【作者】 雷亚国； 李乃鹏； 林京；

【Author】 LEI Yaguo;LI Naipeng;LIN Jing;State Key Laboratory for Manufacturing Systems Engineering,Xi’an Jiaotong University;

【机构】 西安交通大学机械制造系统工程国家重点实验室；

【摘要】 本文提出了一种基于粒子滤波的滚动轴承寿命预测方法。该方法主要包括特征指标提取和剩余寿命预测两个步骤。在特征指标提取部分,首先从轴承原始振动信号中提取大量特征指标,然后经过特征分类、特征选择和加权融合最终获得新的特征指标一分类选择加权融合指标(Classifying Selection and Weighting Fusion Indicator,CSWF)。在剩余寿命预测部分,采用Paris-Erdogan模型来描述轴承的衰退趋势,模型参数首先经最小二乘拟合进行初始化,然后采用粒子滤波对其进行更新,模型参数确定之后,轴承剩余寿命由模型递推预测得到。采用轴承加速寿命实验数据对提出的方法进行验证,将CSWF和寿命预测结果分别与原始特征和自适应神经模糊推理系统的寿命预测结果进行对比,结果表明该方法更适合于滚动轴承的寿命预测。更多还原

【Abstract】 In this paper,a new particle filtering-based approach is proposed to predict the remaining useful life(RUL) of rolling element bearings.The approach includes two steps,i.e.indicator calculation and RUL prediction.In the step of indicator calculation,multiply features are first extracted from original vibration signals of bearings.Then,these features are classified,and representative features are selected and weighting fused.At last,a new health indicator called Classifying Selection and Weighting Fusion indicator(CSWF) is acquired.In the step of RUL prediction,Paris-Erdogan model is utilized to describe the degradation process of bearings.The model parameters are first initialized using least square fitting,and then updated by PF.Once the parameters are determined,the RUL of bearings are acquired using recursive prediction of the model.Vibration data from accelerated degradation tests on rolling element bearings is used to demonstrate the proposed approach.Using the data,the CSWF and the RUL prediction results proposed in this paper are compared with traditional features and an adaptive neuro-fuzzy inference system-based approach,respectively.The comparison results indicate that the proposed approach performs better in predicting the RUL of bearings.更多还原