AmpliSIM is a web service that allows industrial engineering companies to model their environmental impact. As part of the SHAPE programme, Oliver Oldrini and Sylvie Perdriel have been parallelising models to optimise them for HPC clusters so that AmpliSIM can improve its service.
Numerical simulations can be used to assess and forecast the environmental impact of chemical releases or discharges in the air, the sea or the ground. They are used for routine emissions but also in emergencies for accidental releases.
Putting together the pieces needed for an accurate air-quality simulation is a laborious task with many potential pitfalls, as environmental consultant Olivier Oldrini explains: “First of all you need to gather a lot of external data on things like topography, land use, meteorology, buildings and road emissions. This data has to be gathered from many different sources, which also vary depending on the location you are investigating.
“After that, you must then create a graphical projection to bring all of this data together. To do this, a lot of the data has to undergo treatments, including space/time interpolation, file type conversion and geographical re-projection.”
Figure caption: Complex flow of the wind in an urban built-up area
The entire process is cumbersome, time-consuming and prone to error. For this reason, Oldrini and his colleague Sylvie Perdriel decided to create AmpliSIM, a French SME specialising in environmental numerical simulation services. AmpliSIM provides air quality simulations as a commodity, with a web service that allows users to configure, launch and explore air quality numerical simulations without any requirements for software installation or an in-house calculation cluster.
The effects of pollution are often not clearly visible, instead seen far from its source or after an extended period of time. The service that AmpliSIM provides allows businesses to be far more assured about the decisions they make in regard to their environmental impact. “Being well-informed leads to better choices,” says Perdriel. “We are proud to be able to provide people with the tools to make decisions that will ultimately impact the environment in a positive way.”
Figure caption: AmpliSIM web view of a particle matter plume from a storage facility south of Singapore
“For an engineer to independently set up and carry out an environmental simulation takes weeks. With AmpliSIM, it takes just a few days”
AmpliSIM’s customers are not directly involved in any interaction with the computers being used, as all of the calculations are handled as part of its service. “What we have found is that our customers are often not interested in the processors that get the calculations done – they simply want the result of their simulations,” explains Olivier Oldrini.
The company does not simply hand over the raw data from the simulations to the user, however. Value is added through a number of post-processing techniques, making it easier for the user to explore and exploit the results. Other data is also added at this point to provide indications of the health impact of any pollution.
For industrial customers, the main priority is to receive the results as quickly as possible. AmpliSIM as a company does not develop models itself, instead using pre-existing ones. However, most of these models are not made for use on parallel computing architectures, and so Oldrini and Perdriel have been working on a project in which they have been parallelising these models in order to optimise them for HPC clusters.
As part of PRACE’s SME HPC Adoption Programme in Europe (SHAPE), a project called DemocraSIM performed calculations on the IDRIS supercomputer. Thanks to SHAPE and IDRIS expertise, the AmpliSIM toolset, including numerical models and post processing tools, was improved to meet the standards of a parallelised supercomputing architecture. This led to large time reductions in simulation and post processing. Moreover, the ability of the web portal to connect to HPC infrastructure was substantially improved, as well as its ability to support concurrent user access.
Another huge benefit of AmpliSIM’s service is the time and effort that is saved for the customer. For an engineer at a company to independently set up and carry out such simulations would take a couple of weeks. With AmpliSIM, it takes just a few days. The overall engineering time has been reduced by more than 60 per cent, with no time being lost on data retrieval and transformation. This, combined with the excellent quality of data and small error in the modelling process is what makes the service so compelling for users.
The simulations can be used to assess the chronic impact on the environment of a plant over the course of several years. But it can also be used for monitoring and forecasting, combining past data with meteorological data to predict how emissions might spread over the course of the next six to twelve hours. This is done in order to anticipate and prevent any complaints that may interfere with regular activity.
When Oldrini and Perdriel first founded the company, they did not have the technology to offer the web service. After a few years or so of developing it, they now have a full web service up and running and are gaining large companies as customers. Environmental modelling is essential for several global industries, and so the potential for further growth remains huge.
Resources awarded by PRACE: Oliver Oldrini was awarded 15 000 core hours on ADA hosted by Institute IDRIS @ CNRS, France