交叉分段差分进化支持向量回归的气体超声 流量计测量方法

为了进一步提高全量程气体超声流量计的测量精度，基于多通道声波到时和实时温度，提出了一种交叉分段差分进化(Differential Evolution)支持向量回归(Support Vector Regression)DE-SVR模型。考虑到气体在不同流量条件下的流体状态不同，提出了交叉分段处理的方法，采用DE算法优化选取SVR参数。实验结果表明，对于16~1600m3/h全量程，交叉分段DE-SVR和传统积分方法计算气体流量的平均相对误差分别为0.00447和0.02781，前者较后者降低了83.93%；对于16~160m3/h小流量，交叉分段DE-SVR和无分段DE-SVR算法计算结果平均相对误差分别为0.00436和0.03214，前者较后者降低了86.43%。该方法有效避免了声道长度、探头角度以及管道直径等参数不确定性对流量计算的影响，为全量程气体流量的高精度测量提供了保障。

Gas ultrasonic flowmeter measurement method based on cross-segmented differential evolution support vector regression

In order to further improve the measurement accuracy of the full-range gas ultrasonic flowmeter, based on multi-path acoustic arrival time and real-time temperature, a cross-segmented differential evolution support vector regression DE-SVR model is proposed. Considering that the gas is in different fluid states under different flow rates, a cross-segmented processing method is proposed, and the SVR parameters are optimized using the DE algorithm. The results show that for the full range of 16~1600m3/h, the mean relative errors of the cross-segmented DE-SVR and the traditional integration method in calculating the gas flow rate are 0.00447 and 0.02781, respectively, and the former is 83.93% lower than the latter. For a small flow rate of 16~160m3/h, the mean relative errors calculated by the cross-segmented DE-SVR and the unsegmented DE-SVR algorithms are 0.00436 and 0.03214, respectively, and the former is 86.43% lower than the latter. The method effectively avoids the influence of uncertainties in parameters such as acoustic channel length, probe angle and pipe diameter on the flow calculation, and provides high accuracy measurement of the full range of gas flow.