Forecasting satellite thermal-vacuum simulation through grey-box fuzzy systems and swarm intelligence
- Paper number
IAC-06-C2.2.01
- Author
Mr. Ernesto Araujo, Instituto Nacional de Pesquisas Espaciais (INPE), Brazil
- Coauthor
Mr. Ubiratan Freitas, Instituto Nacional de Pesquisas Espaciais (INPE), Brazil
- Year
2006
- Abstract
A grey-box fuzzy modelling working in synergy with swarm intelligence is employed for forecasting space environmental emulation and satellite qualification in a thermal-vacuum chamber. Advantages of identifying a model for environmental simulation unit are, for instance, the ability to detect loss of vacuum, presence of unknown heat sources or sinks, training of thermal-vacuum operators, development of a supervisor decision-support system for helping to control the whole operation, checking the instantaneous operation or even operator's behaviour or performance, and, ultimately design an automatic control for the whole system. Eliciting mathematical models from data to forecast nonlinear behaviour is usually not a simple task mainly when dealing with thermal-vacuum systems. Instead of eliciting models based on purely sampled data (black-box approach) as carried out in previously thermal-vacuum system identification approach, a hybrid semi-mechanistic approach (grey-box approach) is employed here. The objective of this optimized fuzzy model is the identification of a nonlinear thermal-vacuum system based on both particle swarm optimization (PSO) and information about radiation since it is known in advance that the main source of heat transfer and the nonlinear characteristics of this process are naturally represented in absolute temperatures. Fuzzy systems based on Takagi-Sugeno (T-S) model have received particular attention in identification of nonlinear systems due to their potentialities to approximate nonlinear behavior. A fuzzy system is a nonlinear mapping represented by a set of IF-THEN rules and an associated fuzzy inference mechanism in which each element inside a fuzzy region assumes a degree of fulfilment which are associate with uncertain, imprecision and vague information. In the population-based swarm algorithm adopted here individuals learn primarily form the successes of their neighbours. The position of each particle in PSO is a potential solution in the solution space. Each particle is associated to a random velocity and moves through the problem space. At each step, each particle changes its velocity flying toward its best local solution (fitness) each particle has achieved so far as well as its overall best value. The past best position and the best overall position of the group are employed to optimize the solution in order to find out a nonlinear-parameter representation associate to Takagi-Sugeno fuzzy modelling. The approach proposed here explores the ability of PSO to derive the parameters of premise part for generating fuzzy systems for a nonlinear system. Results indicate that this approach was able to generate satisfactory nonlinear models for one-step ahead and infinite-step ahead forecasting.
- Abstract document