DiagRAMS revolutionising predictive maintenance
© Inria Photo Raphaël de Bengy
DiagRAMS is a project primarily aimed at putting predictive maintenance tools in the hands of those that need them the most. It is also a remarkable entrepreneurial adventure, one which came about as a result of a meeting between two partners from the InriaTECH platform and the Modal project team from the Inria Lille - Nord Europe centre. Let’s take a look at it in more detail.
With factories becoming increasingly automated, the ability to deal with breakdowns and malfunctions of all kinds, even those that might be less apparent, has become a constant priority for manufacturers. Just a few years ago, there were only three options available to them: they opted for corrective maintenance (repairing parts each time a breakdown was identified), preventive maintenance, with work carried out on a regular basis in order to ensure that everything was running smoothly, or a combination of the two. With the arrival of the Internet of Things and the rise in the number of connected sensors, a new path has opened up: predictive maintenance. But what does this involve? Detecting the probability of malfunctions in order to be able to predict breakdowns or accidents and to optimise maintenance and repair strategies by moving towards a “just in time” model.
Helping to cut maintenance costs
Companies find this concept particularly attractive. Last year, a study carried out by McKinsey estimated that predictive maintenance should enable manufacturers to save as much as 630 billion dollars between now and 2025 by reducing maintenance costs by between 10% and 40%, by cutting the number of breakdowns in half and by increasing the lifespan of machines. For the time being, however, predictive maintenance services are not always able to meet expectations, particularly when it comes to accessing data that is often highly complex, difficult to read and not tailored to meet the requirements of professionals who are on the go and short on time.
From InriaTECH to company creation
This is where DiagRAMS Technologies came in, a start-up in the process of being launched and which is centred around research undertaken by the Modal* project team at the Inria centre in Lille. But what sets this particular “spin-off” apart in an Inria context is that it was primarily the brainchild of two individuals in charge of partnerships and innovation projects within the InriaTECH platform. With 10 years of experience in the field of industrial machinery for use in the agri-food sector, Jean-François Bouin will occupy the role of president. He will be responsible for running the company, alongside Margot Corréard, whose background is in project management, marketing and communication. “We had supported Modal through partnerships formed with Arcelor Mittal and Alstom, and we felt the tools that they had developed were highly promising and very much in keeping with current industrial needs”, explains Jean-François Bouin. “Margot and myself both have a real business streak, and we enjoy working together, so we decided to go for it, with support from Modal, who were delighted to see their work come to life on the market. Three of the researchers are now our scientific advisors.”
A solution that doesn't involve adding new sensors
“What makes our solution original is that we use existing sensors (process sensors natively installed on the machines), at a time when practically all of the other software options available on the market require additional devices to be fitted. This is more expensive and results in a build-up of data that is not always readable”, explains Margot Corréard,
The machine learning algorithms developed by the research team can be used to process raw industrial data (energy consumption, vibration analysis, temperature curves, etc.) without the need for preprocessing to reduce the information contained to the original signals. This data is compared to “signatures” of breakdowns modelled using predictive algorithms based on the equipment's history of malfunctioning. When a machine stops functioning normally and moves towards a critical state, a real-time alert is launched in order to begin a maintenance operation capable of remedying the malfunction and preventing a total breakdown. “Detection becomes better over time, and it becomes possible to identify problems at an earlier stage, problems that would still be totally undetectable using conventional detection tools - all with a greater degree of precision and without generating false positives”, adds Jean-François Bouin.
This technology can be used to explain the prediction results to users, doing away with the black box effect - manufacturers need to understand what is happening in order to take efficient action. DiagRAMS Technologies work on precise and intuitive visualisation for teams working in the field, whether these are engineers or maintenance staff.
Setting off on an adventure
Although the start-up does not yet exist as a legal entity, the team has already been made a stakeholder in its first major project, with a partnership bringing together Modal, Nokia Germany and Apsys Airbus within the context of the EU programme “EIT Digital - 4.0 Industry”. The aim of this project is to develop data analysis methods for use in predictive maintenance, with a particular focus on analysing the causes of errors. “With this in mind, we recruited a data scientist who will work exclusively on the EIT project”, explains Margot Corréard. We also launched a second recruitment drive to finalise the development of our software solution, which we will be building jointly with our industrial partners before the launch of the subscription-based service, expected sometime in 2020.”
*Modal is an Inria project team that is a joint undertaking with the Paul Painlevé laboratory (CNRS, the University of Lille).