An algorithmic fault-tolerant control architecture without actuator redundancy

---

Tezin Türü: Doktora

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Havacılık ve Uzay Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2018

Öğrenci: ALP MARANGOZ

Danışman: ALİ TÜRKER KUTAY

Özet:

In this thesis work, a novel algorithmic fault tolerant control system architecture against actuator failures is developed. The method is based on injection of perturbations on the controlled states that are connected to healthy actuators, in order to compensate for the failed components and maintain overall stabilization of the system. An adaptive state estimator structure is used for detection of faults and fault mitigation perturbations are generated from a singularly perturbed dynamic system, which is a part of the control architecture. The proposed method is an algorithmic fault tolerant control architecture in a sense that the fault tolerance and reliability is achieved algorithmically and without using any redundant physical components. For the theoretical analysis of the developed control system, the problem is formulated as a nonlinear control problem for interconnected systems and a theorem is structured that includes the assumptions, conditions and stability properties of the proposed architecture. Resultant algorithm can be applied to wide variety of problems including multi-input-multi-output unstable nonlinear systems, provided that the system under consideration is Lipschitz continuous and certain bound conditions are satisfied. Design methodology is explained through theoretical analyses and analytically tractable numerical examples. Applications on more complex systems and limitations of the proposed fault tolerant control system architecture are demonstrated on joint failures of robotic manipulators and propeller loss of quadrotors cases through theoretical analyses and simulation results.