A team of researchers led by UT Professor Marielle Stoelinga is set to receive five million euros in funding from the Dutch Research Council (NWO) for the project ‘PrimaVera: Predictive maintenance for very effective asset management’. As part of the project, big data algorithms will be used to enable more effective prediction of failures in infrastructure and production equipment, making maintenance easier to plan. NWO is providing the funding as part of the Dutch National Research Agenda (Nationale Wetenschapsagenda).
So does this mean an end to delayed trains, power outages or machinery failure? The PrimaVera project certainly marks a major step forward towards that objective. Predictive maintenance, or just-in-time maintenance (carried out just before a system fails), can help boost the reliability of infrastructure and production equipment, reducing maintenance costs.
Together with Prof. Tiedo Tinga, UT Professor Marielle Stoelinga is one of the driving forces behind the project. “Predictive maintenance is a very promising technology”, explains Stoelinga. “Everyone wants to achieve better maintenance and fewer failures at lower cost. But there are some difficult issues that need to be resolved in order to make predictive maintenance a reality, and this is what the PrimaVera project aims to do.”
Existing predictive maintenance techniques are only effective on small-scale systems and are difficult to upscale. Choices made at one place in the chain can have an important impact on other processes elsewhere. The choice of a specific type of sensor or measurement influences the type of predictions that are possible and with that the quality of the predictions. This is why cross-level optimization methods are being developed as part of the PrimaVera project.
Entire chain of maintenance
According to Stoelinga, what makes this project unique is the fact that it tackles the entire chain of maintenance and a multidisciplinary team is working on it. “We will start with better sensors that can do more effective measurements. We will then process the raw data into meaningful information that we can use to make predictions about the condition of a system and its tendency to failure. We will use that condition as our basis for determining when maintenance is required. This is made particularly complex by the fact that you obviously want to cluster your maintenance as effectively as possible to prevent the need to take a machine or section of railway out of action for maintenance on two consecutive occasions.”
Stoelinga and her fellow researchers will also be working on much better and scalable forecasting methods for failures. “The success of any predictive maintenance always depends on the quality of the forecasting methods. Inaccurate predictions on the condition of a bridge or machine can result in more rather than fewer failures. This is what makes data science such an essential part of this project.”
In the researchers’ view, predictive maintenance also calls for a change of mindset in operational processes. If algorithms are to replace maintenance experts in deciding what maintenance needs to be done, this change in mindset will be essential. “Very little is known about how maintenance staff and planners will deal with the recommendations that emerge from big data algorithms. This is uncharted territory and we look forward to exploring it.”
The PrimaVera project is being run by a wide-ranging consortium consisting of a multidisciplinary team of scientists and companies. Stoelinga: “Maintenance not only involves technical factors, such as the quality of the forecasts and the planning algorithms, but also human factors. If you fail to take account of how people deal with maintenance decisions and interpret information, all of your efforts may be pointless.”
The NWA-ORC call for proposals is part of the NWA programme that NWO is running on behalf of the Ministry of Education and Culture & Science (OCW). Currently, the NWO programme for the Dutch National Research Agenda is based around four key themes that together aim to achieve its goals and ambitions.
PrimaVera is part of Research along Routes by Consortia (ORC). It encourages independent research by means of open calls for long-term research projects involving wide-ranging, interdisciplinary and cross-disciplinary consortia on scientific and socially relevant subjects that will clearly benefit from a broad national approach.
Marielle Stoelinga is Professor in the Formal Methods and Tools department of the Faculty of Electrical Engineering, Mathematics & Computer Science (EWI).
Also involved are the Pervasive Systems group (Dr Nirvana Meratnia) and the Dynamics Based Maintenance (Prof. Tiedo Tinga) and Maintenance Engineering (prof. Leo van Dongen) departments of the Faculty of Engineering Technology (ET).