A Maximum Entropy Based Approach to Fault Diagnosis Using Discrete and Continuous Features (original) (raw)
Related papers
Fault Detection and Diagnosis for Gas Turbines Based on a Kernelized Information Entropy Model
Gas turbines are considered as one kind of the most important devices in power engineering and have been widely used in power generation, airplanes, and naval ships and also in oil drilling platforms. However, they are monitored without man on duty in the most cases. It is highly desirable to develop techniques and systems to remotely monitor their conditions and analyze their faults. In this work, we introduce a remote system for online condition monitoring and fault diagnosis of gas turbine on offshore oil well drilling platforms based on a kernelized information entropy model. Shannon information entropy is generalized for measuring the uniformity of exhaust temperatures, which reflect the overall states of the gas paths of gas turbine. In addition, we also extend the entropy to compute the information quantity of features in kernel spaces, which help to select the informative features for a certain recognition task. Finally, we introduce the information entropy based decision tree algorithm to extract rules from fault samples. The experiments on some real-world data show the effectiveness of the proposed algorithms.
Special issue: Data-driven fault diagnosis of industrial systems
Information Sciences, 2014
Special issue: Data-driven fault diagnosis of industrial systems Fault diagnosis (FD) for engineering systems and components has been attracting considerable attention from researchers and engineers in many engineering areas. With the rapid development of Computational Intelligence techniques such as neural networks, fuzzy logic and evolutionary computation, the studies and applications of model-free FD systems have been favourably advanced in the past decades. Such intelligent FD systems have thoroughly detailed the signals and systems methods to describe the dynamics of failure mechanisms in materials, structures, and rotating equipment or even a complex computer program. In the past decades, although model-free FD techniques have been broadly explored in practice, there are still many challenging issues to be further explored in the design of diagnosis systems and algorithms, performance improvement, fault predictability, robust diagnosis of uncertain systems, and framework development for engineering systems. Some new challenges in designing FD systems are also emerging. Some of them are associated with data mining applications, for instance, rule generation and adaptation for rule-based FD expert systems while the discovery of new faults is realized through clustering algorithms and association analysis. This special issue aims to promote the recent advances of data-driven FD systems and algorithms, and address some challenges in designing and developing the state-of-the-art FD systems and algorithms. As a specific application of computing techniques in industrial systems, this special issue has received considerable attention and support from academics, researchers, and engineers. After extensive reviews and revisions, fourteen papers have been accepted. These papers cover a wide variety of applications of fault diagnosis techniques in process industries, for instance, aircraft jet engine and wind turbine diagnosis, case-base reasoning approach for status prediction of a shaft furnace, and multi-sensor data fusion techniques for monitoring system design. It is believed that all papers are of high quality and significance, and such a collection will be useful and valuable to both researchers and engineers. Below, we provide a summary of the papers published in this special issue. The paper entitled ''Fault detection and isolation of a dual spool gas turbine engine using dynamic neural networks and multiple model approach'', by Z.N. Sadough Vaninia, K. Khorasania, and N. Meskinb, presents a fault detection and isolation scheme for an aircraft jet engine, where a multiple model approach and dynamic neural networks are employed to accomplish this goal. Using residuals obtained by comparing each network output with the measured jet engine output and by invoking a properly selected threshold for each network, reliable criteria are established for detecting and isolating faults in the jet engine components. The paper entitled ''Bayesian classifiers based on probability density estimation and their applications to simultaneous fault diagnosis'', by Y.L. He, R. Wang, S. Kwong, and X.Z. Wang, considers a problem of simultaneous fault diagnosis and proposes a new Bayesian classifier for problem solving without assuming the independence among features. Simulation results show that the proposed approach significantly outperforms the traditional ones when the dependence exists among features. The paper entitled ''A fault prediction method that uses improved case-based reasoning to continuously predict the status of a shaft furnace'', by A.J. Yan, W.X. Wang, C.X. Zhang, and H. Zhao, develops an improved CBR-based fault prediction method for a shaft furnace, where a water-filling theory-based weight allocation model and a group decision-making-based revision model are used for improving the prediction accuracy. Also, the optimal allocation mechanism of channel power and the credibility of historical results are used to enhance the predicted results. The paper entitled ''Improved process monitoring and supervision based on a reliable multi-stage feature-based pattern recognition technique'', by L. Al-Shrouf, M. Saadawia, and D. Soffker, presents two new multi-sensor data fusion algorithms for object detection in monitoring of industrial processes. The goals of this work are to reduce the rate of false detection and obtain reliable decisions on the presence of target objects. In this work, the classifier is trained and validated by using the real industrial data. The two algorithms are also tested by using the same data, and their performance and modelling complexity are compared.
Entropy Indices Based Fault Detection
Procedia Manufacturing, 2020
Signal processing-based fault diagnosis is a growing domain in control engineering. As a statistical measure, entropy can measure the complexity of signals, this could be strongly related to the functional status of a system which provides these signals. Therefore, entropy can be a promising non-parametric tool to extract different characteristics of manufacturing system provided signals. Recently, many studies have applied entropy indices in diagnosis, detection and prediction of faults. This paper proposes a theoretical approach to investigate the applicability of entropy indices for the fault characteristics extraction from discrete signals. The study uses synthetic test signals of various structures and properties. At first probability density functions are estimated and entropy indices as the Renyi entropy and sample entropy for different lengths are computed. These are compared and put in relation with fault occurrence. The results show that these indices can be a promising no...
Measurement and Control, 2020
Effective and efficient incipient fault diagnosis is vital to the maintenance and safe application of large-scale key mechanical system. Variable predictive model–based class discrimination is a recently developed multiclass discrimination method and has been proved to be potential tool for multi-fault detection. However, the vibration signals from dynamic mechanical system always present non-normal distribution so that the original variable predictive model–based class discrimination might produce the inaccurate outcomes. An improved variable predictive model–based class discrimination method is introduced at first in this work. At the same time, variable predictive model–based class discrimination will suffer computation difficulty in the case of high-dimension input features. Therefore, a novel feature selection method based on similarity-fuzzy entropy is presented to boost the performance of the variable predictive model–based class discrimination classifier. In this method, the...
New informative features for fault diagnosis of industrial systems by supervised classification
18th Mediterranean Conference on Control and Automation, MED'10, 2010
The purpose of this article is to present a method for industrial process diagnosis. We are interested in fault diagnosis considered as a supervised classification task. The interest of the proposed method is to take into account new features (and so new informations) in the classifier. These new features are probabilities extracted from a Bayesian network comparing the faulty observations to the normal operating conditions. The performances of this method are evaluated on the data of a benchmark example: the Tennessee Eastman Process. Three kinds of fault are taken into account on this complex process. We show on this example that the addition of these new features allows to decrease the misclassification rate.
3 Fuzzy Logic-Based Fault Diagnosis
2009
An intelligent diagnostic scheme using sensor network for incipient faults is proposed using a holistic approach which integrates model-, fuzzy logic-, neural networkbased schemes. In case the system is highly non-linear and there are enough training data available, a neural network based scheme is preferred; where the rules relating the input and output can be derived, a Fuzzy-logic approach is chosen; and where a model is available, a linearized model is employed. These three schemes are integrated sequentially ensuring thereby that critical information about the presence or absence of a fault is monitored in the shortest possible time, and the complete status regarding the fault is unfolded in time. The proposed scheme is evaluated extensively on simulated examples and on a physical system exemplified by a benchmarked laboratoryscale two-tank system to detect and isolate faults including sensor, actuator and leakage ones.
Machine Learning Strategy for Fault Classification Using Only Nominal Data
PHM Society European Conference, 2016
Machine learning methods are increasingly used for rotating machinery monitoring. Usually at set up, only data associated to an engine in a good state, the so called nominal data, are available for the machine learning phase. Nevertheless a classifier requires faulty data to be trained at identifying the causes of the anomalies and this fact has generally limited the usage of data driven approaches to fault detection tasks. The paper suggests a strategy to use machine learning methods even for fault classification purposes and diagnostics. Within the proposed framework three different machine learning methods, Gaussian Mixture Model (GMM), Support Vector Machines (SVM) and Auto Associative Neural Networks (AANN) have been implemented, tested and compared. The idea is to take into account some 'a priori' knowledge about the faults to be classified, to drive the behavior of the machine learning methodology (SVM or AANN or GMM) to be more or less reactive to the different faults. The indicators (features) more sensitive to each kind of fault are firstly selected on the basis of expert knowledge. For each different fault, a set of indicators is defined and computed from nominal data only. Each set is then used to produce training data for one specific fault. Such data sets are then used to train one instance of each method for each different fault. The underlying logic is that fault tuned input data is able to produce fault tuned instances of the methods. For example the instance trained with the indicators associated to a fault 'A' reacts more powerfully in presence of the fault 'A' than the others. Once an anomaly is detected, the comparison among the reactions of the different 'fault tuned' instances allows classifying the fault, not just to detect it. The results show best detection performances for SVM whilst AANN outperforms the other two methods for classification. Gianluca Nicchiotti et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 3.0 United States License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Information Based Fault Diagnosis
Proceedings of the 17th IFAC World Congress, 2008, 2008
Fault detection and isolation, (FDI) of parametric faults in dynamic systems will be considered in this paper. An active fault diagnosis (AFD) approach is applied. The fault diagnosis will be investigated with respect to different information levels from the external inputs to the systems. These inputs are disturbance inputs, reference inputs and auxiliary inputs. The diagnosis of the system is derived by an evaluation of the signature from the inputs in the residual outputs. The changes of the signatures form the external inputs are used for detection and isolation of the parametric faults.
Fault diagnosis of chemical processes with incomplete observations: A comparative study
Computers & Chemical Engineering, 2016
An important problem to be addressed by diagnostic systems in industrial applications is the estimation of faults with incomplete observations. This work discusses different approaches for handling missing data, and performance of data-driven fault diagnosis schemes. An exploiting classifier and combined methods were assessed in Tennessee-Eastman process, for which diverse incomplete observations were produced. The use of several indicators revealed the trade-off between performances of the different schemes. Support vector machines (SVM) and C4.5, combined with k-nearest neighbourhood (kNN), produce the highest robustness and accuracy, respectively. Bayesian networks (BN) and centroid appear as inappropriate options in terms of accuracy, while Gaussian naïve Bayes (GNB) is sensitive to imputation values. In addition, feature selection was explored for further performance enhancement, and the proposed contribution index showed promising results. Finally, an industrial case was studied to assess informative level of incomplete data in terms of the redundancy ratio and generalize the discussion.
Fault diagnosis in physical systems turn out to become very complex as soon as the considered systems are no longer elementary and become more and more complex and sophisticated, it is then legitimate for companies to acquire an effective diagnosis system. In our research work, we are interested in the diagnosis of industrial hybrid systems, which are composed of continuous systems, discrete event systems and an interface that manages the interactions between the two aspects. The aim of this paper is, first, to give an overview of hybrid systems, and second, to review the state of the art of methods and techniques for fault diagnosis of hybrid systems.