1 EXPERIENCES IN MINING DATA FROM COMPUTER SIMULATIONS

Computer simulations enable us to understand complex phenomena through the analysis of mathematical models on high performance computers. By filling the gap between physical experiments and analytical approaches, computer simulations provide both qualitative and quantitative insights into physical phenomena. They are particularly useful when the phenomena is too complex to be analyzed analytically, such as the flow around an airplane, or too expensive, impractical, or dangerous to study experimentally. In the latter case, we include examples such as the effects of a volcano eruption on the earth’s surface temperature, the evolution of stars, car crash tests, the structural integrity of buildings and bridges under various wind conditions and loads, modeling of material behavior at macroand micro-scales, the spread of diseases via entity-based simulations, etc. In this chapter we survey the work being done in the mining of simulation data, describe our experiences in the analysis of data from a fluid mixing problem, and outline the challenges and opportunities in the mining of this relatively new form of data. We first provide a brief introduction to simulation data in Section 1.2, followed by a survey of various applications in Section 1.3. Then, in Section 1.4, we discuss our work in similarity-based object retrieval, and illustrate how it can be used in several problems of interest. Finally, in Section 1.5 we identify open problems in mining simulation data as well as potential solution approaches.