基于音频信号的气固两相流检测方法研究  

Research on Gas⁃Solid Two⁃Phase Flow Detection Method Based on Audio Signal

作　　者：仝卫国[1,2] 门国悦 蔡天娇 崔建昕 TONG Weiguo;MEN Guoyue;CAI Tianjiao;CUI Jianxin(Department of Automation,North China Electric Power University,Baoding,Hebei 071003,China;Hebei Technology Innovation Center of Simulation&Optimized Control for Power Generation,North China Electric Power University,Baoding,Hebei 071003,China)

机构地区：[1]华北电力大学自动化系,河北保定071003 [2]河北省发电过程仿真与优化控制技术创新中心(华北电力大学),河北保定071003

出　　处：《计量学报》2025年第3期383-390,共8页Acta Metrologica Sinica

基　　金：河北省省级科技计划(22567643H)。

摘　　要：利用气固两相流在管道流动中产生的音频信号包含大量流体信息的特点,将音频信号引入气固两相流检测。提出一种基于音频信号的气固两相流分类的检测方法:对音频信号利用小波包分析进行多尺度分析,其去噪效果优于集合经验模态分解重构方法。在重构后的音频信号中选取梅尔频率倒谱系数(MFCCs)作为特征,输入到长短期记忆(LSTM)递归神经网络中。实验结果表明,在气固两相流的弯管处上升段所收集到的音频信号的幅值更大,适合安装采样设备。检测方法对实验中6种流动状态的气固两相流分类效果好,准确率为96.11%,证明了音频信号在气固两相流检测中的可行性。By utilizing the characteristic that the audio signal generated by gas-solid two-phase flow in pipeline flow contains a large amount of fluid information,the audio signal is introduced into gas-solid two-phase flow detection.A detection method for gas-solid two-phase flow classification based on audio signals is proposed,and the feasibility of audio signals in gas-solid two-phase flow detection is demonstrated.Wavelet packet analysis is used for multi-scale analysis of audio signals,and its denoising effect is better than the ensemble empirical mode decomposition reconstruction method.Select Mel frequency cepstral coefficients(MFCCs)as features from the reconstructed audio signal and input them into a Long Short Term Memory(LSTM)recurrent neural network.The experimental results indicate that the amplitude of the audio signal collected in the rising section of the bend in gas-solid two-phase flow is larger,making it suitable for installing sampling equipment.The detection method has a good classification effect on the gas-solid two-phase flow of six flow states in the experiment,with an accuracy rate of 96.11%.

关 键 词：流量计量 气固两相流 小波包分解 音频信号 梅尔倒谱系数 长短期记忆递归神经网络 

分 类 号：TB937[一般工业技术—计量学]