Building Confidence in Model-Based Predictions

Quantifying and reducing uncertainty in physics- and data-driven models

This position paper presents DNV’s perspective on building confidence in model-based predictions by combining physics-based and data-driven modelling with rigorous uncertainty quantification.

Request a copy

How can we build confidence in model-based predictions used in safety-critical systems?

DNV’s latest position paper explores this question in depth, offering a comprehensive framework for understanding, quantifying, and reducing uncertainty in predictive models. The publication outlines:

  • Why combining physics-based and data-driven models is essential for robust, reliable predictions.
  • How uncertainty quantification (UQ) enables better decision-making by distinguishing between randomness and knowledge gaps.
  • Strategic methods for reducing uncertainty, including optimal design of experiments and active learning.
  • Practical tools such as probabilistic machine learning, surrogate modelling, and physics-informed neural networks.
  • Real-world applications in pipeline reliability and offshore wind turbine design.

Download now to discover how uncertainty-aware modelling supports safer, more cost-effective engineering across industries – from energy and maritime to healthcare and autonomous systems.