On November 30, the Steel Process and Data Connection webinar was held, where we told about the project implemented for Feralpi Group: a concrete case of Predictive Quality applied to the manufacturing sector.

The webinar opened with a key question , “How is a technically complex process like manufacturing impacted by data, especially within the steel industry?”

Certainly the main goal is to improve quality within the production of a large steel plant such as Feralpi, a leader in innovation and sustainability.

The path that has been carried out together with Feralpi Group is modeled in what Regesta has called T.I.R. – Intelligent Transformation with Regesta, a path that aims to guide industrial companies in digital transformation by then also introducing intelligent solutions in the production process.

How does it take place? Starting with a digitization of field data, we connect with the vast amount of data that industrial companies and especially steel mills produce. This data is brought into Cloud platforms for collection and supply chain integration, where it can be analyzed with data science tools.

These steps are preparatory to the actual training of predictive models and Machine Learning systems, and then ideally arrive in the actual process of digital and intelligent transformation with the automation of digital systems governed by artificial intelligence.

The project implemented in Feralpi Group was in the field of quality control and predictive quality. In terms of quality integration, a cloud platform was implemented to collect data from the field that would also integrate Feralpi’s SAP management systems and quality control systems. We then developed real-time analysis applications that would allow this information to be analyzed, and data from the different plants and factories within the supply chain to be integrated. Finally, Machine Learning models were trained, thus predictive models, which  allowed to predict defect situations related to process parameter issues.

To give a concrete example, we created the digital model (digital twin) of continuous casting that, by installing sensors, could signal out-of-process casting. With smaller spies, it was also possible to enable predictive models based on trained Machine Learning engines that would signal anomalies in advance, thus preventing possible defects on the product. This allowed us to get to the point of identifying the exact point on the product where these events take place.

Watch the full webinar.