Scientific Machine Learning-assisted Model Discovery from Telemetry Data

2026-03-16

AI Summary

Existing dynamic models often oversimplify physical processes, limiting their ability to accurately fit telemetry data and thereby constraining the predictive performance of digital twins. To address this challenge, this work proposes the Dyad model discovery framework, which leverages symbolic regression to extract interpretable mathematical expressions directly from telemetry data. These expressions are used to semi-automatically augment physics-based model equations, while an “engineer-in-the-loop” workflow integrates domain expertise through human review and validation. This approach establishes, for the first time, an AI-assisted closed-loop engineering design process that synergistically combines scientific machine learning with expert knowledge to dynamically refine model structure. Validation on a transport refrigeration unit digital twin demonstrates that the proposed method significantly improves prediction accuracy against real-world operational data.

Technology Category

Cognitive Modeling & Cognitive Systems: Symbolic RepresentationsHumans and AI: Planning and Decision Support for Human-Machine TeamsSearch and Optimization: Sampling/Simulation-based Search

Application Category

Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMsUser Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalizationGraph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

Calibration of dynamic models to data is an important step in building building digital twins of HVAC equipment, thermal loads and control systems. Sometimes, when a model fails to calibrate to data, a possible cause is that the model has made too many sim- plifying assumptions and is missing physics. In this paper we propose a semi-automated approach, called Dyad Model Discovery, that can augment the physical equations of the model with symbolic expressions discovered from the data. We demonstrate this method on a digital twin of a transportation refrigeration unit to improve its predictive performance, trained using telemetry data. An engineer-in-the-loop workflow is proposed, which provides suggestions to the user which can then be accepted or rejected. This is the first AI-assisted engineering design workflow to our knowledge.

Problem

• digital twin
• model calibration
• physics-informed modeling
• telemetry data
• HVAC systems

Innovation

• Scientific Machine Learning
• Model Discovery
• Digital Twin
• Engineer-in-the-loop
• Symbolic Regression