Font Size: a A A

Hybrid Neural Network-based Fault-tolerant Control Methods In Marine Systems

Posted on:2004-11-13Degree:MasterType:Thesis
Country:ChinaCandidate:G YaoFull Text:PDF
GTID:2208360092481533Subject:Power electronics and electric drive
Abstract/Summary:PDF Full Text Request
As the need of modern production and the advancement of the scientific technology, the process of modern industry is evolving towards the direction of large scale and complex. Once the fault happened in such land of system, great casualty and loss of prosperity will be inevitable. To increase the reliability and the security of life and prosperity, the fault tolerant control (FTC) theory is developed slowly, and gradually became a primary research subject in the modern control theory.As a new inter-discipline, the goal of the FTC is maintain the stability of the whole system at the cost of reducing some performances of the system, when faults happen to several components. It can adapt the system to the obvious change of the environment and avoid the passive effect to the stability and performing of the system caused by the collapse of one or several key components. Since faults are inevitable for any system, the FTC could be regarded as the final defense line to guarantee the safe operation of the system, it offers a new method to improve the stability of the modern complicate dynamic system.In the thirty years of development, traditional FTC gradually shows bigger limitations. It depends on the math model and lacks the general analysis methods for nonlinear system. So the FTC based on the artificial intelligence developed. The intelligent FTC offsets many disadvantages of traditional FTC technology and can deal with the uncertainty of the model and tolerant control of the nonlinear system.In this research background, this paper come up with a FTC method based on the hybrid neural network and apply hi the ship automatic manoeuvre system. The structure of the system and the realization way of its sub-systems are analyzed in detail hi the paper.The principle research contents include:(1) This paper comes up with a new FTC scheme based on the hybrid neural network, which is focused on the sensor failure and actuator failure, thus make use of multi-neural network to realize the fault detection and fault tolerant control of the system.(2) The fault detection technology based on the information fusion is studied. Two steps are designed for the fault detection; firstly, the method of fusion information from multi-sensors based on fuzzy inference through the opinion distance is used for local information fusion, so the stability of omen variants are guaranteed. Second, the fuzzy neural network is used for whole information fusion in order to detect the fault.(3) In the condition of feedback-signal sensor failure, the FTC is also discussed. An output recurrent feedback neural network is designed for estimating the system state. When feedback-signal sensor is in failure, the network could do the real-tune estimation through the detected data of the fault sensors and feedback the signals, thus fault tolerance is achieved.(4) The adaptive control of the control system is studied in the mode of actuator failure. The adaptive neural network is used as system controller in this paper. When faults happen to the actuator but not complete failure, the adaptive neural network could automatically change the weight of the neural network and regulate the control rules on line to accustom to the fault-mode system and make it have the similar performance like the normal system.(5) The integrated intelligent fault tolerant control system is applied in ship automatic manoeuvre system. The simulation experiment is also done.
Keywords/Search Tags:neural network, fuzzy logic, information fusion, adaptive control, fault detection, FTC
PDF Full Text Request
Related items