摘要
针对飞机APU的常见故障,该文提出一种基于粒子群优化的长短期时间记忆网络的飞机APU故障诊断模型。该模型直接使用原始的飞机APU数据作为输入,从QAR数据库中整理出需要的APU故障数据,将其进行归一化和时间序列化的处理并分为训练集和测试集两部分,建立CSV文档数据库。引入自适应学习策略对粒子群进行优化,再用优化后的粒子群对长短期记忆网络模型的隐含层单元数目进行寻优。使用训练集对长短期记忆网络的参数进行训练,并在网络最顶层加入Softmax模型,生成故障检测器,然后再用测试集进行实验。实验结果表明,该模型可以有效地识别APU故障,与LSTM模型、支持向量机模型、循环神经网络模型和极限学习机模型相比,识别准确度有所提高。
In allusion to the common faults of aircraft APU(auxiliary power unit),an aircraft APU fault diagnosis model based on PSO⁃LSTM(particle swarm optimization⁃long short term memory networks)is proposed.In the model,the original aircraft APU data is used as input,and the required APU fault data is sorted out from the QAR database,which is processed by normalization and time serialization,and divided into the training set and testing set,so as to establish the CSV document database.The adaptive learning strategy is introduced to optimize the particle swarm,and then the optimized particle swarm is used to optimize the number of hidden layer units in the LSTM model.The training set is used to train the parameters of the LSTM,and the Softmax model is added to the top layer of the network to generate the fault detector.The experiment is performed with the testing set.The experimental results show that this model can effectively identify the APU fault,and has higher recognition accuracy than LSTM model,support vector machine model,recurrent neural network model and extreme learning machine model.
作者
王坤
高丹妮
WANG Kun;GAO Danni(College of Electronic Information and Automation,Civil Aviation University of China,Tianjin 300300,China)
出处
《现代电子技术》
2021年第6期6-11,共6页
Modern Electronics Technique
基金
国家自然科学基金(U1733119)
国家自然科学基金青年基金(61603395)
中央高校基本科研业务费项目中国民航大学专项资助:APU故障诊断与预测关键技术研究(3122018C001)。
关键词
飞机辅助动力装置
故障诊断
长短期记忆网络
自适应粒子群
参数训练
比较验证
aircraft APU
fault diagnosis
long short term memory networks
adaptive particle swarm optimization
parameter training
comparative validation