摘要
针对深度学习的风电机组齿轮箱诊断方法在噪声环境下的鲁棒性较差且在带标签的样本不足时存在诊断精度较低的问题,提出基于RFECV-GNB风电机组齿轮箱故障诊断方法。该方法结合了交叉验证递归特征消除法(RFECV)在故障数据较少时能有效挖掘故障信号的本质特征,以及高斯朴素贝叶斯(GNB)快速高效的性能进行风电机组齿轮箱的故障诊断。同时,针对RFECV训练时间较长这一问题,提出一种基于CPU并行的任务“打包”算法来提高诊断模型的训练速度。该方法通过超额分配逻辑CPU(LCPU)的方式,实现了LCPU之间工作的有效平衡,以此缩短建模时间。最终,通过多个故障数据集进行实验验证,结果表明在相同故障样本数量下,所提方法与传统方法相比,在诊断精度和建模速度上具有明显优势。
This paper addresses the critical challenges of poor robustness in noisy environments and low diagnostic accuracy when labelled samples are scarce in deep learning-based wind turbine gearbox fault diagnosis.This paper proposes a novel method that integrates RFECV and GNB for wind turbine gearbox fault diagnosis.This method integrates the capability of RFECV to effectively extract essential fault signatures from limited fault data with the high computational efficiency of GNB for wind turbine gearbox fault diagnosis.To mitigate the prolonged training time of RFECV,a CPU parallelized task“packaging”algorithm is introduced,enhancing modeling performance via the over-subscription of LCPUs and efficient load distribution among LCPUs.Extensive experiments on multiple fault datasets demonstrate that the proposed method outperforms traditional approaches with significantly superior diagnostic accuracy and modelling speed under the same number of faulty samples.
作者
王进花
袁山钦
曹洁
Wang Jinhua;Yuan Shanqin;Cao Jie(School of Electrical Engineering and Information Engineering,Lanzhou University of Technology,Lanzhou 730050,China;School of Information Engineering,Lanzhou City University,Lanzhou 730070,China)
出处
《太阳能学报》
北大核心
2025年第4期550-558,共9页
Acta Energiae Solaris Sinica
基金
国家自然科学基金(62063020,61763028)
国家重点研发计划(2020YFB1713600)
甘肃省自然科学基金(20JR5RA463)。
