Developing a framework for generating realistic, but not real, synthetic maintenance records

Informed decisions for life-extension of complex long-life assets require knowledge of the current state of the asset as well as understanding of its maintenance and failure history. A lack of access to sufficient and reliable data for method validation, such as appropriately detailed maintenance logs, inhibits research into better-informed decisions for life-extension of long-life capital-intensive assets. Researchers must negotiate industry partnerships and overcome barriers to data access due to commercial sensitivities. This paper proposes a conceptual framework for generating realistic, but not necessarily real, synthetic maintenance records. Maintenance logs have strong structural synergies with electronic health records. We adapt and map methods for generating synthetic health records using only publicly available data sources to create synthetic maintenance records. Facet strings are used to construct a probabilistic framework for developing brief free notes. The generation of realistic synthetic maintenance records can be achieved using publicly available data, supported by expert engineering knowledge of the design and maintenance of the system.