CODES

Abaqus implementation of a multi-phase-field fracture model for fiber-reinforced composites based on Puck failure theory. The code distinguishes fiber- and inter-fiber-dominated failure mechanisms using separate phase-field variables and supports both AT1 and AT2 formulations. Example input files are provided for several benchmark fracture problems. 

Authors: Pavan Kumar, Rafael Fleischhacker, Aamir Dean, Raimund Rolfes, Heinz Pettermann

Abaqus implementation of a multi-phase-field fracture model for fiber-reinforced composite laminates. The formulation combines Puck failure theory with separate phase-field variables for fiber- and inter-fiber-dominated damage, enabling simulation of damage initiation, interaction, and crack propagation across different ply orientations. Benchmark input files are included for several laminate fracture problems. 

Authors: Pavan Kumar, Rafael Fleischhacker, Aamir Dean, Heinz Pettermann

Abaqus UMAT/UEL implementation of a phase-field fracture model using locking-free solid-shell finite elements. The code supports homogeneous and layered composite structures and can be used with or without fatigue effects, enabling efficient simulation of crack initiation and propagation through the thickness. 

Authors: Pavan Kumar, Aamir Dean, Jose Reinoso, Heinz Pettermann, Marco Paggi 

A practical guide for computing the material tangent numerically in Abaqus UMAT using finite-difference methods. It explains the role of the consistent Jacobian in implicit finite element analysis and provides a step-by-step implementation for approximating the tangent when an analytical derivation is difficult or unavailable. 

Authors: Aamir Dean

Python/TensorFlow implementation of a Deep Energy Method for electromechanical phase-field fracture with piezoresistive self-sensing. The framework couples neural-network-based fracture simulation with an electrical conductivity model, enabling crack evolution to be monitored through changes in electrical resistance. It also includes adaptive refinement and visualization of damage, electric potential, current density, resistance, and energy evolution. 

Authors: Aamir Dean and Betim Bahtiri

A hybrid framework combining electromechanical phase-field simulations with deep learning for fracture prediction in dielectric nanocomposites. The approach uses simulated phase-field and electrical-potential maps to train image-segmentation models, showing that electrical potential can serve as an effective proxy for detecting and tracking crack propagation. 

Authors: Aamir Dean, Jaykumar Mavani, Betim Bahtiri, Behrouz Arash, Raimund Rolfes

A physics-informed geometry-informed neural operator for predicting displacement and stress fields in parametric DogBone specimens under finite-strain hyperelasticity. The model is trained without finite-element field data, using governing physics, boundary conditions, and geometry information to provide fast full-field predictions across varying specimen shapes. 

Authors: Aamir Dean and Betim Bahtiri 

An energy-based physics-informed geometry-conditioned neural operator framework for predicting displacement fields, stress distributions, and stress concentrations in parametric solid mechanics. The framework employs a variational deep-energy formulation, geometry-conditioned attention mechanisms, and exact enforcement of essential boundary conditions to train without labeled finite-element solution data. Implemented for five specimen families—including DogBone, open-hole, rigid-inclusion, double-notch, and single-notch geometries—it enables full-field predictions for previously unseen geometries and supports systematic evaluation of geometric generalization and stress-concentration accuracy.

Authors: Aamir Dean and Betim Bahtiri