| 文娟.应用加权马氏距离与退化模型预测丝杠副寿命[J].高技术通讯(中文),2023,33(9):957~966 |
| 应用加权马氏距离与退化模型预测丝杠副寿命 |
| Prognostics of the ball screw using the weighted Mahalanobis distance methodology and the degradation model |
| |
| DOI：10. 3772/ j. issn. 1002-0470. 2023. 09. 007 |
| 中文关键词: 健康指标； 粒子滤波（PF）； 滚珠丝杠副； 剩余寿命预测； 性能退化模型 |
| 英文关键词: health indicator, particle filter (PF), ball screw, remaining useful life prediction, degradation model |
| 基金项目: |
| 作者 | 单位 | | 文娟 | （浙江工业大学机械工程学院杭州 310023）
（恒丰泰精密机械股份有限公司温州 325000） |
|
| 摘要点击次数: 626 |
| 全文下载次数: 501 |
| 中文摘要: |
| 为了提高数控机床的可靠性、安全性并减少维修费用，提出一种基于模型的滚珠丝杠副剩余寿命预测方法，主要包括健康指标构建与剩余寿命预测。在健康指标构建部分，为了更好地表征滚珠丝杠副的退化状态，通过对多信息域特征进行选择与加权融合，提出一种新的健康指标——加权马氏距离(WTMD)。在寿命预测部分，利用指数模型描述滚珠丝杠副的退化过程，通过粒子滤波（PF）算法结合状态监测信息与模型，更新模型参数，完成剩余寿命预测。采用滚珠丝杠副加速性能退化实验中获取的数据对提出方法进行验证。实际结果表明，相对于传统马氏距离而言，WTMD对滚珠丝杠副的损伤发展更为敏感，能够更好地表征其健康状态。同时，基于指数模型的剩余寿命预测方法能够有效地预测滚珠丝杠副的寿命，其预测误差小于基于线性模型和非线性模型的方法。 |
| 英文摘要: |
| Accurate prognosis of the ballscrew is a critical issue in ensuring the reliability and safety of numerical control machine tool as well as reducing the maintenance cost. This paper proposes a model based method for remaining useful life (RUL) prediction of the ball screw. It consists of two parts, health indicator construction and RUL prediction. In the first part, a new health state indicator, weighted Mahalanobis distance (WTMD) is constructed by fusing multiple features that can correlate to the degradation process of the ball screw. In the second part, an exponential model is used to describe the degradation process. Then, particle filter (PF) is involved to integrate the model with condition monitoring data for state estimation and RUL prediction. To identify the effectiveness of the proposed method, an accelerated life experiment for the ball screw is conducted, and the proposed method is demonstrated using the condition monitoring data during the degradation process. The experimental results show that the WTMD has more apparent sensitivity to deterioration trend than the Mahalanobis distance and the exponential model based method performs better in RUL prediction than both the linear model based method and nonlinear model based method. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
