Duration: 2024-2025
Funding: Dutch Ministry of Defence (MIND)
About the project
The reliability and availability of systems is becoming more and more important in many sectors of industry, where companies are aiming to reduce maintenance costs and increase system performance or output. The large number of sensors on systems and the improved ways of storing their outputs make artificial Intelligence methods attractive to recognize (abnormal) patterns and trends in the collected data. These patterns and trends can help to detect and predict failures, such that predictive maintenance can be implemented. However, real-world data are often noisy, incomplete and of low quality. Due to these data issues, almost any diagnostic or prognostic method proposed in scientific literature is based on publicly available benchmark datasets which are typically generated with simulation models or basic laboratory set-ups. The goal of the Dutch Prognostic Lab is to close the gap between theoretical methods and real applications.
Research approach
As the control over data from fielded systems is very limited (interruption of production processes is often impossible), test set-ups of real systems in a controlled (laboratory) environment are required. As designing, building and operating such experimental set-ups is rather expensive, companies and institutes will typically focus on set-ups for very specific parts or systems. This still limits the possibilities of method development to only that application.
The Dutch Prognostics Lab therefore aims to develop a standard for data collection and documentation of diagnostic and run-to-failure tests. This will involve the development of a standardized data template, enforcing the complete and accurate documentation of each individual test. This will guarantee the relevance and usefulness of the datasets for diagnostic and prognostic model development.
After defining this standard, the Dutch Prognostic Lab will connect a number of existing experimental test benches at different locations and organizations, and in that way creates a distributed testing facility. By sharing the data generated on these set-ups and coordinating the scenarios to be tested, all participating organizations will have access to a much larger amount of high-quality data than generated by their own test bench alone. In a later stage, also means for sharing data with external partners (worldwide) will be explored, which requires policies for access to the data and a pricing / rewarding structure.
Partners
The following partners are involved in the project:
Team
