Neural-network-based scheme for sensor failure detection, identification, and accommodation

Napolitano, Marcello R.; Neppach, Charles; Casdorph, Van; Naylor, Steve; Innocenti, Mario; Silvestri, Giovanni

doi:10.2514/3.21542

Cited by 101 publications

(45 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multiple-model architectures enjoy several advantages [21], such as the fast adaptation, compared with indirect adaptive control (see [22][23][24], and references therein), when the plant dynamics change abruptly (for instance, because of failures), and the ability to provide high levels of performance for different classes of dynamic models. Furthermore, multiple-model approaches have a design modularity that makes them suitable for a number of applications, while being able to provide a natural accommodation of faults.…”

Section: Set-valued Observer-based Fault-tolerant Controlmentioning

confidence: 99%

Fault‐tolerant control of an air heating fan using set‐valued observers: An experimental evaluation

Rosa

Simao

Silvestre

et al. 2015

Adaptive Control & Signal

View full text Add to dashboard Cite

This paper proposes a multiple-model solution to the problem of controlling an air heating fan subject to faults. These faults are modeled by means of an abnormal unknown airflow input rate which the nominal controller is not designed for. Moreover, the average temperature of the air flowing through the system, which can be seen as an offset on the corresponding dynamics, is (slowly) time-varying and highly dependent on the ambient temperature. The fault-tolerant control (FTC) method adopted makes use of set-valued observers (SVOs) to invalidate possible models of the system. Unlike classical fault detection, this approach does not rely on residuals to detect abnormal system operation. This fact allows to reduce the conservatism of the solution and enables a straightforward design from the faulty and nominal models of the plant. Moreover, the absolute distinguishability concept is used to derive input signals that bolster the detection of faults. Although SVOs require heavy real-time calculations that hinder its implementability in systems with low computational power, it is shown that the architecture of the FTC strategy proposed is highly parallelizable and, thus, may take advantage of standard multi-core processing units. Experimental results are presented.

show abstract

Section: Set-valued Observer-based Fault-tolerant Controlmentioning

confidence: 99%

Fault‐tolerant control of an air heating fan using set‐valued observers: An experimental evaluation

Rosa

Simao

Silvestre

et al. 2015

Adaptive Control & Signal

View full text Add to dashboard Cite

show abstract

“…Filter-based methods include observers [3], unknown input observers [4], Kalman filters [5], particle filters [6], sliding mode observers [7], II~, filters [8], and set membership filters [9]. There are also methods based on computer intelligence [10] that include fuzzy logic [11], neural networks [12], genetic algorithms [13], and expert systems [14]. Other methods include those based on Markov models [15], system identification [16], wavelets [17], Bayesian inference [18], control input manipulation [19], and the parity space approach [20].…”

Section: Cardinality Of Yimentioning

confidence: 99%

“…Note that we are not considering whether or not the SSRs exceed their threshold; we are only considering how large the SSRs are relative to their thresholds. The marginal detection rate is given as (12) where (13) If Y'1 is empty, aud Y j is not empty, then (14) If Y J is empty, aud Y'1 is not empty, then (15) Proof Equation (12) cau be obtained using Lemmas 5, and 6, which are in the Appendix. Equations (14), aud (15) follow from (11).…”

Section: B Correct Fault Classification Ratesmentioning

confidence: 99%

Analytic Confusion Matrix Bounds for Fault Detection and Isolation Using a Sum-of-Squared-Residuals Approach

Simon

2010

IEEE Trans. Rel.

View full text Add to dashboard Cite

Given a system which can fail in 1 of n different ways, a fault detection and isolation (FDI) algorithm uses sensor data to determine which fault is the most likely to have occurred. The effectiveness of an FDI algorithm can be quantified by a confusion matrix, also called a diagnosis probability matrix, which indicates the probability that each fault is isolated given that each fault has occurred. Confusion matrices are often generated with simulation data, particularly for complex systems. In this paper, we perform FDI using sum-of-squared residuals (SSRs). We assume that the sensor residuals are s-independent and Gaussian, which gives the SSRs chi-squared distributions. We then generate analytic lower, and upper bounds on the confusion matrix elements. This approach allows for the generation of optimal sensor sets without numerical simulations. The confusion matrix bounds are verified with simulated aircraft engine data.Index Tenns-Aircraft turbofan engine, chi-squared distribu tion, confusion matrix, diagnosis probability matrix, fault detec tion and isolation. TNRTrue negative rate

show abstract

“…A case study chosen for this work is the Sensor Failure Detection, Identification, and Accommodation (SFDIA) flight control scheme [6,7]. The SFDIA scheme is part of an advanced flight control system that uses analytical instead of a physical redundancy to achieve faulttolerance.…”

Section: Case Studymentioning

confidence: 99%

Regressive model approach to the generation of test trajectories

Taylor

Čukić

Proceedings 3rd IEEE Symposium on Application-Specific Systems and Software Engineering Technology

View full text Add to dashboard Cite

One aspect of system safety assurance is the application of large test sets. To aid in the development of test cases, researchers have investigated automated test generation systems which decrease the time and cost of acquiring new tests. Many automated systems currently exist, but few address the generation of trajectories of data.A trajectory is defined as a series of data points with each point relying upon the state of the system and previous point values.Examples of trajectories include an airplane's flight path or data describing a nuclear reaction.This paper describes an original approach to test trajectory generation using statistical regression models. An existing small set of trajectories is utilized to build regressive models. Then, the independent variables of the model can be perturbed, providing the basis for the use of the regressive model in the generation of new trajectories. A case study has been conducted within the scope of testing the sensor failure detection, identification, and accommodation (SFDIA) flight control scheme. We report and evaluate the results of this case study.

show abstract

Neural-network-based scheme for sensor failure detection, identification, and accommodation

Cited by 101 publications

References 6 publications

Fault‐tolerant control of an air heating fan using set‐valued observers: An experimental evaluation

Fault‐tolerant control of an air heating fan using set‐valued observers: An experimental evaluation

Analytic Confusion Matrix Bounds for Fault Detection and Isolation Using a Sum-of-Squared-Residuals Approach

Regressive model approach to the generation of test trajectories

Contact Info

Product

Resources

About