Below is a list of internships. If there are no internships that match your interests, feel free to contact us to discuss whether we can formulate a new project that is interesting both for you and for us.
We can only accommodate a limited number of internships at any one moment. If you are interested to do an internship at VORtech, call us or send us an email to see whether there is a place available.
We are setting up several projects in the field of machine learning. If you are interested in this topic, please contact us to hear what we are working on.
OpenDA voor kalibratie van OpenFOAM modellen
Description of the assignment
OpenFOAM® is free, open source CFD software. It has a large user community in both technology and science and both in the commercial sector and in academia. OpenFOAM has extensive functionality for solving CFD problems: from complex flow situations with chemical reactions to heat transfer to acoustics, solid mechanics and electro magnetism.
OpenDA is free open source software to integrate computational models and observations. It is used for two purposes. First, a computational model can be calibrated automatically with OpenDA: the parameters of the model are chosen automatically to make the model fit the available observations. Second, a computational model can be adapted to the available observations as it is running. This latter aspect is called data assimilation.
OpenDA also has a large user community in a wide variety of application fields.
VORtech has used OpenDA together with OpenFOAM for various applications. Among these are the modelling of the indoor climate of data centers and completing the flow field observed around a wind turbine to fill in the parts that are not observed.
In all these cases, an ad-hoc coupling between OpenFOAM and OpenDA was created. The ad-hoc nature inhibits a distribution of the coupling to the OpenFOAM community at large. Therefore, we want to build a robust and lasting connection between OpenFOAM and OpenDA.
The assignment is the development of an efficient, robust and maintainable coupling between OpenFOAM and OpenDA. This coupling will be made available to the OpenFOAM community, where it is expected to be used primarily for model calibration.
The assignment consist of the following steps:
- Getting to know OpenFOAM and OpenDA,
- Study the ad-hoc implementations,
- Develop an efficient, robust and maintainable coupling between OpenFOAM and OpenDA,
- Testing, documentation and release.
Requirements and support
To perform this assignment, experience with (and enthusiasm for) software development in e.g. Java of C++ is important. Besides, an understanding of numerical mathematics will be very helpful. Knowledge about calibration and data-assimilation will be acquired during the assignment.
You will have daily supervision and support from two experienced scientific software developers from VORtech. Apart from these, there is quite a number of VORtech colleagues available for help.
Data fusion and transport model calibration using OpenDA and OmniTRANS
Ever since the emergence of big data, it plays an increasingly important role in both construction (parameter estimation) and application (forecasting) of strategic, tactic and operational transport model systems. In the Netherlands, the numerical models available in OmniTRANS are widely used to build such transport model systems. Although some parameter estimation methods and tools are readily available within OmniTRANS, most models and applications rely on exogenously estimated sets of parameters and/or interfaces with (big) data using custom Ruby scripts. The OpenDA platform contains several data fusion and parameter estimation methods for numerical models and as such could be a more generic solution for construction and application of OmniTRANS based model systems using big data.
Research focuses on the various OmniTRANS models and applications in which big data (could) play a role. The goal is to define and construct a test case in which big data and OpenDA are used to calibrate and/or apply numerical models within OmniTRANS. Below, a (non-extensive) list of possible data sets, models and estimation methods is displayed:
- Using loop detector data together with some light form of route generation, route choice models and nudging or optimal interpolation to estimate flows, speeds and densities in networks where OD-demand information is poor or unavailable.
- Estimate the representativeness of floating car data (TomTom, Here or Be-Mobile) using loop detector and/or interpolated data together with a traffic assignment model and some data fusion method.
- Using OV-chipcard data to estimate stop choice model and line choice model parameters within the public transport assignment model using a parameter estimation method.
- Using floating car data, loop detector data and other sources (e.g. the current weather forecast) together with a traffic assignment model and e.g. ensemble Kalman filtering for state estimation and short term forecasting of traffic flows, speeds, travel times and densities.
- Research group / information
When interested in this MSc thesis internship on data fusion and transport model calibration, please contact Ir. Luuk Brederode (lbrederode@DAT.nl, 0627369830)
Automatic calibration of traffic models
Developing municipal or regional traffic models is a labour intensive process. Setting up the model, selecting the modelling approach and calibrating the model take a lot of time. During calibration, the model is initially tuned by modifying the parameters in the model through a trial and error process. After that, the Origin-Destination matrix (the main input for the model) is calibrated on the basis of traffic counts. A lot of techniques exist for this last step, but the first calibration step is still much of a craft. Making this step more effective and efficient would mean a significant improvement.
It is likely that mathematical calibration techniques should be able to automate much of this first calibration step. Many of such techniques are implemented in the open source platform OpenDA. This platform is used worldwide in many different fields, including water management, air quality predictions, sewer management and modelling indoor climate. So far, it does not have applications in traffic modelling, but it should be applicable there.
This MSc thesis assignment starts with interviewing model builders. You should determine how they calibrate, which parameters can be adapted and which cannot, what criteria are used and what objective function. Based on this investigation, you will construct a method to automate this process. You will implement your approach (probably in OpenDA) for the Omnitrans traffic model, test it and make it ready for practical application. Some programming skills are required.
OpenDA is a generic open source software framework for data assimilation and model calibration techniques (www.openda.org). Data assimilation methods incorporate observations into a dynamical computer model of a real system in order to improve the quality of the predictions.
OpenDA is under continuous development. There are always tasks to be done that can be the subject of an internship or MSc project. Contact us to hear what the current topics are.