Topic: Causal Latent Space-Based Models for Scientific Learning

Speaker: Joan Borràs-Ferrís 

Date: November 9th at 10:00 a.m. ET 

Abstract

 Industrial processes generate large amounts of data during everyday operation. These records capture variations in raw materials, operating conditions and product quality, offering opportunities to learn about the process. But turning that information into a clear path for solving complex industrial problems remains challenging. This webinar explores how latent variable models, particularly partial least squares (PLS) regression, can support scientific learning from historical production data typical in Industry 4.0. It introduces causal interpretation in the latent space, where proposed changes follow the correlation structure represented by the model, and considers its implications for process improvement. Building on this, the talk will then examine how this framework can support practical decisions: defining multivariate raw material specifications linked to final product quality, expanding these specifications by adjusting process conditions to accommodate greater raw material variability, and comparing suppliers through a latent space-based capability index. Industrial case studies will illustrate the practical use of these methods and the role of process knowledge throughout the analysis. The webinar will close with Dragonet, a software tool developed to make the methods easier to apply. 
Bio: Joan Borràs-Ferrís is a researcher specializing in chemical engineering, applied statistics, and data-driven process modeling in digitalized industrial environments. He holds a PhD in Statistics and Optimization from the Universitat Politècnica de València (UPV) and has developed his career between UPV (Spain) and Université Laval (Canada). He is currently Co-founder and Chief Technology Officer (CTO) at Kensight, a UPV spin-off focused on developing data-driven solutions for industrial processes. In 2024, he received the ENBIS Young Statistician Award for his work in introducing innovative statistical methods and promoting their use in industrial practice. 

Register Now