PHM by Using Multi-Physics System-Level Modeling and Simulation for EMAs of Liquid Rocket Engine

Abstract

The need for condition-based maintenance to improve reusable launch vehicle readiness, reliability and safety, with affordable maintenance cost and quick turnaround time is recognized. But the problem of detecting faults and predicting failure in the components of reusable rocket engine systems is difficult and complex to solve. Since the number of data samples on the fault or failure status during the actual operation of the rocket engine is very small, it is difficult to adapt a data-drive approach of health management. Furthermore, the failure modes for these systems might transcend electrical, mechanical, and fluid systems. Therefore, one of the key concepts of the approach proposed in this study for fault detection and diagnosis is model-based quantitative assessment that considers system-level interactions in the target system. In this approach, multi-physics system-level modeling and simulation for a target system are conducted by using Modelica, an equation-based, object-oriented modeling language that allows acausal modeling for complex cyber-physical systems. Modelica has an important modeling capability for system-level interactions that involve multi-physics phenomena. One advantage of the model-based health-monitoring approach is that faults and failure modes are traced back to physically meaningful information, which is invaluable for the maintainer. Thus, this model-based approach for condition-based maintenance has the potential to provide reliable early fault detection and diagnosis during post-flight investigation for maintenance decision-making. In this study, multi-physics system-level modeling and simulation for a target system under both normal and abnormal conditions have been conducted based on an understanding of the failure mechanism to obtain prior data sets for fault detection and diagnosis. In this proposed approach, the Dynamic Time Warping (DTW)algorithm was utilized to evaluate dissimilarity between the prior data sets and sensor measurement data obtained during the flight, and hierarchical clustering technique was applied for categorization in failure mode based on dissimilarity of these data. In addition, a trial case study has been conducted on electromechanical actuators (EMAs), an important component of a rocket engine, towards the construction of model-based prognostics and health management (PHM)technologies for reusable liquid rocket engines. Based on the trial results of the model-based approach constructed in this study, the possibility of fault detection and diagnosis was demonstrated for virtual EMAs of a liquid rocket engine.