基于条件化证据线性组合更新规则的工业报警器优化设计方法

发布时间：2018-05-30 12:26

本文选题：报警系统设计 + 静态收敛指标　；参考：《杭州电子科技大学》2017年硕士论文

【摘要】：工业过程中主要过程变量的变化可以反映被监控设备的运行状况。报警器的作用是通过对过程变量采样信号的处理,并将其与报警阈值比较,对设备异常状态进行监测。在报警器设计中,学者普遍都把误报率(FAR)、漏报率(MAR)和平均延迟时间(AAD)作为衡量报警器性能的指标。在过程变量统计分布已知的假设下,传统的报警器设计方法通常是基于前两个指标来优化报警器的阈值等参数。由于设备实际运行及状态监测中存在的各种不利因素影响,使得过程变量的统计分布难以准确获取。Dempster-Shafer(DS)证据理论在对不确定性信息的表示、推理和综合处理方面相对于概率论具有其自身的优势。已有学者将信息融合思想引入报警器设计当中,给出了基于报警证据更新/融合规则的报警器设计与优化方法,取得了初步研究成果。本文对报警器设计中的报警证据生成、适用于报警证据的性能指标制定以及报警证据参数优化问题展开更为深入的研究,以增进证据理论在工业报警器设计中的深度应用,主要工作如下:(1)基于Sigmoid函数的报警器证据生成方法。在利用传统分段梯形模糊隶属度函数实现过程变量到相应报警证据的变换时,由于使用了分段函数,难免造成过程变量所含信息的损失。针对此问题,提出基于连续型Sigmoid(S)函数的报警证据生成方法,并通过理论证明和仿真数据统计实验说明该种转换是一种对过程变量所含信息的等价变换。(2)基于静态收敛指标的报警证据优化方法。基于Jousselme证据距离,定义报警证据概率赋值静态收敛指标(SI),并进一步分析证据生成时S函数中的参数与SI的对应关系,以及报警器阈值、FAR/MAR与SI的对应关系;以此为基础,引入对报警证据的精细化折扣,设计关于SI的目标函数,通过对当前时刻所获报警证据的折扣向量的优化及S函数参数的调整提升报警证据的可靠性。(3)基于动态收敛指标的条件化报警证据线性组合更新方法。给出动态收敛指标(DI)的定义,在静态收敛指标优化的基础上,设计基于动态收敛指标的报警证据更新及参数优化方法。通过与传统报警器设计方法和线性组合证据更新方法的对比实验分析,说明本文所提方法的优越性。
[Abstract]:The variation of the main process variables in the industrial process can reflect the operation status of the monitored equipment. The function of the alarm is to monitor the abnormal state of the equipment by processing the process variable sampling signal and comparing it with the alarm threshold. In the design of the alarm system, the false alarm rate, false alarm rate (false alarm rate) and average delay time (AAD) are generally regarded as indicators to measure the performance of the alarm. Under the assumption that the statistical distribution of process variables is known, the traditional alarm design method is usually based on the first two indicators to optimize the alarm threshold and other parameters. Because of the influence of various adverse factors in the actual operation of the equipment and the condition monitoring, it is difficult for the statistical distribution of the process variables to obtain the exact representation of the uncertain information in the evidence theory of .Dempster-Shafern DSs. Reasoning and comprehensive processing have their own advantages over probability theory. Some scholars have introduced the idea of information fusion into the design of alarm device, and presented the design and optimization method of alarm device based on alarm evidence update / fusion rules, and obtained preliminary research results. In this paper, the generation of alarm evidence in the design of alarm system is studied more deeply, which is suitable for establishing the performance index of alarm evidence and optimizing the parameters of alarm evidence, so as to enhance the deep application of evidence theory in the design of industrial alarm. The main work is as follows: 1) the method of alarm evidence generation based on Sigmoid function. When the traditional piecewise trapezoidal fuzzy membership function is used to realize the transformation of process variables to corresponding alarm evidence, the information contained in process variables is inevitably lost because of the use of piecewise functions. In order to solve this problem, an alarm evidence generation method based on continuous Sigmoid function is proposed. It is proved by theory and simulation data statistics that this conversion is a kind of equivalent transformation of information contained in process variables. It is an alarm evidence optimization method based on static convergence index. Based on the evidence distance of Jousselme, the static convergence index of probability assignment of alarm evidence is defined, and the corresponding relation between the parameters of S function and SI, and the corresponding relation between alarm threshold and SI is analyzed. Introducing a refined discount on alarm evidence to design a target function for SI, By optimizing the discounted vector of the alarm evidence obtained at the present time and adjusting the parameters of the S-function, the reliability of the alarm evidence is improved. The linear combination updating method of conditional alarm evidence based on dynamic convergence index is proposed. On the basis of static convergence index optimization, an alarm evidence updating and parameter optimization method based on dynamic convergence index is designed. The advantages of the proposed method are illustrated by comparing with the traditional alarm design method and the linear combined evidence updating method.
【学位授予单位】：杭州电子科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP277

【参考文献】