基于数据驱动的高含硫天然气净化脱硫过程故障检测与诊断

发布时间：2018-04-14 08:06

本文选题：高含硫天然气 + 多变量过程　；参考：《重庆科技学院》2015年硕士论文

【摘要】：随着现代工业系统复杂性和自动化程度不断提高，被控过程同时发生物理化学反应和相位反应，涉及物质转化和能量传递，综合受人、机、环、料、法不确定因素影响。整个生产过程表现不确定性、非线性、强耦合性、动态性等特点。传统基于机理和过程特性的故障检测与诊断方法受到极大限制。基于数据驱动的过程监控以反映系统运行状况的数据为基础，通过各种数据处理与分析手段，挖掘其内在规律，在线检测和识别过程中出现的异常操作和工况，追溯故障发生根本原因，从而为故障排查和系统恢复提供智能决策，最终保证复杂系统运行的可靠性和安全性。目前，，高含硫天然气净化过程主要存在以下三个问题：一是天然气处理量载荷波动会引起净化系统模型参数发生迁移，从而导致静态模型不能够识别正常工况调整而发生监控误报警。二是高含硫天然气净化过程监测数据结构呈现非线性、非高斯性和时序自相关性特点，导致提取驱动净化机理的过程参数显得异常困难。三是提取出净化过程的关键参数无法追溯原始参数贡献度，从而难以实现故障诊断。本文分别讨论主元分析过程监控和独立分量分析过程监控方法，并以美国田纳西-伊斯曼模型为标准测试库检验各种故障检测与诊断方法性能，然后应用这些方法解决实际高含硫天然气净化过程故障检测与诊断的问题。主要取得以下成果：一是针对静态模型无法识别工况调整而导致误报警高问题，提出基于假设检验和动态确定算法的自回归模型时滞阶次确定方法，研究动态主元分析和动态独立分量分析的监控性能。二是针对非线性、非高斯性和动态工业过程，提出基于动态核独立分量分析的故障检测与诊断方法，实现复杂工业过程监控。三是针对故障诊断困难，采用监控统计量对原始参数的一阶偏导数度量贡献度，提出基于统计量的一阶偏导数贡献图的故障诊断方法。最后以高含硫天然气净化过程为研究对象，采用动态核独立分量分析的故障检测与诊断方法，达到很好的监控性能效果；并针对故障诊断的异常参数，提出相应的安全控制措施。
[Abstract]:With the increasing complexity and automation of modern industrial system, the controlled process takes place in both physical and chemical reactions and phase reactions, which involve material transformation and energy transfer, and are comprehensively affected by uncertain factors such as human, machine, ring, material, and method.The whole production process is characterized by uncertainty, nonlinearity, strong coupling and dynamics.Traditional fault detection and diagnosis methods based on mechanism and process characteristics are greatly limited.The data-driven process monitoring is based on the data which reflects the system running condition. Through various data processing and analysis methods, the inherent rules are mined, and the abnormal operations and working conditions that appear in the process of on-line detection and identification are detected and identified.The root cause of the failure is traced back, which provides intelligent decision for fault troubleshooting and system recovery, and finally ensures the reliability and security of complex system operation.At present, there are three main problems in the purification process of high-sulfur natural gas: first, the fluctuation of natural gas treatment load will cause the migration of the model parameters of the purification system.As a result, the static model can not identify the normal condition adjustment and the monitoring error alarm occurs.The other is that the monitoring data structure of the purification process of high sulfur containing natural gas is nonlinear, non- and time series autocorrelation, which makes it very difficult to extract the process parameters that drive the purification mechanism.Third, the key parameters of the purification process can not be traced back to the original contribution, so it is difficult to achieve fault diagnosis.This paper discusses the principal component analysis (PCA) process monitoring and independent component analysis (ICA) process monitoring methods, and uses Tennessee Eastman model as the standard test library to test the performance of various fault detection and diagnosis methods.Then these methods are used to solve the problem of fault detection and diagnosis in the purification process of high-sulfur natural gas.The following results have been achieved:First, aiming at the problem of high false alarm caused by the adjustment of static model's unidentifiable working condition, a method of determining the time-delay order of autoregressive model based on hypothesis test and dynamic determination algorithm is proposed.The monitoring performance of dynamic principal component analysis and dynamic independent component analysis is studied.Secondly, aiming at nonlinear, non- and dynamic industrial processes, a fault detection and diagnosis method based on dynamic kernel independent component analysis (DKICA) is proposed to realize the monitoring of complex industrial processes.Thirdly, aiming at the difficulty of fault diagnosis, a fault diagnosis method based on the first order partial derivative contribution graph based on statistics is proposed by using the first order partial derivative of the monitoring statistics to measure the contribution degree of the original parameter.Finally, the method of fault detection and diagnosis based on dynamic kernel independent component analysis (DKIA) is used to study the purification process of high sulfur containing natural gas, which achieves a good monitoring performance effect, and aiming at the abnormal parameters of fault diagnosis,Put forward the corresponding safety control measures.
【学位授予单位】：重庆科技学院
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE64

【参考文献】