基于声学层析成像的炉内温度场重建算法研究

摘要

为了获取快速准确的锅炉温度场在线监测信息,提出了一种基于声学层析成像(AT)测量的代数重建-神经网络(ART-NN)温度场重建算法,该算法结合了代数重建方法(ART)和BP神经网络方法的优势.采用该算法对多种典型的温度场模型进行数值仿真,分析了算法的重建结果和稳健性,并利用实验研究进一步评估该算法的性能.结果表明:ART-NN算法的重建质量和稳健性要优于Tikhonov正则法、Landweber迭代法和ART方法,为提高锅炉炉膛温度场重建质量提供了一种有效方法.%To fast and acurately obtain the temperature distribution information in real-time monitoring of a boiler furnace,an ART-NN temperature distribution reconstruction algorithm was proposed for the acoustic tomography (AT) measurement by integrating the advantages of algebraic reconstruction technique (ART) and back propagation neural network (BPNN),which was subsequently used to reconstruct a variety of typical temperature distribution models,and then the reconstruction results and robustness were analyzed.Meanwhile,to further evaluate the performance of the proposed algorithm,experimental studies were conducted.Results show that both the reconstruction quality and the robustness of ART-NN algorithm are superior to that of the Tikhonov regularization,Landweber iteration and the ART method,which therefore may serve as a reference for quality improvement of temperature distribution reconstruction of boiler furnaces.