Research on the BP Neural Network Temperature Compensation Algorithm in Flow Metrology
崔晓志 王 翥
(哈尔滨工业大学(威海)信息与电气工程学院,山东 威海 264209)
流量计量中BP神经网络温度补偿算法研究
Research on the BP Neural Network Temperature Compensation Algorithm in Flow Metrology
崔晓志王翥
(哈尔滨工业大学(威海)信息与电气工程学院,山东 威海264209)
摘要:针对时差法计量流量时受温度影响而存在的非线性问题,提出了基于BP神经网络的温度补偿算法。该算法通过引入动量因子和改善数据敏感度,提高了BP神经网络的预测能力,通过建立温度与流量之间的非线性映射关系来补偿流量计量。仿真分析可知,该算法表现出较好的数据融合及预测能力。实验验证进一步表明,相对于现有查表修正算法,该算法补偿性能稳定,最大误差在±2.0%以内,最大绝对误差方差为0.48,达到2级表水平,具有重要的工程应用价值。
关键词:超声波流量神经网络仪器仪表测量温度补偿数据融合建模
Abstract:In view of the nonlinear problem caused by temperature effect existing in using time difference method to measure flow rate,the temperature compensation algorithm based on BP neural network is proposed.This algorithm improves the prediction performance of BP neural network by introducing momentum factor and improving data sensitivity,and compensates flow measurement by establishing nonlinear mapping relationship between temperature and flow.Simulation analysis shows that the algorithm features better capability of data fusion and prediction.Furthermore,experimental verification indicates that the compensation performance of this algorithm is more stable than that of existing table lookup correction algorithm,the maximum error is within ± 2.0% and the maximum absolute error variance is 0.48,reaches grade II level.So the compensation algorithm has good value of engineering application.
Keywords:UltrasoundFlowNeural networkInstrumentationMeasurementTemperature compensationData fusionModeling
0引言
时差法超声波热量表通过测量超声波沿顺水流与逆水流方向传播时间的差值来计量流量,在此基础上计算出热量,因此流量计量是决定超声波热量表计量特性的关键环节[1]。为提高超声波流量测量精确度,国内外学者开展了大量研究。Wang B等对测量管道内径及超声波换能器安装倾角等参数进行优化,通过减小管道内压力损失提高流量测量的准确性[2]。陈子静等在流体力学修正原理和实验仿真基础上,提出不同流场下的流量修正方法,降低了流量测量误差[3-4]。但以上研究未考虑温度对整个流量测量过程的影响。李志浩等采用神经网络算法对15~50 ℃内流量数据进行理论补偿,但训练数据有限,且未进行实验验证[5]。……