I enjoy translating the research that I do into software that can be used by engineers, consultants and practitioners.
Here you can find some of the open-source projects that I have been recently working on.
Numerical models
STEM
STEM is a numerical model that simulates railway induced vibrations. STEM accounts for the train-track interaction and the propagation of the vibrations through the subsurface. The model is able to compute the vibration levels at the ground surface taking into account the presence of irregularities in the track geometry, the type of train and train speed, and the spatial variability of the track and soil properties. The STEM model is based on the finite element method and it is powered by Kratos Multiphysics.
STEM consists of the following components:
- STEM
- Geometry generator utils (based on gmsh)
- Vehicle models
- Random field generator (based on GS Tools)
- Kratos Multiphysics
STEM can be installed via pip. The STEM documentation constains instructions on how to install STEM and constains several tutorials.
Train traveling on random media:
Scatter
Scatter is a 3D FEM model that models wave propagation on spatial varying subsurface. The spatial variability is modelled by means of Random Fields.
Subsurface with spatial variability:
Wave propagation on subsurface with spatial variability:
ROSE
ROSE is a train/track interaction model. It can simulate the passage of trains, changes in track stiffness and the presence of discontinuities along the railway track.
Train traveling over a transition zone:
Data models
ROSE network analyses
Dashboard to visualise the results obtained by combining ROSE with railway track data. The data is incorporated in the model by using Kalman filtering, clustering and inverse analysis.
Railway track settlement at network level:
Data Fusion Tools
The Data Fusion Tools are a framework to combine different data sets to perform subsurface schematisations and parametrisations, using AI.
The general idea behind data fusion is that the combination of results from multiple data sources, leads to more and better information, than using a single data source. Multiple data sources providing redundant information increase reliability and reduce uncertainty. Complementary information from distinct sources enhances the quality of the output information, by either increasing the spatial and temporal coverage, increase robustness, reduce noise and increase estimation accuracy, which would not be possible to achieve by using just information from a single individual data source.
Prediction of CPT data using data fusion of CPT, resistivity and geomorphological data:
Visualisation of data fusion data:
Optimisation of testing location for subsurface mapping
Super learn makes use of reinforcement learning to determine the optimal locations for the execution of subsurface testing, in order to minimise the costs and errors.
Visualisation of the training:
Random Layers
Random Layers generates Random Fields and Conditional Random Fields for multi-layered systems.
| Overview 3D | Slice view |
|---|---|
 |  |
OURS
OURS is the Dutch calculation method for railway induced vibrations. This model allows the estimation of the vibration level in dwellings nearby railway tracks.
Subsurface schematisation in OURS:
Tools and utilities
BRO reader
CPT connection to the BRO to request and parse CPT data.
Signal processing
Signal processing tools for time signals (filtering, integration, windowing, FFT, PSD).
Dynamic stiffness
Dynamic stiffness following Wolf & Deeks (2004).
Dynamic solvers
Numerical solvers for dynamic equilibrium equations.