Denoising of pipeline leak signals is of great significance to improve the accuracy of pipeline leak detection. Variational mode decomposition (VMD) has the function of signal denoising. However, the inaccurate setting of VMD parameters will affect the result of signal decomposition. This paper proposes a method for optimizing VMD parameters using particle swarm optimization (PSO-VMD). The ratio of the mean and variance of permutation entropy is used as the fitness function of the particle swarm optimization algorithm to search for the optimal number of signal decomposition layers K and penalty factors α. The signal is decomposed using the VMD with the best parameters. Finally, permutation entropy (PE) is used to select the intrinsic modal functions (IMFs) which contains signal characteristics, and these IMFs are used for reconstruction, so as to complete the pipeline signal denoising and leak detection. Experiments show that, compared with the other three denoising methods, the SNR of pipeline signal denoised by the proposed method is increased by 2.1127 on average, MSE and MAE are reduced by 0.000 35 and 0.0043 respectively, and the recognition accuracy of SVM is improved. 5.5%. Therefore, the proposed method has better denoising performance and higher leak detection rate.

管道泄漏信号的去噪对提高管道泄漏检测的准确性具有重要意义。变模分解（VMD）具有信号降噪功能。但是，VMD参数设置不正确会影响信号分解的结果。本文提出了一种利用粒子群算法（PSO-VMD）优化VMD参数的方法。排列熵的均值和方差之比用作粒子群优化算法的适应度函数，以搜索信号分解层的最佳数量K和惩罚因子α。使用具有最佳参数的VMD分解信号。最后，利用置换熵（PE）选择包含信号特征的固有模态函数（IMF），并利用这些IMF进行重构，从而完成流水线信号的去噪和泄漏检测。实验表明，与其他三种去噪方法相比，该方法去噪后的流水线信噪比平均提高2.1127，MSE和MAE分别降低0.000 35和0.0043，提高了SVM的识别精度。5.5％。因此，该方法具有更好的去噪性能和较高的检漏率。