Development of an optimization theory based tool (AutoPOM) for the Bell 427 helicopter simulation model validation

This thesis presented here is a part of work for Bell 427 helicopter simulator research project. The purpose of the research was to obtain a global simulation model which was a composition of several distinct simulation models of the Bell 427 helicopter in various flight conditions. The main tasks of our university team were the corrections of the raw flight test data and the simulation model validation.; The corrections of the raw flight test data were to reorganize the flight test data in standardized formats so that all partners could easily use it. The validation was to evaluate and approve the global simulation model in accordance with Advisory Circular AC 120-63. Our team used the software POM developed by NRC to validate the simulation model. Usually, one flight case validation process took about 4-6 hours or more by use of POM. In this research project, the software AutoPOM based on POM was developed. By use of AutoPOM, one flight case validation process took about 20-30 minutes in the computer (AMD Athlon(TM) XP 1800+, 1.53GHz, 512MB). This was 12 times faster than POM.; This thesis presents the helicopter mathematical model, corrections of the raw flight test data and the simulation model validation. The contribution of this thesis was the development of the software AutoPOM.; Validation process for flight cases by use of POM can be described as a mathematical multi-objective optimization problem. This problem can be expressed by a utility function of Least Square cost functions with their weights. Because the decision vector is different from the optimization vector, the expressions for the first partial derivatives of the objective function with respect to the decision vector can not be found. Thus the optimization problem can not be solved by indirect method, but it can be by direct methods and genetic algorithms. Three algorithms (Hooke and Jeeves' method, Nelder and Mead's method, and Genetic algorithm) were selected and applied to the software POM to develop the software AutoPOM. The AutoPOM interface is created by use of Matlab Graphical User Interface (GUI) techniques, which is clear and easy to use. This is an important criterion for engineers in industry.; Experimental results show that the three algorithms can solve the optimization problem. The Hooke and Jeeves' method starts from one point, its failure exploration steps delay the convergence speed. The Nelder and Mead's method starts from (n + 1) points; it uses more information from (n + 1) points and each step is forward to the convergence zoom. The Genetic Algorithm (GA) convergence speed varies due to random choice of the crossover sites and the mutation sites, the offspring may be worse than their parents, the convergence speed can be either fast or slow.; Based on the convergence speeds of the three algorithms, we recommend the use of the Nelder and Mead's method.