경험적 모델과 실험 자료 기반 하이브리드 로켓 신뢰성 분석

Reliability evaluation is of paramount importance in the design of critical systems such as rocket motors. Numerical analysis technique that can predict the reliability during the design stage has been a viable solution, from which possible design changes to attain higher reliability, can be made in a fast manner. In the real practice, however, the approach relying solely on the numerical codes have found only a limited success due to the lack of knowledge to model the complex phenomena of propulsion or prohibitive computational cost for underlying physics. In this study, reliability analysis procedure is addressed for a hybrid rocket motor, which is currently developed by our research group. The advantages of hybrid rocket over the other systems are the on-off capability, lower development cost, less environmental impact, and more importantly, greater safety and reliability. The performance is characterized by the fact that the oxidizer and the fuel are in different phases; the fuel is a solid of cylindrical configuration, while the oxidizer is injected into its port as a liquid spray. In the hybrid rocket motor, high fidelity codes for the propulsion simulation are not readily available or almost useless unless validated via costly instrumentation. Instead, design procedure has been established based on the low-fidelity or empirical model supported and validated by greater number of experimental data for input parameters and responses. Despite the obvious presence of uncertainties associated with this procedure, however, the performance has mostly been assessed by a primitive regression technique. In this study, method for quantification of various uncertainties such as ignorance of probability distribution, limited number of data, bias in the validation experiment and inherent measurement error is explored. Bayesian approach plays the central role to this end, which can take care of various uncertainties in an integrated manner. Consequently, the propulsion performance will be given in the form of probability distribution, from which the reliability is computed and used for design evaluation.