Perform Fault Detection Research Articles

Multiobjective Two-Dimensional CCA-Based Monitoring for Successive Batch Processes With Industrial Injection Molding Application

Successive batch processes generally involve within-batch and batch-to-batch correlations, and monitoring of such batch processes is imperative. This paper proposes a multiobjective two-dimensional canonical correlation analysis (M2D-CCA)-based fault detection scheme to achieve efficient monitoring of successive batch processes. First, three-way historical batch process data are unfolded into two-way time-slice data. Second, for each time-slice measurement, CCA is performed between the current measurement and previous measurements from both time and batch directions, which takes the within-batch and batch-to-batch correlations into account. To determine the involved measurements and eliminate the influence of unrelated variables, multiobjective evolutionary optimization is performed, which tries to maximize the preserved canonical correlation coefficients and minimize the number of involved variables. Finally, based on the established M2D-CCA model, an optimal fault detection residual is generated for each time-slice measurement. The M2D-CCA fault detection scheme performs fault detection using the current measurement and the information provided by its previous samples and batches, and therefore exhibits a superior monitoring performance. The M2D-CCA fault detection approach is tested on a numerical example and an industrial injection molding process. Monitoring results verify the feasibility and superiority of the proposed monitoring scheme.

State space neural networks and model-decomposition methods for fault diagnosis of complex industrial systems

Reliable and timely fault detection and isolation are necessary tasks to guarantee continuous performance in complex industrial systems, avoiding failure propagation in the system and helping to minimize downtime. Model-based diagnosis fulfils those requirements, and has the additional advantage of using reusable models. However, reusing existing complex non-linear models for diagnosis in large industrial systems is not straightforward. Most of the times, the models have been created for other purposes different from diagnosis, and many times the required analytical redundancy is small. The approach proposed in this work combines techniques from two different research communities within Artificial Intelligence: Model-based Reasoning and Neural Networks. In particular, in this work we propose to use Possible Conflicts, which is a model decomposition technique from the Artificial Intelligence community to provide the structure (equations, inputs, outputs, and state variables) of minimal models able to perform fault detection and isolation. Such structural information is then used to design a grey box model by means of state space neural networks. In this work we prove that the structure of the Minimal Evaluable Model for a Possible Conflict can be used in real-world industrial systems to guide the design of the state space model of the neural network, reducing its complexity and avoiding the process of multiple unknown parameter estimation in the first principles models. We demonstrate the feasibility of the approach in an evaporator for a beet sugar factory using real data.

Fault Detection on Big Data: A Novel Algorithm for Clustering Big Data to Detect and Diagnose Faults

With computer technology improving exponentially, data will grow incomprehensibly in size, complexity, and noise. However, latent within the data, valuable signals are hidden that, if discovered, can offer abundant information, such as fault detection. Traditionally, principal component analysis has been used to perform fault detection in large, multivariate systems. However, these methods often struggle to find the true origin, as they are susceptible to contribution smearing. In this work, a chemical plant system was analyzed and a novel cluster and detect method for fault detection utilizing machine-learning clustering algorithms was created in aim to improve fault detection time and diagnosis. Plant data containing complex variables were simulated, clustered into groups through a unique algorithm based upon correlations, and analyzed through principal component analysis as individual groups. This approach often resulted in quicker identification and more accurate diagnosis than the traditional principal component analysis method.

A General Modeling Technique for a Triple Redundant 3 × 3-Phase PMA SynRM

A general modeling technique is proposed for a triple redundant 3 × 3-phase permanent-magnet-assisted synchronous reluctance machine (PMA SynRM). The magnetomotive force (MMF) of the machine is divided into three parts, each associated with one three-phase set. The MMF of each three-phase set can be described by four variables: d - and q -axis components of the currents, the rotor angle, and an MMF offset component that captures the mutual coupling between three three-phase sets. Therefore, the complete machine behavior in all operating conditions can be captured by means of 4-D tables, which store the flux linkage and torque information. The 4-D tables are produced by finite-element (FE) analysis for one three-phase set. As a result, the machine behavior can be predicted by interpolating the 4-D tables. The model is capable of representing healthy and fault operations, including unequal current operation in three three-phase sets, and offers great flexibility for performance assessment, postfault control, and fault detection. Its effectiveness is verified by extensive FE simulation and experimental tests in different operation modes.

Shipboard Fault Detection Through Nonintrusive Load Monitoring: A Case Study

As crew sizes aboard maritime vessels shrink in efforts to reduce operational costs, ship operators increasingly rely on advanced monitoring systems to ensure proper operation of shipboard equipment. The nonintrusive load monitor (NILM) is an inexpensive, robust, and easy to install system useful for this task. NILMs measure power data at centralized locations in ship electric grids and disaggregate power draws of individual electric loads. This data contains information related to the health of shipboard equipment. We present a NILM-based framework for performing fault detection and isolation, with a particular emphasis on systems employing closed-loop hysteresis control. Such controllers can mask component faults, eventually leading to damaging system failure. The NILM system uses a neural network for load disaggregation and calculates operational metrics related to machinery health. We demonstrate the framework’s effectiveness using data collected from two NILMs installed aboard a U.S. Coast Guard cutter. The NILMs accurately disaggregate loads, and the diagnostic metrics provide easy distinction of several faults in the gray water disposal system. Early detection of such faults prevents costly wear and avoids catastrophic failures.

Fault Detection and Isolation of the Brake Cylinder System for Electric Multiple Units

Air brake systems are crucial systems for safe and stable operation of electric multiple units (EMUs). The brake cylinder system, which includes brake cylinders, corresponding pressure sensors, and connection pipes, plays a vital role in the EMU air brake system. This is because brake cylinder pressures directly affect the brake operation. Currently, brake cylinder pressures are monitored by univariate control charts, i.e., the brake cylinder system will be considered as faulty if a certain pressure goes beyond its allowed range. Besides, serious sensor hardware faults such as open circuit or short circuit can also be detected by system self-inspection circuits. However, three kinds of faults, including brake cylinder component fault, soft sensor fault, as well as gas leakage fault, cannot be well handled by current monitoring methods if these faults are not very serious. In this paper, a fault detection index called intervariable variance (IVV) is first presented to perform fault detection for these faults. Fault detectability analysis is provided, and the IVV statistic is also compared with the univariate control chart approach. Then, a fault isolation strategy is proposed to distinguish different kinds of faults and determine the location of the occurred fault. Finally, the effectiveness of the proposed fault detection and isolation method is demonstrated via extensive experimental studies that are carried out on the EMU brake test bench of Qingdao Sifang Rolling Stock Research Institute Co., Ltd., China.

A M-EKF fault detection strategy of insulation system for marine current turbine

Abstract Marine current power is a new renewable energy source, it is significant to keep the marine current turbine healthy. But harsh marine conditions, such as regular thermal cycling, high humidity and heavy salt mist, bring severe challenges to the marine current turbine, especially to the winding insulation system. Therefore, it is of great importance to perform fault detection on the insulation system. In this paper, a Modified Extended Kalman Filter (M-EKF) fault detection strategy based on the electrical parametric model derived from the RLC network modeling of Roebel bars is proposed. The proposed fault detection strategy mainly includes two parts. The first step is to establish reasonable state space equations for the model structure extracted to monitor the state of the insulation system. In the second step, the conventional continuous EKF is modified to follow more accurately the state of the marine current turbine. Simulation and experiment results show that the proposed method can realize timely monitoring of the winding insulation system of marine current turbine with good performance.

Mixed fault diagnosis scheme for satellite formation

PurposeThe purpose of this paper is to present a solution for the uncertain fault with the propulsion subsystem of satellite formation, using the Lur’e differential inclusion linear state observers (DILSOs) and fuzzy wavelet neural network (FWNN) to perform fault detection and diagnosis.Design/methodology/approachThe uncertain fault system cannot be described based on the accurate differential equations. The set-value mapping is introduced into the state equations to solve the problem of uncertainty, but it will cause output uncertainty. The problem can be solved by linearization of Lur’e differential inclusion state observers. The Lur’e DILSOs can be used to detect uncertain fault. The fault isolation and estimation can be performed using the FWNN.FindingsThe mixed approach from fault detection and diagnosis has featured fast and correct to found the uncertain fault. The simulation results to indicate that the methods of design are not only effective but also have the advantages of good approximation effect, fast detection speed, relatively simple structure and prior knowledge and realization of adaptive learning.Research limitations/implicationsThe hybrid algorithm can be extensively applied to engineering practice and find uncertain faults of the propulsion subsystem of satellite formation promptly.Originality/valueThis paper provides a fast, effective and simple mixed fault detection and diagnosis scheme for satellite formation.

An Observer-Based Fault-Tolerant Controller for Flexible Buildings-Based on Linear Matrix Inequality Approach

Since failures in sensors will degrade the performance of Active Mass Damper (AMD) control systems, a dynamic filter design method, a state observer design method, and a robust control strategy are developed and presented in this paper to overcome this deficiency. The filter design method will be transformed into a H2/H∞ control problem that can be solved by Linear Matrix Inequality (LMI) approach. Thus, it can be used to perform fault detection and isolation (FDI) for the control systems. And, the state observer design method uses the acceleration responses as the feedback signal. The detected and isolated fault signals in accelerometers are used to estimate the whole states that are used to calculate the control force though a robust control strategy based on regional pole-assignment algorithm. Then, the active fault-tolerant control (FTC) will be accomplished. To verify its effectiveness, the proposed methodology is applied to a numerical example of a ten-storey frame and an experiment of a single span four-storey steel frame. Both numerical and experimental results demonstrate that the performances of FTC controller and the control system will be improved by the designed dynamic FDI filter and that it can effectively detect and isolate fault signal.

Canonical variable analysis and long short-term memory for fault diagnosis and performance estimation of a centrifugal compressor

Centrifugal compressors are widely used for gas lift, re-injection and transport in the oil and gas industry. Critical compressors that compress flammable gases and operate at high speeds are prioritized on maintenance lists to minimize safety risks and operational downtime hazards. Identifying incipient faults and predicting fault evolution for centrifugal compressors could improve plant safety and efficiency and reduce maintenance and operation costs. This study proposes a dynamic process monitoring method based on canonical variable analysis (CVA) and long short-term memory (LSTM). CVA was used to perform fault detection and identification based on the abnormalities in the canonical state and the residual space. In addition, CVA combined with LSTM was used to estimate the behavior of a system after the occurrence of a fault using data captured from the early stages of deterioration. The approach was evaluated using process data obtained from an operational industrial centrifugal compressor. The results show that the proposed method can effectively detect process abnormalities and perform multi-step-ahead prediction of the system’s behavior after the appearance of a fault.

Sensor Fault Diagnosis Observer for an Electric Vehicle Modeled as a Takagi-Sugeno System

A sensor fault diagnosis of an electric vehicle (EV) modeled as a Takagi-Sugeno (TS) system is proposed. The proposed TS model considers the nonlinearity of the longitudinal velocity of the vehicle and parametric variation induced by the slope of the road; these considerations allow to obtain a mathematical model that represents the vehicle for a wide range of speeds and different terrain conditions. First, a virtual sensor represented by a TS state observer is developed. Sufficient conditions are given by a set of linear matrix inequalities (LMIs) that guarantee asymptotic convergence of the TS observer. Second, the work is extended to perform fault detection and isolation based on a generalized observer scheme (GOS). Numerical simulations are presented to show the performance and applicability of the proposed method.

Fault detection and classification using artificial neural networks

Process monitoring is considered to be one of the most important problems in process systems engineering, which can be benefited significantly from deep learning techniques. In this paper, deep neural networks are applied to the problem of fault detection and classification to illustrate their capability. First, the fault detection and classification problems are formulated as neural network based classification problems. Then, neural networks are trained to perform fault detection, and the effects of two hyperparameters (number of hidden layers and number of neurons in the last hidden layer) and data augmentation on the performance of neural networks are examined. Fault classification problem is also tackled using neural networks with data augmentation. Finally, the results obtained from deep neural networks are compared with other data-driven methods to illustrate the advantages of deep neural networks.

Sequential fault detection for sealed deep groove ball bearings of in-wheel motor in variable operating conditions

Sealed deep groove ball bearings (SDGBBs) are employed to perform the relevant duties of in-wheel motor. However, the unique construction and complex operating environment of in-wheel motor may aggravate the occurrence of SDGBB faults. Therefore, this study presents a new intelligent diagnosis method for detecting SDGBB faults of in-wheel motor. The method is constructed on the basis of optimal composition of symptom parameters (SPOC) and support vector machines (SVMs). SPOC, as the objects of a follow-on process, is proposed to obtain from symptom parameters (SPs) of multi-direction. Moreover, the optimal hyper-plane of two states is automatically obtained using soft margin SVM and SPOC, and then using multi-SVMs, the system of intelligent diagnosis is built to detect many faults and identify fault types. The experiment results confirmed that the proposed method can excellently perform fault detection and fault-type identification for the SDGBB of in-wheel motor in variable operating conditions.

Adaptive soft sensors for quality prediction under the framework of Bayesian network

Soft sensor is widely used to predict quality-relevant variables which are usually hard to measure timely. Due to model degradation, it is necessary to construct an adaptive model to follow changes of the process. Adaptive models—moving windows (MW), time difference (TD), and locally weighted regression (LWR) under the framework of Bayesian network (BN) are proposed in this paper. BN shows great superiorities over other traditional methods, especially in dealing with missing data and the ability of learning causality. Furthermore, the acquisition of variances in BN makes it possible to perform fault detection, on the basis of 3-sigma criterion. A debutanizer column and CO2 absorption column are provided as two industrial examples to validate the effectiveness of our proposed techniques. In a debutanizer column, RMSE of MW-BN is decreased by 40% in comparison to MW-PLS. In a CO2 absorption column, the largest absolute prediction error of TD-BN is reduced by approximate 7% when compared with that of TD-PLS. Furthermore, about 38% improvements of prediction precision can be achieved in LW-BN in contrast to LW-PLS.

Kernel dynamic latent variable model for process monitoring with application to hot strip mill process

Dynamic models are preferred rather than static models in the process monitoring of modern manufacturing. Compared with static models, dynamic models can reflect not only correlations but also causality among measurements and manipulated variables. Linear dynamic models are very common due to the simplicity of representation and parameter estimation. However, because of natural nonlinearity of a dynamic process, it is ineffective to apply linear models within a long term and varying condition. Nonlinear dynamic models are hence desired under such a circumstance. In this paper, a kernel dynamic latent variable (KDLV) model is proposed to describe the nonlinearity between original measurements and dynamic latent variables. This model is an extension of dynamic latent variable model in the aspect of nonlinearity, and keeps all merits of it. In order to build such a model, a KDLV search algorithm is proposed to acquire key model parameters from data, then a KDLV modeling procedure is derived to complete the whole model. After the KDLV model is trained from data, corresponding detection strategy is also developed to perform fault detection. The KDLV based fault detection is applied to the monitoring of hot strip mill process and comparison study is also conducted on both DLV and DKPCA models.

Performance monitoring of non-gaussian chemical processes with modes-switching using globality-locality preserving projection

In this paper, we propose a novel performance monitoring and fault detection method, which is based on modified structure analysis and globality and locality preserving (MSAGL) projection, for non-Gaussian processes with multiple operation conditions. By using locality preserving projection to analyze the embedding geometrical manifold and extracting the non-Gaussian features by independent component analysis, MSAGL preserves both the global and local structures of the data simultaneously. Furthermore, the tradeoff parameter of MSAGL is tuned adaptively in order to find the projection direction optimal for revealing the hidden structural information. The validity and effectiveness of this approach are illustrated by applying the proposed technique to the Tennessee Eastman process simulation under multiple operation conditions. The results demonstrate the advantages of the proposed method over conventional eigendecomposition-based monitoring methods.

Multivariable Adaptive Artificial Pancreas System in Type 1 Diabetes.

The review summarizes the current state of the artificial pancreas (AP) systems and introduces various new modules that should be included in future AP systems. A fully automated AP must be able to detect and mitigate the effects of meals, exercise, stress and sleep on blood glucose concentrations. This can only be achieved by using a multivariable approach that leverages information from wearable devices that provide real-time streaming data about various physiological variables that indicate imminent changes in blood glucose concentrations caused by meals, exercise, stress and sleep. The development of a fully automated AP will necessitate the design of multivariable and adaptive systems that use information from wearable devices in addition to glucose sensors and modify the models used in their model-predictive alarm and control systems to adapt to the changes in the metabolic state of the user. These AP systems will also integrate modules for controller performance assessment, fault detection and diagnosis, machine learning and classification to interpret various signals and achieve fault-tolerant control. Advances in wearable devices, computational power, and safe and secure communications are enabling the development of fully automated multivariable AP systems.

Slow feature analysis based on online feature reordering and feature selection for dynamic chemical process monitoring

This study considers the insufficiency of traditional monitoring methods to eliminate dynamics, and proposes a novel online feature reordering- and feature selection-based slow feature analysis (SFA) algorithm. The SFA algorithm explores the process dynamics from the view of inner variation of data to extract the slowly varying features. The extracted SFs are considered as the representations of steady- and dynamic-state processes. Online feature reordering and feature selection strategies maximize online fault information and can be used to perform fault detection operation. The proposed method is applied to two simulated processes. Monitoring results show that the proposed method has better monitoring results than those of traditional methods.

Distributed Diagnosis of Actuator and Sensor Faults in HVAC Systems

This paper presents a model-based methodology for diagnosing actuator and sensor faults affecting the temperature dynamics of a multi-zone heating, ventilating and air-conditioning (HVAC) system. By considering the temperature dynamics of the HVAC system as a network of interconnected subsystems, a distributed fault diagnosis architecture is proposed. For every subsystem, we design a monitoring agent that combines local and transmitted information from its neighboring agents in order to provide a decision on the type, number and location of the faults. The diagnosis process of each agent is realized in three steps. Firstly, the agent performs fault detection using a distributed nonlinear estimator. After the detection, the local fault identification is activated to infer the type of the fault using two distributed adaptive estimation schemes and a combinatorial decision logic. In order to distinguish between multiple local faults and propagated sensor faults, a distributed fault isolation is applied using the decisions of the neighboring agents. Simulation results of a 5-zone HVAC system are used to illustrate the effectiveness of the proposed methodology.

High Performance and Reliable Fault Detection Scheme for the Secure Hash Algorithm

Background/Objectives: Among the powerful techniques against the protected hash functions (SHA) the method of fault injection attacks. For obtain the confidential information, the method consist to inject this attacks during the process of the hash algorithm. In literature, they proposed a many methods of countermeasures to secure the SHA implementation against these attacks. Methods/Statistical Analysis: In this paper, we proposed a new scheme of fault detection; this scheme is based on the combines of two redundancies named hybrid redundancy for the hash algorithm. The weaknesses and the strengths of our proposed method against the fault attacks are discussed. Findings: our schemes proposed reaches 99.999% fault coverage. Moreover, we are implemented our proposed scheme on Xilinx Virtex-5 and Virtex-II Pro FPGA. Its area overhead fault coverage, frequency and throughput degradation have been compared and it is shown that our proposed scheme allows a trade-off between the security of the SHA and hardware overhead. Application/Improvements: Compared to other work, our fault detection scheme has the important performances in terms of frequency, occupied slices, fault coverage and throughput.