Fault Detection of Parallel Duty Rotating Machinery Subject to Variable Duty Cycles

This paper describes the development and validation of a strategy for vibration-based online detection of faults in a particular class of machinery commonly found in industry. This machinery is defined by two basic characteristics that preclude it from the application of standard online condition monitoring approaches. The first characteristic is that the machine is in a constant state of transient operation, thus making it difficult to differentiate between changes in transducer measurements that are due to variable duty and those that are due to machine malfunction or fault. The second characteristic is the fact that the machinery involves multiple sub systems that are subject to identical and simultaneous loading, operating environments and duty cycles. An example of such machinery is a system of multiple hydraulic pumps driving a common load. The monitoring strategy presented here employs an anomaly detection scheme together with methods of signal processing and feature extraction that exploit the parallel duty arrangements of this equipment. Preliminary data is collected from a pair or motor-gearbox assemblies used in a material haulage system in the mining industry. Testing and validation was performed on a mechatronic test bench with both healthy and faulted components. Experiments were carried out to investigate how various configurations of the system were able to detect different types of faults. The results indicate that this approach is effective and merits further investigation.