Statistical Fault Detection Research Articles

Multi-scale adaptive multivariate state estimation fault detection enhancement for time-varying industrial system based on multi-output Gaussian process autoregression

Most chemical processes are time-varying and nonlinear due to load changes, product demand transitions or other causes, causing unsatisfactory performance of conventional non-parametric multivariate statistical fault detection method. In this work, a probabilistic multivariate state estimation fault detection method based on multi-output Gaussian process autoregression is first developed to overcome the linear limitation and noise information ignorance of multivariate state estimation technology-based fault detection methods. In multi-output Gaussian process regression framework, the Leave-One-Sample-Out strategy is developed to construct the dataset by leaving each historical sample from the memory matrix out as one training output in turn while regarding the corresponding remaining historical samples as one training input. Then, we design the multi-output Gaussian process autoregression model to estimate the current state of the system. Utilizing the estimation covariance information, a probabilistic index based on multivariate Kullback-Leibler divergence is developed for fault detection. Moreover, a multi-scale adaptive fault detection enhancement based on discrete multi-objective Archimedes optimization and just-in-time learning is further proposed to handle the time-varying problem. In this scheme, the memory matrix is batch-adaptively updated at monthly intervals based on discrete multi-objective Archimedes optimization to handle the slow change of system characteristics caused by equipment degradation. On the other hand, to cope with the rapid change of system characteristics caused by operation transition, the just-in-time learning is employed to select samples from the memory matrix that best match to the current system state for sample-adaptive updating. Finally, two industrial cases are used to verify the performance of the proposed method.

“Won’t get fooled again”: statistical fault detection in COVID-19 Latin American data

BackgroundClaims of inconsistency in epidemiological data have emerged for both developed and developing countries during the COVID-19 pandemic.MethodsIn this paper, we apply first-digit Newcomb-Benford Law (NBL) and Kullback-Leibler Divergence (KLD) to evaluate COVID-19 records reliability in all 20 Latin American countries. We replicate country-level aggregate information from Our World in Data.ResultsWe find that official reports do not follow NBL’s theoretical expectations (n = 978; chi-square = 78.95; KS = 4.33, MD = 2.18; mantissa = .54; MAD = .02; DF = 12.75). KLD estimates indicate high divergence among countries, including some outliers.ConclusionsThis paper provides evidence that recorded COVID-19 cases in Latin America do not conform overall to NBL, which is a useful tool for detecting data manipulation. Our study suggests that further investigations should be made into surveillance systems that exhibit higher deviation from the theoretical distribution and divergence from other similar countries.

A nonlinear process monitoring strategy with a 2-phase fault diagnosis approach

The article delves into the development of Statistical Fault Detection and Diagnosis Strategies for an Integrated Steel Plant (ISP) taking into account the nonlinear relationship amongst the monitored Process and Feedstock Characteristics. The strategies being devised is based on Neural Network Fitting model cum Principal Component Analysis based technique (NNF-PCA) and Kernel Principal Component Analysis (KPCA) based technique. For detection of fault(s) Hotelling T2 chart based on (Principal Component Analysis) PCA and KPCA scores were employed and an ensemble strategy amalgamating KPCA and Self Organizing Map neural network has also been proposed for the detection of the out-of-control observations or faults. The article also proposes a 2-phase Fault Diagnosis approach christened as Preliminary Diagnosis phase and Specific Diagnosis phase. The Preliminary Diagnosis phase is based on Pattern Analysis of the control chart monitoring statistic observations and the Specific Diagnosis phase is based on the employment of appropriate Fault Diagnostic Statistic. The Preliminary Diagnosis reveals the broader source of assignable cause for the onset of the fault(s) and the Specific Diagnosis reveals the relative contribution of the individual Process and Feedstock characteristics. An in-depth comparative analysis between the NNF-PCA based strategy and KPCA based strategy w.r.t. three comparative aspects and four comparative parameters were carried out with their findings being duly highlighted which revealed the slight effectiveness of the KPCA based strategy with respect to the NNF-PCA based counterpart.

A Proportional Digital Controller to Monitor Load Variation in Wind Turbine Systems

Monitoring the variation of the loading blades is fundamental due to its importance in the behavior of the wind turbine system. Blade performance can be affected by different loads that alter energy conversion efficiency and cause potential safety hazards. An example of this is icing on the blades. Therefore, the main objective of this work is to propose a proportional digital controller capable of detecting load variations in wind turbine blades together with a fault detection method. An experimental platform is then built to experimentally validate the main contribution of the article. This platform employs an automotive throttle device as a blade system emulator of a wind turbine pitch system. In addition, a statistical fault detection algorithm is established based on the point change methodology. Experimental data support our approach.

A novel scheme for multivariate statistical fault detection with application to the Tennessee Eastman process

<p style='text-indent:20px;'>Canonical correlation analysis (CCA) has gained great success for fault detection (FD) in recent years. However, it cannot preserve the prior information of the underlying process. To cope with these difficulties, this paper proposes an improved CCA-based FD scheme using a novel multivariate statistical technique, called sparse collaborative regression (SCR). The core of the proposed method is to take the prior information as a supervisor, and then integrate it with CCA. Further, the <inline-formula><tex-math id="M1">\begin{document}$ \ell_{2,1} $\end{document}</tex-math></inline-formula>-norm is employed to reduce redundancy and avoid overfitting, which facilitates its interpretability. In order to solve the proposed SCR, an efficient alternating optimization algorithm is developed with convergence analysis. Finally, some experimental studies on a simulated example and the benchmark Tennessee Eastman process are conducted to demonstrate the superiority over the classical CCA in terms of the false alarm rate and fault detection rate. The detection results indicate that the proposed method is promising.</p>

A Hybrid Approach for Process Monitoring: Improving Data-Driven Methodologies With Dataset Size Reduction and Interval-Valued Representation

Kernel principal component analysis (KPCA) is a well-established data-driven process modeling and monitoring framework that has long been praised for its performances. However, it is still not optimal for large-scale and uncertain systems. Applying KPCA usually takes a long time and a significant storage space when big data are utilized. In addition, it leads to a serious loss of information and ignores uncertainties in the processes. Consequently, in this paper, two uncertain nonlinear statistical fault detection methods using an interval reduced kernel principal component analysis (IRKPCA) are proposed. The main objective of the proposed methods is twofold. Firstly, reduce the number of observations in the data matrix through two techniques: a method, called IRKPCA $_{ED}$ , is based on Euclidean distance between samples as dissimilarity metric such that only one observation is kept in case of redundancy to build the reduced reference KPCA model, and another method, called IRKPCA $_{PCA}$ , is established on the PCA algorithm to treat the hybrid correlations between process variables and extract a reduced number of observations from the training data matrix. Secondly, address the problem of uncertainties in systems using a latent-driven technique based on interval-valued data. Taking into account sensors uncertainties via IRKPCA ensures better monitoring by reducing the computational and storage costs. The study demonstrated the feasibility and effectiveness of the proposed approaches for faults detection in two real world applications: Tennessee Eastman (TE) process and real air quality monitoring network (AIRLOR) data.

Fault Detection using Principal Component Analysis and Mean Value Modeling in a 2 MW gas engine

This paper describes the combination of statistical techniques and mathematical modeling in order to developed a fault detection system in a 2 MW natural gas engine under actual operation conditions. The Mixing chamber, turbochargers, intake and exhaust manifolds, cylinders, throttle and bypass valves, and the electric generator, which are the main components of the gas engine, were studied under a mean value engine to complement the statistical analysis. Objective: The main objective of this paper is to integrate two approaches in order to relate the faults with the changes of mean thermodynamic values of the system, helping to sustain the engine in optimal operating conditions in terms of reliability. The Principal Component Analysis (PCA), a multivariate statistical fault detection technique, was used to analyze the historical data from the gas engine to detect abnormal operation conditions, by means of statistical measures such as Square Prediction Error (SPE) and T2. These abnormal operation conditions are categorized using cluster techniques and contributions plots, to later examine its causes with the support of the results of a mean value mathematical model proposed for the system. The integration of the proposed methods allowed successfully identify which component or components of the engine might be malfunctioning. Once combined, these two methods were able to accurately predict and identify faults as well as shut downs of the gas engine during a month of operation. Statistical analysis was used to detect faults on a 2 MW industrial gas engine, also the result were compared with a mean value model in order to detect variations of the thermodynamic properties of the system at abnormal conditions.

A New Nonlinear Model-Based Fault Detection Method Using Mann–Whitney Test

A well-established theory of statistical inference enables the generation of statistical fault detection approaches. Previous works mainly focus on parametric tests that assume that probabilistic distributions of both healthy and faulty residuals can be parameterized. However, such assumptions may be quite limited for general nonlinear stochastic systems because those residuals are usually with unknown distribution. In this article, a new fault indicator is defined to replace the traditional residual, and a multistep fault detection method is developed via the standard Mann-Whitney (MW) test. Moreover, with weak assumptions on faults and systems, a one-step fault detection approach is proposed by means of the modified MW test. Finally, the effectiveness of the proposed fault detection method is verified by a simulation of three water tank system.

Statistical sensorless short-circuit fault detection algorithm for photovoltaic arrays

One of the main challenges for the conventional protection system of a photovoltaic (PV) array is the occurrence of light fault conditions including a low location mismatch fault, a fault with a high fault path resistance, and a fault under low solar irradiance because the fault current increment is not enough for triggering current-based protective devices. The operation of the maximum power point tracking system and utilizing blocking diodes may also result in a light fault condition. An undetected fault condition causes a potential fire hazard and energy loss. This paper proposes a waveshape based statistical fault detection algorithm for light fault detection. The proposed algorithm quantifies the waveshape tailedness of superimposed PV array power by the kurtosis function. The proposed algorithm is able to discriminate the light fault condition from severe partial shading and is also effective for open-circuit faults. In addition to no need for additional sensors, it does not require a training dataset and prior information about the PV array. The merits of the proposed algorithm are corroborated through several case studies on a simulation model of a test PV array considering the parameter uncertainty and the presence of noise in the signals.

Generating optimal overlapping subsystems for distributed statistical fault detection subject to constraints

Fast and accurate detection of faults is essential for the safe and cost-effective operation of chemical plants. Large scale process systems can potentially be affected by a wide variety of faults. An effective approach for detecting faults in such systems is to first divide or decompose the system's sensors into subsystems and to then implement a statistical monitoring method in each of the subsystems. The detection performance of such distributed multivariate statistical process monitoring methods depends strongly on the decomposition of the system's sensors into subsystems. In our previous work, we developed a simulation optimization method, called Performance Driven Agglomerative Clustering, which uses a greedy search strategy, based on clustering algorithms from graph theory, to find a decomposition of the system's sensors into subsystems for which the detection performance of a distributed monitoring method is near optimal. It is possible that the detection performance can be further improved by allowing a sensor to be part of multiple subsystems in the decomposition. User defined requirements may also place constraints on the decomposition of the sensors into subsystems. In this work, we propose the Extended Performance Driven Agglomerative Clustering method which allows to incorporate constraints and sensors to be allocated to multiple subsystems. To demonstrate its effectiveness, the proposed method is applied to the benchmark Tennessee Eastman Process.

Global-and-local-structure-based neural network for fault detection

A novel statistical fault detection method, called the global-and-local-structure-based neural network (GLSNN), is proposed for fault detection. GLSNN is a nonlinear data-driven process monitoring technique through preserving both global and local structures of normal process data. GLSNN is characterized by adaptively training a neural network which takes both the global variance information and the local geometrical structure into consideration. GLSNN is designed to extract the meaningful low-dimensional features from original high-dimensional process data. After nonlinear feature extraction, Hotelling T2 statistic and the squared prediction error (SPE) statistic are adopted for online fault detection. The merits of the proposed GLSNN method are demonstrated by both theoretical analysis and case studies on the Tennessee Eastman (TE) benchmark process. Extensive experimental results show the superiority of GLSNN in terms of missed detection rate (MDR) and false alarm rate (FAR). The source code of GLSNN can be found in https://github.com/htzhaoecust/glsnn.

A Correlation-Based Distributed Fault Detection Method and Its Application to a Hot Tandem Rolling Mill Process

In a hot tandem rolling mill (HTRM) process, the operating performance of mill stands can determine the quality of steel products, thus, should be properly monitored. Unlike previous work, which focused on a single stand or multiple stands together, this paper proposes an improved statistical fault detection (FD) method for real-time process monitoring of a target stand by considering its correlation with neighboring stands. First, it is verified that the detection performance of the traditional statistical FD method can be improved by reducing the influence of noises in process data. Then, a distributed canonical correlation analysis (CCA)-based FD method is developed to reduce the influence at the target stand by means of the compressed data transmitted from neighboring stands through a communication network. Furthermore, the proposed method can decrease the transmission and computational costs incurred by the centralized CCA method, which makes it applicable to the HTRM process. A hardware-in-the-loop simulation is finally carried out to validate the proposed method, where better monitoring performance is shown compared with the existing methods.

An effective statistical fault detection technique for grid connected photovoltaic systems based on an improved generalized likelihood ratio test

This paper proposes an improved statistical failure detection technique for enhanced monitoring capabilities of PV systems. The proposed technique offers reduced false alarm and missed detection rates compared to the generalized likelihood ratio test (GLRT) by taking into consideration the nature variance of the GLRT statistics and applying a multiscale representation. The multiscale nature of the data provides better robustness to noises and better monitoring quality. The effectiveness of the proposed multiscale weighted GLRT (MS-WGLRT) method in detecting failures is evaluated using a set of synthetic and simulated PV data where the developed chart is used for detecting single and multiple failures (e.g., Bypass, Mix and Shading failures). Moreover, a set of real-data was used in order to prove the effectiveness of the proposed technique in detecting partial shading faults. All results show that the MS-WGLRT method offers better fault detection performances compared to the classical WGLRT and conventional GLRT charts.

Development of a Fault Monitoring Technique for Wind Turbines Using a Hidden Markov Model.

Regular inspection for the maintenance of the wind turbines is difficult because of their remote locations. For this reason, condition monitoring systems (CMSs) are typically installed to monitor their health condition. The purpose of this study is to propose a fault detection algorithm for the mechanical parts of the wind turbine. To this end, long-term vibration data were collected over two years by a CMS installed on a 3 MW wind turbine. The vibration distribution at a specific rotating speed of main shaft is approximated by the Weibull distribution and its cumulative distribution function is utilized for determining the threshold levels that indicate impending failure of mechanical parts. A Hidden Markov model (HMM) is employed to propose the statistical fault detection algorithm in the time domain and the method whereby the input sequence for HMM is extracted is also introduced by considering the threshold levels and the correlation between the signals. Finally, it was demonstrated that the proposed HMM algorithm achieved a greater than 95% detection success rate by using the long-term signals.

Monitoring of a simulated milling circuit: Fault diagnosis and economic impact

The early detection and root cause identification of abnormal events in industrial processes is important, to allow for timely corrective actions, ensuring continued economic operation. This paper investigates the application of statistical fault detection methods, in conjunction with process topology data-driven techniques for root cause analysis, to a simulated milling circuit. Two faults (faulty particle size analyser and rapid mill liner wear) were simulated, and the statistical monitoring techniques tested. Fault detection proved accurate, and variables closely associated with the faults were identified by the root cause analysis. The need to further formalise the selection of data for process topology generation for root cause analysis was highlighted. The milling circuit simulation and fault data has been made available as a resource for future research. Economic performance factors were developed to quantify the impact of the faults and motivate for fault detection and diagnosis.

Model‐Based Analysis of Spur Gears’ Dynamic Behavior in the Presence of Multiple Cracks

Early detection of tooth cracks is crucial for effective condition‐based monitoring and decision making. The scope of this work was to bring more insight into the vibration behavior of spur gears in the presence of single and multiple simultaneous tooth cracks. The investigation was conducted in both time and frequency domains. A finite element analysis was performed to determine the variation in stiffness with respect to the angular position for different combinations of crack lengths. A simplified nonlinear lumped parameter model of a one‐stage gearbox with six degrees of freedom was then developed to simulate the vibration response of faulty external spur gears. Four different multiple‐crack scenarios were proposed and studied. The performances of various statistical fault detection indicators were considered and investigated. The simulation results obtained via MATLAB indicated that, as the severity of a single crack increases, the values of the time domain statistical indicators increase also, but at different rates. Moreover, the number of cracks was found to have a negative effect on the values of all the performance indicators, except for the RMS. The number and amplitude of the sidebands in the frequency spectrum were also considered, while assessing the severity of the faults in each scenario. It was observed that, in the case of consecutive tooth cracks, the number of spectrum peaks and the number of cracks were consistent in the frequency range of 4‐5 kHz. The main finding of this study was that the peak spectral amplitude was the most sensitive indicator of the number and severity of cracks.

Fault Diagnosis for an Industrial High Pressure Leaching Process with a Monitoring Dashboard

Although fault detection and diagnosis are extensively studied and demonstrated on synthetic benchmarks, very few demonstrations of industrial application exist. In this paper, an investigation of faulty conditions in an industrial base metal refinery is conducted using standard statistical fault detection and diagnosis methods, and the IntelliSenzo™ monitoring dashboard. A workflow for offline fault investigation is described, and demonstrated on the industrial data, highlighting the need for careful record keeping, a structured approach, and iterative investigative steps. Potential propagation paths established links between the identified fault, process variables of interest, and engineering hypotheses, guiding the investigation and leading to recommendations of corrective action to the operating plant. The root cause of the fault (leaking and choked feed valves) were identified as potential causes in the investigation. Plant personnel confirmed that this structured approach, using process data, statistical techniques, and engineering knowledge, could speed up fault investigations, and reduce the costs of production losses and plant downtime.

Enhanced data validation strategy of air quality monitoring network

Quick validation and detection of faults in measured air quality data is a crucial step towards achieving the objectives of air quality networks. Therefore, the objectives of this paper are threefold: (i) to develop a modeling technique that can be used to predict the normal behavior of air quality variables and help provide accurate reference for monitoring purposes; (ii) to develop fault detection method that can effectively and quickly detect any anomalies in measured air quality data. For this purpose, a new fault detection method that is based on the combination of generalized likelihood ratio test (GLRT) and exponentially weighted moving average (EWMA) will be developed. GLRT is a well-known statistical fault detection method that relies on maximizing the detection probability for a given false alarm rate. In this paper, we propose to develop GLRT-based EWMA fault detection method that will be able to detect the changes in the values of certain air quality variables; (iii) to develop fault isolation and identification method that allows defining the fault source(s) in order to properly apply appropriate corrective actions. In this paper, reconstruction approach that is based on Midpoint-Radii Principal Component Analysis (MRPCA) model will be developed to handle the types of data and models associated with air quality monitoring networks. All air quality modeling, fault detection, fault isolation and reconstruction methods developed in this paper will be validated using real air quality data (such as particulate matter, ozone, nitrogen and carbon oxides measurement).

Improved Statistical Fault Detection Technique and Application to Biological Phenomena Modeled by S-Systems.

In our previous work, we have demonstrated the effectiveness of the linear multiscale principal component analysis (PCA)-based moving window (MW)-generalized likelihood ratio test (GLRT) technique over the classical PCA and multiscale principal component analysis (MSPCA)-based GLRT methods. The developed fault detection algorithm provided optimal properties by maximizing the detection probability for a particular false alarm rate (FAR) with different values of windows, and however, most real systems are nonlinear, which make the linear PCA method not able to tackle the issue of non-linearity to a great extent. Thus, in this paper, first, we apply a nonlinear PCA to obtain an accurate principal component of a set of data and handle a wide range of nonlinearities using the kernel principal component analysis (KPCA) model. The KPCA is among the most popular nonlinear statistical methods. Second, we extend the MW-GLRT technique to one that utilizes exponential weights to residuals in the moving window (instead of equal weightage) as it might be able to further improve fault detection performance by reducing the FAR using exponentially weighed moving average (EWMA). The developed detection method, which is called EWMA-GLRT, provides improved properties, such as smaller missed detection and FARs and smaller average run length. The idea behind the developed EWMA-GLRT is to compute a new GLRT statistic that integrates current and previous data information in a decreasing exponential fashion giving more weight to the more recent data. This provides a more accurate estimation of the GLRT statistic and provides a stronger memory that will enable better decision making with respect to fault detection. Therefore, in this paper, a KPCA-based EWMA-GLRT method is developed and utilized in practice to improve fault detection in biological phenomena modeled by S-systems and to enhance monitoring process mean. The idea behind a KPCA-based EWMA-GLRT fault detection algorithm is to combine the advantages brought forward by the proposed EWMA-GLRT fault detection chart with the KPCA model. Thus, it is used to enhance fault detection of the Cad System in E. coli model through monitoring some of the key variables involved in this model such as enzymes, transport proteins, regulatory proteins, lysine, and cadaverine. The results demonstrate the effectiveness of the proposed KPCA-based EWMA-GLRT method over Q , GLRT, EWMA, Shewhart, and moving window-GLRT methods. The detection performance is assessed and evaluated in terms of FAR, missed detection rates, and average run length (ARL1) values.

Statistical fault detection in photovoltaic systems

Faults in photovoltaic (PV) systems, which can result in energy loss, system shutdown or even serious safety breaches, are often difficult to avoid. Fault detection in such systems is imperative to improve their reliability, productivity, safety and efficiency. Here, an innovative model-based fault-detection approach for early detection of shading of PV modules and faults on the direct current (DC) side of PV systems is proposed. This approach combines the flexibility, and simplicity of a one-diode model with the extended capacity of an exponentially weighted moving average (EWMA) control chart to detect incipient changes in a PV system. The one-diode model, which is easily calibrated due to its limited calibration parameters, is used to predict the healthy PV array’s maximum power coordinates of current, voltage and power using measured temperatures and irradiances. Residuals, which capture the difference between the measurements and the predictions of the one-diode model, are generated and used as fault indicators. Then, the EWMA monitoring chart is applied on the uncorrelated residuals obtained from the one-diode model to detect and identify the type of fault. Actual data from the grid-connected PV system installed at the Renewable Energy Development Center, Algeria, are used to assess the performance of the proposed approach. Results show that the proposed approach successfully monitors the DC side of PV systems and detects temporary shading.