Resource Conscious Fault Classifiers

As condition monitoring of systems continues to grow in both complexity and application an overabundance of data is amassed and computational capabilities are unable to keep abreast of the subsequent processing requirements. A means of establishing computable prognostic models to accurately reflect process condition, whilst alleviating computational burdens, is essential. This is achievable by quantifying a parameter’s explanatory power and restricting input of redundant information to modelling algorithms.

This paper will focus on methods of reducing input parameter volume without sacrificing model efficiency. Underlying parameter variances are assessed by deterministic power to inform selection. A number of approaches are discussed including variable clustering to reorganise the harmonics of common diagnostic features into a smaller number of heterogeneous groups with representatives from each selected to reflect conditions with minimal information redundancy. Furthermore, confirmatory factor analysis enables parameters with superior deterministic power to be identified alongside complimentary, uncorrelated, variables. Variables with little explanatory capacity can be eliminated leading to further variable reductions. Theoretical techniques are demonstrated via application to predictive maintenance of industrial rotating machinery and large-scale industrial processes.