Custom High-Performance-Computing Software Justification: Advanced Thermoelectric Prediction Model (ATPM)

Custom high-performance-computing software is expensive and time consuming to develop and verify; therefore, expending the resources to develop such software needs to be rationalized. Predicting radioisotope thermoelectric generator (RTG) performance is such a case where the software development effort can be justified. Some of the demands of such a modeling tool are time and temperature dependent material properties, anisotropic materials, quasi-steady state analysis, thousands of components, multi-physics (thermal conduction, thermal radiation, electrical conduction, thermoelectric effect, sublimation, etc.), and variable dimensions. Each of the above will slow down the computation of a single simulation, combining that slow down with the need to perform multi-parameter non-linear optimization, Monte Carlo analysis, and parametric studies; thousands to millions of simulations are often required. Current commercial general-purpose modeling tools meet the RTG modeling demands, but run times are impractical and/or require expensive computational resources (cost prohibitive or scheduling prohibitive). This leads to the effort to design a custom modeling tool for RTG's. The financial costs, personnel costs, and performance are compared for a variety of modeling solutions including other custom options and commercial software. Benefits of configuration management options, verification options, validation options, and long-term support are discussed. Lastly, the versatility of the custom model versus the commercial solutions are compared.