary to establish different monitoring models for each process mode, especially the transition process between stable modes. The biggest feature is the dynamic characteristics of the variables. This feature can be better reflected in this transition using a time-varying covariance instead of a fixed covariance during the transition phase. In this paper, a new strategy for batch process sub-phase partition and process monitoring is proposed. Firstly, the three-dimensional data matrix is expanded into a new two-dimensional data according to the time slice expansion strategy. Secondly, the data of each time slice is transformed by Kernel Entropy Component Analysis (KECA), and then the production process is divided into phases according to the spatial angle of the kernel entropy. The production operation process is divided into a stable phase and a transition phase, and monitoring models are respectively established to monitor the production process; Finally, the application of the penicillin fermentation simulation platform shows that the Sub-MKECA phase partition results can reflect the mechanism of the batch process well, and the fault monitoring of the process shows that it can detect faults in time and accurately, and has high practicality value.

Tensor dynamic neighborhood preserving embedding algorithm for fault diagnosis of batch process

Batch process data has close relation in time series and dynamic characteristics. Traditional diagnosis algorithms often ignore process correlation of time series and dynamic characteristics, which would lead to larger errors of monitoring results. Aiming at dynamic characteristics of batch process, a tensor dynamic neighborhood preserving embedded (TDNPE) algorithm is proposed in this paper. Firstly, batch process data is regarded as a kind of second order tensor. The tensor factorization method is used to model that can avoid destroying internal structures of data. Then dynamic neighborhood preserving embedded algorithm is used to extract process feature information by considering local features of space and time in tensor space and that can effectively deal with process dynamic characteristics. The contribution plot method is used to diagnose fault variables when faults are detected. The simulation results of penicillin fermentation process verified the effectiveness of the proposed algorithm.

MGNPE‐LICA algorithm for fault diagnosis of batch process

AbstractBatch production process data usually contains mixture ingredients of the Gaussian and non‐Gaussian distribution. Traditional monitoring methods require data to meet the requirements of Gaussian distribution and do not take into account the global and local features for feature extraction. An ICA algorithm can deal with process diagnosis of non‐Gaussian distribution but cannot deal with process diagnosis of Gaussian distribution. So a multi‐way global neighbourhood preserving embedding, local independent component analysis (MGNPE‐LICA) algorithm, is proposed in this paper. Firstly, raw data is divided into Gaussian and non‐Gaussian spaces by the D‐test. For the Gaussian space, the MGNPE algorithm is used to fully extract local structural features and global structural features of data. For the non‐Gaussian space, the MLICA algorithm is used to solve non‐Gaussian problems and at the same time reserve global and local information of data. Then the monitoring index of two spaces synthesizes a joint monitoring indicator to monitor the process. The contribution plot method is used to diagnose fault variables after detecting faults. The simulation results of a penicillin fermentation process verified the effectiveness of the proposed algorithm.

A kernel‐based approach for fault diagnosis in batch processes

J. Chemometrics 2014; 28: S697–S707 The third dataset under study does not relate to a spray drying process, but to a batch drying process. On page 698 the sentence following Formula n°1 should read: “…where xiT and xjT are two row vectors…” (transpose symbols are missing). On page 703 the caption of Figure 7 should read: “PLSDA scores plot of the model built on the BWU kernel data matrix...” (the word “kernel” is missing). On page 707 the sentence following Formula n°B2 should read “Rewriting formula (B2)...”.

A kernel‐based approach for fault diagnosis in batch processes

This article explores the potential of kernel‐based techniques for discriminating on‐specification and off‐specification batch runs, combining kernel‐partial least squares discriminant analysis and three common approaches to analyze batch data by means of bilinear models: landmark features extraction, batchwise unfolding, and variablewise unfolding. Gower's idea of pseudo‐sample projection is exploited to recover the contribution of the initial variables to the final model and visualize those having the highest discriminant power. The results show that the proposed approach provides an efficient fault discrimination and enables a correct identification of the discriminant variables in the considered case studies. Copyright © 2014 John Wiley & Sons, Ltd.

Fault diagnosis for batch processes based on two-dimensional principal component analysis

Fault diagnosis for batch processes based on two-dimensional principal component analysis

Fault diagnosis of a benchmark fermentation process: a comparative study of feature extraction and classification techniques

This paper investigates fault diagnosis in batch processes and presents a comparative study of feature extraction and classification techniques applied to a specific biotechnological case study: the fermentation process model by Birol et al. (Comput Chem Eng 26:1553-1565, 2002), which is a benchmark for advanced batch processes monitoring, diagnosis and control. Fault diagnosis is achieved using four approaches on four different process scenarios based on the different levels of noise so as to evaluate their effects on the performance. Each approach combines a feature extraction method, either multi-way principal component analysis (MPCA) or multi-way independent component analysis (MICA), with a classification method, either artificial neural network (ANN) or support vector machines (SVM). The performance obtained by the different approaches is assessed and discussed for a set of simulated faults under different scenarios. One of the faults (a loss in mixing power) could not be detected due to the minimal effect of mixing on the simulated data. The remaining faults could be easily diagnosed and the subsequent discussion provides practical insight into the selection and use of the available techniques to specific applications. Irrespective of the classification algorithm, MPCA renders better results than MICA, hence the diagnosis performance proves to be more sensitive to the selection of the feature extraction technique.

LW-MPCA-SVM-based Fault Diagnosis of Batch Process

An ensemble approach of fault diagnosis based on lifting wavelet, multi-way principal component analysis and support vector machines (LW-MPCA-SVM) for batch process is developed in this paper. Firstly, data are preprocessed to remove noise by lifting scheme wavelets, which are faster than first generation wavelets; then multi-way principal component analysis is used to extract feature for high-accurate diagnosis of the batch process, and SVM is used to diagnose faults. To validate the performance and effectiveness of the proposed scheme, LW-MPCA-SVM is applied to diagnose the faults in the simulation benchmark of fed-batch penicillin fermentation process. The results of simulation clearly demonstrate the effectiveness and feasibility of the proposed method, which diagnoses faults more accurately with desirable reliability.

Design of Simulation System for Batch Process

Process monitoring and fault diagnosis of batch process is a research focus in the industrial control field. In this paper, penicillin fermentation is taken as the research background, a visual batch process simulation system is designed based on mathematical models of an actual production process. By introducing different fault signals to the penicillin fermentation simulation process, the designed system can be used to simulate the real penicillin fermentation production process clearly. In the end, an ideal experimental simulation data for batch process fault diagnosis is provided.

Fault Diagnosis for Batch Processes by Improved Multi-model Fisher Discriminant Analysis

Since there are not enough fault data in historical data sets, it is very difficult to diagnose faults for batch processes. In addition, a complete batch trajectory can be obtained till the end of its operation. In order to overcome the need for estimated or filled up future unmeasured values in the online fault diagnosis, sufficiently utilize the finite information of faults, and enhance the diagnostic performance, an improved multi-model Fisher discriminant analysis is represented. The trait of the proposed method is that the training data sets are made of the current measured information and the past major discriminant information, and not only the current information or the whole batch data. An industrial typical multi-stage streptomycin fermentation process is used to test the performance of fault diagnosis of the proposed method.

FAULT DIAGNOSIS OF BATCH PROCESS BASED ON CMPCA USING DTW

This paper presents fault diagnosis of batch process based on consistent multi-way principle component analysis (CMPCA) using dynamic time warping (DTW). MPCA can rearrange the 3-D array into a 2-D matrix that is used for fault diagnosis in batch process using PCA. Due to the inconsistent of batch process, fault diagnosis usually has some errors based on MPCA. DTW has the ability to reconcile two batch processes and make their trajectories to be consistent. Data from an industrial polymerization reactor are used to illustrate that fault diagnosis based on consistent MPCA is different from MPCA.

Some aspects of fault diagnosis in batch processes

The aim of this paper is to explore alternatives for implementing Fault Diagnosis tools for complex dynamic systems, like complete batch processes. A hybrid, modular, hierarchical architecture is proposed. It is based on a temporal division of the process evolution and the task sequences.