Fault diagnosis based on black-box models with application to a liquid-level system

Palma, Lúıs Brito; Coito, Fernando; Silva, R.N.

doi:10.1109/etfa.2003.1248772

Cited by 5 publications

(7 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Equations (11) and (12) shows step number 7. t is the current learning step number, T is the total learning number. N z (0) is the initial value, INT[x] is function to round the x to the nearest integers towards zero.…”

Section: Neural Networkmentioning

confidence: 99%

“…The FDI methods can be classified into two major groups: model-free methods and model-based methods [11]. Scheme of FDI is shown in Fig.…”

Section: Design Of Fault Detection and Identificationmentioning

confidence: 99%

“…neural networks, fuzzy logic, neuro-fuzzy schemes, evolutionary programming, etc. [11]. Knowledge-based methods are sometimes called modelfree or qualitative methods.…”

Section: Design Of Fault Detection and Identificationmentioning

confidence: 99%

See 2 more Smart Citations

Soft computing and signal processing based active fault tolerant control for benchmark process

Postalcioglu

Erkan

2007

Neural Comput & Applic

View full text Add to dashboard Cite

Ideally, when faults happen, the closed-loop system should be capable of maintaining its present operation. This leads to the recently studied area of fault-tolerant control (FTC). This paper addresses soft computing and signal processing based active FTC for benchmark process. Design of FTC has three levels: Level 1 comprises a traditional control loop with sensor and actuator interface and the controller. Level 2 comprises the functions of online fault detection and identification. Level 3 comprises the supervisor functionality. Online fault detection and identification has signal processing module, feature extraction module, feature cluster module and fault decision module. Wavelet analysis has been used for signal processing module. In the feature extraction module, feature vector of the sensor faults has been constructed using wavelet analysis, sliding window, absolute maximum value changing ratio and variance changing ratio as a statistical analysis. For the feature cluster module, the self-organizing map (SOM), which is a subtype of artificial neural network has been applied as a classifier of the feature vector. As a benchmark process three-tank system has been used. Control of the three-tank system is provided by fuzzy logic controller. Faults are applied to three level sensors. Sensor faults represent incorrect reading from the sensors that the system is equipped with. When a particular fault occurs in the system, a suitable control scheme has been selected on-line by supervisor functionality to maintain the closed-loop performance of the system. Active FTC has been achieved by switch mode control using fuzzy logic controller. Simulation results show that benchmark process has maintained acceptable performance with FTC for the sensor faults. As a result, when the system has sensor faults soft computing and signal processing based FTC helps for the best performance of the system.

show abstract

Section: Neural Networkmentioning

confidence: 99%

“…The FDI methods can be classified into two major groups: model-free methods and model-based methods [11]. Scheme of FDI is shown in Fig.…”

Section: Design Of Fault Detection and Identificationmentioning

confidence: 99%

See 1 more Smart Citation

Soft computing and signal processing based active fault tolerant control for benchmark process

Postalcioglu

Erkan

2007

Neural Comput & Applic

View full text Add to dashboard Cite

show abstract

“…Reference [7] has succeeded in describing these distributions by means of continuous approximation under CUSUM detection, whereas the results for the GLR still has much room for improvement. Regarding the integrated FD, a general type of approaches is the integration of filters (or innovation nodes as in neural network) and detection algorithms: relevant research has been developed concerning CUSUM [8] and GLR [9]. This paper firstly attempts computing the probability distribution of detection delay with the GLR method, concerning the discrete random walk formed by taking summation of the original Gaussian i.i.d.…”

Section: Introductionmentioning

confidence: 99%

Statistical characterization of the GLR based fault detection

Yang

Zhao

2011

Proceedings of the 2011 American Control Conference

View full text Add to dashboard Cite

This paper is mainly focused on providing the probabilistic performance criteria for generalized likelihood ratio (GLR) test based fault detection schemes. Analytical expressions regarding probability distribution of the detection delay and time between false alarms are presented, which are validated by simulation. The results can be applied to important industrial application, such as abnormal signal monitoring (magnitude or energy), targeting and navigation, as well as the design of fault tolerant control systems (FTCS).

show abstract

“…To circumvent this problem one alternative is to use black-box models, like adaptive ARX models, and/or neural networks. FDI techniques based on neural networks are gaining more and more interest, mainly due to their ability to deal with nonlinear systems, and their robustness to noise, [7]- [8]- [12]. Most applications of neural observers are related to neural observers-predictors; some examples are the papers, [2]- [7], and the thesis, [9].…”

Section: Introductionmentioning

confidence: 99%

Process Fault Diagnosis Approach based on Neural Observers

Palma

Coito

Silva

2005 IEEE Conference on Emerging Technologies and Factory Automation

Self Cite

View full text Add to dashboard Cite

This paper presents an approach to process fault diagnosis (FDI) in nonlinear dynamical systems, based on a bank of neural observers. Each neural observer is tuned to a particular fault and predicts, using its embedded model, the expected values for the sensor readings. The residuals, the difference between the sensor readings and the predicted readings, are used as fault indicators. Each neural observer is based on a multi-layer perceptron feed-forward neural network with externalfeedback connections, and an adjustable gain.The focus of this work is on diagnosis ofparametric faults on process components, by means of analyzing the residuals. The proposed FDI technique has been implemented on a simulation model ofa DC motor under closed-loop control. Results from experiments are presented and discussed.

show abstract

Fault diagnosis based on black-box models with application to a liquid-level system

Cited by 5 publications

References 12 publications

Soft computing and signal processing based active fault tolerant control for benchmark process

Soft computing and signal processing based active fault tolerant control for benchmark process

Statistical characterization of the GLR based fault detection

Process Fault Diagnosis Approach based on Neural Observers

Contact Info

Product

Resources

About