Induction motor parameters estimation and faults diagnosis using optimisation algorithms.

Abstract

Induction motors are the most widespread rotating electric machines in industry due to their efficient and cost-effective performance. Induction motors are used to mainly operate at the constant speed since the rotor speed depends on the supply frequency. The development of power electronic devices and converter technologies has revolutionized the adjustable-speed induction motor drives. For most high-performance control methods, the effective motor control requires precise knowledge of the motor’s parameters, which are usually obtained from manufacturers. However, the manufacturers describe these parameters under starting or full-loading condition only, instead of the normal operating conditions. It is well known that motor parameters are influenced by not only the load level but also environmental factors, such as temperature, humidity and lubricant viscosity. The first part of the thesis describes the application of the sparse grid optimisation method in solving the induction motor parameter estimation problem. Kernel of the method is the efficient search in minimising the cost function on the grid created by using the Hyperbolic Cross Points (HCPs). The cost function quantifies the difference between simulation results and measurement results. Within model reference adaptive system (MRAS) framework, a global optimisation algorithm, HCP algorithm (HCPA), runs the motor model and finds the best parameters to minimise the value of the cost function. Since the proposed method requires only voltage and current signals, it is compatible with sensorless control methods, which have the benefits of increasing system reliability and reducing cost. The presented experimental validation shows that the relative errors of the estimated parameter values are less than 10% under various load levels. The estimated parameters can be further refined by applying local search method using global search result as a start point. On the other hand, an induction motor failure results in severe damage not only to the motor itself but also to motor related equipment devices in an industrial plant. Consequently, motor condition monitoring and fault diagnosis are of great necessity to detect motor faults at the early stage in order to reduce unscheduled downtime, repair costs, and increase life span of machines. Emergence of a fault will cause a gradual drift of fault-related characteristic model parameters. Therefore, a generic method to detect motor faults developed in this research is based on monitoring these parameters. In the second part of this thesis, the proposed parameter estimation technique based on the sparse grid optimisation is utilised to detect stator short circuit faults by monitoring two characteristic parameters: fault level and fault location. Experimental results show that the proposed diagnosis method is capable of detecting stator short circuit fault levels and location under different load conditions. Compared to the genetic algorithm, the HCPA shows improved robustness in the case of unbalanced voltage supply. This non-invasive diagnosis method only needs a short length of voltage and current signals recorded at switch board without disrupting the machine’s normal operation. The third part of this thesis demonstrates a multi-motor condition monitoring scheme which can substantially reduce implementation cost for some industrial plants. The proposed multi-motor condition monitoring scheme builds on top of the technology implemented in modern Intelligent Electronic Devices (IEDs) for motor protection and control. The backbone of this scheme is the broadly accepted Ethernet technology and the IEC 61850 communication standard. Due to the widespread use of IEC 61850 in various industries, cost of the technology is significantly reduced while reliability has been improved. Based on the proposed systems, various applications can be developed to achieve remote condition monitoring of induction motors.