Machine Performance and Condition Monitoring Using Motor Operating Parameters Through Artificial Intelligence Techniques

Condition monitoring (CM) of gearboxes is a necessary activity due to the crucial importance of gearboxes in power transmission in most industrial applications. There has long been pressure to improve measuring techniques and develop analytical tools for early fault detection in gearboxes.
This thesis develops new gearbox monitoring methods by demonstrating that operating parameters (static data) obtained from machine control processes can be used, rather than parameters obtained from vibration and acoustic measurements. Such a development has important implications for the future of CM techniques because it could greatly simplify the measurement process.
To monitor the gearbox under different operating and fault conditions based on the static data, three artificial intelligence (AI) techniques: a general regression neural network (GRNN), a back propagation neural network (BPNN), and an adaptive neuro-fuzzy inference system (ANFIS) have been used successfully to capture nonlinear variations of the electric motor current and control parameters such as load settings and temperatures.
The three AI systems are taught the expected values of current; load and temperature for the gearbox in a given condition, and then measured values obtained from the gearbox with a known fault introduced are assessed by each of the AI models to indicate the presence of this abnormal condition. The experimental results show that each of GRNN, BPNN and ANFIS are adequate and are able to serve as an effective tool for gearbox condition monitoring and fault detection.
The main contributions of this study is to examine the performance of a model based condition monitoring approach by using just operating parameters for fault detection in a two stage gearbox. A model for current prediction is developed using an ANFIS, GRNN and BPNN which captures the complicated inter-relations between measured variables, and uses direct comparison between the measured and predicted values for fault detection.