Process Fault Detection Research Articles

Distributed incipient fault detection with causality-based multi-perspective subblock partitioning for large-scale nonlinear processes

Detecting incipient faults is a crucial task in complex large-scale nonlinear industrial processes. However, conventional distributed frameworks encounter challenges in identifying appropriate subblock partitioning methods. To address this problem, we propose a novel distributed incipient fault detection method, called CMPSP-MKPCA, based on causal relationship multi-perspective subblock partitioning (CMPSP) and modified kernel principal component analysis (MKPCA), facilitating incipient fault detection in large-scale nonlinear processes. Firstly, considering the property of incipient faults to undergo limited propagation among process flow connections, we obtain fine-grainehe direction of potential incipient fault information propagation. Secondly, the design of multi-perspective detection blocks provides real-time selection of sub-block delineation results to adapt to the different needs of incipient faults during latency and transition periods. Thirdly, local detection is realized by combining the Modified KPCA method and Kullback-Leibler divergence (KLD), while the global detection statistics is obtained by Bayesian inference fusion. Simulation experiments based on the continuous stirred tank reactor (CSTR) system and the Tennessee Eastman (TE) system verify the effectiveness and superiority of the method.

Gas turbine aero engine fault detection using Geo-TLSVM and digital twin with multimodal data analysis

DT is the digital representation of physical systems to simulate real situations in a digital version. Here, the DT of the GaT-AE is created for the Fault Detection (FD) process. The GaT is a type of internal combustion engine that uses the air as the working fluid to propel the flight. However, the incorporation of big data analysis in the DT model is challenging in previous works. So, a multimodal GaT-AE fault identification system is proposed. First, the 3D print of the GaT-AE is created. Then, the data is collected from both built-in sensors and DT. Afterward, the multimodal data is pre-processed and balanced by using HT-QNN and QADASYN, respectively. Next, the features are processed by GT-PFS, and finally, the faults are detected by Geo-TLSVM. The analysis proved that the developed model outperformed the other state-of-the-art model with its effective FD rate of 97.24%.

Multi-node knowledge graph assisted distributed fault detection for large-scale industrial processes based on graph attention network and bidirectional LSTMs

Modern industrial processes are characterized by extensive, multiple operation units, and strong coupled correlation of subsystems. Fault detection of large-scale processes is still a challenging problem, especially for tandem plant-wide processes in multiple fields such as water treatment process. In this paper, a novel distributed graph attention network-bidirectional long short-term memory (D-GATBLSTM) fault detection model is proposed for large-scale industrial processes. Firstly, a multi-node knowledge graph (MNKG) is constructed using a joint data and knowledge driven strategy. Secondly, for large-scale industrial process, a global feature extractor of graph attention networks (GATs) is constructed, on the basis of which, sub-blocks are decomposed based on MNKG. Then, local feature extractors of bidirectional long short-term memory (Bi-LSTM) for each sub-block are constructed, in which correlations among multiple sub-blocks are considered. Finally, a multi-subblock fusion collaborative prediction model is constructed and the comprehensive fault detection results are given by the grid search method. The effectiveness of our D-GATBLSTM is exemplified in a secure water treatment process case, where it outperforms baseline models compared, with 27% improvement in precision, 15% increase in recall, and overall F-score enhancement of 0.22.

Dual-attention LSTM autoencoder for fault detection in industrial complex dynamic processes

Complex dynamic characteristics resulting from multi-system coupling and closed-loop control are ubiquitous in modern industrial process data, presenting significant challenges for process fault detection. However, conventional data-driven fault detection methods assume the data to be static or slightly dynamic. Addressing the complex dynamic characteristics and nonlinearity inherent in industrial processes, this paper proposes a dual-attention long short-term memory autoencoder (DALSTM-AE) for fault detection in dynamic processes. Long short-term memory (LSTM) and autoencoder (AE) are combined into a special encoder-decoder LSTM architecture to learn both dynamic features and deep representations of variables in an unsupervised manner. Then, a dual-attention module is embedded in the decoder to properly learn the temporal dependencies associated with long input sequences and retain the most critical information. In addition, based on the reconstruction results of the DALSTM-AE model, two monitoring statistics are designed for fault detection. Finally, the effectiveness and superiority of the proposed method are fully demonstrated through case studies on a numerical simulation example, the Tennessee Eastman (TE) benchmark process, and practical coal pulverizing systems in power plants.

Gap-MK-DCCA-Based Intelligent Fault Diagnosis for Nonlinear Dynamic Systems

In intelligent process monitoring and fault detection of the modern process industry, conventional methods mostly consider singular characteristics of systems. To tackle the problem of suboptimal incipient fault detection in nonlinear dynamic systems with non-Gaussian distributed data, this paper proposes a methodology named Gap-Mixed Kernel-Dynamic Canonical Correlation Analysis. Initially, the Gap metric is employed for data preprocessing, followed by fault detection utilizing the Mixed Kernel-Dynamic Canonical Correlation Analysis. Ultimately, fault identification is conducted through a contribution method based on the T2 statistic. Furthermore, a comparative analysis was conducted using Canonical Variate Analysis, Dynamic Canonical Correlation Analysis, and Mixed Kernel-Dynamic Canonical Correlation Analysis on the Tennessee Eastman process. Experimental results indicate varying degrees of improvements in the detection rate, false alarm rate, missed detection rate, and detection time compared to the comparative methods, demonstrating the industrial value and academic significance of the method.

Accurate Liquid Level Measurement with Minimal Error: A Chaotic Observer Approach

This paper delves into precisely measuring liquid levels using a specific methodology with diverse real-world applications such as process optimization, quality control, fault detection and diagnosis, etc. It demonstrates the process of liquid level measurement by employing a chaotic observer, which senses multiple variables within a system. A three-dimensional computational fluid dynamics (CFD) model is meticulously created using ANSYS to explore the laminar flow characteristics of liquids comprehensively. The methodology integrates the system identification technique to formulate a third-order state–space model that characterizes the system. Based on this mathematical model, we develop estimators inspired by Lorenz and Rossler’s principles to gauge the liquid level under specified liquid temperature, density, inlet velocity, and sensor placement conditions. The estimated results are compared with those of an artificial neural network (ANN) model. These ANN models learn and adapt to the patterns and features in data and catch non-linear relationships between input and output variables. The accuracy and error minimization of the developed model are confirmed through a thorough validation process. Experimental setups are employed to ensure the reliability and precision of the estimation results, thereby underscoring the robustness of our liquid-level measurement methodology. In summary, this study helps to estimate unmeasured states using the available measurements, which is essential for understanding and controlling the behavior of a system. It helps improve the performance and robustness of control systems, enhance fault detection capabilities, and contribute to dynamic systems’ overall efficiency and reliability.

Reliability of river fleet vessels hull elements

Improving the process of fault detection of metal ship hull elements by using a probabilistic approach is considered in the paper. The most labor-intensive part of the defect detection report, namely, measurement and assessment of hull structures individual elements wear, accounting for more than 90 % of the total measurements, is reviewed. Defect detection of a ship hull means the examination, measurement and assessment of defects in each element of the ship hull in order to establish method and volumes of repairs that ensure its reliable operation until the next survey under specified operating conditions. The necessity of carrying out a preliminary calculation stage for defect detection is substantiated. It allows you to significantly reduce the number of measurements of hull structures elements, which will have a positive effect on reducing the cost of repair both due to a smaller number of measurements and the duration of repairs. It has been proved that using the probability approach to predict the residual thickness of individual elements also takes into account the effect on the total strength, which guarantees the reliability of the ship hull in operation. Tables to assess the technical condition of individual hull elements based on their level of wear have been developed, and a method for filling them has improved. The tables are made in two versions, namely, paper and electronic, the most convenient for a specific user — the designer of the defect detection report tables. A test example of using the electronic tables for assessing technical condition by wear level, which proves that the predicted residual thicknesses obtained by calculation will in all cases, be less, than similar ones obtained from measurements results, is carried out using a representative vessel. This proves the reliability and feasibility of using the preliminary defect detection stage, which is essentially an analogue of the concept of the zero stage in ship repair. Using the probability approach to determine the residual thicknesses will allow you not only to reduce the costs of defect detection of the hull itself, but also at early stages (before placing the ship on a slip or dock) to reduce the volume of hull structures repairs, and to choose the most appropriate and effective method of repair.

An Open-Source Software Reliability Model Considering Learning Factors and Stochastically Introduced Faults

In recent years, software development models have undergone changes. In order to meet user needs and functional changes, open-source software continuously improves its software quality through successive releases. Due to the iterative development process of open-source software, open-source software testing also requires continuous learning to understand the changes in the software. Therefore, the fault detection process of open-source software involves a learning process. Additionally, the complexity and uncertainty of the open-source software development process also lead to stochastically introduced faults when troubleshooting in the open-source software debugging process. Considering the phenomenon of learning factors and the random introduction of faults during the testing process of open-source software, this paper proposes a reliability modeling method for open-source software that considers learning factors and the random introduction of faults. Least square estimation and maximal likelihood estimation are used to determine the model parameters. Four fault data sets from Apache open-source software projects are used to compare the model performances. Experimental results indicate that the proposed model is superior to other models. The proposed model can accurately predict the number of remaining faults in the open-source software and be used for actual open-source software reliability evaluation.

Deep Probabilistic Principal Component Analysis for Process Monitoring.

Probabilistic latent variable models (PLVMs), such as probabilistic principal component analysis (PPCA), are widely employed in process monitoring and fault detection of industrial processes. This article proposes a novel deep PPCA (DePPCA) model, which has the advantages of both probabilistic modeling and deep learning. The construction of DePPCA includes a greedy layer-wise pretraining phase and a unified end-to-end fine-tuning phase. The former establishes a hierarchical deep structure based on cascading multiple layers of the PPCA module to extract high-level features. The latter builds an end-to-end connection between the raw inputs and the final outputs to further improve the representation of the model to high-level features. After constructing the model structure of DePPCA, we first present the detailed training processes of the pretraining and fine-tuning stages, then clarify the theoretical merits of the proposed model from the perspective of variational inference. For process monitoring purposes, we develop two statistics based on the established DePPCA. The monitoring performance of these two statistics can remain superior even if the features extracted by DePPCA are significantly compressed to univariate. This makes the feature extraction process and online monitoring procedure of DePPCA quite fast. In other words, the proposed DePPCA can achieve accurate and efficient process monitoring by only extracting one feature for each sample. Finally, the effectiveness of DePPCA is evaluated on the Tennessee Eastman (TE) process and the multiphase flow (MPF) facility.

A Self-Attention Mechanism Integrating Adaptive Double Subspace for Fault Detection in Industrial Processes

A Self-Attention Mechanism Integrating Adaptive Double Subspace for Fault Detection in Industrial Processes

Full Decoupling High-Order Dynamic Mode Decomposition for Advanced Static and Dynamic Synergetic Fault Detection and Isolation

Real industrial processes often present coupled static and dynamic characteristics, leading to significant challenges for fault detection and isolation. However, traditional dynamic modeling methods may lead to the coupling problem of statics and dynamics, which provide ambiguous process status descriptions and incorrect fault isolation results. In this work, a novel full decoupling high-order dynamic mode decomposition (FDHODMD) method is developed for fault detection and isolation of dynamic processes. Different from the existing dynamic methods, the proposed FDHODMD can separate the high-order dynamic information from static information. First, the static characteristics are separated by discarding the decomposed features corresponding to smaller singular values. Then, a high-order dynamic model is established to present the temporal relationships between variables. In this way, the effect of static characteristics on dynamics analysis can be eliminated. Accordingly, from both static and dynamic perspectives, multiple statistics are designed to comprehensively detect the anomalies and provide an explicit status identification. In addition to the static fault isolation strategy, a dynamic fault isolation strategy is designed to recognize the fault variables after detecting the deviation, which divides the complex dynamic system into several individual dynamic modes to avoid the interaction of dynamic characteristics. Thereupon, the contributions of different variables to each dynamic mode can be clearly presented to recognize the fault variables. Finally, the validity of the proposed method is illustrated through both a numerical case and a real industrial process. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —In industrial processes, the static and dynamic characteristics of data are often coupled. This work presents a FDHODMD method, which can be considered the high-order version of DMD, to achieve the decoupling of dynamics and statics while extracting high-order dynamic characteristics. Then the proposed fault detection strategy designs multiple statistics for monitoring the process from both dynamic and static perspectives, which ensures the safe operation of industrial processes. After detecting the anomaly, the proposed fault isolation strategy can recognize the fault variables separately dominating the dynamic and static deviation, which may help engineers locate the fault source.

Time-slice dynamic prediction and multiway serial PCA for batch industrial process monitoring

This paper introduces an innovative method for batch industrial process monitoring and fault detection. By combining time-slice dynamic prediction with an enhanced adaptive multiway principal component analysis (PCA) technique, it aims to balance the trade-off between high fault detection rates and low false alarm rates in monitoring batch industrial data. The method extracts dynamic features from 3D batch data time slices and utilizes a serial PCA framework with variable control limits for real-time monitoring. Comparative evaluations against traditional adaptive multiway PCA and dynamic time-slice canonical correlation analysis approaches demonstrate its superior performance. A case study using a novel cold rolling mill simulation model underscores its effectiveness for batch industrial process monitoring and fault detection.

Fault detection and diagnosis for chemical processes using dynamic global–local preserving projections

AbstractMost traditional multivariate statistical monitoring methods require an assumption that the observation values at a certain moment and a past moment are statistically independent. However, in actual chemical and biological processes, the sample at a certain moment is often affected by the previous moment. Therefore, given the problem of more false alarms and poor detection ability based on the traditional principal component analysis, this article proposes a dynamic global–local preserving projections (DGLPP) algorithm. Unlike dynamic local preserving projections (DLPP) and dynamic principal component analysis (DPCA), DGLPP controls the global and local information retained in the dimensionality reduction data by introducing weight coefficients, which makes the algorithm applicable to more types of industrial processes. Moreover, new parameter determination methods are also proposed for improved detection and diagnosis. Through the improved contribution graph method, we can see the influence degree of each variable on the fault, to monitor and isolate the fault. Finally, by verifying the operation of the multivariable process and two practical cases, the results show that compared with DPCA, DLPP, and global local retained projection (GLPP) methods, the performance under this method has been significantly improved.

Improved Fault Detection in Chemical Engineering Processes via Non-Parametric Kolmogorov–Smirnov-Based Monitoring Strategy

Fault detection is crucial in maintaining reliability, safety, and consistent product quality in chemical engineering processes. Accurate fault detection allows for identifying anomalies, signaling deviations from the system’s nominal behavior, ensuring the system operates within desired performance parameters, and minimizing potential losses. This paper presents a novel semi-supervised data-based monitoring technique for fault detection in multivariate processes. To this end, the proposed approach merges the capabilities of Principal Component Analysis (PCA) for dimensionality reduction and feature extraction with the Kolmogorov–Smirnov (KS)-based scheme for fault detection. The KS indicator is computed between the two distributions in a moving window of fixed length, allowing it to capture sensitive details that enhance the detection of faults. Moreover, no labeling is required when using this fault detection approach, making it flexible in practice. The performance of the proposed PCA–KS strategy is assessed for different sensor faults on benchmark processes, specifically the Plug Flow Reactor (PFR) process and the benchmark Tennessee Eastman (TE) process. Different sensor faults, including bias, intermittent, and aging faults, are considered in this study to evaluate the proposed fault detection scheme. The results demonstrate that the proposed approach surpasses traditional PCA-based methods. Specifically, when applied to PFR data, it achieves a high average detection rate of 98.31% and a low false alarm rate of 0.25%. Similarly, when applied to the TE process, it provides a good average detection rate of 97.27% and a false alarm rate of 6.32%. These results underscore the efficacy of the proposed PCA–KS approach in enhancing the fault detection of high-dimensional processes.

A novel ensemble learning approach for fault detection of sensor data in cyber-physical system

Cyber-physical systems (CPS) play a pivotal role in various critical applications, ranging from industrial automation to healthcare monitoring. Ensuring the reliability and accuracy of sensor data within these systems is of paramount importance. This research paper presents a novel approach for enhancing fault detection in sensor data within a cyber-physical system through the integration of machine learning algorithms. Specifically, a hybrid ensemble methodology is proposed, combining the strengths of AdaBoost and Random Forest with Rocchio’s algorithm, to achieve robust and accurate fault detection. The proposed approach operates in two phases. In the first phase, AdaBoost and Random Forest classifiers are trained on a diverse dataset containing normal and faulty sensor data to develop individual base models. AdaBoost emphasizes misclassified instances, while Random Forest focuses on capturing complex interactions within the data. In the second phase, the outputs of these base models are fused using Rocchio’s algorithm, which exploits the similarities between faulty instances to improve fault detection accuracy. Comparative analyses are conducted against individual classifiers and other ensemble methods to validate the effectiveness of the hybrid approach. The results demonstrate that the proposed approach achieves superior fault detection rates. Additionally, the integration of Rocchio’s algorithm significantly contributes to the refinement of the fault detection process, effectively leveraging the strengths of AdaBoost and Random Forest. In conclusion, this research offers a comprehensive solution to enhance fault detection capabilities in cyber-physical systems by introducing a novel ensemble framework. By synergistically combining AdaBoost, Random Forest, and Rocchio’s algorithm, the proposed methodology provides a robust mechanism for accurately identifying sensor data anomalies, thus bolstering the reliability and performance of cyber-physical systems across a multitude of critical applications.

Shared Parameter Network: An efficient process monitoring model

The use of deep learning methods for fault detection and diagnosis (FDD) has continued to gain research attention due to the continuous availability of data and the need for reliable FDD methods. In our previous work, we proposed a probabilistic bidirectional recurrent network (PBRN) for process prediction and detection of faults. A limitation of PBRN is the model size. In this paper, we reduce the computational requirements of the original PBRN by introducing a shared parameter network (SPN). The SPN uses a shared parameter space to learn relevant features for process prediction and fault detection. This shared parameter setting reduces the model size and training time significantly. Using the SPN, we achieved a 76 percent reduction in model size and a 48 percent reduction in training time. The training validation loss profile and fault detection performances of both models also demonstrate the superiority of the SPN.

A derivative, integral, and proportional features extractor for fault detection in dynamic processes

Detecting faults in dynamic systems is challenging due to temporal dependencies and signal correlations. Feature extraction from time-series data is a common step in fault detection, which is usually performed according two main approaches: knowledge-based and statistical. Knowledge-based methods provide interpretable features but require significant design time; also, they usually lack generality. Statistical methods are faster but lead to dimensionality issues and lack of interpretability. To address these challenges, we propose an interpretable and automated feature extraction method. It combines the benefits of knowledge-based and statistical methods, offering a fast solution for extracting effective and interpretable indicators without requiring prior domain knowledge. Also, this method consistently computes a fixed set of interpretable features describing the process’s dynamic behavior.We extensively compare our approach with state-of-the-art methods considering nine open-source datasets spanning various domains. Our results show that classifiers trained on features extracted with the proposed method achieve performance comparable to those provided by state-of-the-art automatic features extractors, according to F1-score, true positive rate, and false positive rate. In addition, our approach proves to be more robust to dimensionality issues and enhances interpretability, extracting a reduced set of features effective at providing insight into the detected anomalies’ characteristics. Additionally, we demonstrate that the selected features maintain consistent performance across different classifiers, showcasing their versatility.

A novel fault detection scheme for a nonlinear dynamic process based on generalized non‐negative matrix projection‐maximum mean discrepancy: Application on the DAMADICS benchmark process

AbstractIn order to address the issue of minor fault detection in nonlinear dynamic processes, this paper proposes a fault detection method based on generalized non‐negative matrix projection‐maximum mean discrepancy (GNMP‐MMD). Firstly, the GNMP is employed to acquire the residual scores of the samples. Subsequently, a sliding window approach is integrated with MMD for real‐time monitoring of sample status within the residual subspace. In this study, GNMP is utilized to mitigate the impact of non‐Gaussianity in data distribution, while MMD serves to alleviate autocorrelation among samples. A numerical case and experimental data collected from the DAMADICS process are utilized to simulate and validate the proposed method. Compared to traditional principal component analysis (PCA), dynamic principal component analysis (DPCA), dynamic kernel principal component analysis (DKPCA), non‐negative matrix factorization (NMF), GNMP, and MMD, the experiment results clearly illustrate the feasibility of the proposed method.

A fault detection method based on sparse dynamic canonical correlation analysis

AbstractFault detection based on canonical correlation analysis (CCA) has received increased attention due to its efficiency in exploring the relationship between input and output. However, traditional CCA may generate redundant features in both the input and output projections while maximizing the correlations. In this paper, sparse dynamic canonical correlation analysis (SDCCA) is developed for dealing with the fault detection of dynamic processes. Through posing sparsity in the extraction of features, the interpretability of canonical variates is enhanced attributed to the sparsity of canonical weights. Based on the SDCCA model, the monitoring metric is established for fault detection. Moreover, the upper control limit (UCL) based on monitoring metrics is determined by the kernel density estimation (KDE) method to avoid the violation of the Gaussian assumption. The superiority of the proposed SDCCA‐based fault detection method is illustrated through a comparative study of the Tennessee Eastman process benchmark.

Sensitivity analysis of an automated fault detection algorithm for residential air-conditioning systems

The state of the art of fault detection and diagnosis (FDD) for residential air-conditioning systems is expensive and not yet amenable to widespread implementation. FDD for homes can significantly reduce utility costs, and increase the lifespan of the equipment. The cost barriers currently, however, make FDD for homes economically unviable for large scale implementation. In prior work, we offered a solution to reduce FDD costs by proposing an automated fault detection algorithm to serve as a screening step before more expensive FDD tests can be conducted. The algorithm uses only the home thermostat and local weather information to identify thermodynamic parameters and detect high-impact air-conditioning faults, including those that occur during equipment installation. We had tested the algorithm on a single EnergyPlusTM model of a home in Orlando, Florida. The thermodynamic parameter identification process is highly nonconvex involving several local optimal solutions. In this paper we propose a novel method to select the best model for fault detection from among the list of local optimal solutions to make the algorithm more robust to homes of different construction, without which the fault detection process would be infeasible. Another unique contribution of the paper is implementing the solution on real-world data. We also bring the algorithm closer to market by testing it on real-world data. We implement the algorithm on data obtained from experiments conducted by the Florida Solar Energy Center (FSEC) on a laboratory home equipped with a heat pump where faults were intentionally added for a period of seven months. The algorithm successfully detected an undercharge fault with 70.6% accuracy, concurrent duct leakage and undercharge faults with 85.2% accuracy, and duct leakage faults with 69.1% accuracy. A sensitivity analysis is also performed on EnergyPlus models of nine types of homes that vary in construction to demonstrate the robustness of the algorithm. The algorithm achieves an average accuracy of 71% for no-fault condition, 77% for 40% undercharge fault, and 76% for duct-leak fault.