DIADICA Synthetic Core
Modelling capabilities
Sixteen domains of computational mechanics, probability and physics-informed artificial intelligence. The methods of each domain are implemented with in-house Python code on open tools and applied in projects and research.
Video · 53 s
Modelling capabilities in one minute
Four verification problems with a known solution, solved with in-house code: a beam optimized by SIMP, with initial compliance equal to the solid beam's divided by 0.5³; a resistance and load-effect limit state, with FORM and four million Monte Carlo simulations against the exact probability of failure; a centre-cracked plate, with the J-integral stress intensity factor within 0.25 % of the reference solution; and a manufactured solution with second-order convergence and Richardson extrapolation.
Domains
Methods by domain
01
Continuum mechanics
- Stress and strain tensors
- Linear and nonlinear elasticity
- Viscoelasticity
- Green's functions
- Beams and piles on variable-stiffness elastic foundations
02
Plasticity and constitutive models
- Theory of plasticity
- Computational plasticity and return mapping
- Constitutive models for steel and concrete
- Soil models: Mohr-Coulomb and Hardening Soil
03
Finite element method
- Linear and nonlinear formulation
- Geometric nonlinearity and large displacements
- Second-order analysis and buckling
- Computational geomechanics
- Mesh convergence
04
Boundary element method
- BEM formulation
- Boundary singularities
- Application to soil media
- Hierarchical BEM with Barnes-Hut trees
05
Fracture mechanics and damage
- Linear elastic and elastoplastic fracture
- Nonlocal continuum damage
- Fracture energy regularization
- Meshless methods
06
Nonlinear structural analysis
- Matrix stiffness method
- Fibre-section frame elements
- Nonlinear static analysis and capacity spectrum
- Moving loads and influence lines
07
Structural dynamics
- Modal and time-history analysis
- Damping and substitute structure
- Seismic isolation and dampers
- Incremental dynamic analysis
08
Performance-based earthquake engineering
- Probabilistic seismic hazard and ground motion models
- Direct displacement-based design
- Loss estimation (FEMA P-58)
- Pounding, torsion and irregularity
- Wind
09
Computational geotechnics
- Soil-structure interaction
- Equivalent linear and nonlinear site response
- Modulus and damping degradation
- Liquefaction
- Slope stability by strength reduction
10
Fluid mechanics and hydraulics
- Computational fluid dynamics
- Large eddy simulation (LES)
- OpenFOAM
- Advection, diffusion and transport
- Hydrological and hydraulic modelling
11
Optimization
- Topology optimization (SIMP)
- Multiscale topology optimization
- Bayesian optimization
- Design of experiments
12
Numerical methods and computational geometry
- Numerical solution of partial differential equations
- Finite differences
- Time integration
- Computational geometry algorithms
13
Probability and reliability
- Structural and system reliability
- Monte Carlo and quasi-Monte Carlo
- Subset simulation
- Global sensitivity: Sobol indices and Morris method
14
Uncertainty quantification
- Bayesian inference and inverse problems
- Model calibration (Kennedy-O'Hagan)
- Aleatory and epistemic uncertainty
- Multifidelity models
15
Physics-informed artificial intelligence
- Physics-informed neural networks (PINN)
- DeepONet and FNO neural operators
- Gaussian processes and polynomial chaos
- Reduced-order models
- Quantized models for field devices
16
Verification and credibility
- Verification and validation of computational models
- Credibility per NASA-STD-7009
- Manufactured solutions and Richardson extrapolation
- Conformal prediction and neural network calibration
Verification
Verification and validation
Numerical verification. The model is verified against closed-form solutions, manufactured solutions or published benchmark cases, and the convergence of the discretization is checked before the result is used. Surrogate models are evaluated on cases that were not used to build them, and their error is reported with the result.
Validation. Numerical verification is carried out by DIADICA. Validation, which compares the model with the real phenomenon, requires field data and the judgement of whoever provides it. The credibility of each model is documented following NASA-STD-7009.
Signature. Design decisions and the signature of the deliverable rest with the engineer in charge of the project. Each result keeps the record of its calculation.
Tools
Computing environment
Python, NumPy, SciPy, OpenSeesPy, PyTorch, JAX, scikit-learn, GPy, SALib and OpenFOAM. High-volume runs are executed on on-demand cloud computing infrastructure.