基于SSA-CNN-CBAM的超声气固两相流浓度信息识别

气力运输在某些工业生产流程中一直是重要的一环,对于金属管道内的气固两相流流量的精确实时检测随着工业环境对检测技术要求的不断提高一直是一大挑战。针对该问题提出了一种结合麻雀优化算法(SSA)、卷积神经网络(CNN)、注意力模块(CBAM)的混合模型,用于识别气固两相流浓度。利用SSA对CNN-CBAM模型的参数完成寻优,构建SSA-CNN-CBAM混合模型进行超声后向散射波形识别,得到对浓度的分类结果。通过反复实验,将混合模型实验结果与SSA-CNN模型、CNN-CBAM模型、CNN模型的实验结果进行对比,结果表明此模型对气固两相流浓度的分类识别结果更加理想。试验结果表明该方法可以实现对气固两相流输送的实时浓度检测,通过该试验为工业上的气固两相流输送检测提供参考,为相关技术的发展和应用提供新的思路和方法。

Effect of selected signals of interest on ultrasonic backscattering measurement in cancellous bones

This study examined how the signals of interest （SOI） effect on the backscattering measurement numerically based on 3-D finite-difference time-domain （FDTD） method. High resolution microstructure mappings of bovine cancellous bones provided by micro-CT were used as the input geometry for simulations. Backscatter coefficient （BSC）, integrated backscatter coefficient （IBC） and apparent integrated backscatter （AIB） were calculated with changing the start （L1） and duration （L2） of the SOl. The results demonstrated that BSC and IBC decrease as L1 increases, and AIB decreases more rapidly as L1 increases. The backscattering parameters increase with fluctuations as a function of L2 when L2 is less than 6 mm. However, BSC and IBC change little as L2 continues to increase, while AIB slowly decreases as L2 continues to increase. The results showed how the selections of the SOI effect on the backscattering measurement. An explicit standard for SOl selection was proposed in this study and short L1 （about 1.5 mm） and appropriate L2 （6 mm-12 mm） were recommended for the calculations of backscattering parameters.