面向系统层级多工况下带钢热轧全流程质量监测与故障诊断方法研究

Associated with the continuously increasing system automation, integration and complexity degrees for modern hot strip mill processes, safety and stability of product quality are essential for the economic benefit. Due to the complex operation mechanism, large scale, multiple operation conditions and multi-level cooperative associations, it’s generally difficult to realize process monitoring effectively based on traditional fault diagnosis methods. As a result, fault detection and diagnosis approaches for plant-wide processes as alternatives become more and more popular. Motivated by these observations, the main objectives of this project are to study advanced theories and methodologies of quality-related fault detection, root cause diagnosis, propagation mechanism analysis, and quantitative evaluation based on data-driven and knowledge supervision as supplementary methods for plant-wide hot strip mill process. To be specific, the issues addressed in this project include: (1) Decomposition and condition division for plant-wide processes on system levels; (2) Quality monitoring for plant-wide processes under three-dimensional perspectives of multi-level, plant-wide and multi-condition; (3) Root cause diagnosis, propagation path identification and quantitative evaluation for quality-related fault under multiple conditions; (4) Application verification. The project seeks to construct a quality monitoring and fault diagnosis framework, which will provide theoretical basis and technical support for safety monitoring and independent security of hot strip mill process as the representative of process industries.

现代带钢热轧生产过程日趋规模化、复杂化、集成化，产品质量的安全、稳定是企业实现经济效益的关键。由于带钢热轧过程机理复杂、规模庞大、工况复杂多变、多层级协作关联等原因，传统的故障诊断方法难以实现有效的过程监控，使得全流程故障检测与诊断技术日益受到关注。为此，本项目以多层级多工况下带钢热轧全流程为背景，以数据驱动为主，知识监督为辅的思路研究带钢热轧全流程质量相关故障检测、故障根源诊断、故障传播机理解析与故障定量评估等理论与技术，拟研究和解决以下几方面问题：1）面向系统层级的带钢热轧全流程分解与工况划分；2）“多层级、全流程、多工况”三维度带钢热轧全流程质量监测；3）多工况模式下带钢热轧全流程质量相关故障的根源诊断、传播路径识别与定量诊断；4）应用验证。旨在建立一套全流程质量监测与故障诊断理论与方法，为以带钢热轧为代表的流程工业过程的安全监控及自主保障提供理论支撑与技术保证。

为适应市场对多品种、多规格、高质量的定制化生产需求，现代流程工业正朝着高效、大型和集成化方向发展。通过过程监测来保障流程工业的安全可靠运行以及连续稳定的产品质量逐渐成为过程控制的首要任务。复杂流程工业具有规模庞大、工序级联、层级协作、工况频变、故障多源多征兆、质量多环节耦合的特点。单一数据驱动的建模与监测方法难以深度解析质量相关全景监测，且无法准确及时判定质量相关故障的原因。本项目以带钢热连轧这一典型的流程工业过程为研究背景，采用数据驱动思路从“三个维度”（系统层级、全流程、多工况）研究了全流程质量的全景监测、质量异常成因追溯、故障传播路径识别及故障程度评估理论与技术。主要解决了：1）如何在系统层级与工艺知识指导下实现全流程分解与工况划分，2）如何设计面向系统层级与多工况的全流程质量全景监测方法，3）如何利用定量技术追溯质量异常成因和故障传播路径识别以及多源故障诊断，4）如何将基于数据的定量分析与基于知识的定性分析相结合，建立兼顾不同层级的不同因素和层级信息的质量在线定量诊断与评估的关键技术问题，从而为以带钢热轧为代表的流程工业质量相关的过程监测与故障诊断提供了一套新技术和解决方案。项目执行期间共发表SCI论文40篇、EI论文19篇，授权发明专利3件。培养博士后4人、博士生10人、硕士生21人。上述研究成果在鞍山钢铁集团公司带钢热连轧生产线和河北视窗浮法玻璃生产线上进行了实际应用验证，取得了良好的过程监测效果。在应用前景方面，研究成果将为以带钢热轧为代表的流程工业过程的安全监控及自主保障提供理论支撑与技术保证。