针对双激光位移传感器测量大型壳段厚度过程中噪声对检测精度的影响，提出利用变分模态分解来实现对厚度信号的自适应去噪，利用相邻固有模态函数之间的离散Hellinger距离来获取最佳的模态数。该方法将变分模态分解算法引入到激光信号的自适应滤波过程中，分析并改进了变分模态分解算法的过分解、欠分解以及能量泄露的问题。然后，对改进的变分模态分解与希伯特振动分解和自适应噪声总体集合经验模态分解进行性能对比，提出了固有模态函数的相对瞬时能量概率的概念。最后，结合离散Hellinger概率分布距离理论判断固有模态之间的信噪分界点，实现了对信号的重构及滤波处理。仿真和实验结果表明，该方法对壳段厚度信号处理的信噪比为39.27 dB，比自适应噪声总体集合经验模态分解方法提高了10 dB，具有良好的自适应性，无需先验条件便能快速有效地识别并分离激光信号中的噪声成分。

The adaptive denoising of shell thickness signal by Variational Mode decomposition method was proposed aimed at the influence of noise on laser detection precision in the measurement process by a double laser displacement sensor.The discrete Hellinger distance between intrinsic mode functions was used to obtain the optimal modal number. Firstly, the VMD algorithm was introduced into adaptive filtering process of laser signal; excessive decomposition, insufficient decomposition and energy leakage problems of the VMD algorithm was analyzed and improved.Then,the performance of the improved VMD algorithm was comparied with those of Hiebert vibration decomposition and overall ensemble empirical mode decomposition of adaptive noise, the conception of relative instantaneous energy probability of intrinsic mode function was proposed.Finally, Boundary point of signal and noise among intrinsic mode was judged combined with the theory of discrete Hellinger probability distribution distance to realize reconstruction and filter processing of signals. Simulation and experimental results indicate that signal to noise ratio of shell thickness signal processing of the method is 39.27 dB,which has improved 10 dB compared with that of overall ensemble empirical mode decomposition method of adaptive noise. It has good adaptability, and identifies and separates noise components in laser signals quickly and effectively without priory conditions.