排序方式： 共有14条查询结果，搜索用时 859 毫秒
11.
针对航空发动机轴承故障诊断过程中预测精度不足以及过拟合的问题,提出基于迁移学习的半监督集成学习器（SSIT）用以发动机轴承故障预测。首先,训练改进的基于迁移学习的极限学习机（TELM）以及基于迁移学习的支持向量机（TSVM）,分别迁移不同目标空间的高相似度样本加入到源样本空间进行训练。然后,与对应的基学习器集成同簇学习器来识别未标记样本,构成半监督学习器不断调整初始基学习器权重,并继续集成半监督基学习器的识别结果到SSIT中。通过此学习机识别提取特征后的,用以故障识别。实验结果清楚地表明:此种方法可以有效降低迁移学习中的负迁移效果,提升迁移精度10%左右,降低机器学习中的过拟合效果,提高半监督学习稳定性,与现有的预测方法相比可以提高精度9%以上。 相似文献
12.
A monitoring and comparison experiment with two types of sensors on a turbojet engine is carried out. Compared with a probe-typed sensor, which is designed successfully before, signals are collected to verify the va- lidity and better feasibility of the circular sensor. According to the signals monitored over 131 h, the typical signals of 125--129 phases are analyzed. The results show that the unusual exhaust particles are carbon depositions from fuel spray nozzle. Therefore, with the electrostatic sensor, early warning can be provided for initial fault condition, as well as real-time reference for the condition-based maintenance. 相似文献
14.