Improved global model for predicting gas radiative properties over a wide range of conditions

Abstract

Calculating radiative transfer in gas mixtures, especially those containing CO2 and H2O, is important for high temperature applications, such as boilers and furnaces. Of all the models that can perform such calculations, the weighted-sum-of-gray-gas (WSGG) model is currently the most widely used one for solving engineering problems, due to its excellent computational efficiency. However, the validity range of a WSGG model is usually limited because its accuracy strongly depends on the conditions at which the model coefficients are generated. In this work, a general model is developed for predicting the total emissivities of gas mixtures containing CO2, H2O, and any non-absorbing gases. The model can be used for a wide range of conditions: temperatures 300–2900 K, pressures 1–60 bar and path lengths 0.01–100 m for any mixture compositions. The model is developed based on artificial neural networks (ANN) trained using around 4.5 million samples generated by a statistical narrow band model. Two ANNs were selected—considering both accuracy and computational speed—and were compared against existing WSGG models. The first ANN model has more than an order of magnitude smaller mean error than those of existing WSGG models, and its computational speed is comparable to the WSGG model. The second ANN model has even higher accuracy than the first model, but with slightly lower computational speed.