Modeling and analysis of factors affecting repair effectiveness of repairable systems using Bayesian network

Abstract

Imperfect maintenance modeling and analysis of repairable systems involves an additional parameter called repair effectiveness index (REI) ‘q’, along with shape and scale parameters in generalized renewal process (GRP), which is a measure of the quality of repair or repair effectiveness (RE). The quantitative measure of this parameter as proposed by Kijima through his virtual age models attracted enough attention and is extensively used by the researchers. But, RE could be dependent on many subjective factors and also requires qualitative analysis as well for better understanding of repair effects and performance. Quantitative assessment of RE is necessary but not sufficient to analyze it completely. The paper brings out the limitations of the quantitative assessment of RE and highlights the need for further examining it qualitatively. This paper conducts an extensive study with the help of field experts on selection and analysis of various factors affecting RE and proposes eleven primary factors and their sub-factors (total 55 factors/sub-factors) which affect it the most. After due selection of the factors and sub-factors, the paper then proposes a Bayesian network (BN) to model their dependency on each other and with RE. As a result, the proposed BN model provides the measurable effect of all the selected factors and sub factors on RE in percentage form. The results are demonstrated with the help of two examples inspired by practical industrial applications. The presented work could be extremely useful for industries in undertaking reliability improvement of repairable systems by analyzing their repair quality in detail. The proposed methodology can also be used as fundamental guidelines to develop basic understanding of the repair effectiveness and how it can be improved for a particular system leading to an overall improvement in the reliability of the system.