A Concept for the Dynamic Adjustment of Maintenance Intervals by Analysing Heterogeneous Data

Abstract

Efficient operation and maintenance processes are important factors to reduce the operating costs of industrial facilities and components. Therefore, both research and industry developed maintenance strategies ranging from fixed time intervals to condition-based activities. However, due to unpredictable events, disturbances and unknown processing times for maintenance activities, many strategies do not meet the requirements of real-world industrial systems. In this paper, a new data-driven concept is presented where data analysis is used to support the dynamic adjustments of maintenance intervals. An overall strategy is developed in which the analysis of data is an integral part for standard maintenance processes, considering their particular workflow and their constraints. The analysed data come from different systems such as Enterprise Resource Planning, Condition Monitoring Systems, and internal service logs of the components or from maintenance activities. The concept encompasses the aligned application of different methods for aggregating these data and for predicting the component’s condition and its remaining useful life. In particular, it is exemplarily shown how the Weibull distribution, the Wiener process, and neural networks are combined to support decisions regarding the dynamic adjustment of the maintenance intervals in industrial facilities. This leads to a better utilisation of components, avoids failures and breakdowns and saves cost. The capability and applicability of these methods is illustrated by applying them to generators of an offshore wind farm.