Proper maintenance is crucial for the reliability, availability, safety, and cost-effectiveness of high tech systems. At the same time, maintenance is very expensive, requiring specialized personnel and equipment. In the Netherlands alone, maintenance of capital goods costs €300 million euros annually, whereas malfunctions due to poor maintenance cost another €80 million, not even mentioning the significant societal burden of defects: accidents, injuries, and malfunctioning of public infrastructure.
The holy grail in maintenance is predictive maintenance (PM): by exploiting recent advances in the Industrial Internet-of-Things, sensor technology, data analytics, and optimization, we can predict failures better and perform just-in-time-maintenance. By repairing or renewing the system just before it fails, maintenance cost are lowered, while the up-time increases.
Despite significant effort in industry and academia, realizing just-in-time maintenance remains challenging. It requires very accurate predictions of the system health and failure times —mispredictions may lead to more, rather than fewer failures— as well as operational ways to turn these predictions into effective and usable maintenance decisions. These challenges encompass multiple phases of the PM work flow, and therefore demand a holistic multidisciplinary approach.
With a truly multidisciplinary consortium, we bring together the required expertise to enforce scientific breakthroughs: we will develop novel combinations of model-driven and data-driven failure prediction techniques, equipping (black box) data analyses with pivotal domain knowledge; multi-scale optimization techniques enabling optimization across different levels of the PM workflow; and integral approach to health predictions and maintenance optimization, which also consider human and organizational factors.
In this way, the PrimaVera project will not only lead to better asset performance and lower cost. We will also lay the foundations for autonomous maintenance, where assets continuously monitor themselves and initiate maintenance decisions themselves.
