The methods used at SCHEME mainly call for virtual experimentation, as well as the use of advanced data analysis techniques with a high level of added value.
We develop three areas in particular:
Landscape / water quality
There is a strong demand for expert assessment, particularly concerning water resource management related to land-use and landscape organization. More specifically, we have noticed an interest coming from local authorities and state services, for our catchment model used for diffuse pollution issues: TNT2. This model was used in the context of the “nitrate infringement” issue in Brittany (infraction to the water framework directive n° 75/440/CEE of 16/06/75 regarding the water quality needed in surface waters intended for drinking water production in Member States). It allowed the scientific assessment of the proposed action programmes, and therefore permitted France to avoid financial penalties (28 million € and a daily fine of 117,882 €).
Winning back water quality over nitrates : TNT2
Furthermore, hydraulic and flood issues are constantly growing. The use of one dimensional models have shown their limitations in certain situations. This is why we have chosen to work with two and three dimensional models such as GERRIS and OPENFOAM allowing a spatial exploration and thus, a more accurate model.
Porous media and process engineering
Another growing field today is the modelling of reactive transport and transfer in porous media. In other words, it is the simulation of fluid flows from various chemical origins in porous media (e.g. soil, windrow, etc.) and the chemical and biological alteration caused by interactions between components during the different stages of porous media (gas, liquid, solid, biofilm, etc.) or components of the fluid considered. The mathematical models resulting from the study of these strongly coupled phenomena are solved with complex simulators called « Multi-physical Computational Fluid Dynamics ». There is an increasing demand for research and transfer of expertise for industrial applications related to the environment (industrial composting, groundwater transfers and pollution, diffuse pollution of agricultural origins or punctual pollution due to oil products such as – cf. Shale gas-, biofilm studies...). Our added value is based on our capacity to constantly develop new solvers and digital solutions (i.e. simulators, e.g. using the OpenFOAM solver), and proposing high quality expertise assessments in the field.
Productivity gain for process engineering : OpenFOAM
Data mining & Machine Learning
The Machine learning approach can be found today in diverse working fields such as pattern recognition (character recognition, objects) and development of artificial intelligence. The demand for this approach is on the rise especially in the industrial and environmental field and remains an affordable solution capable of providing potent calculations and measurements.
The objective of the Machine Learning approach is to generalize through a learning process. In other words, it produces a model that comprehensively represents the characteristics and interactions contained in the database on which the model was built. The model built allows a detailed analysis of the information contained in the database yet very powerful and easy to use. These approaches are particularly interesting to use when links between explanatory variables and processes are unknown or complex. Machine Learning can often go further than conventional statistical approaches that require prior knowledge about the links and which can often restrict working hypotheses.
A fourth «Living» branch will eventually be developed.
