The potential of fleet based analytics for data-driven operation and maintenance optimisation

More and more companies are starting proactive or predictive maintenance projects and they are looking how to optimise their operations thanks to the enormous amount of data they collect. However, these companies mainly focus on the individual machine level, leaving a huge potential for fleet-based analytics untapped.  

That is why on 13 December 2017 in the framework of the HYMOP project, Sirris together with the other project partners Vrije Universiteit Brussel, imec, University of Antwerp and KU Leuven organised the second edition of the seminar “Fleet-based analytics for data-driven operation and maintenance optimisation”. With about 70 participants from both academia and industry, the event shows the growing attractiveness and interest in fleet-based analytics for data-driven operation and maintenance optimisation.   

A set of (nearly-)identical machines is defined as a fleet. The analysis of data collected on such a fleet represents a real potential for optimising operation and maintenance of those machines:  

  • Data and knowledge leverage across the fleet: analysing data at fleet level will allow you to build more representative datasets, as you will be able to combine data coming from several machines. The dataset will be covering a richer set of operational contexts and device types compared to a dataset including the data from one single machine. Your dataset will potentially also include more occurrences of rare events. Finally, you will be able to construct complete and qualitative datasets: being able to detect and remove outliers and impute missing values. Having a representative and complete dataset is the first step towards a successful data analysis!

  • Performance benchmarking across the fleet and identification of cross-fleet performance degradation and underperformance: analysing data at fleet level will allow you to develop a reliable real-life reference baseline for different contexts. In other words, when a machine is behaving normally in its normal operational conditions. You can then use this reference baseline to determine abnormal behaviour. In turn, this serves as the basis to predict failures and detect anomalies.

  • Composite modelling and simulation: analysing the data at fleet level allows you to break down your problem into parts that can be modelled separately. Considering the individual components, the machine and the different assets separately will reduce the complexity of your analysis and allow specialisation in different contexts. Subsequently, the individual components can be recomposed in an overall model by means of so-called hyper-modelling techniques, which is a non-trivial research task. 

Interested in discovering the potential of fleet based analytics in your specific context? You would like to have further information on the HYMOP project? Then get in touch and e-mail us