CIMPLO - Cross-Industry Predictive Maintenance Optimization Platform

Almost every enterprise has to deal with it: maintenance of machines and infrastructure. Traditional maintenance concepts rely on a 'fixed interval approach', taking into account a considerable safety margin. As a result, maintenance almost always occurs either too early or, in worse cases, too late. This arguably makes it one of the most inefficient and simultaneously critical activities in the industry. The CIMPLO platform aims to optimize the perfect moment (both cost and safety-wise) to perform maintenance on engines.

Project Partners

CIMPLO is a project of the Leiden Institute of Advanced Computer Science (LIACS), Centrum voor Wiskunde en Informatica (CWI), and industrial partners KLM Air France (related to aircraft engines) and Honda Research Institute Europe (related to electric vehicle fleets).

We spoke in the spring of 2024 with Dr. Niki van Stein, a researcher at the Natural Computing Group of LIACS and manager of the applied data science lab. She obtained her Ph.D. in Computer Science from Leiden University in 2018. Niki's research interests lie in eXplainable AI for automated machine learning, global (Bayesian) optimization, and neural architecture search. She is primarily involved in research with direct applications in industry, such as predictive maintenance, optimization of automotive and ship design, and planning optimization. The project team was advised by a user committee. Participants in this committee include other industry partners such as Volkswagen, Damen, Tata Steel, and DAF Trucks.

Objective and Outcome

The CIMPLO project aims to develop a cross-sector predictive maintenance optimization platform that meets the real requirements for dynamic, scalable maintenance planning with multiple criteria. To realize the full business benefits in terms of safety, time, and financial savings, the CIMPLO project combines predictive maintenance with dynamic multi-objective planning, allowing maintenance events and required assets to be dynamically (re)scheduled. Niki proudly states: "A significant outcome of the project is that we were able to identify with a reasonably explainable model which sensors were good at measuring the condition of an aircraft engine and this was subsequently confirmed by a domain expert. So, not only could we find a model that performed well, but the domain experts also saw that it was logical and made sense. I am also proud of the fact that we delivered a comprehensive package - a real product - that partners could immediately use in a testing environment."

Sensors and Data

Nowadays, all components of industrial machines are monitored with numerous sensors. Together, they provide a vast amount of data, which is potentially useful for maintenance. For project partners KLM and Honda, examples of these machines are aircraft and passenger car engines, respectively. Built-in sensors provide the ability to model the degradation of components in these types of engines. The data can be used for predictive modeling and dynamic planning.

A challenge in the project was the availability of well-labeled data. Niki explains: "Good data is not always available; fortunately, airplanes do not break down so quickly. So, we often deal with a large imbalance in the data. This limits us quite a bit in terms of methods, especially in the area of sensor data. Obtaining data from different industries is also not easy - partly due to confidentiality - so it's difficult to test methods on different user groups. We also had to adjust methods midway because we could receive additional data by streaming more flight data."


After CIMPLO, the researchers looked at the feasibility of bringing this to the market. A feasibility study to determine if the market is ready for this and what the necessary steps are to start this, for example, as a spin-off company. This also involves looking at possible obstacles and the required funding for such a step. "The advice that came out of it is to first further develop the software in the current XAIPre project, and then to see with one of the partners in the project how we can integrate this into their production environment," says Niki.

Follow-up Project XAIPre

The XAIPre project aims to develop Explainable Predictive Maintenance (XPdM) algorithms that not only provide engineers with a prediction but also:

  1. give a risk analysis if maintenance is delayed
  2. provide the criteria or indicators used by the algorithm to make that analysis.

By providing more insight into the state of the machine, technicians gain more power and control over their maintenance process. XAIPre focuses on maintenance in the maritime industry, with Heerema Marine Contractors as a project partner. Niki adds: "We also try to design the methods developed during XAIPre as generalistic as possible, so that other industries can also make use of them."

Further Development

New NWO applications are being prepared together with other industries. This year, a PhD student started on this topic, but for the maintenance of locomotives.

More information about CIMPLO and follow-up projects can be found at

Gerelateerde content

Bekijk hieronder de calls gerelateerd aan het thema CIMPLO - Cross-Industry Predictive Maintenance Optimization Platform