FUSION Issue 1 2026 DIGITAL SINGLE PAGES - Flipbook - Page 9
FUTURE OF
MODELLING
“ Use physics
to produce
trustworthy
data, then
apply AI where
it adds value ”
Prof. Wenbin Yu,
Purdue University
From physics to
intelligence: the
mechanics of the
structure genome
can act as instant “surrogate
models,” reproducing complex
multiscale analyses in seconds
rather than hours.
At the heart of Professor Yu’s
research lies a framework known
as the Mechanics of the Structure
Genome (MSG). This multiscale
modelling approach takes a
top-down view of structures
such as beams, plates, and shells,
identifying a minimal “structure
gene” that captures essential
physical behaviour. By mapping
material properties into this
reduced description, engineers
can predict performance
accurately while avoiding
unnecessary computational cost.
INSIGHT: Machine learning can
uncover unknown relationships
within materials, inferring hard-tomeasure behaviours from limited
experimental data.
MSG has proven versatile – capable
of modelling everything from
classic laminates to advanced
metamaterials. By linking structural
mechanics with micromechanics, it
bridges the gap between material
innovation and system-level
design, a crucial step for modern
composite structures used in
aerospace applications.
Why AI matters
Professor Yu identifies four
major constraints on traditional
structural modelling: computing
power, knowledge gaps,
integration challenges, and
accessibility. AI provides tools to
overcome each of these.
SPEED: Neural networks trained
on high-quality simulation data
INTEGRATION: Large language
models can orchestrate multiple
simulation tools through naturallanguage interfaces, helping
engineers streamline crossdisciplinary workflows.
ACCESSIBILITY: By simplifying
advanced computational processes,
AI makes high-fidelity modelling
achievable even for teams without
specialist coding expertise.
Yu’s team has demonstrated that
deep-learning surrogates can
replicate thermal-conductivity
simulations of woven composites
almost instantly – a leap that opens
new opportunities for design
iteration, uncertainty quantification,
and digital prototyping.
The rise of hybrid
modelling
The next frontier is not pure AI, but
hybrid physics–AI modelling. This
approach combines the predictive
rigour of traditional mechanics with
the adaptability of data-driven
methods. In practice, that means
using physics-based models to
generate reliable training data, then
deploying AI to accelerate and
extend those insights.
This synergy is already shaping
tools such as Composites AI,
an emerging expert system
that can interpret models,
automate workflows, and even
perform optimisation studies
autonomously. It represents a
step towards intelligent simulation
environments capable of learning
from past analyses while maintaining
physical accuracy – a development
that could redefine design
efficiency in the aerospace sector.
AT A GLANCE:
AI IN AEROSPACE
MODELLING
Accelerates
simulation
workflows
Balancing rigour and
innovation
Despite rapid progress, Professor
Yu stresses that physics must
remain the foundation. “AI is
pragmatic,” he notes. “Use physics
to produce trustworthy data, and
then apply AI where it adds value.”
This pragmatic balance between
theory and computation ensures
that future models retain scientific
rigour while embracing the speed
and flexibility that AI enables.
For aerospace engineers, the
message is clear: the future lies in
partnership, not replacement. By
uniting data-driven intelligence
with decades of physical
understanding, the industry can
achieve breakthroughs in safety,
efficiency, and innovation –
ushering in a new paradigm of
intelligent structural design.
Based on insights shared by Professor Wenbin Yu, Purdue University, during the AIAA Journal Seminar Series
webinar “Shifting the Paradigm of Aerospace Structural Modelling with AI,” October 2025.
Improves data
integration
Enables hybrid
physics–AI
design
Expands
accessibility for
engineers
dtc-uk.com
8
