Fault Detection And Diagnosis Problem Research Articles

Semi-supervised learning for industrial fault detection and diagnosis: A systemic review

The automation of Fault Detection and Diagnosis (FDD) is a central task for many industries today. A myriad of methods are in use, although the most recent leading contenders are data-driven approaches and especially Machine Learning (ML) methods. ML algorithms fall into two main categories: supervised and unsupervised methods, depending on whether or not the instances are labeled with the expected outputs. However, a new approach called Semi-Supervised Learning (SSL) has recently emerged that uses a few labeled instances together with other unlabeled instances for the training process. This new approach can significantly improve the accuracy of conventional ML models for industrial environments where labeled data are scarce. SSL has been tested as a promising solution over the past few years for several FDD problems, although there have been no systemic reviews of this sort of approach up until the present review. In this study, an attempt to organize the existing literature on SSL for FDD using the taxonomy of van Engelen & Hoos is reported. The most and the least frequently used SSL algorithms are identified and considered in terms of different fault detection tasks and their most common dataset structure. Moreover, a set of best practices are proposed in the conclusions of this work for implementation under real industrial conditions, so as to avoid some of the most common faults.

Fault diagnosis using Interpolated Kernel Density Estimate

Fault detection and diagnosis (FDD) methods often face the challenge that there is no prior knowledge about the distribution of the monitored variables and features. Data-driven methods based on statistics often incorporate a simplifying assumption that the data can be described by a Gaussian distribution. Such a simplification might cause deterioration in the accuracy of fault detection. Kernel Density Estimation is a powerful non-parametric method. Its main advantage lies in the fact that no prior assumption is necessary about the distribution of the data, and that it is able to handle multimodal distributions in an effective way. However, it is very computationally expensive.In this paper, Interpolated Kernel Density Estimate (IKDE) is proposed to reduce the computational cost and improve processing times. IKDE is based on interpolated KDE functions using Chebyshev polynomials and the barycentric formula to obtain the posterior probabilities. The method is used with a Naive Bayes classifier for fault detection for a non-Gaussian distributed, multimodal induction motor dataset. The performance and computational cost compared with the standard KDE implementation show the superiority of IKDE, making it suitable for online FDD problems where reduced computational costs are highly beneficial. The main advantage of this approach is the reduction in the stored data as well as the reduced number of multiplications needed for a single evaluation.

Data augmentation for bearing fault detection with a light weight CNN

Bearings are vital part of rotary machines. A failure of bearing has a negative impact on schedules, production operation and even human casualties. Therefore, in prior achieving fault detection and diagnosis (FDD) of bearing is ensuring the safety and reliable operation of rotating machinery systems. However, there are some challenges of the industrial FDD problems. First, there are unbalanced samples because industrial faults rarely occur. Conse-quently, the labeled data which can refer to failure information are limited in the industry and data augmentation methods are critical pre-processing be-fore training data driven models. Second, due to many learnable parameters in model and data of long sequence, both lead to time delay for FDD. There-fore, this paper proposes various data preprocessing methods and Light-Convolutional Neural Network (LCNN).

Handling Incomplete Sensor Measurements in Fault Detection and Diagnosis for Building HVAC Systems

Due to the development of sensor networks and information technology, data-driven fault detection and diagnosis (FDD) has been made possible with real-time multiple sensor measurements. However, due to inevitable sensor errors or communication failures, the raw data are usually incomplete with corrupted values, lost values, or undetected missing values. In practice, the incomplete data are usually dealt with by directly excluding incomplete measurements and abnormal spikes. In addition, some preprocessing methods, which naively impute data though averaging or smoothing, have also been widely applied. In this article, we address the building FDD problem with incomplete data by proposing a new approach, the adjacent information recovery (AIR) filter. The AIR filter is utilized to deal with the FDD for a typical air handling unit (AHU) system with incomplete data based on the ASHRAE Research Project 1312. Experimental results show that the proposed method improves FDD performance by recovering missing sensor measurements and outperforms the state-of-the-art methods. Note to Practitioners —Fault detection and diagnosis (FDD) for smart buildings by addressing the fact that FDD systems are of great importance for saving energy and improving occupancy comfort levels and building safety levels. Existing FDD methods are mainly based on the assumption that sensor data are complete and reliable, which are rarely true in real practice. To solve the building FDD problem with incomplete data, in this article, the adjacent information recovery (AIR) filter is proposed to recover the missing data before applying FDD methods. The AIR filter takes the time series adjacency information into consideration via hidden Markov model (HMM) and includes the channel adjacency information with the collaborating filtering technique.

Fault detection and diagnosis for delay-range-dependent stochastic systems using output PDFs

This paper investigates a new fault detection and diagnosis(FDD) scheme for delay-range-dependent stochastic systems. Compared with classical FDD problem, the measurable information in this paper is supposed to be the output probability density function(PDF), rather than the output itself. By using the square root B-spline approximation technique, the dynamic weight model of the output PDFs is established and the considered problem is converted into a nonlinear FDD problem for stochastic systems with delays. The main objective of this paper is to construct a filter based residual generator such that the fault can be detected and estimated. The FDD criteria is provided on the basis of linear matrix inequalities(LMIs). Besides, to improve the FDD performance, the tuning parameters, slack variables as well as the free-weighting matrices are applied to optimize the FDD criteria. Finally, the simulations are given to demonstrate the effectiveness of the proposed method.

Fault detection and diagnosis of chillers using Bayesian network merged distance rejection and multi-source non-sensor information

Applying the fault detection and diagnosis (FDD) techniques to chillers is beneficial to reduce building energy consumption and to enhance the energy efficiency of buildings. The purpose of this study is to propose a chiller FDD method with better performance for field implementation. The technological paths are as follows: (i) in order to identify new types of faults and to update the FDD fault libraries, a distance rejection (DR) technique is merged into the Bayesian network (BN) by transforming the chiller FDD problem into a single-class classification problem. Furthermore, the DR can be tuned to obtain an adjustable false alarm rate (FAR); (ii) to increase the diagnostic accuracies of known (or existing) faults and the identification accuracies of new types of faults, multi-source non-sensor information (MI) is merged into the BN, i.e., maintenance records and repair service history, healthy states of related equipment and on-site observed information. A novel chiller FDD method based on BN merged DR and MI (DR-MI-BN) is proposed in this study. The performance of this proposed method is evaluated by using the experimental data from ASHRAE RP-1043. Test results show that the FAR can be tuned for different users’ requirements, and that merging the MI significantly improves the diagnostic accuracies of known faults from 77.2% to 99.8% at most (for refrigerant leakage) and the identification accuracies of new types of faults from 56.6% to 99.6% at most (for NF7).

Model Based Fault Diagnosis of Low Earth Orbiting (LEO) Satellite using Spherical Unscented Kalman Filter

Abstract Model based fault detection and diagnosis (FDD) using a non-linear estimation technique is presented here. The non-linear estimation technique namely spherical Unscented Kalman Filter (UKF) has been applied to other kinds of estimation problems but has never been applied to the FDD problem of a Low Earth Orbiting (LEO) satellite. It has been shown in this work that compared to the standard UKF, which is a derivative free estimation technique unlike the popular Extended Kalman Filter (EKF), the spherical UKF can perform better in terms of computational savings without sacrificing accuracy. Hence it is better suited for real-time fault diagnosis. A planar model of the satellite is used to demonstrate the technique.

A robust pattern recognition-based fault detection and diagnosis (FDD) method for chillers

A new chiller fault detection and diagnosis (FDD) method is proposed in this article. Different from conventional chiller FDD methods, this article considers the FDD problem as a typical one-class classification problem. The fault-free data are classified as the fault-free class. Data of a fault type are regarded as a fault class. The task of fault detection is to detect whether the process data are outliers of the fault-free class. The task of fault diagnosis is to find to which fault class does the process data belong. In this study, support vector data description (SVDD) algorithm is introduced for the one-class classification. The basic idea of the SVDD-based method is to find a minimum-volume hypersphere in a high dimensional feature space to enclose most of the data of an individual class. The proposed method is validated using the ASHRAE RP-1043 (Comstock and Braun 1999) experimental data. It shows more powerful FDD capacity than multi-class SVM-based FDD methods and PCA-based fault detection methods. Four potential applications of the proposed method are also discussed.

Aircraft Nonlinear AFTC based on Geometric Approach and Singular Perturbations in case of actuator and sensor faults

The Singular Perturbation (SP) theory seems to be a very advantageous framework to describe the longitudinal aircraft dynamic and for designing advanced Flight Controls. In this paper, thanks to SP theory, a novel Active Fault Tolerant Flight Control (AFTFC) has been developed in order to accommodate faults occurring on all actuators and sensors. The proposed AFTFC is composed by two main subsystems: a Fault Detection and Diagnosis (FDD) module, that provides fault estimation, and a Controller that, exploiting this estimation, tracks a reference trajectory, even in presence of fault. The jointly use of the NonLinear Geometric Approach and SP allows the solution of the FDD problem, otherwise not possible for the considered fault scenario. The effectiveness of the proposed AFTFC system is highlighted by extended simulations exploiting a detailed wide–body aircraft flight simulator.

Fault detection and diagnosis for non-Gaussian stochastic distribution systems with time delays via RBF neural networks

A new fault detection and diagnosis (FDD) problem via the output probability density functions (PDFs) for non-gausian stochastic distribution systems (SDSs) is investigated. The PDFs can be approximated by radial basis functions (RBFs) neural networks. Different from conventional FDD problems, the measured information for FDD is the output stochastic distributions and the stochastic variables involved are not confined to Gaussian ones. A (RBFs) neural network technique is proposed so that the output PDFs can be formulated in terms of the dynamic weighings of the RBFs neural network. In this work, a nonlinear adaptive observer-based fault detection and diagnosis algorithm is presented by introducing the tuning parameter so that the residual is as sensitive as possible to the fault. Stability and Convergency analysis is performed in fault detection and fault diagnosis analysis for the error dynamic system. At last, an illustrated example is given to demonstrate the efficiency of the proposed algorithm, and satisfactory results have been obtained.

Delay-dependent fault detection and diagnosis using B-spline neural networks and nonlinear filters for time-delay stochastic systems

A new fault detection and diagnosis (FDD) scheme is studied in this paper for the continuous-time stochastic dynamic systems with time delays, where the available information for the FDD is the input and the measured output probability density functions (PDFs) of the system. The square-root B-spline neural networks is used to formulate the output PDFs with the dynamic weightings. As a result, the concerned FDD problem can be transformed into a robust FDD problem subjected to a continuous time uncertain nonlinear system with time delays. Delay-dependent criteria to detect and diagnose the system fault are provided by using linear matrix inequality (LMI) techniques. It is shown that this new criterion can provide higher sensitivity performance than the existing result. Simulations are given to demonstrate the efficiency of the proposed approach.

Observer-Based Optimal Fault Detection and Diagnosis Using Conditional Probability Distributions

A new optimal fault detection and diagnosis (FDD) scheme is studied in this paper for the continuous-time stochastic dynamic systems with time delays, where the available information for the FDD is the input and the measured output probability density functions (pdf's) of the system. The square-root B-spline functional approximation technique is used to formulate the output pdf's with the dynamic weightings. As a result, the concerned FDD problem can be transformed into a robust FDD problem subjected to a continuous time uncertain nonlinear system with time delays. Feasible criteria to detect and diagnose the system fault are provided by using linear matrix inequality (LMI) techniques. In order to improve FDD performances, two optimization measures, namely guaranteed cost performance and H <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">infin</sub> performance, are applied to optimize the observer design. Simulations are given to demonstrate the efficiency of the proposed approach

Filter-based fault detection and diagnosis using output PDFs for stochastic systems with time delays

In this paper, a fault detection and diagnosis (FDD) scheme is studied for general stochastic dynamic systems subjected to state time delays. Different from the formulation of classical FDD problems, it is supposed that the measured information for the FDD is the probability density function (PDF) of the system output rather than its actual value. A B-spline expansion technique is applied so that the output PDF can be formulated in terms of the dynamic weights of the B-spline expansion, by which a time delay model can be established between the input and the weights with non-linearities and modelling errors. As a result, the concerned FDD problem can be transformed into a classic FDD problem subject to an uncertain non-linear system with time delays. Feasible criteria to detect the system fault are obtained and a fault diagnosis method is further presented to estimate the fault. Simple simulations are given to demonstrate the efficiency of the proposed approach. Copyright © 2006 John Wiley & Sons, Ltd.

A fast training neural network and its updation for incipient fault detection and diagnosis

Fast incipient fault diagnosis is becoming one of the key requirements for safe and optimal process operations. There has been considerable work done in this area with a variety of approaches being proposed for incipient fault detection and diagnosis (FDD). Incipient FDD problem is particularly difficult in the case of chemical processes as these processes are usually characterized by complex operations, high dimensionality and inherent nonlinearity. Neural networks have been shown to solve FDD problems in chemical processes as they develop inherently non-linear input—output maps and are well suited for high dimensionality problems. In this work, to enhance the neural network framework, we address the following three issues, (i) speed of training; (ii) introduction of time explicitly into the classifier design; and (iii) online updation using a mirror-like process model.