基于极限学习机的船舶柴油机故障诊断

Fault Diagnosis of Marine Diesel Engine Based on Extreme Learning Machine

针对传统深度核极限学习机网络仅利用端层特征进行分类导致特征不全面,以及故障诊断分类器中核函数选择不恰当等问题,提出基于多层特征表达和多核极限学习机的船舶柴油机故障诊断方法。利用深度极限学习机网络提取故障数据的多层特征;将提取出的各层特征级联为一个具有多属性特征的故障数据特征向量;使用多核极限学习机分类器准确地实现柴油机的故障诊断。在标准分类数据集和船舶柴油机仿真故障数据集上的实验结果表明,与其他极限学习机算法相比,该方法能够有效提高故障诊断的准确率和稳定性,且具有较好的泛化性能,是柴油机故障诊断一个更为优秀实用的工具。

The traditional deep kernel extreme learning machine network only uses the end-layer feature and pays little attention to optimizing the choice of kernels in fault diagnosis of marine diesel engine. To solve the problem, a fault diagnosis of marine diesel engine based on multi-layer feature expression and multiple kernel extreme learning machine is proposed. The multi-layer features of faults are extracted using deep extreme learning machine network. A multi-attribute fault feature vector is combined by the extracted feature vectors of each layer. The multiple kernel extreme learning machine classifier is used to realize the diagnosis of faults. Comparing the proposed method with other extreme learning machine methods on some benchmark classification datasets and artificial faults sets of marine diesel engine, experimental results show that the proposed method is effective to improve the accuracy and stability and it has strong generalization ability in fault diagnosis, it proves to be a good tool for fault diagnosis of marine diesel engine.