基于复杂网络和深度学习的两相流可视化与动力学建模研究

This project focuses on gas-liquid two-phase flow and complex network theory. In particular, we first will design a new three-dimensional distributed conductance and capacitance dual-mode sensor and conduct gas-liquid two-phase flow experiments to obtain multi-source measurement signals; then we will develop complex network and deep learning theory to realize multi-source sensor information fusion, aiming to visualize complicated flow structure, measure flow parameters and reveal the dynamical behaviors governing the transitions of different flow patterns. The research works include: Firstly, based on the sensitivity field information, we will develop a novel optimization solution for the design of three-dimensional distributed dual-mode sensors; Secondly, we will propose a wavelet multi-resolution multilayer network method to analyze the multivariate signals from distributed conductance sensor, with the purpose of studying the dynamical mechanics underlying the evolution and transitions of gas-liquid bubble flow, slug flow, churn flow and annular flow; Thirdly, we intend to propose a three-dimensional weighted network to analyze the multivariate signals from distributed capacitance sensor, which allows us to visualize complicated flow structure and probe into the instability mechanism of liquid film between gas slugs and the intermittent oscillation dynamical mechanism of the liquid slugs and gas slugs; Finally, we will develop a complex network-deep learning based soft-sensing model to measure the water volume fraction and phase flow rate in two-phase flow. In this project, based on the optimization design of three-dimensional distributed sensor and experimental data acquisition, we develop and apply the cutting-edge complex network theory to carry out interdisciplinary research, with the purpose of providing important information for the energy-efficient and incremental exploitation of gas field.

本项目以气液两相流为研究对象，在全新设计三维分布式电导电容双模态传感器并获取实验信号基础上，发展复杂网络和深度学习理论融合多源测量信息，旨在实现两相流三维可视化、复杂流动参数测量、流型演化动力学机制揭示。研究内容主要包括：建立基于灵敏场信息的三维分布式双模态传感器优化设计新方案；提出小波多分辨率分层复杂网络理论融合分布式电导传感器测量信息，研究气液泡状流、段塞流、混状流、环状流的生成及演化动力学机制；提出三维加权复杂网络理论融合分布式电容传感器测量信息，实现对两相流复杂流态的三维可视化，并研究气塞之间液膜的失稳机制及液塞与气塞间歇性振荡动力学机制；建立基于复杂网络和深度学习的复杂流动参数软测量模型，实现对油气井中不同地层段含水率和分相流量的测量。本项目以三维分布式传感器优化设计和动态实验为信息采集基础，以复杂网络前沿理论为主线，开展多学科交叉研究，为油气田节能增产开采提供重要指导信息。

本项目针对垂直井气液两相流，通过有限元分析方法设计了双模态传感器、三维分布式扇区电导传感器、Wire-Mesh传感器，该传感器系统可以有效捕获空间场域内复杂流动信息。研发了与传感器相配套的测量系统，开展动态实验获取不同流型下的多源测量信号。提出小波分层复杂网络分析理论对传感器多源信息进行融合，揭示了气液泡状流、段塞流、混状流、环状流的生成及演化动力学机制；提出加权复杂网络理论对分布式传感器多源信息进行融合，基于Wire-Mesh传感器测量数据和深度学习超分辨率技术，实现了对两相流复杂流动形态的可视化，揭示了气塞之间液膜失稳机制以及含泡液塞与大气塞间歇性振荡动力学机制；提出了多任务时间通道融合网络和分阶段密集连接网络两种复杂流动参数软测量模型，实现了对油气井中不同地层段复杂流动参数的精确测量。该研究为解决油气井中复杂流态三维信息获取、复杂流动机制揭示、流型可视化和流动过程参数测量问题提供了新途径。围绕上述研究成果，作为第一/通讯作者在IEEE Transactions on Industrial Informatics、Chemical Engineering Journal、IEEE Transactions on Instrumentation and Measurement、IEEE Transactions on Neural Networks and Learning Systems、IEEE Sensors Journal、IEEE Transactions on Circuits and Systems II、Petroleum Science、Chaos等期刊上发表SCI检索论文35篇；作为第一发明人，已获得13项国家发明专利授权。项目负责人获得国家优秀青年科学基金项目资助和天津市杰出青年科学基金项目资助，作为负责人获批金额为1600万元的国家重点研发计划课题，项目负责人入选全球高被引科学家、中国高被引学者、天津市中青年科技创新领军人才，获评2021年首届强国青年科学家提名（全国40人）。本项目研发的分布式传感器测量系统和提出的复杂网络深度学习多源信息融合方法，在天津大学模拟井装置上取得了很好的实验效果后，已在中石化胜利油田多个油区进行应用，并取得了显著的经济和社会效益。

内容获取失败，请点击重试