Systems and methods for data set completion for use in chemical process settings
By integrating physics-based models with machine learning models, the systems and methods address the challenges of data set completeness and efficiency in chemical processes, enabling real-time adjustments and improved process efficiency.
Patent Information
- Application Number
- PCT/EP2024/075833
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-09-18
- Filing Date
- 2024-09-17
- Publication Date
- 2025-03-27
AI Technical Summary
Existing methods for generating data sets in chemical processes, such as polymerization operations, face challenges in completeness, interpretability, and efficiency, particularly when relying on machine learning models that lack interpretability and first-principle-based models that require expert knowledge and are time-consuming.
The proposed systems and methods integrate physics-based models with trained machine learning models to generate substantially complete or augmented data sets. This involves preprocessing real-time data, applying it to physics-based models to produce new parameters, and then using these outputs within machine learning models to predict and complete the data set, allowing for faster adjustments to polymerization operation settings and polymeric compositions.
This approach enables faster and more accurate analysis and completion of data sets, allowing for real-time adjustments to polymerization operations, which can lead to improved product quality and process efficiency by optimizing settings such as temperature, pressure, and catalyst usage.