The present work is a study on the computational approach to the health monitoring of a particular class of machinery:​ machinery that operates in an unsteady fashion with little prior information on potential faults. In particular, the objective is to detect the onset of a rapidly progressing serious fault occurring during machine operation. The strategy that is developed is to be applicable to an automated online monitoring system. A general computational method is developed for initial detection of anomalous conditions emanating from this class of machines. The strategic approach that is proposed for the diagnostic routine involves using novelty detection augmented by a separate classification of the machine’s mode of operation from mechanical vibration measurements.

The application domain of the experimental work is the rotating machinery associated with a critical subsystem of an electromechanical excavator employed in the mining industry. This application is a typical example of an unsteadily operating machine for which standard fault detection monitoring techniques have proven ineffective.

A series of laboratory and field experiments on mobile mining equipment were conducted to further define and understand the problem at hand as well as to determine the feasibility of the proposed approach. The field experiments involved the instrumentation of swing machinery of an electromechanical excavator operating in an open pit mine. Vibration and other data from these experiments were analyzed in order to understand the nature of the system’s behaviour, further understand and define the problem and finally validate a portion of the system.

Due to various practical constraints encountered when conducting experimental work in an operational mining environment, portions of the work were carried out in the laboratory. An apparatus was created to simulate the operation and duty of the system in the field. Using this apparatus, experiments were conducted to develop and validate the fault detection routines as well as to validate the overall feasibility of the approach.

The results of the experimental work suggest that the approach is a feasible one. The conclusion discusses the potential implementation of the approach and recommends further work to build on the findings of this research.