Simulation of liquid rocket engine failure propagation using self-evolving scenarios

Traditional probabilistic risk assessment approaches often require failure scenarios to be explicitly defined through event sequences that are then quantified as part of the integrated analysis. This approach becomes difficult when failure propagation paths change as a function of the system state and Mission Elapsed Time (MET). Additionally, if the propagation paths represent interactions among even a modest number of components, the number of possible scenarios becomes combinatorially intractable. This paper presents an alternate approach for quantifying failure propagation probabilities in such a case. Rather than explicitly defining scenario sequences, simple physical models are created for each of the components. In this way, only the physical states and rules of component interactions must be defined, rather than event sequences for each individual scenario. Initiating failures are introduced into the system (either randomly or as defined by relative likelihood) and the failures cascade through the system via the interaction rules. This process is repeated using Monte Carlo methods, allowing the most probable scenarios to “self-evolve” in terms of both sequence path and frequency. This approach was applied to failures occurring in the engine compartment of a space launch vehicle with four liquid rocket engines and four high-pressure helium tanks. Each engine was modeled with key components, such as turbomachinery, combustion chamber, propellant lines, and add itional support systems. Three test cases were conducted with different high-energy engine failure initiators. End results of interest include loss of only a single additional engine and tank burst, which represent the loss-of-mission (LOM) and loss-of-crew (LOC) end states, respectively. We show that most scenario outcomes are unique and that many scenarios involve complex chain reactions that are difficult to predict, let alone enumerate. This demonstrates the importance of the modeling approach in capturing the majority of the involved risk, which can be used to assess the overall risks to the crew during a launch vehicle abort.