LogoDrihm

  • 1957990414.jpg
  • alluvione_brasile_800_800.jpg
  • fotoArticolo_3177.jpg
  • stazionebrignole2011_1.jpg
  • stazionebrignole2011_2.jpg
  • viaxxsettembre_1.jpg

DRIHM ICT-Video

DRIHM presents an interesting video explaining the objectives and best practices of the project

frame video

Login Form

On March 21, 2014, Professor Dragan Savic (University of Exeter, http://emps.exeter.ac.uk/engineering/staff/dsavic) gave the following talk at CIMA Foundation:

Machine Learning Methods for Modelling of Environmental Phenomena

Recent years have seen considerable interest in the environmental applications of machine intelligence, since, if sufficient observation
data is available, they could obviate the need to build and calibrate physically-based models, for example based on hydrodynamic equations. This
talk will present two examples of intelligent systems developed to utilise machine intelligence methods for environmental applications. The first,
called RAPIDS, deals with urban drainage systems and the utilisation of rainfall data to predict flooding of urban areas in near real-time. The
RAPIDS framework is based on a 2-layer, feedforward MLP (Multi-Layer Perceptron) Artificial Neural Network (ANN), used to relate incoming
rainstorm data to the extent of flooding present at each manhole in a urban drainage system. The ANN system has the potential to provide early
warning and scenario testing for decision makers within reasonable time.
The second system is a two-dimensional cellular automata (CA) model aimed at achieving fast 2D flood modelling for large-scale problems. This CA model employs simple transition rules and a weight-based system rather than complex Shallow Water Equations. The simplified feature of CA allows the model to be implemented in parallel computing environments, resulting in significantly improved modelling efficiency. Both the ANN and CA models are tested on hypothetical case studies and real world examples, and the outputs compared to those from traditional physically-based hydraulic models. The results demonstrate that computational methods that require hours or days to run will not be able to keep pace with fast-changing situations such as pluvial flooding and thus the systems developed are able to react in close to real time.

The slides of the talk are available here.

We use cookies to improve our website and your experience when using it. To find out more about the cookies we use and how to delete them, see the Cookie Policy page.

I accept cookies from this site.