(625a) Distributed Fault Diagnosis for Networked Nonlinear Uncertain Systems
AIChE Annual Meeting
2017
2017 Annual Meeting
Computing and Systems Technology Division
Process Monitoring & Fault Detection
Wednesday, November 1, 2017 - 3:15pm to 3:32pm
These realizations have motivated the design of FDI schemes with distributed architecture for networked systems. There exist a plethora of research in designing fault detection and isolation framework for monitoring of engineering systems (see e.g., [2], [3], [4], [5], [6], [7], [8], [9], [10] and [11]). However, most of the existing results in the literature present a centralized FDI scheme that might not be applicable to networked systems due to computational capa- bilities, communication infrastructure and reliability issues as discussed above. There also exist several results for FDI design in networked systems. In [12], a distributed data based actuator fault identification scheme is presented for linear networked process systems. To this end, local FDI agents are designed for each subsystem using moving horizon estimator (MHE)s to identify the fault. In [13], a distributed actuators FDI scheme is proposed for a class of interconnected uncertain nonlinear systems. To this end, a local FDI framework is designed for each subsystem. In each local FDI scheme, first a fault detection scheme is designed which simply uses output estimation error as residual. Then, a bank of fault isolation estimators is designed using adaptive estimation techniques. In [14], a distributed actuators fault detection and isolation framework is presented for nonlinear uncertain large-scale discrete-time dynamical systems. To overcome the scalabil- ity issues, the large scale system under considerations is modeled as several interconnected subsystems. For each subsystem a local FDI scheme is designed using adaptive estimation techniques. Decisions made by the local fault diagnoser are gathered by a higher level global fault diagnoser to determine health of the overall system. A consensus-based estimator is designed to enhance the ability of FDI schemes in detection and isolation of faulty shared information between the subsystems. In [15], a distributed sensor fault diagnosis for a network of interconnected cyber-physicalsystem (CPS)s (see e.g., [16]) is presented. The FDI module consists of two layers. The first layer is built using local FDI schemes to isolate multiple local sensor faults. The detection is carried out using adaptive estimation techniques while the isolation is done using logic rules. The second layer is realized by applying a global decision logic designed to isolate multiple sensor faults that may propagate in the network. However, none of the existing results have addressed the problem of isolation of multiple actuator faults or simultaneous actuator and sensor faults in uncertain nonlinear networked systems. Also, the results addressing actuator faults are based on the assumption that full state measurements are available.
Motivated by the above considerations, in this work, we present a distributed fault diag- nosis framework for a generalized class of networked uncertain nonlinear systems. The idea is to design a bank of FDI schemes based on the robust FDI methodology presented in [11] in a distributed manner. In particular, the nature of the interconnection is first determined and appropriate FDI schemes are developed. To this end, at first boundedness of estima- tion error in the presence of uncertainty and exchange of information is established. Then a bank of residuals is designed for each subsystem using an appropriate subset of the available measurements (and associated state observers), to determine the expected behavior of the system and compare with the observed evolution, where each residual is sensitive to a subset of faults and insensitive to the rest, in the absence of uncertainty. To achieve FDI in the presence of uncertainty, thresholds are defined in a way that they account for the impact of the uncertainty on the estimation error and the prediction of the expected system behavior. In this way, each residual is still insensitive to a subset of faults in the presence of uncertainty and sensitive to the rest if the fault functionality satisfies a detectability condition. Then the detectability and isolability conditions for single and simultaneous faults are derived, when the detectability analysis establishes that the sensitive property of residuals is retained in the presence of uncertainty (see [11] for more on this). The local FDI schemes communi- cate to diagnose the faults in the shared interconnections between subsystems. Results are demonstrated using a networked CSTR example.
 References:
 [1] Xing-Gang Yan and Christopher Edwards. Robust decentralized actuator fault detection and estimation for large-scale systems using a sliding mode observer. International Journal of control, 81(4):591–606, 2008.
[2] Paul M Frank. Fault diagnosis in dynamic systems using analytical and knowledge-based redundancy: A survey and some new results. Automatica, 26(3):459–474, 1990.
[3] Mogens Blanke, Michel Kinnaert, Jan Lunze, Marcel Staroswiecki, and J Schr¨oder.
 Diagnosis and fault-tolerant control, volume 2. Springer, 2006.
 [4] Claudio De Persis and Alberto Isidori. A geometric approach to nonlinear fault detection and isolation. Automatic Control, IEEE Transactions on, 46(6):853–865, 2001.
[5] Xing-Gang Yan and Christopher Edwards. Nonlinear robust fault reconstruction and estimation using a sliding mode observer. Automatica, 43(9):1605–1614, 2007.
[6] Prashant Mhaskar, Charles McFall, Adiwinata Gani, Panagiotis D Christofides, and James F Davis. Isolation and handling of actuator faults in nonlinear systems. Auto- matica,44(1):53–62,2008.
[7] Xiaodong Zhang, Marios M Polycarpou, and Thomas Parisini. Fault diagnosis of a class of nonlinear uncertain systems with lipschitz nonlinearities using adaptive estimation. Automatica, 46(2):290–299, 2010.
[8] Xiaodong Zhang, Qi Zhang, and Niharika Sonti. Diagnosis of process faults and sensor faults in a class of nonlinear uncertain systems. Systems Engineering and Electronics, Journal of, 22(1):22–32, 2011.
[9] Miao Du and Prashant Mhaskar. Isolation and handling of sensor faults in nonlinear systems. Automatica, 50(4):1066–1074, 2014.
[10] Miao Du, James Scott, and Prashant Mhaskar. Actuator and sensor fault isolation of nonlinear process systems. Chemical Engineering Science, 104:294–303, 2013.
[11] Hadi Shahnazari and Prashant Mhaskar. Simultaneous actuator and sensor fault isola- tion of nonlinear systems subject to uncertainty. In 2016 American Control Conference (ACC), pages 6857–6862. IEEE, 2016.
[12] Di Peng, Nael H El-Farra, Zhiqiang Geng, and Qunxiong Zhu. Distributed data-based fault identification and accommodation in networked process systems. Chemical Engi- neering Science, 136:88–105, 2015.
[13] Xiaodong Zhang and Qi Zhang. Distributed fault diagnosis in a class of interconnected nonlinear uncertain systems. International Journal of Control, 85(11):1644–1662, 2012.
[14] Riccardo MG Ferrari, Thomas Parisini, and Marios M Polycarpou. Distributed fault detection and isolation of large-scale discrete-time nonlinear systems: An adaptive ap- proximation approach. IEEE Transactions on Automatic Control, 57(2):275–290, 2012.
[15] Vasso Reppa, Marios M Polycarpou, and Christos G Panayiotou. Distributed sensor fault diagnosis for a network of interconnected cyberphysical systems. IEEE Transactions on Control of Network Systems, 2(1):11–23, 2015.
[16] Panos J Antsaklis, Bill Goodwine, Vijay Gupta, Michael J McCourt, Yue Wang, Po Wu, Meng Xia, Han Yu, and Feng Zhu. Control of cyberphysical systems using passivity and dissipativity based methods. European Journal of Control, 19(5):379–388, 2013.