基于RVM的液体火箭发动机试验台故障预测方法

RVM-based fault prediction method for liquid rocket engine test stand

液体火箭发动机试验台故障预测问题实际上是与试验台相关的参数预测问题,通过预测相关参数在试验台运行过程中的变化趋势,可以判断试验台未来某一时刻是否可能发生故障。由于液体火箭发动机试验台系统复杂、不易建模,提出了一种相关向量机(relevancevector machine,RVM)故障预测模型。在模型的训练阶段,根据数据序列的特征,分别采用单参量、相空间重构和多参量的方法进行了模型的训练,然后利用训练好的模型对试验台总体健康度和启动过程推力进行了趋势预测。预测结果表明,该方法能有效地跟踪试验台可能发生的故障及故障发展趋势。

The fault prediction of liquid rocket engine test stand is actually the prediction of parameters associated with the test rig. By predicting the variation trends of those parameters, whether the test rig will get fault at a certain time in the future can be judged. As liquid rocket engine test-stand system is complex and difficult to model, a model based on relevance vector machine （RVM） is proposed in this paper. At the training stage of the model, the single-parameter method, phase space reconstruction method and multi-parameter method are used respectively to train the model according to the features of the data sequence, and then the trend of overall health degree and start-up thrust of the test stand is predicted by the trained model. The prediction result shows that this method based on RVM can effectively predict the possible fault and its trend.