Ph. D. Project
Fault detection and isolation method design with reliability assessment in distributed networks
2016/10/01 - 2019/09/30
Other supervisor(s):
Dr. Lizeth TORREZ, Universidad Nacional Autónoma d (
Objective: To establish a theoretical foundation of a fault detection and isolation framework with reliability assessment to detect and to isolate sensor and actuator faults/failures for large distributed networks.

Motivation: The complexity of large distributed systems (such as telecommunication, air conditioning, electrical or drinking water distribution networks) and the huge amount of information carried by them have caused an increase in demand for network management systems for higher levels of network availability and reliability. In particular, the area of network fault management requires a lot of expertise because is becoming critical due to the dynamic nature and heterogeneity of networks.

Breakdowns of networks cause huge financial losses. Consequently it is crucial to diagnose a fault or failure with efficiency and/or to predict a fault. Actuator and sensor lifetime and system reliability provide some indicators of component prediction degradation and/or probability of fault occurrence. To increase the performance of fault diagnosis such knowledge should be considered. For these reasons, there is a real need and a challenge for establishing a theoretical foundation of network fault management with reliability assessment.

Methodology: The first step is to design an approach to detect and isolate faults in sensors and actuators conforming a distributed network by generating residual sets [1]. A fault represents a malfunction event in a system (say the network). A residual is a signal (constructed from sensor measurements) that reacts to a chosen fault or subset of the considered faults. And by generating a suitable set of such residuals, fault detection and isolation can be achieved. A common thread is the development of systematic design and analysis methods for residual generators based on a number of different model classes, namely deterministic and stochastic linear models on state-space, descriptor, or transfer function form, and non-linear polynomial systems. In addition, it is considered important that there exist readily available computer tools for all design algorithms.

To detect possible faults in sensors and actuators in the distributed network, alarms must be generated by designing residuals constructed form the measurements of the network sensors and by establishing thresholds for each of them [4]. The alarms can be generated for a sub-sector, a sector or for the global network.

To isolate the faults, signature matrices can be constructed from residual sets such that each signature can permit the isolation of a possible fault or a fault set in the network [2]. Normally, once the isolation is performed a decision can be taken. However, the residuals are constructed from the available measurements provided by the network sensors, which can be faulty [3]. This fact complicates the isolation decision. Therefore, a set of signatures must be evaluated with probabilistic tools to take the best decision. In other words, it is necessary to assess the reliability of each sensor and actuator to supplement the residual sets allowing the enhancement of the detection and isolation decision [5]. The reliability assessment will provide the fault probability of each sensor and actuator and/or also based on the prediction of the remaining useful life of each component based on the probability of possible degradation [6]. In order to develop the FDI approach, the main idea is to take into account through an appropriate fusion between the reliability and the residuals in order to increase the fault isolation efficiency i.e. (i) the probability of simultaneous and sequential faults and the possibility to isolate them, (ii) the total number of sensors and actuators and their neighbours (redundancy) to keep the network operating.

Novel/Additive Information: The novelty of this thesis proposal as well as the expected results is the fusion of two research disciplines in a very important application: the diagnosis and the prognosis in distribution networks for detecting and isolating present and future faults. There are some results of both disciplines separately for specific distribution networks, therefore, the reason of this work is to establish a unified theory

[1] A. Rosich and V. Puig. (2013, July). Model-based leakage localization in drinking water distribution networks using structured residuals. In Control Conference (ECC), 2013 European (pp. 410-415). IEEE.
[2] L. Torres, C. Verde, R. Carrera and R. Cayetano. "Algoritmos de diagnóstico para fallas en ductos", Tecnología y Ciencias del Agua, Volume 5, Issue 4, 2014

[3] P. Weber, C. Simon, D. Theilliol and V. Puig (2012, August). Fault-tolerant control design for over-actuated system conditioned by reliability: a drinking water network application. In 8th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes, SAFEPROCESS 2012.
[4] L. Torres, G. Besançon, C. Verde and O. Gonzalez. "High gain observers for leak location in subterranean pipelines of liquefied petroleum gas", International Journal of Robust and Nonlinear Control , Volume 24, Issue 6, Pages 1127 -1141, 2014.
[5] C. S. Hood and C. Ji. (1997). Proactive network-fault detection [telecommunications]. Reliability, IEEE Transactions on, 46(3), 333-341.
[6] P. Weber, D. Theilliol and C. Aubrun. (2008). Component reliability in fault-diagnosis decision making based on dynamic Bayesian networks. Proceedings of the Institution of Mechanical Engineers, Part O: Journal of Risk and Reliability, 222(2), 161-172.
model-based fault diagnosis method, reliability assessment, component degradation, distributed netwo
Control Identification Diagnosis
Bourse Mexicaine (CONACYT)