Particle-Filter-Based Fault Diagnosis for the Startup Process of an Open-Cycle Liquid-Propellant Rocket Engine

Area of Science:

• Aerospace Engineering
• Mechanical Engineering
• Control Systems

Background:

• Open-cycle liquid-propellant rocket engines (LPREs) experience over 30% of failures during startup.
• Existing fault detection and diagnosis (FDD) algorithms for LPRE startup require improvement for reliability.

Purpose of the Study:

• To develop and validate a novel, model-based fault diagnosis algorithm for LPRE startup.
• To enhance the accuracy and performance of FDD systems in critical aerospace applications.

Main Methods:

• Utilized a particle filter (PF), a theoretically optimal nonlinear filter, for residual generation.
• Integrated the PF with a modified cumulative sum (CUSUM) algorithm and a multiple-model (MM) approach for fault detection and diagnosis.
• Employed numerical simulations and Monte Carlo analysis for algorithm validation and comparison.

Main Results:

• The PF-based FDD algorithm successfully detected and diagnosed faults during simulated LPRE startup.
• Numerical confirmation using CUSUM and MM methods validated the algorithm's efficacy.
• Comparative analysis demonstrated superior performance of the PF-based FDD algorithm over previous nonlinear filter-based methods.

Conclusions:

• The proposed particle filter-based fault diagnosis algorithm offers a significant advancement for LPRE startup safety.
• This enhanced FDD approach provides a more reliable method for identifying and addressing engine failures.
• The study highlights the potential of advanced nonlinear filtering techniques in aerospace system monitoring.