WSC 2004 Final Abstracts
Sunday 3:00:00 PM 4:30:00 PM
Chair: Rogene Eichler West (Pacific Northwest National Laboratory)
A Probabilistic Total System Approach to the Simulation of Complex Environmental Systems
Rick Kossik and Ian Miller (GoldSim Technology Group LLC)
GoldSim is a powerful and flexible Windows-based computer program for carrying out probabilistic simulations of complex systems to support management and decision-making in engineering, science and business. The program is highly graphical, highly extensible, able to directly represent uncertainty, and allows you to create compelling presentations of your model. Although GoldSim can be used to solve a wide variety of complex problems, it is particularly well-suited (and was originally developed) to support evaluation of complex environmental systems. Powerful contaminant transport features allow nearly any kind of natural or man-made environmental system to be simulated. This paper provides a brief overview of GoldSim, with special emphasis on environmental applications.
Factors Affecting the Expectation of Casualties in the Virtual Range Toxicity Model
José A. Sepúlveda, Luis Rabelo, Jaebok Park, Fred Gruber, and Oscar Martínez (University of Central Florida)
The Virtual Range (VR) is an environment that integrates in a seamless fashion several models to improve complex systems visualization. A complex system is a non-linear system of systems whose interactions bring together inter-esting emergent properties that are very difficult to visual-ize and/or study by using the traditional approach of de-composition. The VR Toxicity Model as described here represents the different systems that interact in the deter-mination of the expectation of casualties (Ec) resulting from the toxic effects of the gas dispersion that occurs after a disaster affecting a Space Shuttle within 120 seconds of liftoff. We present a detailed description of the VR and the factors affecting Ec. The system will help local authorities to estimate the population at risk in order to plan for areas to evacuate and/or for the resources required to provide aid and comfort and mitigate damages in case of a disaster.
Adaptive Wavelet Neural Network for Prediction of Hourly NOx and NO2 Concentrations
Zhiguo Zhang and Ye San (Harbin Institute of Technology)
Adaptive neural network is a powerful tool for prediction of air pollution abatement scenarios. But it is often difficult to avoid overfit during the training of adaptive neural network. In this paper, based on the wavelet theory, a new algorithm is proposed to improve the generalization of adaptive neural network during online learning. The new algorithm trains adaptive wavelet neural network to model hourly NOx and NO2 concentrations of variance of emission sources. Results show that the new algorithm improves the generalization and the convergence velocity of adaptive wavelet neural network during online learning. The simulations also illustrate that adaptive wavelet neural network is capable of resolving variance of emission sources.