基于BPNN的微小反应器温度测量

Temperature measurement for microreactor based on BPNN

针对常规测温方式应用于微小反应器的局限性以及温度精确测量的需求,提出基于反向传播神经网络(BPNN)的温度测量方法,通过应用目标温度预测值和目标温度测量值的差值处理降低并消除空白噪声引起的附加温度影响,分别进行了微小反应器内稳态、非稳态和温度分布的温度测量研究。结果表明:BPNN的预测值与测量值的吻合度高,差值均控制在±0. 05℃,其中25℃和40℃典型稳态工况的差值波动范围为±0. 01℃和±0. 02℃,远优于目标温度测量值的波动范围±0. 05℃和±0. 2℃;非稳态工况(过氧化氢酶反应)可分辨温度信号峰的过氧化氢浓度从0. 01%降低到0. 001%,温度测量水平提升了1个数量级;径向动态温度分布例证了温度的高辨析度能精准表达微小反应器内的传热传质特性。提出的方法能提高微小反应器温度测量的信噪比、分辨率和精度,降低热绝缘要求和成本,具有应用价值。

Aiming at limitation of routine temperature measurement mode applied to microreactor and requirements of accurate measurement of temperature,a temperature measurement method based on back propagation neural network（ BPNN） is put forward by applying difference of the predicted values and measured values of target temperature to process and eliminate effect of additional temperature caused by white noises. The method is studied for temperature measurement of steady condition,unsteady condition and temperature distributions in micoreactor respectively. The results show that the predicted value and measurement value fitness is high,difference is controlled within ± 0. 05,and the wave range is ± 0. 01 ℃ at 25 ℃,± 0. 02 ℃ at 40 ℃under steady condition,far superior to fluctuation range of ± 0. 05 ℃ and ± 0. 2 ℃ of target,temperature measurement value,under unsteady condition,catalase reaction,the concentration of hydrogen peroxide which can distinguish temperature signal peak reduce from 0. 01% to 0. 001%,temperature measurement level has an improvement of about one order of magnitude. Radial dynamic temperature distribution verify that with high resolution of temperature can accurately reflect heat and mass transfer properties in microreactor. The method can improve the signal-to-noise ratio,resolution and precision of temperature measurement in microreactor,meanwhile because it lower requirements of heat insulation and reduce the cost,it has application value.