基于自组织神经网络的燃烧诊断研究

Simulation on Combustion Diagnosis Based on Self-Organized Neural Networks

自组织神经网络原理被尝试应用到燃烧诊断系统中．网络的输入是从稳定和非稳定燃烧工况下获取的火焰辐射信号的频谱估计值．经过自组织训练后，网络对不同燃烧工况下的输入具有明显不同的输出，通过验证证实了这种方法能对检测到的燃烧火焰信号进行有效的处理，从而获取燃烧状态稳定与否的信息．

In this paper, the self-organized neural networks were applied into a diagnostic system for combustion.The input signal of the neural networks was the power spectrum estimation of the flame signal from stable and unstable combustion states. Through trained by self-organization , the networks had different output maps for flame signals detected from stable and unstable combustion states. Verification showeed that this way can process efficiently the combustion flame signal detected, and such information can be obtained that whether the combustion state is stable or not.