Harmonizing Heterogeneous Diagnostic Data of a Vehicle Fleet for Data-Driven Analytics

Abstract

Data-driven technologies, such as predictive maintenance, become increasingly important to today's automotive industry due to advancements of connected cars and Over-the-Air technologies. A data source that has barely been used in the literature so far is diagnostic data, which is obtained by sending requests to the electronic control units of a vehicle. Diagnostic data can be collected cost-effectively and is already available on a large scale to car manufacturers today. However, the use of diagnostic data is associated with some difficulties. The set of measured variables differs greatly between different vehicles of the same type due to different configurations and therefore differences in the electronic control units. In this contribution, we show how diagnostic data can be harmonized for the use of data-driven modeling. An heuristic three-step procedure is introduced to identify similar measured variables. Finally, our approach is verified on a synthetic data set. Future data-driven technologies are able to use larger and more cost-efficient data sets this way.