基于LSTM的燃料电池发动机寿命预测

Life Prediction of Fuel Cell Engine Base on LSTM

随着工业4.0和数据科学的发展,利用数据对氢燃料电池发动机剩余寿命进行预测,将有利于提前发现发动机性能退化问题,进而及时采取维护保养措施,对发动机安全运行以及延长发动机运行寿命至关重要。传统发动机寿命预测一般基于机理模式的经验判断或者数理统计,但在氢燃料电池发动机这个技术尚不稳定成熟的发展阶段,传统手段无法保证相对较低的误差。本文在不依赖于机理模式的情况下,利用传感器收集的数据,基于神经网络深度学习的模式,构建一种基于数据驱动的长短期记忆神经网络(LSTM)的剩余寿命预测模型,通过机器的训练与学习,分析预测氢燃料电池发动机的寿命衰减情况,为预测性维护提供数据支持。

With the development of Industry 4.0 and data science,using data to predict the remaining life of hydrogen fuel cell engine will help to detect the degradation of engine performance in advance,and then timely maintenance measures can be taken,which is very important for the safe operation of the engine and the extension of the operating life of the engine.The traditional engine life prediction is generally based on the empirical judgment or mathematical statistics of the mechanism model,but in the development stage of hydrogen fuel cell engine,the technology is unstable and immature,and the traditional means cannot guarantee the relatively low error.Without depending on the mechanism mode,this paper uses the data collected by sensors,and based on the model of neural network deep learning,the paper constructs a remaining life prediction model based on data-driven long short-term memory(LSTM).Through machine training and learning,the life decay of hydrogen fuel cell engine is studied,providing data support for predictive maintenance.