基于LFSS和改进BBA的滚动轴承在线性能退化评估特征选择方法

Feature selection method for rolling bearings' online performance degradation assessment based on LFSS and FSBBA

在滚动轴承性能退化评估中,不同工况会影响振动信号特征对故障程度的敏感性,在早期有限样本中选择适用于状态评估的有效特征是实现在线评估轴承性能退化程度的关键。首先提出一种基于均方根的早期有限样本判定方法 Limited Feature Select Sample(LFSS),其次提出一种针对性能退化评估特征选择的改进Binary Bat Algorithm(BBA)算法——Feedback Seeking Binary Bat Algorithm(FSBBA),将其应用于滚动轴承早期有限样本中进行故障特征选择,克服了原始BBA容易陷入局部寻优的缺点。基于LFSS与FSBBA算法,构建了滚动轴承在线状态评估模型,并将其运用于两例滚动轴承全寿命数据特征选择,性能退化评估指标分析结果表明了所提出方法的有效性。

Different working conditions affect the sensitivity of vibration signals’ features to fault level in rolling bearings’ performance degradation assessment. Selecting effective features applicable for condition assessment in early limited samples is the key to realize the online assessment of rolling bearings’ performance degradation level. Here,an early limited sample determination method named the limited feature select sample（ LFSS） was proposed based on root mean square,then an improved binary bat algorithm（ BBA） named the feedback seeking BBA（ FSBBA） aiming at feature selection of performance degradation assessment was proposed. They overcame the disadvantage that BBA is easy to fall into local optimization when it is applied to select fault features in bearings’ early limited samples. The model for rolling bearings’ online condition assessment was constructed based on LFSS and FSBBA. This model was applied in two examples of rolling bearings ’ whole life data feature selection. The analysis results of their performance degradation assessment indexes verified the effectiveness of the proposed method.