Detecção e diagnóstico de falhas em motores de indução

• Ano de defesa: 2005
• Autor(a) principal: Lane Maria Rabelo Baccarini
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Minas Gerais
• Palavras-chave em Português: Maquinas eletricas de indução; Engenharia elétrica; Localização de falhas (Engenharia)
• Link de acesso: https://hdl.handle.net/1843/HSAA-6MJPHW

This main purpose of this work is to design and implement a system to detect and diagnose electrical faults (stator inter-turn short circuit and broken rotor bars) and mechanical faults (unbalance and shaft misalignments) for three phase induction machines. This approach starts by obtaining the best patterns for fault detection and symmetrical models for the machine, in order to simulate electrical and mechanical faults. Further simulations are carried out to check the validation of these computational models. Then, a strategy to detect and diagnose faults was developed and tested on the rig. A method that utilizes the torque residual is proposed to diagnose and identify rotor broken bars. The residual is determined using the machine flux obtained through the machine dynamic equations, which are rotor's faults sensitive, and also through a sliding mode observer for unknown inputs, that are rotor's faults robust. To determine an occurrence of inter-turn short circuit, it is proposed the monitoring of the negative sequence impedance values. The method has shown to be robust to inherent problems occurring on the measurement system and is capable of detecting short-circuits in the initial stages. The experimental technique for mechanical imperfections detection is based on the machine stator's current spectrum analysis. However, this technique is not useful to classify the type of these imperfections, and further information is required. In this work, it is proposed the use of artificial neural networks and support vector machines for the mechanical imperfections classification. The vibration signals are the inputs for the networks. Experimental data obtained from previous works were utilized to validate this classification process. The rig is composed by a three phase induction machine, a DC generator used as load, transducers and signal conditioner, data acquisition boards and a personal computer. The machine was designed to allow non-destructive tests of broken bars and short-circuits among stator inter-turn. The computational implementations for the detection and fault diagnosis techniques run in a LabView environment. The simulation and experimental results show that the proposed approach appears to be a suitable tool in on-line induction machine fault detection and diagnosis, and can also to become an interesting tool for predictive engineering maintenance.