| Any system has a failure rate.When faults occur,fault diagnosis and tolerance technology ensures systems operate continuously with the faults.Therefore,the technology is necessary to high reliable systems.Converters modulate input currents of motors,and sensors measure variables and feedback them to control units.They are essential parts for motor systems and also have a high failure rate.To enhance faulttolerance of the converter and the current sensor in switched reluctance motor(SRM)systems,in-depth research was made in this thesis.Fault diagnosis is a necessary pre-step of active fault tolerance.Firstly,for transistors of asymmetric half bridge(AHB)converter,a fault diagnosis method based on current features is proposed.In a phase of the AHB converter,there are four loops that are one positive voltage loop,one negative loop and two zero voltage loops.When faults of transistors occur,conducting loops will change.Different changes of conducting loops correspond to different fault types.Thus,if the conducting loop is known,the faulty transistors are known.The positive voltage loop,the negative loop and the zero voltage loop can be distinguished by current slope.To tell apart the two zero voltage loops,current sensors are reconfigured,making signs of current measurements reverse when different zero voltage loops are conducted.Conditions that may lead to misdiagnosis are illustrated and avoided in online execution.Comparisons of the proposed method and existing relevant methods of fault diagnosis are made according to cost,efficiency,application scope and so on.The proposed fault diagnosis method is demonstrated under all possible fault types of the transistor by simulations and experiments and the results verify its effectiveness.Secondly,topologies of the converters were studied.To enhance the reliability of converters,many fault-tolerant converters have been proposed in literatures.However,design of fault-tolerant topologies is lack of regularity,and a universal design rule has not been proposed.As a kind of switch circuit,the converter can be described by the matrix.Therefore,the design of fault-tolerant converters can be transformed to the matrix design.The fault tolerance of a circuit is mainly determined by its alternative loops.Thus,aiming to construct as many alternative loops as possible under given condition,a design rule of fault-tolerant converters is proposed.Take a converter of the SRM as an example to illustrate the design rule.A AHB converter is chosen as a basic converter.First,the converter is described by a connection matrix,under a condition of adding a switch to each phase,a group of candidate fault-tolerant converters is constructed.To establish a relationship between the matrix and the alternative loop,minus-1 Boolean determinant is defined,then,a following proposition is proven: for a connection matrix of a circuit,its minus-1 Boolean determinant represents sum of loops in the circuit,where each term of the minus-1 Boolean determinant represents a loop of the circuit.According to the proposition,alternative loops of candidate fault-tolerant converters are calculated and analyzed,then the optimal one is selected among them.Subsequently,the fault tolerance performance of the derived converter is tested by simulations and experiments,validating the design rule.Thirdly,to increase fault tolerance of current measuring part of SRM systems,in terms of hardware fault tolerance,with an extra current sensor,a current estimation method is proposed,which can be looked as a redundancy of the traditional current measuring scheme.The current estimation model of the method consists of two equations.The extra current sensor is used to measure sum of phase currents.One of the equations is based on relationship between measurement of the current sensor and phase currents,another is the mechanical equation of motors.Torque,a variable in the mechanical equation,has complex nonlinear relation with phase currents.This makes it difficult to solve the current estimation model online.To deal with the problem,a piecewise modelling method is used to establish a cubic polynomial torque model.Accuracy of the established torque model is better than that of traditional torque models after comparisons.To avoid identifications of moment of inertia and friction coefficient,a parameter of angular speed and acceleration is defined and it can compensate errors of the torque model.On the basis of features of the current estimation model and phase currents of the SRM,the Newton’s method is chosen to solve the current estimation model and a practical stop condition of iteration is given.Accuracy of the current estimation is tested by experiments.Finally,in terms of analytic fault tolerance,a current sensorless control method is proposed.When all of the current sensors are invalid,SRMs can keep operating well by using the control method.Double closed loops of angular speed and acceleration are employed in the control method,and thus,both of the feedback variables are obtained from a position sensor.Firstly,feasibility of the proposed method is analyzed in two points: 1.Its convergence is proven;2.Approximate linear relation between electromagnetic torque and angular acceleration in steady state is proven and equivalence of torque control and angular acceleration control is analyzed further.According to flux features of the SRM and structural features of the converter,a hysteresis control method with three bounds is proposed to track reference angular acceleration.Given both motor efficiency and torque ripple,turn-on and turn-off angle are selected properly.Voltage chopping control(VCC)method,direct torque control(DTC)method and the proposed method are applied to operate the SRM respectively,comparisons of their current,torque and speed waves are made.The results indicate that the proposed method is comparable to the DTC and better than the VCC in control performance.In addition,complex conversion from the current to the electromagnetic torque is needed in DTC.Due to approximate linear relation between electromagnetic torque and angular acceleration,the proposed method does not depend on the electromagnetic torque,and thus has lower computational cost. |