Assets are complex mixes of complex systems, built from components which, over time, may fail. The ability to quickly and efficiently determine the cause of failures and propose optimum maintenance decisions, while minimizing the need for human intervention is necessary. Thus, for complex assets, much information needs to be captured and mined to assess the overall condition of the whole system. Therefore the integration of asset information is required to get an accurate health assessment of the whole system, and determine the probability of a shutdown or slowdown. Moreover, the data collected are not only huge but often dispersed across independent systems that are difficult to access, fuse and mine due to disparate nature and granularity. If the data from these independent systems are combined into a common correlated data source, this new set of information could add value to the individual data sources by the means of data mining. This paper proposes a knowledge discovery process based on CRISP-DM for failure diagnosis using big data sets. The process is exemplified by applying it on railway infrastructure assets. The proposed framework implies a progress beyond the state of the art in the development of Big Data technologies in the fields of Knowledge Discovery algorithms from heterogeneous data sources, scalable data structures, real-time communications and visualizations techniques.