Authors

Abstract

The power take-off (PTO) shaft of a tractor is subjected to complex and variable random loads during field operations, and the accuracy of the resulting load spectrum directly affects the reliability of fatigue life prediction. To address the limitations of traditional parametric extrapolation methods, particularly their inadequate fitting performance, a non-parametric extrapolation method based on DBSCAN clustering is proposed. Field test validation demonstrates that the proposed method effectively captures the multi-modal distribution characteristics of the load. The extrapolated results show good agreement with the measured data in terms of cycle count distribution and pseudo-damage indicators. Compared with the fixed-bandwidth method, the proposed approach increases the coefficient of determination by 3.401% and reduces the mean squared error (MSE) by 39.154%, while yielding pseudo-damage values closer to 1. These findings provide a novel technical approach for the construction of load spectra under complex operating conditions of agricultural machinery.

Abstract in Chinese

拖拉机动力输出轴（PTO）在田间作业过程中承受复杂多变的随机载荷，其载荷谱编制的准确性直接影响疲劳寿命预测的可信度。针对PTO载荷在多工况下呈现多模态、非平稳分布特征，传统参数化外推方法难以精确拟合的问题，提出一种基于DBSCAN聚类的非参数外推方法。首先，对实测PTO转矩信号进行预处理，采用雨流计数法提取均值-幅值载荷循环样本；其次，引入DBSCAN密度聚类算法对样本进行划分，自动识别不同工况对应的自然簇群并剔除噪声点；然后，对各簇分别采用基于样本协方差的自适应带宽核密度估计，建立二维概率密度函数；最后，通过蒙特卡洛模拟结合分层抽样生成外推载荷循环，获得覆盖全寿命周期的PTO载荷谱。田间实测数据验证表明，该方法能有效刻画载荷的多模态分布特征，外推结果在统计参数、循环计数分布及伪损伤等指标上与实测数据具有良好的一致性，克服了全局固定带宽核密度估计的局限性。研究结果为拖拉机PTO疲劳强度设计与台架耐久性试验提供了更精确的载荷输入，也为农业装备复杂工况下载荷谱编制提供了新的技术途径。