Exploitable results Simulating sintering Sintering is an important technique in the manufacture of precision metal parts for a range of industries. Computer models are used to optimize the design of new parts by simulating the compression stage of manufacture. By using parallel computing techniques, the speed of these simulations can be greatly increased, shortening the design process, improving quality. Software called PowCom is used to model the compression phase of the sintering process. A typical model involves 5000 elements and 20 time steps, and runs sequentially in around 20 h. A run time of 8 h is required, to allow models to be run overnight. A graphical tool has been incorporated to show the evolution of material densities during the compression process. The original sequential code was parallelized, with the innermost computational steps being solved via a Schur complement method with a GMRES solver. The cost of implementation has been kept to a minimum by designing the system to run on a network of personal computers (PC). Studies are being undertaken on architectures with more and faster processors to examine the potential performance of the code at larger scales. The tool results have also been compared to experimental data to calibrate it, and to find the tuning parameters for different raw materials. AMES makes sintered precision metal parts for the automotive industry. The system is now being used not only for the design of new pieces but also for the analysis and improvement of existing manufactured parts. Urban traffic management As congestion on the roads increases, PETRI (Parallel Environment for a real-time Traffic management and Information system) makes traffic control safer and more efficient. This cost-effective, high performance computing technology provides traffic controllers with a tool to monitor traffic, simulate various traffic control options and implement changes. Messages can be sent out using the available communications media, ensuring that the public is kept informed about current and predicted road conditions. Through parallel processing the PETRI system is able to provide real-time control of traffic networks. On-line traffic forecasting is also now possible. Traffic management departments will be able to use this system to improve traffic conditions in urban areas. Parallel processing is used to achieve better results in less time. The PETRI system consists of software that runs on a symmetric multiprocessor platform. The software comprises three modules that have been integrated and parallelized. The demand prediction module uses a neural network parallel algorithm; the flow distribution module uses a master-slave approach; and the detailed simulation module uses clustering and colouring techniques to minimize synchronization. A graphical interface helps the operator visualize and control the urban network and an output system sends current and predicted traffic information to the public through several media including radio, personal computers and variable message signs (VMS). The municipality of Barcelona has been testing a prototype of the PETRI system to manage traffic on the city's ring road, a 43 km motorway containing about 20% of Barcelona's urban network. Ocean environmental modelling Industrial pollutant spills and accidents at sea require immediate decisions to minimize detrimental effects. TRIMODENA uses parallel processing to provide environmental assessors with the detailed models needed to simulate the effects of winds and tides on currents and the dispersion of particles in the sea. Predictions of hazardous consequences can be made quickly, allowing resources to control the problem to be allocated in the most efficient way. TRIMODENA's flexibility also makes it suited to longer term studies on environmental impacts and fish farming. It can be used on different platforms, ranging from personal computer (PC) to supercomputer, depending on the size of the project concerned. The key to the power of the TRIMODENA system is the parallelisation of quasi-tridimensional finite element codes used to analyse this type of problem. These codes perform many calculations to simulate patterns and dispersion processes in a complex marine environment with irregular boundaries. The quasi-3-dimensional approach uses an interpolation technique that allows 3-dimensional solutions to be found with less computational effort than with conventional pure 3-dimensional models. The application comprises three codes. ECADIS models the residual currents induced by the wind and boundary conditions corresponding to the general currents of the area and the measured data. MAREAS computes the elevation and currents generated by tidal waves, while RECODE simulates pollutant dispersion processes by computing the distribution at fixed time increments. Simple models can be run on local PCs while supercomputers, available via the Internet, are used for larger problems. È in corso la ricerca di dati su OpenAIRE... Si è verificato un errore durante la ricerca dei dati su OpenAIRE Nessun risultato disponibile
