Fault Detection Of Chemical Processes Research Articles

Intelligent prediction of incipient fault in vinyl chloride production process based on deep learning

With the development of industrial information technology, deep learning (DL) has been successfully applied in chemical process fault detection. However, the features of incipient faults could not be more evident at the initial stage, which makes it difficult for deep learning to fully mine the feature probability information, resulting in poor performance of incipient fault monitoring. This paper proposes an intelligent predictive model based on mechanistic modeling to predict incipient faults and determine optimal response times for vinyl chloride production (VCP) processes. Firstly, a dynamic simulation of the VCP production process is performed to obtain datasets of early faults. Secondly, data dimensionality reduction is performed by cleverly combining spearman ranking correlation coefficient (SRCC) and slow feature analysis (SFA). Then, a long short-term memory network (LSTM) with attention mechanism (AM) is built to predict the future trends of key variables. Finally, the optimal response time for different types of incipient faults is determined by comparing the effectiveness of various control schemes. Compared with traditional methods, the R2 of the proposed prediction model corresponding to different step sizes can reach 99.7%, 99.5%, 99.2%, and 98.7%. In addition, the response times for single and parallel faults are 2 and 2.5 h, which helps to control and eliminate potential incidents in advance.

A comprehensive fault detection and diagnosis method for chemical processes

Deep learning methods based on supervised learning have various applications in chemical process fault detection and diagnosis (FDD). However, these methods as black boxes are unable to provide guidance for process recovery, and the labels and fault samples available in actual chemical processes are insufficient for ideal supervised learning. In this study, eMixDTFCN containing FDD backbone network, semi-supervised learning framework, and explainable method is proposed for the first time to address the above challenges simultaneously. Dense temporal feature convolutional network (DTFCN) combining process variable embedding and dense temporal convolution is taken as the backbone network, where the sparse features from the high-dimensional embedded space are learned efficiently through the dense temporal convolution. MixMatch based on consistent regularization and entropy minimization enables DTFCN with semi-supervised learning and imbalanced learning. SHAP method trains an explainer to provide fault-related variables for process recovery. The high effectiveness and comprehensiveness of eMixDTFCN are validated in two typical chemical processes.

Chemical process fault detection and trend analysis based on KESN

AbstractFault detection has great significance for chemical process safety with the development of science and technology. The conventional echo state network‐based fault detection method does not highlight key fault features, and cannot forecast future fault trend after the occurrence of faults. For the above problems, a chemical process fault detection and trend analysis strategy based on key feature enhanced echo state network (KESN) is proposed. First, dynamic features are extracted by a detecting echo state network. Then, a weighting strategy is designed to enhance key features and increase fault detection rates. After detecting a fault, independent component analysis is utilized to extract independent key features. Future fault trend is forecasted based on the forecasting multi‐KESN. Simulation results on the Tennessee Eastman process demonstrate the effectiveness of the proposed method.

Robust Fault Detection in Monitoring Chemical Processes Using Multi-Scale PCA with KD Approach

Effective fault detection in chemical processes is of utmost importance to ensure operational safety, minimize environmental impact, and optimize production efficiency. To enhance the monitoring of chemical processes under noisy conditions, an innovative statistical approach has been introduced in this study. The proposed approach, called Multiscale Principal Component Analysis (PCA), combines the dimensionality reduction capabilities of PCA with the noise reduction capabilities of wavelet-based filtering. The integrated approach focuses on extracting features from the multiscale representation, balancing the need to retain important process information while minimizing the impact of noise. For fault detection, the Kantorovich distance (KD)-driven monitoring scheme is employed based on features extracted from Multiscale PCA to efficiently detect anomalies in multivariate data. Moreover, a nonparametric decision threshold is employed through kernel density estimation to enhance the flexibility of the proposed approach. The detection performance of the proposed approach is investigated using data collected from distillation columns and continuously stirred tank reactors (CSTRs) under various noisy conditions. Different types of faults, including bias, intermittent, and drift faults, are considered. The results reveal the superior performance of the proposed multiscale PCA-KD based approach compared to conventional PCA and multiscale PCA-based monitoring methods.

Fault Detection and Diagnosis in Tennessee Eastman Process with Deep Autoencoder

Data-driven modeling has been considered as an attractive approach for fault detection in chemical processes. Of special interest to industry are methods that represent nonlinear phenomena and detect complex faults. In this paper, a semi-supervised deep learning method - deep autoencoder for fault detection in Tennessee Eastman Process (TEP) is proposed. The TEP process is a simulated benchmark for evaluating process control and monitoring methods. The performance of the proposed method is evaluated and compared to Principal Component Analysis (PCA). The experimental results demonstrate that the proposed optimized five-layers DAE model for fault detection outperforms the standard PCA. Of special importance to real-world applications is its capability for automatic variable selection. In comparison to PCA it demonstrated higher prediction accuracy for most of the generated faults. Deep autoencoder has the potential to become an excellent approach for process monitoring and fault detection in chemical processes.

Robust fault detection for chemical processes based on dynamic low-rank matrix and optimized LSTM

The data collected by sensors in modern chemical process systems are always contaminated by industrial noise, so robust fault detection is an important technology to ensure process safety. However, the existing robust detection techniques have limitations in extracting dynamic and time series features of chemical process data. In this paper, a robust fault detection method based on dynamic low-rank matrix and optimized LSTM is proposed for dynamic chemical processes under noise background. First, a new low-rank matrix decomposition method, dynamic principal component pursuit (DPCP), is proposed for the dynamic characteristics of process data containing noise. The PCP is improved using the time delay excursion method, i.e., the DPCP is constructed by embedding an augmented dynamic matrix with a time lag factor on the PCP and solved using the alternating direction multiplier method. The purpose is to strip the useless noise from the useful detection information and extract the dynamic low-rank matrix. Second, an optimized LSTM (OPLSTM) is proposed for the time series features of dynamic low-rank matrix. vector factors for feature selection of memory cells in the LSTM are reweighted to balance the time series features of industrial processes and avoid heavy reliance on erroneous fault information. In addition, support vector data description (SVDD) is used to describe the hypersphere with clear boundary, and distance-based statistic is constructed to achieve fault detection with high fault detection rate and low false alarm rate. Finally, to assess the effectiveness of the proposed method, we performed extensive experiments on the Tennessee Eastman Process and the electrolytic aluminum process, and compared them with multiple methods. The experimental results show that the proposed method is effective and robust for chemical processes fault detection in the background of industrial noise.

High-fidelity positive-unlabeled deep learning for semi-supervised fault detection of chemical processes

With the rapid development of the modern chemical process industry, process monitoring techniques have been investigated to enhance the loss prevention capability. Fault detection which attempts to detect whether an abnormal condition has happened is an essential step of process monitoring. Supervised learning based fault detection methods usually are not feasible in practical industry situations, because their most important prerequisite that adequate data labels are accessible is extremely hard to meet. In real industrial situations, a common scenario is that a few normal state data are labeled while no fault state data are labeled, and this is a natural fit for the positive-unlabeled (PU) learning. In this work, a three-step high-fidelity PU (THPU) approach based on deep learning is proposed for semi-supervised fault detection of chemical processes. A self-training nearest neighbors (STNN) algorithm is designed in Step 1 to conservatively generate a small reliable positive dataset by extracting data from the unlabeled training dataset. A positive and negative data recognition (PNDR) algorithm is designed in Step 2 to augment the reliable positive dataset and generate a reliable negative dataset based on the remaining unlabeled dataset. A convolutional neural network is trained in Step 3 based on the above reliable datasets as the binary classifier for online fault detection. The benchmark simulated Tennessee Eastman process dataset and a real industrial hydro-cracking process dataset are utilized to illustrate the effectiveness and robustness of the proposed THPU approach. The experimental results demonstrate the superiority of the proposed approach compared to the other competing PU learning approaches and supervised fault detection models.

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.

Novel variation mode decomposition integrated adaptive sparse principal component analysis and it application in fault diagnosis

The sparse principal component analysis (SPCA) is widely used in the fault detection for nonlinear complex chemical processes in recent years. However, insufficient data processing, fixed models and fault type single classification cannot be used in the time-varying process. Therefore, a novel adaptive sparse principal component analysis (ASPCA) algorithm fused with improved variation mode decomposition (IVMD) (ASPCA-IVMD) is proposed for fault detection in chemical processes. The bat algorithm is innovatively integrated to optimize the parameters of the variable modulus decomposition. Then the optimized parameters are used for data preprocessing to suppress noise. In addition, based on the traditional SPCA, the threshold calculation is fused to realize the adaptive selection of principal components. After the principal components are determined, T2 and Q statistics are used for fault detection. Finally, the proposed method is verified by the Tennessee Eastman process case. The results demonstrate that the proposed method can select the principal components adaptively according to the data for having the real-time property of chemical process. Meanwhile, compared with traditional methods (principal component analysis, sparse principal component analysis, deep belief network integrating dropout, adaptive unscented Kalman filter integrating radial basis function and sparse deep belief network), the detection rate of the ASPCA-IVMD method is more than 99%, which shows superiority.

Fault detection in Tennessee Eastman process with temporal deep learning models

Automated early process fault detection and prediction remains a challenging problem in industrial processes. Traditionally it has been done by multivariate statistical analysis of sensor readings and, more recently, with the help of machine learning methods. The quality of machine learning models strongly depends on feature engineering, that in turn heavily relies on expertise of the process engineers and model developers. With the recent advent of deep learning neural network methods and abundance of available sensor data, it became possible to develop advanced approaches to early fault detection and prediction that do not require feature engineering and provide more accurate and timely results.In this paper we investigate a wide range of recurrent and convolutional architectures on the publicly available simulated Tennessee Eastman Process extended TEP dataset for the fault detection in chemical processes. We have selected the best architecture for the task and proposed a novel temporal CNN1D2D architecture that achieves overall better performance on the dataset than any referenced method. We have also proposed to use Generative Adversarial Network GAN to extend and enrich data used in training.

Fault detection and identification using Bayesian recurrent neural networks

In the processing and manufacturing industries, there has been a large push to produce higher quality products and ensure maximum efficiency of processes, which requires approaches to effectively detect and resolve disturbances to ensure optimal operations. While many types of disturbances can be compensated by a control system, it cannot handle some large process disruptions. As such, it is important to develop monitoring systems to effectively detect and identify those faults such that they can be quickly resolved by operators. This article proposes a novel probabilistic fault detection and identification method which adopts a newly developed deep learning approach using Bayesian recurrent neural networks (BRNNs) with variational dropout. The BRNN model is general and can model complex nonlinear dynamics. Moreover, compared to traditional statistic-based data-driven fault detection and identification methods, the proposed BRNN-based method yields uncertainty estimates which allow for simultaneous fault detection of chemical processes, direct fault identification, and fault propagation analysis. The performance of the method is demonstrated and contrasted to (dynamic) principal component analysis, which is widely applied in the industry, in the benchmark Tennessee Eastman process (TEP) and a real chemical manufacturing dataset.

Machine Learning-Based Reduced Kernel PCA Model for Nonlinear Chemical Process Monitoring

Principal component analysis (PCA) is a popular tool for linear dimensionality reduction and fault detection. Kernel PCA (KPCA) is the nonlinear form of the PCA, which better exploits a complicated spatial structure of high-dimensional features, where a kernel function implicitly defines a nonlinear transformation into a feature space wherein standard PCA is performed. Despite its success and flexibility, conventional KPCA might not perform properly because the use of KPCA for a large-sized training dataset imposes a high computational load and a significant storage memory space since the required elements used for modelling have to be saved and used for monitoring, as well. To address this problem, a reduced KPCA (RKPCA) for fault detection of chemical processes is developed. RKPCA is a novel machine learning tool which merges dimensionality reduction, supervised learning as well as kernel selection. This novel method is used to reduce the size of recorded measurements while maintaining the most relevant data features. The removed observations, including redundant samples that are linearly correlated in the collected measurements, are described by only one sample. The obtained uncorrelated observations via PCA technique are then employed to identify the reduced KPCA model by which Hotelling $$T^2$$ and squared predictive error or Q statistics are extracted for detection purposes. Besides, their combination is also used as a detection index. The performance of the proposed process monitoring technique is illustrated through its application to Tennessee Eastman process. The obtained results demonstrate the effectiveness of the developed RKPCA technique in detecting various faults with remarkably reduced computation time and memory storage space.

The Extended Kalman Filter for State Estimation of Dynamic UV Flash Processes

We present an extended Kalman filter for state estimation of semi-explicit index-1 differential-algebraic equations. It is natural to model dynamic UV flash processes with such differential-algebraic equations. The UV flash is a mathematical statement of the second law of thermodynamics. It is therefore important to thermodynamically rigorous models of many phase equilibrium processes. State estimation of UV flash processes has applications in control, prediction, monitoring, and fault detection of chemical processes in the oil and gas industry, e.g. separation, distillation, drilling of oil wells, multiphase flow in oil pipes, and oil production. We present a numerical example of a UV flash separation process. It involves soft sensing of vapor-liquid compositions based on temperature and pressure measurements.

Covariance eigenpairs neighbour distance for fault detection in chemical processes

AbstractThis paper presents a new data‐driven fault detection method called covariance eigenpairs neighbour distance (CEND) for monitoring chemical processes. It processes measured data in one‐step sliding windows to estimate covariance, and the eigenpairs of each sample covariance matrix are recursively calculated using rank‐one modification. The eigenvalues and selected elements of eigenvectors are stacked into reference vectors. For each window data containing the latest measurement vector, the neighbour distance of eigenpairs from the reference matrix can be effectively calculated by virtue of k‐d tree. The neighbour distance serves as the detection index, whose control limit can be determined by validation dataset with assigning a significance level. A high neighbour distance exceeding the control limit implies the occurrence of fault. Simulations on the continuous stirred tank reactor (CSTR) and the Tennessee Eastman process (TEP) both indicate the superior fault detectability of the proposed method, compared with conventional multivariate statistical process monitoring (MSPM) methods such as principal component analysis (PCA) and independent component analysis (ICA).

Data Visualization and Visualization-Based Fault Detection for Chemical Processes

Over the years, there has been a consistent increase in the amount of data collected by systems and processes in many different industries and fields. Simultaneously, there is a growing push towards revealing and exploiting of the information contained therein. The chemical processes industry is one such field, with high volume and high-dimensional time series data. In this paper, we present a unified overview of the application of recently-developed data visualization concepts to fault detection in the chemical industry. We consider three common types of processes and compare visualization-based fault detection performance to methods used currently.

Multiscale PLS-based GLRT for fault detection of chemical processes

Fault detection is necessary to assure safe operations of plant and productivity at desired level. Most of the industrial process data are highly correlated, which causes an issue with analyzing huge process data. To address this issue, the Partial least squares (PLS) method has been used successfully in process monitoring. However, measured process data are usually contaminated with errors that mask the important features in the data and reduce the effectiveness of any fault detection method in which these data are used. Unfortunately chemical process data (as in the case of most practical data) usually possess multiscale characteristics, meaning that they contain features and noise that occur at varying contributions over time and frequency. Wavelet-based multiscale representation of data has been shown to be a powerful data analysis, modeling, and feature extraction tool due to its ability to provide efficient separation of deterministic and stochastic features. Hence, the objective of this paper is to extend our previous work (Botre et al., 2016) to deal with autocorrelated and noise in the dataset with multiscale representation and present a new fault detection method of chemical processes using Multiscale Partial Least Square (MSPLS)-based generalized likelihood ratio test (GLRT) technique. The efficiency and robustness of the proposed method are demonstrated through two illustrative examples, one using simulated continuous stirred tank reactor (CSTR) and the other using Tennessee Eastman process problem (TEP) data. The results demonstrate the effectiveness of developed MSPLS-based GLRT over the conventional PLS-based techniques for fault detection in CSTR and TEP processes.

Data-based fault detection in chemical processes: Managing records with operator intervention and uncertain labels

Developing data-driven fault detection systems for chemical plants requires managing uncertain data labels and dynamic attributes due to operator-process interactions. Mislabeled data is a known problem in computer science that has received scarce attention from the process systems community. This work introduces and examines the effects of operator actions in records and labels, and the consequences in the development of detection models. Using a state space model, this work proposes an iterative relabeling scheme for retraining classifiers that continuously refines dynamic attributes and labels. Three case studies are presented: a reactor as a motivating example, flooding in a simulated de-Butanizer column, as a complex case, and foaming in an absorber as an industrial challenge. For the first case, detection accuracy is shown to increase by 14% while operating costs are reduced by 20%. Moreover, regarding the de-Butanizer column, the performance of the proposed strategy is shown to be 10% higher than the filtering strategy. Promising results are finally reported in regard of efficient strategies to deal with the presented problem.

Kernel PLS-based GLRT method for fault detection of chemical processes

Fault detection is essential for proper and safe operation of various chemical processes, and it has recently become even more important than ever before. In this paper, we extended our previous work (Mansouri et al. (2016)), which addresses the problem of fault detection of chemical systems using kernel principal component analysis (KPCA)-based generalized likelihood ratio test (GLRT), to widen its applicability for processes represented by input-output models. Specifically, hypothesis testing fault detection technique that are based on linear and nonlinear partial least squares (PLS) models are developed. For nonlinear PLS models, a kernel PLS (KPLS) modeling framework is utilized. KPLS has been widely used to model various nonlinear processes, such as distillation columns and reactors. Thus, in the current work, a KPLS-based GLRT fault detection method is developed, in which KPLS is used as a modeling framework and the KPLS model generated residuals are evaluated using a GLRT statistic. The fault detection performance of the developed KPLS-based GLRT method is illustrated through a simulated example representing a continuously stirred tank reactor (CSTR). The simulation results show that the KPLS-based GLRT method outperforms its linear PLS-based version, and that both of the aforementioned techniques provide clear advantages over the conventional linear and nonlinear PLS based statistics, i.e., T2 and Q.

Kernel PCA-based GLRT for nonlinear fault detection of chemical processes

Fault detection is often utilized for proper operation of chemical processes. In this paper, a nonlinear statistical fault detection using kernel principal component analysis (KPCA)-based generalized likelihood ratio test (GLRT) is proposed. The objective of this work is to extend our previous work (Harrou et al. (2013) to achieve further improvements and widen the applicability of the developed method in practice by using the KPCA method. The KPCA presented here is derived from the nonlinear case of principal component analysis (PCA) algorithm and it is investigated here as modeling algorithm in the task of fault detection. The fault detection problem is addressed so that the data are first modeled using the KPCA algorithm and then the faults are detected using GLRT. The detection stage is related to the evaluation of detection indices, which are signals that reveal the fault presence. Those indices are obtained from the analysis of the difference between the process measurements and their estimations using the KPCA technique. The fault detection performance is illustrated through two simulated examples, one using synthetic data and the other using simulated continuously stirred tank reactor (CSTR) data. The results demonstrate the effectiveness of the KPCA-based GLRT method over the conventional KPCA method through its two charts T2 and Q for detection of single as well as multiple sensor faults.

Finding a trade-off between observability and economics in the fault detection of chemical processes

This paper presents a methodology to quantitatively gauge the potential economical loss due to unobserved faults when standard statistical monitoring charts are used. It is shown that in closed loop operation, a shorter time for detection may result from retuning the controller at the expense of higher product variability. Accordingly, an optimization approach is proposed for finding a trade-off between the economic losses resulting from lack of detection and losses resulting from higher product variability. In order to account for faults with different frequency contents, the method is applied in the frequency domain. The proposed optimization based methodology is later validated in the time domain.