Advanced Factories is the meeting that annually brings together the most innovative companies specialized in industrial automation, robotics, machine tools and digital manufacturing, along with the technologies that allow boosting industrial competitiveness thanks to new business models, new production processes and the implementation of Industry 4.0. This category rewards the development in the field of artificial intelligence (AI) in industrial and productive processes.

Fersa Bearings (FERSA) during the last years has carried out several research projects for the application of Artificial Intelligence (AI) to improve the operational efficiency of its processes to increase the profitability of its product. The focus of the work is on achieving a reduction in the number of rejects and improving quality.

Along with this strategy, the final implementation of a pilot has been carried out in the Z3 line of the Zaragoza plant, validating different technologies tested since 2018. During this period, fundamental research work has been developed as technological validations in such strategic issues for the company as the digitization of machines and manufacturing assets, unit traceability of the product or the exploitation of process data.

The Artificial Intelligence (AI) tool, FANDANGO, improves the operational efficiency of the processes and makes it possible to predict in real time possible defects and deviations in the manufacturing process and to propose changes in the key control parameters to correct these deviations at the head of the line. This prediction makes it possible to anticipate corrective actions and adjust the production process efficiently. This is achieved thanks to a metamodel that relates the process data with the online quality control measures and the identification of quality defects in the process, and which is integrated by different Machine Learning models that reproduce the real behavior of the machines. FERSA has developed a tool that collects both manufacturing data and online dimensional control data and sends them in real-time to the station models, which generate a series of indicators and adjustment suggestions. These models are automatically retrained with historical data when the quality of their predictions falls below certain thresholds.

The results of the process models and the descriptive manufacturing data have been integrated into a web interface that allows, on the one hand, to have a digital twin of the manufacturing line and, on the other hand, to consult the entire manufacturing history with the data associated with each component of the final bearing reference.

Anticipating deviations in this process and the associated quality defects has a direct impact on the operation of the line, avoiding the need for rework that significantly reduces efficiency. On the other hand, having this support also reduces the percentage of scrap parts by minimizing the number of parts that are out of tolerance at the end of the line and cannot be reworked. FERSA has carried out these developments within the framework of the CIEN 2018 call of the CDTI within the FANDANGO project in collaboration with two research organizations, Tecnalia Research & Innovation and the University of Zaragoza. It has been necessary to count on the collaboration of a multidisciplinary team formed by researchers, technologists and different departments of Fersa such as engineering, systems, quality and production. This collaboration has allowed FERSA to acquire the necessary knowledge for the integration of AI in the manufacturing processes.