基于虚拟传感与故障机理的油气设备安全预测理论及模型研究

The harsh operating environment and complex structure of key equipment in petrochemical industry pose the limitations on the number and installation locations of physical sensors to timely acquire the equipment status, and also cause the difficulty of effective fault diagnosis and non-linear performance degradation prediction. Thus, the equipment presents high operational risk and downtime rate. To improve fault diagnosis and prognosis, a key equipment in petrochemical industry - flue gas turbine, is selected as research object in this study. First, virtual sensing technology is investigated to effectively acquire equipment status and improve fault identification through configuration optimization and information selection of physical sensors. Next, a fault prognosis model with the focus of uncertainty analysis is presented based on particle filter by coupling failure mechanism and hybrid sensing (e.g. virtual sensing and physical sensing). The key research questions are identified as: dynamics modeling and defect propagation mechanism of fault components, improved particle filter algorithm for long-term prediction. By means of the presented virtual sensing method and the fault prognosis model, equipment safety prediction theory is built based on virtual sensing and fault mechanism. The presented model and theory are then validated though the experimental data from simulating fault, natural progression fault and field test. It is anticipated to provide new idea and new method for early fault diagnosis and prognosis of petrochemical equipment. It is of significance to ensure the equipment safety and make maintenance strategy in petrochemical industry.

石化关键设备结构复杂、工况恶劣，运行风险高、故障频发，而物理传感安装受限造成监测信息获取难、故障征兆发现难和非线性趋势预测难。针对这一问题，本申请课题以故障频发炼化设备烟气轮机为研究对象，在优化物理传感布局与信息选择的基础上，开发适用于设备安全预测的虚拟传感技术，提升有效监测信息的获取手段和故障征兆的辨识方法；利用粒子滤波器融合故障部件失效机理和混合传感信息（虚拟传感和物理传感），提出针对趋势发展不确定性分析的故障预测模型。重点研究粒子滤波模型构建中的故障部件动力学建模与故障发展机理，以及面向长周期预测的粒子滤波改进算法等关键科学问题。最终通过模拟故障、自然损伤和现场测试三种方式，构建基于虚拟传感与故障机理的设备安全预测理论及模型，从而提升安全检测手段和故障预测技术。项目预期为油气设备早期故障全方位预示提供新思路、新方法，对保证设备安全可靠运行和维修决策具有重要的理论和现实意义。

本项目针对油气关键设备中存在样本小、非线性、运行工况复杂、有效监测数据获取困难等特点，重点研究了基于虚拟传感与故障机理的设备故障预测方法，并成功应用于油气设备关键部件的状态预警及故障预测中。首先，针对油气关键设备运行过程中退化样本小的特点，通过整合直接传感与间接传感技术，提出了基于极限学习机的虚拟传感监测框架，研究了易测在线辅助变量与难测关键变量之间的关联关系，从而实现设备运行状态的实时在线监测。其次，建立了转子系统的三维仿真模型，研究了转子系统在发生不平衡故障时，不平衡质量变化以及不平衡位置变化对仿真结果的影响，并通过试验用于转子系统不平衡故障的识别与研究，验证了仿真分析的有效性。再者，提出了一种整合粒子滤波与虚拟传感技术于一体的增强预测方法，以一种更泛化的解析形式有效表征设备退化状态与融合观测特征之间的非线性关系，从而改善设备状态预测的精度；最后，提出了基于最大期望算法的非线性在线辨识方法，可根据最新可用观测对预测方法中模型参数进行自适应更新与修正；此外，进一步研究了模型参数的分布范围对长周期预测中不确定性的影响因素，并通过置信分布与统计方法量化不确定性，从而为关键设备的预测预警及维修决策分析提供支持。