一种改进的DDTFA方法及其在齿轮箱故障诊断中的应用

针对数据驱动时频分析(DDTFA)方法的初始相位函数选取问题,提出一种可准确、快速且自适应优选初始相位函数的改进DDTFA方法。引入数学中函数求极值的思想,将信号的初始相位函数选取问题转换为初始解集的连续寻优问题,通过对DDTFA中的高斯牛顿迭代算法进行精简,以初始解集中的初始相位函数迭代一次斜率的变化量为导数获得初始解集的连续导数集,进而求得局部极大值,并以局部极大值对应信号分量的能量最强为准则优选信号的初始相位函数,进而完成信号分解。仿真分析与齿轮箱故障诊断实例表明,该方法可准确、快速且自适应地优选初始相位函数,并有效提取故障特征,且具有一定抗噪性。

An improved DDTFA method and its application in gearboxes’ fault diagnosis

Aiming at initial phase function selection of the data-driven time-frequency analysis (DDTFA) method, an improved DDTFA method was proposed to optimize initial phase function accurately, quickly and adaptively.Introducing the idea of solving extreme values of a function in mathematics, a signal’s initial phase function selection problem was converted into a continuous optimization of an initial solution set.Through simplifying Gauss-Newton iterative algorithm in DDTFA, the slope variation of initial phase function in an initial solution set before and after first iteration was taken as the derivative to obtain the continuous derivative set of the initial solution set, and further get local extreme values.Then the energy of signal components corresponding to local extreme values being the maximum was taken as the criterion to optimally choose initial phase function of a signal, and further complete signal decomposition.Simulation analysis and actual examples of gearboxes’ fault diagnosis showed that the proposed method can correctly, rapidly and adaptively optimize initial phase function, and effectively extract fault features; it has a certain anti-noise ability.