Building a differentiable AI model of apple tissues

This project aims to develop a fully differentiable, physics-based digital twin of apple tissue that links microstructural, biochemical, and mechanical processes to macroscopic fruit quality outcomes. By integrating artificial intelligence with computational solid mechanics and post-harvest biology, the project will establish a new modelling paradigm for predicting and optimising the mechanical integrity of fleshy fruit during storage and transport.

At its core, the research will combine differentiable physics frameworks (e.g., JAX-FEM) with advanced machine learning approaches such as Fourier Neural Operators to overcome key limitations of traditional finite element modelling. The project will leverage statistical microCT-derived surrogate meshes to represent tissue microstructure, enabling efficient, resolution-invariant simulations. These models will be trained to infer relationships between biochemical markers (e.g., ATR-FTIR measurements) and mechanical properties, thereby providing a mechanistic bridge between cellular-scale processes and bulk fracture behaviour.

A central objective is to solve challenging inverse problems in biomechanics, allowing the model to predict how internal tissue properties evolve under different post-harvest conditions. In particular, the research will focus on simulating the effects of interventions such as controlled atmosphere storage and 1-MCP treatment on cellular degradation and softening processes. By embedding uncertainty quantification within a probabilistic modelling framework, the digital twin will support robust predictions of fruit firmness, failure, and shelf-life.

The expected outcome is a scalable and deployable simulation tool capable of replacing computational bottlenecks associated with conventional modelling approaches. Beyond advancing fundamental understanding of plant tissue biomechanics, the project will deliver practical impact by enabling data-driven optimisation of post-harvest handling and global horticultural supply chains.
Overall, this interdisciplinary project sits at the interface of bioengineering, applied mathematics, and machine learning, offering a unique opportunity to develop next-generation modelling tools for complex biological systems.

• A strong academic background (Honours or Master’s degree) in Bioengineering, Computational Mechanics, Computer Science, Applied Mathematics, or a closely related quantitative discipline.
• Demonstrable proficiency in scientific programming using Python. 
• Solid understanding of continuum mechanics, solid mechanics, or finite element analysis.
• Strong analytical and problem-solving skills, with the ability to work rigorously on complex, multidisciplinary problems.
• Excellent written and verbal communication skills, with the ability to engage across engineering, computational, and biological domains.

MBIE-NSF

Page expires: 17 December 2026