论文部分内容阅读
为改善集输立管系统流型识别实时性,本文初步提出了一种基于非完整周期信号的识别方法。从严重段塞流过程中液塞生长、液塞流出、气液喷发、液塞回落的某一阶段或相邻两阶段选取较短时长的压力、压差波动信号序列,采用计算速度较快的方法进行特征提取。通过自组织神经网络融合局部流动特征,描述集输立管系统空间流动特性,实现流型识别。本文实验流速范围内,该方法对稳定流型与不稳定流型的平均识别率达到90%以上,初步验证了该方法的可行性。
In order to improve the real-time identification of manifold-type riser system, this paper presents a preliminary identification method based on non-complete periodic signal. In the process of serious slug flow, the liquid slug growth, liquid slug flow out, gas-liquid eruption, liquid slug down or the adjacent two stages of the pressure and pressure fluctuation signal sequences of relatively short duration are selected, Method for feature extraction. The local flow characteristics are fused by using self-organizing neural network to describe the spatial flow characteristics of riser system and realize the flow pattern recognition. The experimental results show that the proposed method achieves an average recognition rate of more than 90% for both steady and unstable flow patterns, which proves the feasibility of the proposed method.