神经网络在标准热电偶测温中的应用

热电偶在工业测温中有着广泛应用，但其温度和热电势之间呈非线性关系，实际应用中需要采用标准化热电偶并存储和查找分度表来得到实测温度，这样既限制了测温精度，又使其在热电特性发生变化后很难再用。本文利用人工神经网络自学习、自适应的特点以及多层前向神经网络的函数逼近能力来解决热电偶温度计算的问题，并给出了两层前向神经网络和BP算法计算热电偶温度的详细过程。结果表明，这种计算方法准确性高、实用性强。

Application of Neural Networks in Measuring Temperature of Standard Thermocouple

Thermocouple is widly applied in industry temperature measurement, but it is not a linear relation between its temperature and potential. Temperature is got by using the standard thermocouple and looking up temperature-voltage table in application, which not only limits measure precision but also makes it very difficult to use its after thermoelectric characteristic changed. To solve above problem this paper uses the studying and adaptive characteristic of artificial neural network and ability of multilayer feedforward networks with backpropagation algorithm, provides a detailed course of calculating temperature with twolayers feedforward networks and BP algorithm. The fact shows that this computing method has good reliability and practicability.