In many industrial sectors, material performance depends on precise control of raw materials and manufacturing parameters. Teams must deal with increasingly complex processes involving a large number of variables in order to guarantee the quality, reliability and efficiency of production.
Traditionally, these processes have been optimised through manual adjustments, modifying one parameter at a time and relying heavily on the experimenter’s experience. This approach, although historically widespread, is time-consuming, resource-intensive and difficult to transpose to new areas where the parameter space becomes too vast to be explored effectively. It also limits the ability to take full advantage of the data generated by industrial systems.
To overcome these limitations, mRT introduced an approach based on machine learning, combined with the continuous use of online measurements. The aim was to implement automated, intelligent exploration of process parameters, capable of operating autonomously once the optimisation campaign had been launched. This approach was developed gradually, building on existing measurement systems and integrating naturally into the industrial environment, while allowing the solution to learn and adapt in real time.
The deployed solution is based on a machine learning algorithm that continuously analyses data from process sensors. The system automatically adjusts parameters, tests the most relevant combinations and interprets the results as it runs. The architecture implemented connects the production equipment directly to the analysis module, creating an autonomous learning loop. This ability to quickly explore a wide range of variables makes it possible to identify the settings that lead to the best performance much more efficiently than a conventional human approach.
The benefits are immediate and measurable. Optimisation campaigns that previously required several months of work can now be carried out in just a few hours. The company gains speed, improves control over its processes and significantly reduces the costs associated with testing phases. The solution also provides better reproducibility of results and facilitates the exploration of new technological areas.
This project is a concrete illustration of the potential of machine learning to transform the optimisation of complex processes. It paves the way for similar applications in many industrial contexts where speed, precision and automation are becoming key factors in competitiveness and innovation.
