Data-driven Fault Detection Research Articles

On Real-Time Fault Detection in Wind Turbines: Sensor Selection Algorithm and Detection Time Reduction Analysis

In this paper, we address the problem of real-time fault detection in wind turbines. Starting from a data-driven fault detection method, the contribution of this paper is twofold. First, a sensor selection algorithm is proposed with the goal to reduce the computational effort of the fault detection method. Second, an analysis is performed to reduce the data acquisition time needed by the fault detection method, that is, with the goal of reducing the fault detection time. The proposed methods are tested in a benchmark wind turbine where different actuator and sensor failures are simulated. The results demonstrate the performance and effectiveness of the proposed algorithms that dramatically reduce the number of sensors and the fault detection time.

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.

Data-driven Fault Detection and Diagnosis for HVAC water chillers

Faulty operations of Heating, Ventilation and Air Conditioning (HVAC) chiller systems can lead to discomfort for the users, energy wastage, system unreliability and shorter equipment life. Faults need to be early diagnosed to prevent further deterioration of the system behaviour and energy losses. Since it is not a common practice to collect historical data regarding unforeseen phenomena and abnormal behaviours for HVAC installations, in this paper, a semi-supervised data-driven approach is employed for fault detection and isolation that makes no use of a priori knowledge about abnormal phenomena. The proposed method exploits Principal Component Analysis (PCA) to distinguish anomalies from normal operation variability and a reconstruction-based contribution approach to isolate variables related to faults. The diagnosis task is then tackled by means of a decision table that associates the influence of faults to certain characteristic features. The Fault Detection and Diagnosis (FDD) algorithm performance is assessed by exploiting experimental datasets from two types of water chiller systems.

Weighted Data-Driven Fault Detection and Isolation: A Subspace-Based Approach and Algorithms

Well-established theory of subspace system identification and model-based fault detection and isolation (FDI) enable the birth of subspace-based data-driven FDI approach. In this paper, we develop subspace-based FDI approach with a scheme of weighted historical and operating data. We propose two kinds of weighted data-driven fault detection algorithms and present fault isolation algorithm and its modified version incorporated with forgetting factors. Analysis of sensitivity and precision shows the weighted algorithms can obtain more accurate results without loss of sensitivity. Effectiveness and improvements of the proposed algorithms are validated on the widely used benchmark platform of Tennessee-Eastman process (TEP).

Differential feature based hierarchical PCA fault detection method for dynamic fault

By sensor accuracy degradation or unwanted alternating current signals, sensor fault with zero cross point (ZCP) may occur in real systems and conventional data-driven fault detection methods could be invalid. In this regard, this paper proposes a hierarchical principal component analysis (PCA) fault detection method based on the differential features of dynamic faults to detect the fault with ZCPs. The main contribution of this work are as follows: (1) A new differential based feature extraction method is first proposed to well character the dynamic trend of the observation; (2) then, a hierarchical detection criterion is proposed according to the detection ability of each round of PCA anomaly detection; (3) it is convenient to extend the proposed method to other statistical based fault detection techniques whose detection criteria are also a distance defined by fault amplitude.

Orthogonal nonnegative matrix factorization based local hidden Markov model for multimode process monitoring

Traditional data driven fault detection methods assume that the process operates in a single mode so that they cannot perform well in processes with multiple operating modes. To monitor multimode processes effectively, this paper proposes a novel process monitoring scheme based on orthogonal nonnegative matrix factorization (ONMF) and hidden Markov model (HMM). The new clustering technique ONMF is employed to separate data from different process modes. The multiple HMMs for various operating modes lead to higher modeling accuracy. The proposed approach does not presume the distribution of data in each mode because the process uncertainty and dynamics can be well interpreted through the hidden Markov estimation. The HMM-based monitoring indication named negative log likelihood probability is utilized for fault detection. In order to assess the proposed monitoring strategy, a numerical example and the Tennessee Eastman process are used. The results demonstrate that this method provides efficient fault detection performance.

A Data-driven Fault Detection Method Based on Dissipative Trajectories

Fault detection is becoming increasingly important as the complexity of industrial process develops. In this paper, a data-driven fault detection method is proposed. The dissipativity theory is adopted to find the appropriate dissipativity properties for the process input output trajectory. The dissipativity properties can be viewed as an Abstract energy property, and the dissipativity properties of input output trajectories represent process dynamic features. As faults occur, the dissipative trajectories will change thus allow fault detection to be performed based on these dissipativity properties. A training algorithm is developed to search for the related properties using input output data. A prior knowledge of the process can be incorporated into the algorithm to facilitate the training. The proposed fault detection method is illustrated on a case study of a mono-chlorobenzene plant simulated using VMGSim.

Data-driven Method in Detecting CAN Network Intermittent Connection Fault

The reliability of the network is critical to the performance and the safety of manufacturing systems. Controller area network (CAN) has been widely adopted due to its electrical robustness, low price and deterministic access delay. However, there are numerous factors that can sabotage the reliability of networks. The fault at intermittent connection (IC) of the network cable is one of the major issues. When multiple IC faults occur in the network, especially on the trunk cable, the troubleshooting for these IC faults becomes very difficult. IC problems may result in delayed or deceptive information of machines and production status which can severely impede manufacturing normal operation. To improve the network reliability and communication quality, a novel data-driven fault detection methodology is developed in this paper, which can locate multiple IC faults in the network. By analyzing IC induced network errors, the characteristics of error frames can be captured. The information extracted from the error frames is utilized to filter the network error and facilitate a two-level procedure to identify the IC faults. Case studies are performed to demonstrate the proposed method. This research provides a multiple IC faults location method which is accurate and easy to implement.

A novel process monitoring and fault detection approach based on statistics locality preserving projections

Data-driven fault detection technique has exhibited its wide applications in industrial process monitoring. However, how to extract the local and non-Gaussian features effectively is still an open problem. In this paper, statistics locality preserving projections (SLPP) is proposed to extract the local and non-Gaussian features. Firstly, statistics pattern analysis (SPA) is applied to construct process statistics and grasp the non-Gaussian statistical property using high order statistics. Then, locality preserving projections (LPP) method is used to discover local manifold structure of the statistics. In essence, LPP tries to map the close points in the original space to close in the low-dimensional space. Lastly, T2 and squared prediction error (SPE) charts of SLPP model are used to detect process faults. One simple simulated system and the Tennessee Eastman process show that the proposed SLPP method is more effective than principal component analysis, LPP and statistics principal component analysis in fault detection performance.

Data-driven fault detection and isolation scheme for a wind turbine benchmark

This paper investigates a new scheme for fault detection and isolation based on time series and data analysis. This scheme is applied in wind turbines which are used to tap the potential of renewable energy. The proposed scheme is performed in two steps and it is based on process data without using any kind of physical modeling. The first step, the fault detection, is based on an alternative method based on the Gibbs sampling algorithm in which the occurrence of a sensor fault is modeled as a change point detection in a time series. The second step, the fault isolation, is handled via a Fuzzy/Bayesian network scheme classifying the kind of fault. The proposed fault detection and isolation (FDI) strategy offers as main contribution the independence from any kind of dynamical modeling and the unprecedented usage of the Gibbs sampling. Furthermore, this work offers a novel data driven FDI approach based on Fuzzy-Bayesian inference and suitable for the wind turbines systems. This approach presents a good performance for detection and diagnostics of sensor faults in a standard wind turbine benchmark.

Fault-tolerant optimal control of a building HVAC system

This article presents the development and application of a fault-tolerant control technology, its online implementation, and results from several tests conducted for a large-sized HVAC system. By integrating model-based model predictive control and data-driven fault detection and diagnosis algorithms, the technology automatically adapts the HVAC control laws to a set of subsystem faults and can therefore reach and maintain the largest energy consumption reduction levels that are achievable at any point throughout a building lifecycle. The model predictive control algorithm generates optimal set-points that minimize energy consumption for the HVAC actuator loops while meeting equipment operational constraints and occupant thermal comfort constraints. The fault detection and diagnosis algorithm uses probabilistic graphical models to detect and classify in real time potential faults of the HVAC actuators based on data from multiple sensors. The fault-tolerant control system is realized by executing the two algorithms on the same platform, within the same framework, and by using the fault detection and diagnosis algorithm's output to continuously update the model predictive control algorithm constraints. The proposed integrated technology is executed at the supervisory level in a hierarchical control architecture as an extension of a baseline building management system. The performance and limitations of the fault detection and diagnosis, model predictive control, and fault-tolerant control algorithms are illustrated and discussed using measurement data recorded from multiple tests.

An Improved Detection Statistic for Monitoring the Nonstationary and Nonlinear Processes

The objective of this paper is to address a data-driven fault detection design for the nonstationary and nonlinear processes. Firstly, an improved statistic is proposed for fault detection, which fits the data using the design functions. The fitted parameters are then used for computing the trend of the fault-free data, based on which the prediction residual is generated and the improved statistic is constructed. This method can cope with the limitations of the standard Hotelling statistic in the sense of adaptation and condition number. Secondly, based on the formula of the inverse of the calibration covariance matrix, an incremental and decremental algorithm is proposed for updating the improved statistic. Compared with the brute force algorithm, it can reduce the computational complexity significantly, which benefits the online detection. The effectiveness of the improved statistic is validated by a nonstationary and nonlinear numerical case. Also it is used for monitoring the satellite attitude control system. The results show that the improved statistic, compared with the standard Hotelling statistic, is more sensitive to the additive fault.

Data driven fault detection and isolation: a wind turbine scenario

Uno de los mayores inconvenientes presentes en la generación de energía eólica son los altos costos de mantenimiento asociados a fallas mecánicas. Este problema se hace más evidente en las turbinas de viento de escala industrial, en donde incrementos en el tamaño y la capacidad nominal traen consigo problemas adicionales asociados a vibraciones estructurales y efectos aeroelásticos en las hojas. Debido al incremento en la capacidad de operación, es imprescindible detectar de manera eficiente fallas y degradaciones en el sistema, garantizando la integridad, su fiabilidad y reduciendo los costos de operación. Este artículo presenta un sistema para la detección y aislamiento de fallas (FDI), basado en técnicas "Guiadas por los datos " (Data driven, abreviado DD). La arquitectura propuesta es una estrategia de varios niveles en donde: (i) el primer nivel detecta la ocurrencia de una falla (detección), mientras que (ii) el segundo identifica su origen (aislamiento). Se estudiaron y compararon cuatro técnicas de clasificación para cada uno de los niveles (Máquinas de Vectores de Soporte, Redes Neuronales Artificiales, K vecinos cercanos y Mezcla de Gaussianas). La mejor estrategia en cada nivel fue seleccionada para construir el sistema FDI. El rendimiento del sistema propuesto se evalúa en un modelo de referencia de una turbina eólica de escala comercial.

Data-Driven Multimode Fault Detection for Wind Energy Conversion Systems

In this paper a data-driven fault detection scheme for wind energy conversion system is proposed. The method uses the offline measurements of the wind turbine in a wide range of operating points and builds a monitoring system which is able to detect faults with higher detection rate compared to the classical data-driven techniques. The nonlinear characteristics of wind turbine are approximated by multiple piece-wise linear systems and hence the monitoring scheme is robust against model uncertainties which may arise due to nonlinearity of the system. Moreover, the proposed method is able to discriminate the faults which are affecting the power production in the wind energy system. The effectiveness of the monitoring scheme is demonstrated using the data collected from different 2MW wind turbines. The superior performance of the proposed method is further compared with the classical data-driven methods and the results are discussed.

Multimode Process Fault Detection Using Local Neighborhood Similarity Analysis

Traditional data driven fault detection methods assume unimodal distribution of process data so that they often perform not well in chemical process with multiple operating modes. In order to monitor the multimode chemical process effectively, this paper presents a novel fault detection method based on local neighborhood similarity analysis (LNSA). In the proposed method, prior process knowledge is not required and only the multimode normal operation data are used to construct a reference dataset. For online monitoring of process state, LNSA applies moving window technique to obtain a current snapshot data window. Then neighborhood searching technique is used to acquire the corresponding local neighborhood data window from the reference dataset. Similarity analysis between snapshot and neighborhood data windows is performed, which includes the calculation of principal component analysis (PCA) similarity factor and distance similarity factor. The PCA similarity factor is to capture the change of data direction while the distance similarity factor is used for monitoring the shift of data center position. Based on these similarity factors, two monitoring statistics are built for multimode process fault detection. Finally a simulated continuous stirred tank system is used to demonstrate the effectiveness of the proposed method. The simulation results show that LNSA can detect multimode process changes effectively and performs better than traditional fault detection methods.

Data-driven fault detection with process topology for fault identification

In this paper a fault diagnosis framework based on detection with feature extraction methods and identification based on data-driven process topology methods was investigated. A simulation of a simple system consisting of two tanks with heat exchangers was used to generate data for normal operating conditions and a number of faults. Fault detection methods included principal component analysis and kernel principal component analysis feature extraction with Shewhart, cumulative sum and exponentially weighted moving average monitoring charts. Process topology information was extracted with linear cross-correlation, partial cross-correlation and transfer entropy. Connectivity maps were constructed to identify possible fault propagation paths to aid root cause analysis and changes in connectivity structure due to faults were exploited for fault identification. Kernel principal component analysis with a CUSUM chart gave the best detection performance, while connectivity graphs based on partial correlation gave an accurate representation of the system and assisted fault identification.

Data-driven modelling, control, and fault detection of wind turbine systems

In this paper, data–driven system modelling, control, and fault detection technique for wind turbine systems are researched. The developed algorithm is to recursively update system parameters using predictor–based system identification (SID) technique. Updated system parameters are used to design subspace predictive controller for wind turbine systems with constraint on pitch angle and generator torque. The usefulness of this application is highlighted through simulation on a benchmark example, 5 MW NREL/Upwind reference turbine. It shows that developed algorithm which streamlines controller design process from identification to fault detection is useful for making wind turbine systems being tolerant to faults.

Process-Quality Monitoring Using Semi-supervised Probability Latent Variable Regression Models

In order to sustain safety and high product quality, the data-driven fault detection tools are increasingly used in industrial processes. The quality variables are the key index of the final product. Obtaining them in high frequency is time-consuming in the laboratory because they require the efforts of experienced operators. Meanwhile, process variables such as the temperature, the pressure, and the flow rate can be readily sampled in high frequency; hence the sample size between the process and the quality data is quite unequal. To effectively integrate two different observation sources, the high-rate process measurements and low-rate quality measurements, a semi-supervised regression model with probabilistic latent variables is proposed in this article to enhance the performance monitoring of the variations of the process and the quality variables. The corresponding statistics are also systematically developed and a TE benchmark problem is presented to illustrate the effectiveness of the proposed method.

An Improved Detection Statistic for Systems with Unsteady Trend

The object of this paper is to address data-driven fault detection design for systems with unsteady trend, which shows cyclicity, monotonicity and non-zero mean. Firstly, mean theorem and covariance theorem are proposed and proved. The former is the mean property of projection matrix, and the latter is the recursive formula for covariance matrix of regression residual. Secondly, an improved fault detection statistic, called Least Square T2 (LST2), is proposed. It can partly solve the detection problem for systems with unsteady trend. The improvement can also partly cope with the limitations of the traditional multivariate detection methods, such as Principal Component Analysis (PCA). Thirdly, based on the two theorems, a recursive algorithm and a moving window algorithm of LST2 are given, thus both time and space complexity are greatly reduced for online detection. The effectiveness of the presented detection statistic is evaluated with an application of monitoring satellite attitude control system. The case study result shows that the false alarm rate of LST2 is much lower than that of T2 based on PCA, while LST2 is more sensitive to fault.

Data-driven fault detection filter design for time-delay systems

In this paper, robust fault detection (FD) problems for time-delay LTI systems with unknown inputs are studied. This paper is proposed to evaluate the robustness as well as sensitivity of residual signals to the unknown inputs as well as to the faults in terms of L2. First, the weighting matrix is selected for an appropriate design of filter, then fault detection filter design with Lyapunov-Krasovskii function (LKF) is designed with time delay. The main results include the detailed derivation of these steps followed by its implementation on an open-loop time-delay system for chemical reactor example.