传统故障诊断算法用于舰船蒸汽动力系统非均衡类样本问题存在准确率低且误分类的情况，本文提出一种基于支持向量机-BP神经网络集成学习的故障诊断模型。使用主层次分析法降低原始数据集的特征维度以改善数据量之间的冗余问题，然后设计3个并行的支持向量机分类器进行故障诊断，将分类器的结果数据融合作为BP神经网络的输入，进行二次诊断，得到最终的结果。用仿真平台采集的样本数据进行验证并重复运行30次获取结果，最终模型平均召回率为84%、平均准确率为95.04%。对比传统诊断算法及常用集成算法，该方法的准确率以及少数类样本的召回率明显高于其他诊断方法。
Traditional fault diagnosis algorithms are used for the problem of unequilibrium samples of ship steam power system with low accuracy and misclassification, and this paper proposes a fault diagnosis model based on support vector machine-BP neural network ensemble learning. First, the main hierarchy method is used to reduce the feature dimension of the original dataset to improve the redundancy between data volumes. Then, three parallel support vector machine classifiers are designed for fault diagnosis, and the result data of the classifier is fused as the input of the BP neural network for secondary diagnosis to obtain the final result. The sample data collected by the simulation platform was verified and repeated 30 times to obtain the results, and the average recall rate of the final model was 84%,the average accuracy was 95.04%. Compared with traditional diagnostic algorithms and commonly used ensemble algorithms, the accuracy of this method and the recall rate of a few samples are significantly higher than those of other diagnostic methods.
2023,45(5): 97-101 收稿日期：2022-10-27
DOI：10.3404/j.issn.1672-7649.2023.05.018
分类号：U676.1
基金项目：国家重点研发计划项目(2019YFE0104600)；国家自然科学基金资助项目(51909200)
作者简介：郑鑫(1997-),男,硕士研究生,研究方向为舰船动力系统健康管理与智能运维
