HEALTH: Hazard Estimation and Analysis of Lifelong Truck Histories

Important results from the project

Unplanned downtime can be avoided by accurate prediction of the failure through continuously monitoring of vehicles’ health status. However, to reveal patterns behind failures in a system as complex as a modern truck, new methods for analysing the data need to be developed. The HEALTH project created sequence models capturing the lifelong history of a truck, and used it to explain the relations between different events such as failures, repairs, fault codes - leading to better maintenance.

Expected long term effects

HEALTH enhanced current uptime promise of Volvo Trucks by deploying data-driven predictive maintenance solutions in production environment. Several Machine Learning methods such as HMM, LSTMs, GANs for representing lifelong histories of trucks are used to identify vehicles that are likely to fail soon, and corrective actions are suggested based on the specification of the component. Overall effect of the project is minimizing the downtime for the customer, i.e. minimizing cost and maximizing uptime.

Approach and implementation

The HEALTH was a two years project, starting in October 2017. The work is carried out in close collaboration between Volvo Trucks Aftermarket and Halmstad University. The project is contributed in data aggregation, fully and partially observable sequence modelling, causal analysis. Implementation includes research and development of new machine learning methods, their deployment, and finally evaluation in real business settings. Some of the methods are integrated to Volvo production environment and some initiated further investigations at Volvo.