化工过程故障监测与状态评估的集成化方法研究

For complex chemical process, strong production continuity、complicated process flow and multi-monitored-variables make the operating condition always unstable, fault monitoring and state assessment are major means to assure safety in complex chemical process, ensure the quality of products and reduce energy consumption. Also, studying for the theory and methods of integration would become an effective approach to improve the system performance of fault monitoring and state assessment in chemical process. In this project, methods of feature extraction for space and time characteristics, monitoring models for coordinating local and global structures and monitoring statistics for description of data multi-peak distribution are developed to improve the accuracy of fault monitoring in varying operation process. Constructing intelligent computing methods based relational model of contribution and variables, exploring fault location and decision based on the fusion of traditional statistical technique and mechanism knowledge are to better the efficiency of fault location and decision. This project researches for strategy of operating state assessment in multivariable system, obtaining the off-line guidance information for controller/actuator adjustment to keep the benefit during the working of chemical process.

复杂化工过程的生产连续性强、工艺流程复杂、监控变量多，运行工况并不总是稳定不变的，故障监测与状态评估是保障复杂化工过程运行安全，保证产品质量，降低能耗的重要手段。研究集成化的理论与方法是提高化工过程故障监测与状态评估系统性能的有效途径。本课题研究空间与时间特征提取方法，建立协调局部结构与全局结构的监控模型，构建描述多峰分布数据的监控指标，以期提高变工况过程故障监测的准确性。基于智能计算方法构造贡献度与变量的关系模型，探索传统数据统计方法与化工过程机理知识相融合的故障定位及决策方法，以期提高故障定位的精度与决策的效能。研究多变量系统的运行状态评估策略，获取控制器/执行机构调整的离线指导信息，以期保持化工过程运行的效益。

本项目通过系统分析化工过程中生产数据的特征，深入研究时间与空间特征提取方法、结合数据集局部结构与全局结构的故障监测模型构建方法、变工况过程监控指标构建方法，对具有变工况的过程进行精确有效的故障检测；深入研究将定性知识与定量知识融合进行准确定位导致系统故障变量的方法，并对不同类型的故障变量制定相应的决策方案；深入研究建立多变量过程运行状态评估指标的方法以实现过程运行状态评价。通过深入研究分析，取得了一些具有创新性的研究成果，总结整理后已分别在国内外专业期刊、学术会议发表。此外，本项目的研究成果已经成功应用于宝武钢铁集团的炼焦生产过程。