基于神经网络的燃烧二维温度场测量模型

A Neural Network Based Two-dimensional Model to Be Used in Combustion Temperature Field Measurements

分析了彩色CCD摄像机的测量机理,推导了温度测量模型。由于测量模型的复杂性和非线性,因此基于BP神经网络在非函数逼近方面的特性,提出了用BP神经网络建立燃烧二维温度场的测量模型。由于CCD摄像机RGB值直接输入神经网络时测量误差大,同时考虑到比色测温的优点,所以采用了一种比值输入的神经网络测温方法。应用结果表明:以BP神经网络为测量模型进行燃烧二维温度场测量是可行的,选择合适的神经网络结构和相应的训练精度,测量模型具有较高的测量精度,具有工业推广价值。

A two-dimensional model, to be used in combustion temperature measurements with a CCD video camera, is being treated. The measuring mechanism of CCD coloured video cameras is analyzed and a model for treating temperature measurements derived. Due to the complexity and nonlinearity of measurement models, by taking advantage of BP neural network's specific property of approximation of nonlinear functions, a proposal is presented to use BP neural networks to makeup two dimensional measurement models for combustion temperature fields. Since considerable measuring errors may occur if RGB magnitudes of CCD video cameras were directly fed into neural networks, and in consideration of the advantages of colorimetic temperature measurements, a BP neural network with proportional input is used. Application results show that it is feasible to use BP neural networks for modeling purposes in two-dimensional matched by a corresponding training precision, the model can be made to acquire a relatively high precision of measurement, and thus be worthy of industrial application.