PT Unknown
AU Wang, Y
TI Fault Estimation Schemes of Wireless Networked Control Systems for Real-Time Industrial Applications
PD 05
PY 2014
LA en
AB With the rapid growth of microelectronics, information and communication technologies, much attention has been paid on the research of wireless networked control systems (W-NCSs). The development of W-NCSs raises new challenges in fault estimation (FE) technology regarding to the imperfect data transmission, such as transmission delay, packet loss, jitter and so on. To ensure the system safety and reliability, an effective FE approach over networks is of prime importance to be developed.
On the other hand, aiming for the applications on real-time industrial automation, the specific characteristics of network should be properly considered. Since the transmission tasks of measurements and control commands are normally deterministic over a period of time, a deterministic transmission mechanism and the relevant FE scheme should be proposed. Motivated by the widespread popularity of centralized and decentralized structures for industrial processes, development of both centralized and decentralized FE schemes for W-NCSs, which can be applied on industrial automation, is the primary objective of this thesis.
This thesis is first dedicated to the modeling of communication and process. For the communication modeling, time division multiple access (TDMA) based medium access control (MAC) protocol is modified to guarantee the real-time performance. The process model is built considering multirate sampling based on the hierarchical structure of W-NCSs. By observing the uncertainty of networks and effects of faults, a linear periodic (LP) system model, which is the integration of communication model and process model, is presented as a basis for the later developments.
The further study focuses on the development of FE schemes for both centralized and decentralized W-NCSs. To reach an enhanced robustness against unknown disturbance and initial state estimate error, the centralized FE approach is proposed with the help of stochastic model in Krein space. For decentralized FE, the algorithm is implemented by every sub-system, and the coupling relations between sub-systems should be properly considered. Based on it, the FE approaches are presented with two kinds of residual signals, i.e., non-shared residuals and shared residuals, respectively.
To illustrate the effectiveness of the derived FE approaches, an industrial platform WiNC integrated with three-tank system is utilized in this thesis. The FE algorithms have been verified for three data transmission cases, i.e., sampling-based, delay and packet loss, so that the robustness against imperfect communication is demonstrated. Moreover, the performances of sensor and actuator FE have also been tested well on WiNC platform.
ER