针对发动机缸内进气流场由于不可视和动态而导致其难以无损在线测量的难题，提出了柴油发动机进气流场的层析图像诊断方法． 分析了运用工业计算机层析成像系统获取进气过程中的柴油发动机缸内进气流场的层析图像原理，讨论了缸内进气流场中的气体迹点、密度、流速及其变化在层析图像上的行为规律，研究了最大平均互信息的缸内进气系统层析图像分割预处理方法，论述了基于随机变量互信息的缸内进气流场多目标特征图像提取方法． 通过单缸立式柴油机缸内进气流场的高能工业计算机层析成像诊断实验，无损和实时地获取受限空间内的气缸内流场气体迹点、密度、流速及其变化状态信息，发现了流动过程中气体与气缸、气体与气体之间相互作用而形成的层析链结构及其形态． 与现有缸内进气流场的其他诊断方法相比，文中方法具有无损、在线和多维诊断的特点．

In order to overcome the difficulty in the online and non-destructive measurement of engine cylinder's intake airflow field caused by its invisibility and dynamic characteristics,a tomographic image diagnosis method is proposed.In the investigation,first,the principles of industrial computer tomography ( ICT) for obtaining the tomographic image of diesel engine cylinder's intake airflow field are analyzed.Then,the behaviors of airflow field's trace point,density,velocity and velocity variation in the tomographic image are discussed,and the segmentation pre-treatment methods of the cylinder intake system's tomographic image are explored based on the maximum average mutual information.Moreover,an extraction method of multi-target image features ofthe intake airflow field is presented based on the mutual information of random variables.Finally,the information of the airflow field's trace point,density,flow velocity and velocity variation status inside confined spaces are non-destructively extracted in realtime from the high-energy ICT images of a single-cylinder vertical diesel engine,and the tomographic chain structure and its shape formed by the gas-cylinder and gas-gas interactions in the flow process are discovered.It is found that the proposed tomographic image diagnosis method is more suitable for the nondestructive,online and multi-dimension diagnosis than the existing methods.