Context-driven decisions for railway maintenance

Railway assets suffer wear and tear during operation. Prognostics can be used to assess the current health of a system and predict its remaining life, based on features that capture the gradual degradation of its operational capabilities. Prognostics are critical to improve safety, plan successful work, schedule maintenance, and reduce maintenance costs and down time. Prognosis is a relatively new area; however, it has become an important part of condition-based maintenance of systems. As there are many prognostic techniques, usage must be tuned to particular applications. Broadly stated, prognostic methods are either data driven, or rule or model based. Each approach has advantages and disadvantages, depending on the hierarchical level of the analysed item; consequently, they are often combined in hybrid applications. A hybrid model can combine some or all model types; thus, more-complete information can be gathered, leading to more-accurate recognition of the impending fault state. However, the amount of information collected from disparate data sources is increasing exponentially and has different natures and granularity; therefore, there is a real need for context engines to establish meaningful data links for further exploration. This approach is especially relevant in railway systems where the maintainer and operator know some of the failure mechanisms, but the sheer complexity of the infrastructure and rolling stock precludes the development of a complete model-based approach. Hybrid models are extremely useful for accurately estimating the remaining useful life (RUL) of railway systems. This paper addresses the process of data aggregation into a contextual awareness hybrid model to obtain RUL values within logical confidence intervals so that the life cycle of railway assets can be managed and optimized.