(183b) Advancements in Machine Learning for Chemical Process Analytics: The Eigen IC Analyzer and ML-Assisted Raman Spectroscopy

The Eigen IC Analyzer employs transformer-based models to predict properties of products from distillation columns with high accuracy. These predictions are essential for the effective functioning of closed-loop control strategies. The discussion will cover the architecture of these ML models, emphasizing their role in real-time process monitoring and the substantial improvements observed in production environments.

The recent incorporation of ML-assisted Raman spectroscopy represents a significant advancement, enhancing the onstream analysis of chemical compositions. Integrating ML models with Raman spectroscopy enables the automatic interpretation of complex spectral data, crucial for close-loop process control. This presentation will explore the experimental validation of these ML models and discuss their potential impact on pilot projects aimed at commercial applications.

An in-depth discussion on the deployment of the Eigen IC Analyzer within an NGL fractionation process will be presented, focusing on how ML inferential models are integrated at various points—such as the deethanizer, depropanizer, debutanizer, and deisobutanizer—to predict and control the composition of product streams effectively. This integration highlights the Analyzer’s role in enhancing the precision and efficiency of the fractionation process, proving essential for continuous process optimization.

Furthermore, the benefits of the Eigen IC Analyzer and ML-assisted Raman spectroscopy in enhancing closed-loop control and providing critical inputs as Control Variables (CVs) to Model Predictive Control (MPC) systems will be elaborated. These technologies improve the responsiveness of MPC systems by providing faster and more accurate predictions of process variables, enabling more dynamic adjustments to process conditions. The capacity of these ML tools to handle large volumes of data and adapt to process changes underscores their value in managing complex, multivariable control scenarios within chemical processing plants.

In conclusion, this presentation will outline the integration of the Eigen IC Analyzer and its extension to ML-assisted Raman spectroscopy, demonstrating significant advancements in ML applications in chemical engineering. These technologies offer substantial improvements in process control, efficiency, and safety, addressing the complex challenges of modern industrial processes. The final remarks will stress the importance of continuous innovation and adaptation in leveraging AI technologies to meet modern industrial challenges and improve overall process efficiencies.