基于改进的集成经验模态分解法的润滑油磨粒检测研究

Research of lubricating oil particle sensor modeling based on improved integrated empirical modal decomposition method

为了提高油液磨粒检测问题的准确率,本文提出了改进的集成经验模态分解方法,其能够将带有噪声干扰的传感器输出信号分解为多个有真实物理意义的固有模态函数,分离出有效输出信号与噪声干扰,同时探究了改进的集成经验模态分解方法的参数值,实现了其超参数的自适应设定;然后确立了磨粒半径与分离出的有效信号的波峰值、波谷值、峰峰值等特征值的数学模型,之后采用多项式拟合方法拟合出磨粒半径与这些特征值的模型曲线,分析获取了磨粒的尺寸、数量、磁性等信息,并通过实验验证了拟合曲线的准确性。

In order to solve the problem of oil particle detection,the integration of empirical mode decomposition(IIEMD)method is improved.First the sensor output signal with noise is decomposed into limited physical significance of intrinsic mode functions(IMFs),output signal and the noise are isolated effectively,the parameter value of the improved IIEMD method is explored and the adaptive setting of its hyperparameter is realized;then the mathematical models of the abrasive radius and the characteristic values such as the peak,trough,peak-to-trough of the effective signal are established,after the model curves of the radius of abrasive particles and these characteristic values are fitted by polynomial fitting method,finally the size,quantity,magnetism and other information of the abrasive particles are obtained,and the accuracy of the fitting curve is verified by experiments.