Kateryna Morozovska : Discovering Partially Known ODEs for Chemical Kinetics

Value of the Information & Strength of the Argument

The talk provides valuable insights into applying PINNs to real-world chemical kinetics problems, demonstrating the method’s capability to handle scarce and noisy data. The argumentation is solid, supported by simulation and experimental results, and the speaker clearly explains the methodology and challenges. The use of symbolic regression to discover unknown functions is a notable contribution, though the talk acknowledges limitations in the current implementation.

Scientific Rigor, Source Quality, Title Accuracy

The talk references several peer-reviewed papers and provides a clear description of the methods. The sources are credible and relevant to the topic. The title accurately reflects the content, and the presentation is well-structured. The speaker also mentions the summer school website for further resources. The talk does not include a formal discussion of limitations or potential biases, but overall, the scientific rigor is adequate for a seminar presentation.

The title accurately reflects the content, which focuses on discovering partially known ODEs for chemical kinetics using PINNs.

7/10

The talk presents original research with clear methodology, references to peer-reviewed papers, and practical applications. However, it is a seminar presentation without full peer review, and some claims lack detailed validation.

Markers derived by PSI from the transcript: the creator did not define chapters.

• 00:00Introduction and overview of the talk
• 02:00Background on PINNs and inverse problems
• 05:00Case study 1: Cellulose degradation - problem setup
• 10:00Parameter estimation results for cellulose degradation
• 15:00Discovery of unknown function using symbolic regression
• 20:00Case study 2: Epoxy curing - problem description
• 25:00Results for epoxy curing and future work
• 30:00Conclusion and acknowledgments
• 35:00Q&A session and discussion

• Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations — Reference [1] in the description, foundational paper on PINNs.
• A reassessment of the low temperature thermal degradation of cellulose — Reference [2], discusses cellulose degradation mechanisms.
• Ageing of cellulose in mineral-oil insulated transformers — Reference [3], CIGRE report on transformer insulation aging.
• On the kinetics of degradation of cellulose — Reference [4], introduces Emsley's system of ODEs.
• Symbolic regression analysis — Reference [5], foundational work on symbolic regression.
• Discovering a reaction–diffusion model for Alzheimer’s disease by combining PINNs with symbolic regression — Reference [6], recent work combining PINNs and symbolic regression.
• Discovering Partially Known Ordinary Differential Equations: a Case Study on the Chemical Kinetics of Cellulose Degradation — Reference [7], the speaker's own paper on the cellulose degradation case.
• Summer school on PINNs — Website for the summer school on PINNs, mentioned in the talk.

• Physics-informed neural networks: A deep learning framework for solving forward and inverse problems involving nonlinear partial differential equations — The foundational PINN paper supports the methodology used in the talk.
• Discovering a reaction–diffusion model for Alzheimer’s disease by combining PINNs with symbolic regression — Recent work combining PINNs and symbolic regression, supporting the approach.