中文摘要：

针对变分模态分解在实际应用过程中需要根据先验知识确定惩罚函数和分量分解个数这一缺陷,提出了一种改进方法,即广义变分模态分解方法。该方法减少了人为因素对分解结果造成的主观影响,将信号分解转化为非递归、变分模态分解方式,能够有效分离频率成分相近的谐波分量,且对信噪比较小的信号有着良好的鲁棒性。将该方法应用于齿轮箱复合故障诊断中,仿真和实验的结果验证了该方法的有效性。

英文摘要：

According to the defects of the VMD that its penalty parameters and number of components were based on the prior knowledges in the processes of the actual applications, based on the VMD an improved method, namely, the GVMD was proposed herein. This method held the potentials to overcome the deficiencies of VMD, and it might reduce the subjective influences on the decomposition results. By decomposing the signal into non recursive and variational modal, the method might effectively separate the harmonic frequency components that were similar to each other and had good robustness. It was applied in the composite fault diagnosis of the gearboxex, and the simulation results and test verify＇the validity of this method.