作为液体火箭发动机推进剂输送系统的关键部件,涡轮泵的运行状态直接影响着整个运载系统的性能,然而,现有的故障诊断方法往往面临特性参数选择片面及计算复杂度高等问题.针对上述局限,提出了面向涡轮泵的轻量化故障诊断框架.所提方法利用Hessian局部线性嵌入算法对信号时域、频域及时频特征进行降维,并引入一种轻量化的深度学习模型MLP-Mixer作为分类器,进而实现不同故障状态的辨识.采用某型号涡轮泵试车数据验证了所提方法的有效性,结果表明,该方法能够在保障诊断精度的同时有效降低计算复杂度,提高诊断效率.
As the key component of a liquid rocket engine propellant delivery system,the operation state of the turbopump directly affects the performance of the entire launch system.However,the existing fault diagnosis methods often suffer from the problems of one-sided selection of feature parameters and high computational complexity.Aiming at the above limitations,a lightweight fault diagnosis framework towards the turbopump was proposed.In the proposed method,a Hessian local linear embedding(HLLE)algorithm was used to reduce the dimensions of the signals in time-domain,frequency-domain and time-frequency features,and a lightweight deep learning model MLP-Mixer was introduced as a classifier to realize the identification of different fault states.The validity of the proposed method was verified by the test run data of some type of turbopump.The results show that the proposed method can effectively reduce the computational complexity and improve the diagnostic efficiency while ensuring the diagnostic accuracy.