| With the rapid development of modern industrial technology,the safe and reliable operation of the system is an inevitable trend.In most application area,faults are usually intermittent.Intermittent fault is widely present in electronic,mechanical equipment and power systems,and often appears in modern cutting-edge technology fields.In addition,intermittent fault occurs repeatedly and randomly,making it difficult to diagnose and isolate it quickly and accurately.In the research results of intermittent fault,most of the data-driven methods for specific objects focus on whether intermittent fault occurs in the system,which fail to detect all its appearance and disappearance times,while the method based on analytical model can effectively solve this problem.The interval observer,as a powerful supplement to observer theory,can avoid the residual evaluation and threshold selection steps of the traditional observer-based fault diagnosis method,and provide a new idea for the real-time diagnosis of intermittent fault.However,the cooperative constraint and conservative estimation of interval observer are still problems that need to be studied at present.At the same time,for most real systems,it is almost impossible to model a system with fully known parameters and without considering disturbance and noise.In view of these difficulties and problems existing in the research,this paper mainly includes the following contents.For a class of uncertain systems with unknown disturbance and noise,a robust interval observer design method is proposed.The coordination of the dynamic error system is guaranteed by coordinate transformation,and the design of the observer gain is transformed into a Sylvester equation solving problem,avoiding the assumption of the existence of the gain matrix.Then,based on the design of the interval observer structure,taking full advantage of the abundant degrees of freedom in solving the Sylvester equation,an interval observer gain with H_∞performance is designed,and a more compact interval estimation is obtained.In view of the difficulties in intermittent fault detection and adaptive threshold selection,an intermittent fault detection method is proposed based on the upper and lower bound estimations of interval observer.The use of interval observer to study the problem of intermittent fault diagnosis is considered for the first time in this article.The state calculation of interval observer is given at every moment,and the generation of its residuals will not be affected by historical data.The estimation value of interval observer can be directly used for intermittent fault detection,which is a natural adaptive threshold.At the same time,the identification scheme for intermittent fault,transient fault and permanent fault is designed based on sliding time window and counter,which can provide help for subsequent maintenance and treatment.Aiming at the isolation and location problem of the sensor intermittent fault,an intermittent fault isolation method based on multi-interval observers is proposed.It can not only detect sensor intermittent fault,but also locate the fault.Considering that the failure behavior of the sensor will lead to the error control of the controller,which will affect the stability of the closed-loop feedback system,an output feedback fault-tolerant control strategy is proposed.When the intermittent fault occurs,the feedback signal can be reconstructed to ensure the control performance.Finally,a Simulink model of DC motor control system DR300 is built for experimental verification.Multiple robust interval observers are designed for intermittent fault detection and isolation,and the output is reconstructed as a new feedback after the fault isolation,so as to ensure the stable operation of the system. |
