CAJ | 학술논문

针对超声波热量表采用时差法测量流量时，因受温度影响而存在的非线性问题，提出了分别基于曲面拟合和BP神经网络的温度补偿算法。两种算法通过建立温度与流量之间的非线性映射关系，达到补偿流量测量的目的。建模与仿真可知， BP神经网络补偿算法表现出更好的数据融合及预测能力。验证实验表明，相对于现有查表修正算法和曲面拟合补偿算法，BP神经网络补偿算法补偿效果更佳，补偿后流量测量误差在±2.2%以内，绝对误差方差最大值为0.68，补偿效果显著，具有较高的工程应用价值。
침대초성파열량표채용시차법측량류량시，인수온도영향이존재적비선성문제，제출료분별기우곡면의합화BP신경망락적온도보상산법。량충산법통과건립온도여류량지간적비선성영사관계，체도보상류량측량적목적。건모여방진가지， BP신경망락보상산법표현출경호적수거융합급예측능력。험증실험표명，상대우현유사표수정산법화곡면의합보상산법，BP신경망락보상산법보상효과경가，보상후류량측량오차재±2.2%이내，절대오차방차최대치위0.68，보상효과현저，구유교고적공정응용개치。
When using transit-time ultrasonic heat meter for the flow measurement,there is a nonlinear problem af?fected by temperature. In order to solve it,this paper proposed two kinds of temperature compensation algorithms re?spectively based on curve fitting algorithm and BP neural network. These two algorithms compensated flow measure?ment by establishing mapping relationship between temperature and flow. After modeling and simulation analysis , BP neural network compensation algorithm showed better ability of data integration and prediction. Spot test proved that BP neural network compensation algorithm had superior correction effect than the existing look-up table correc?tion algorithm and curve fitting compensation algorithm. Flow measurement error was limited within ± 2.2%and the maximum absolute error variance was 0.68 after BP neural network compensating. BP neural network compensation algorithm had great value of application with significant compensation effect.