Abstract: As the core component of aerospace vehicles, the safe operation of engines are critical. Ablation of engine pipelines may lead to performance degradation and even sudden failures. Traditional diagnostic methods mostly rely on contact-based detection, which has limitations including delayed response and susceptibility to high-temperature interference, making it difficult to meet the requirements of real-time safety monitoring. To address the safety hazards caused by pipeline ablation in engines under extreme operating conditions, this study proposesed a novel fault diagnosis method based on exhaust flame spectral detection. A controllable experimental model of pipeline ablation was established by adding simulated ablated metal powder in a specific proportion to the fuel. An imaging system and a high-resolution spectrometer were used in synergy to achieve real-time flame monitoring and spectral feature extraction. Experimental results show that this method can effectively identify metal atoms in the flame and quickly locate the ablated area, providing a monitoring approach for promoting the safe application of engines.