Application of artificial neural networks in the prediction of maximum absorbed dose during a major solar particle event

A major source of concern for planners involved in preparing for future human exploration of Earth's moon and Mars is the occurrence of random, large Solar Particle Events (SPEs) consisting of large numbers of energetic protons and heavy ions emitted by the Sun. During the events radiation exposures of astronauts performing extravehicular activities or extraterrestrial surface operations can reach levels which are mission-threatening over time periods ranging from hours to days. The total dose received by astronauts during one of these events can be estimated from the measured intensity and energy dependence of the particle fluxes. Typical profiles of dose versus time, calculated for past large SPEs display a Weibull-type mathematical form. Artificial Neural Networks are ideal approximators for this type of nonlinear function. In this work an innovative Sliding Time-Delay Neural Network (STDNN) is developed and applied to the problem of predicting maximum doses received by astronauts from large SPEs as they are occurring. The STDNN is trained and tested using astronaut organ doses calculated from actual data collected by satellites during previous large SPEs. Results of this testing indicate that the STDNN is able to predict total doses to within 5% of the organ dose calculations for SPEs which are not part of the training sets. This is the first time that Artificial Neural Networks have been used successfully to predict doses from these events.