Font Size: a A A

Research On Active Decentralized Fault Tolerant Control For Reconfigurable Manipulator With Multiple Concurrent Failures

Posted on:2015-09-29Degree:DoctorType:Dissertation
Country:ChinaCandidate:B ZhaoFull Text:PDF
GTID:1228330428984042Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Reconfigurable manipulator system can fulfill different tasks by changing its degree offreedom and task spaces, along with various combinations of modules. By virtue of thereconfiguration technique and theorem, the reconfigurable manipulator system becomessmarter, furthermore, it can save resources and manufacturing space. In other words, themain motive to design reconfigurable manipulator lies in its strong robustness, flexibleand low cost. In the presence of these advantages, it has a great deal of potentialapplications in a variety of fields, such as aerospace and deep space exploration, precisionmachining, military and battle fields, rescue, teleoperation, high risk task, even theentertainment. Therefore, the reconfigurable manipulator system has important researchvalue and engineering significance. Meanwhile, this research can promote a rapiddevelopment of other disciplines. With the progress in bionic science, artificialintelligence and computational technique, the reconfigurable manipulator system canprovide a platform for their experiments, and the research on reconfigurable manipulatoris inseparable from the support of these disciplines.As well known, the reconfigurable manipulator is often utilized in long distant,unknown and dangerous environments, faults will inevitably occur in the actuators,sensors and other units. In these situations, direct maintenance is difficult as well. If it cannot be repaired timely, the reconfigurable manipulator will work in unpredictable manner,this not only will shorten its using life, but even the worse could lead to disastrousconsequences. Thus, it becomes an urgent task to improve safety and reliability. In thecomputer control system which is applied to the manipulators, the reliability of thehardware and software has reached a quite high level. The plenty of failures occur inactual systems showed that the failure in actuator and sensor is the main reason cause theloss of effectiveness. In this viewpoint, it is valuable to research on the fault tolerantcontrol of actuator and sensor without any doubt. From the information available currently,there is only a few outcomes have been carried out for the fault tolerant control of thereconfigurable manipulator. This thesis is concerned with the fault tolerant control designfor reconfigurable manipulator. The research mainly covers the control scheme based onthe local module information, fault detection based on decentralized observer technique,effectiveness factor integration based actuator fault tolerant control, active decentralizedfault tolerant control for sensor fault based on signal reconstruction and active substitutingtechniques, active fault tolerant control for concurrent failures, as well as thedecentralized cascaded position/force control scheme, etc.. And this thesis is organized as follows, in which the main research works are included.1. The research background and significance have been presented first, then overviewthe research status of reconfigurable manipulator, as well as the fault diagnosis and faulttolerant control. Furthermore, some research problems on fault tolerant control ofreconfigurable manipulator have been pointed out.2. In order to sufficiently reflect the modular property of reconfigurable manipulator,the entire dynamic model of reconfigurable manipulator system derived by Newton-Euleriterative algorithm has been divided into a set of subsystems coupling withinterconnection term. Based on this perspective, the failure models are given for thesubsystem in actuator failure, sensor failure and concurrent failure.3. The chattering phenomenon which is brought by the traditional sliding mode controlcould shorten the working life of the actuator. For the purpose of smoothing the controlsignal, a decentralized controller has been presented by combining the backsteppingtechnique and terminal sliding mode. Meanwhile, it can guarantee the convergence infinite time. However, some difficulties such as control parameter tuning, practicalengineering applications and singularity appear. Hereby, a satisfactory controlperformance achieved by employing integral second order sliding mode to insteadterminal sliding mode.4. For the subsystem in actuator failure, a support vector machine observer isestablished to detect the fault. Then based on the terminal reaching law, a decentralizedfault tolerant controller is designed by integrating the effectiveness factor into thesubsystem dynamic model. In this fault tolerant controller, the uncertainty andinterconnection term are estimated by RBF neural network, and the estimated error iscompensated adaptively. The simulation results show that the actual trajectories can stillfollow the desired trajectories even though the system in failure.5. In the presence of position and velocity failures, an active decentralized fault tolerantcontrol scheme based on signal reconstruction technique is proposed. An adaptive fuzzydecentralized control system is adopted to track the trajectories of joints in the normalworking status. When the position or velocity sensor fault is detected, the reconstructedsignals obtained by numerical integrator and differential tracker respectively are utilizedto substitute the relevant fault signals for feedback control to realize active fault tolerantcontrol. Yet, this scheme can tolerant the failure in position or velocity sensor, thus anactive substituting decentralized fault tolerant control is carried out in the following. Thesensor failure can be transformed into pseudo-actuator scenario by constructing anonlinear transformation for subsystem structure with diffeomorphism theory. Adecentralized sliding mode observer is designed to detect multiple sensor failure in real time, whose output signals are employed to substitute the relevant fault signal to realizethe active fault tolerant control of the reconfigurable manipulator with multisensor failure.6. Focus on the system with the multiple concurrent failures, the decentralized fuzzysliding mode observer is investigated to estimate the fault functions in real time, and thenonlinear term of the subsystem is approximated by fuzzy logic system, the activedecentralized fault tolerant control is achieved along with nonsingular fast terminaltechnique, it improves the control performance when the system in multiple concurrentfailures. All the proposed methods have been investigated with simulations byreconfigurable manipulators with different configurations, and the results illustrate theeffectiveness.7. In many practice applications, the tasks, such as open or close a door usingreconfigurable manipulator, install and replace the panels in space station, search andrescue, as well as rehabilitation, the reconfigurable manipulator endpoint contacts with theoperate environment, the existing research can not undertake the demand. Considering thereconfigurable manipulator system with task constraint, the cascaded control is introduced,then transform the system model into the chained form through diffeomorphism. For thedevelopment of decentralized control, the chained system is considered as a set ofsubsystems interconnected by coupling torques of force control and position control. Byvirtue of backstepping and LMI techniques, a decentralized cascaded controller isdesigned to guarantee the asymptotic stability.In the final, the conclusion and the perspective of future research are given at the end ofthis thesis.
Keywords/Search Tags:Reconfigurable manipulator, Active fault tolerant control, Decentralized control, Decentralized observer, Multiple concurrent failures, Signal reconstruction, Activesubstituting, Task constraint
PDF Full Text Request
Related items