Combining FAP,MAP and correlation analysis for multivariate alarm thresholds optimization in industrial process

In the real industrial process, alarm threshold optimization is an important part of alarm system rationalization. If the design of alarm threshold is unreasonable, it would result in nuisance alarms, among which the critical alarms are overwhelmed. In order to alleviate this phenomenon, we propose a method of multivariate alarm thresholds optimization to reduce the nuisance alarms. Firstly, causal relationship between process variables is constructed based on the time delay estimation method, thus we can determine the alarms propagation path and then select the optimized variables. Secondly, in order to guarantee both the process safety and correlation consistency, three factors - false alarm probability (FAP), missed alarm probability (MAP), and the correlation between the alarm information and process information – are combined to establish the objective function of the optimization process for the first time. Then, the optimal thresholds are obtained by the genetic algorithm. Finally, the validity and effectiveness of the developed method are illustrated by the Tennessee Eastman process.     

A data similarity based analysis to consequential alarms of industrial processes

The analysis of consequential alarms is beneficial to avoiding alarm flooding and finding out root alarms in an industrial process. In this context, a novel similarity computation method taking into account of correlation delays between process alarms is introduced firstly. Subsequently, the Granger causality method is suggested to further clarify mutual impacts of similar alarm variables based on process data. Through the combination of alarm data similarity analysis and process data causality analysis, the consequential alarms can be effectively identified along with their evolution paths. An industrial case is employed to illustrate the benefits of the contribution.     

基于数据驱动的石化装置静态报警抑制策略研究

由生产状态变化引起的误报警频发,为解决现有针对基于状态报警的抑制方法缺乏完整性与定量分析的问题,提出了新的报警抑制策略。通过分析报警记录等数据库,对引起报警的生产条件进行结构化整定并量化,利用分类筛选与数据过滤找出基于状态的报警,再结合关联性分析和概率判断,建立了基于数据驱动的静态报警抑制策略。使用现场数据的试验证明了该方法的有效性。     

Integrating probabilistic signed digraph and reliability analysis for alarm signal optimization in chemical plant

The alarm system given in industrial plants are massive and complex. Under such condition, critical alarms are overwhelmed by false and unnecessary alarms and thus result in severe safety issues. To address the problem, this paper proposes a probabilistic signed digraph (PSDG) based alarm signal selection method that requires achieving maximal system reliability. In this method, a PSDG model is firstly constructed to visualize the causal relations between process variables. Then the criteria of observability and identifiability are imposed to determine the candidate alarm variables that can qualitatively distinguish all assumed faults. Instead of selecting the minimum number of combinations of candidate variables, the alarm variables are optimized by a reliability formulation that takes into account the missed alarm and false alarm probabilities of the system; this formulation is solved by the receiver operating characteristic (ROC) graph. Finally, the developed methodology is illustrated using a Tennessee Eastman process.     

基于贝叶斯估计的多探测器火警判定方法研究*

为应对感烟探测器的大量误报对消防应急响应带来的挑战,考虑目前以感烟探测器为主的火警设施误报率高且短期内难以全部更换的特点,提出基于贝叶斯估计的多探测器火警判定方法,通过多个探测器的报警时间间隔计算火源位置的后验概率分布,并提出火警真实度概念,为火警判定提供依据。结果表明:使用多探测器耦合模型时每增加1个探测器可将误报率降低约4个数量级,该方法在探测器正常、部分失效、误报的情景下均能有效判别火警。     

Alarm management optimization in chemical installations based on adapted HAZOP reports

Most current alarm systems used in chemical installations show poor performance due to alarm flooding. This study focuses on alarm management systems optimization using the deviation propagation relationship hidden in the hazard and operability study (HAZOP) report, which can be transformed into a critical information source for alarm optimization management. More concretely, this means matching the alarm tag number with the process deviations in the deviation column, possible cause column, and consequence column. Furthermore, a backtracking method and a reasoning method were established to identify the initial alarm and associated alarms. Besides, a root fault diagnosis was carried out. A method of detecting hardware faults and unreasonable alarm thresholds is established using alarm causality corresponding to the deviation causality and associated alarm generation-skipping tracing method. According to the severity of the consequence corresponding to the deviation, a determined alarm priority method is constructed. The results show that the deviation propagation relationship in the HAZOP report is clear, and the topological relationship is easy to build based on the deviation propagation relationship. With comprehensive and in-depth HAZOP analysis reports in China, the alarm management optimization technology based on adapted HAZOP reports shows good prospects for application and promotion.     

基于BP神经网络的受限空间火灾联合探测方法

针对航空受限空间火灾探测高误报的问题,在现有技术成果基础上对多种火灾探测方式进行研讨,并提出1种基于BP神经网络技术的飞机机身内部受限空间火灾联合探测报警方法。该方法结合现有烟雾感应、气体传感器探测等常用火灾探测技术,以红外热成像探测为辅助手段,采用神经网络实现数据融合,对模拟实验舱火灾烟雾进行联合探测,在单一火灾探测方式基础上提高了探测准确率。     

Data classification and performance evaluation for the most commonly-used univariate alarm systems

Alarm systems are critically important for safe and efficient operations of industrial plants, but many industrial alarm systems are suffering from too many nuisance alarms. This paper proposes a method to classify normal and abnormal data segments and evaluate performance indices for the most commonly used univariate alarm systems. The proposed method consists of three steps. First, piece-wise linear representations are exploited in separating historical data samples of an analog process variable configured with alarms into data segments with same qualitative trends. Second, data segments are classified into normal, abnormal and unclassified conditions via a mean hypothesis test; a required assumption is that data segments in normal and abnormal conditions have different mean values being distinguishable from alarm thresholds. Third, based on the normal and abnormal data, performance indices of univariate alarm systems are calculated, including two newly formulated ones as the false alarm duration ratio and the missed alarm duration ratio. The effectiveness of the proposed method is illustrated by numerical and industrial examples.     

Fault diagnosis of pipeline and pump unit systems using status coupling analysis

Earlier studies on fault diagnosis of the pipeline and pump unit systems (PPU) relied mainly on independent equipment analyses, which usually lead to false alarms because of the loss of information fusion. The aim of this study is to utilize the status coupling relationship to improve fault detection sensitivity and reduce false alarm rate. A real-time status identification of related equipment step is added between capturing abnormal signals and listing out diagnosis results. For example, when the pipeline pressure fluctuation is found abnormal, a status analysis of pump units is performed immediately, if the pump units are proven to be operational normally, then the pipeline leak alarm is acknowledged valid. The logical reasoning algorithm is used to capture abnormal conditions of pipeline pressures. The pump unit faults are captured by combining information from multiple sources. Field applications show that the proposed method significantly improves the PPU fault detection capability on fault detection sensitive and accuracy.     

Model-based information fusion investigation on fault isolation of subsea systems considering the interaction among subsystems and sensors

The offshore oil industry has expanded to deep water and Arctic. The harsh operating conditions (e.g., ice and strong wind) and increasing complicated system raise the occurrence likelihood of system faults. This requires timely fault isolation and management in the subsea system. However, the offshore oil industry mainly relies on humans to isolate faults based on alarms. With harsh operating conditions and increasing complicated system, this industry urgently needs research on more efficient fault isolation and cause diagnosis methods. Unfortunately, limited research is conducted on fault isolation method in the offshore oil industry. Furthermore, in industry 4.0 era, large amounts of information are obtained. This provides precondition for the application of information fusion technique which aims to improve diagnosis results. However, to the authors’ knowledge, information fusion has not been much studied in the fault isolation of the offshore oil industry. Moreover, the interaction of different subsystems contains valuable information. How the interaction of different subsystems can influence the fault diagnosis has not been explored. This paper proposes a Bayesian network (BN) based method for timely fault isolation and cause diagnosis for the offshore oil industry. The work fuses different information, and it also includes the dependency among different subsystems in the fault diagnosis. As an important alarm source, false alarms are also taken into account in the model. A case study on the subject of the subsea wellhead and chemical injection systems is conducted to demonstrate the functions and merits of the proposed method.     

A risk-based approach to design warning system for processing facilities

Alarm flooding is a major safety issue in today's processing facilities. Important recommendations are available for alarm management; however, they are often violated in practice, especially in the alarm systems implemented through the distributed control system. An effective process alarm prioritization and management system is desired for a safe and effective operation of a process facility.In present work, authors address two main issues related to an alarm system – the reliability and the prioritization of the alarms. The main objective is to deal with the alarm-flooding problem in process facilities. A multi alert voting system based on sensor redundancy approach is proposed to improve the reliability. A quantitative risk-based alarm management approach is proposed to address the flooding issue. In the risk-based approach, an integrated model consisting of the probability (P), the impact (I) of the potential hazards, and the process safety time is proposed to prioritize these raised alarms.The proposed approach is further explained by a reactor system with pressure and temperature variable monitoring and controls, where the hazards associated with two alerts caused by over high pressure and over high temperature are analyzed and integrated with response time for alarms generation and prioritization.     

多传感器耦合区域火灾报警技术研究

早期预警与低误报率一直是建筑火灾探测面临的挑战与难题。已有研究多针对特殊场所特定燃烧产物或多种传感器耦合,普适性不强,探测设备成本高,无法大规模应用。通过将火灾烟气蔓延规律与探测器信号时空分布融合,在不增加探测器数量和分布的情况下,提出了一种基于建筑结构微元的多传感器耦合区域火灾报警模型。对典型火灾场景的烟气蔓延情况进行了模拟分析,狭长结构中探测器信号强度变化具有一定的规律性。应用区域火灾报警模型后,报警时间较传统模式提前了14.7%,基本杜绝了单个探测器误报引发建筑物火灾报警的问题。结果表明:多传感器耦合探测模式显著缩短了火灾报警时间,降低了火灾探测误报率,实现了火灾的早期准确识别与预警。     

基于数字锁相技术的飞机货舱火灾烟雾光学参数方法研究*

为解决飞机货舱火灾光学探测易受环境杂光和空气中灰尘、水汽干扰,误报率高的问题,提出1种基于数字锁相的双波长火灾烟雾探测方法。利用数字锁相技术调制双波长光源,选择性地提取微弱光散射信号,并设计前、后双向散射光探测光路。采用常见火灾材料及干扰源进行真假火源实验,依据米氏散射理论及不同粒径气溶胶颗粒对不同波长光的散射效应差异,以不对称比（AR）和双波长光强比（DWIR）作为火灾检测参数对不同粒径颗粒进行区分,根据2种火灾检测参数建立数据样本,利用BP神经网络进行分析,并加入复合式传感器实验作为对比。研究结果表明:基于数字锁相技术的火灾烟雾光学探测方法能够在一定程度上区分火灾烟雾和干扰源,误报率小于3.5%。     

水蒸气引起感烟火灾探测器误报研究

针对水蒸气这种典型的火灾探测误报源．在自行研制的火灾探测综合模拟实验平台中,通过收集水沸腾形成的水雾模拟水蒸气干扰源,研究了冬季与夏季,即不同环境温度务件下,水蒸气对散射型光电感烟火灾探测器的误报影响。通过比较感烟火灾探测器的输出值与其附近湿度变化曲线,分析了误报发生的原因。结果表明在气温较低的冬季,水蒸气较易形成液态小水珠,容易导致感烟火灾探测器误报;而在夏季由于气温较高,这种误报不易发生。     

基于化工生产过程历史数据的报警优化方法研究

为了有效降低报警装置漏报率和误报率,基于化工生产的过程历史数据,探讨了报警优化的两种方法。以误报率和漏报率为优化指标,建立报警阈值目标函数,从最小化目标函数的角度出发,采用数值优化的方法进行报警阈值的优化设计。基于报警时长的变异系数,应用自适应报警延时方法来降低报警数量。实例研究表明,应用基于化工生产过程历史数据的方法,可有效减少无效报警数量,降低误报率和漏报率,提高报警质量,对于提高石化生产过程的安全性具有积极意义。     

Evaluation and comparison of alarm reset modes in advanced control room of nuclear power plants

An automation function has been widely applied in main control room of nuclear power plants (NPPs). The alarm system of fourth nuclear power plant (FNPP) in Taiwan is also going to be developed with automatic technology that is expected to support the operators’ performance and reduce the number of alarms. In this study, an experiment with a training simulator as an advanced alarm system was conducted to compare the effects of different alarm reset modes on performance and subjective ratings. The objective was to evaluate the practicability of alarm system with only auto-reset function in FNPP. Results revealed that, using the auto-reset mode, participants had lower task load index (TLX) on effort in the first test trial and was more satisfied under multi-task condition. In contrast, using manual reset mode, participants were more satisfied on alarm handling, monitoring, and decision making. In other words, both reset modes are necessary to assist the operator in different aspects, but with only single reset mode is insufficient. The reset function in advanced alarm system therefore should be very flexible.     

潮湿引起感烟火灾探测器误报研究

针对潮湿现象成为典型的火灾探测误报源 ,在火灾探测综合模拟实验平台中 ,通过收集水沸腾形成的水雾进行加湿 ,研究了冬季与夏季潮湿环境 ,对散射型光电感烟火灾探测器的发生误报影响。通过比较探测器的输出值与其附近湿度变化曲线 ,从原理上分析了误报发生的原因。实验结果表明 :在气温较低的冬季 ,水蒸气较易形成液态小水珠 ,容易导致感烟探测器误报警 ,而在夏季由于气温较高 ,这种情况较难发生。对相对湿度较高或达到饱和湿度 ,以及非火灾因素的雾气或微细水珠 ,对感烟探测器可靠性的影响 ,开展下进一步的工作进行了展望     

A design method of a plant alarm system for first alarm alternative signals using a modularized CE model

Management of a plant alarm system has been identified as one of the key safety issues because of disasters caused by alarm floods. When a chemical plant is at abnormal state, an alarm system must provide useful information to operators as the third layer of an independent protection layer (IPL). Therefore, a method of designing a plant alarm system is important for plant safety. Because the plant is maintained in the plant lifecycle, the alarm system for the plant should be properly managed through the plant lifecycle. To manage changes, the design rationales of the alarm system should be explained explicitly. This paper investigates a logical and systematic alarm system design method that explicitly explains the design rationales from know-why information for proper management of changes through the plant lifecycle. In the method, the module structure proposed by Hamaguchi et al. (2011) to assign a fault origin to be distinguished is extended. Using modules to investigate the sets of alarm sensors and the alarm limits setting for first alarm alternative signals to distinguish the fault origin, an alarm system design method is proposed. Also, the completeness of fault propagation for a branch of the cause–effect model as the plant model is explained. Using the modules and the set of fault origins to be distinguished by the alarm system, we try to explicitly explain the design rationales of the alarm system.     

三高气田钻完井重大事故现场监测数据采集管理系统的设计与实现

把信息化技术用于油气田安全领域,用采集的现场数据评估安全生产环境,实现油气田安全信息化是近年来的新课题。为了有效地监测和预防油三高气田（高含硫、高产量、高压力）井喷事故造成的灾害,讨论了基于网络架构的重大事故现场监测数据采集管理系统软件的研究与设计,包括气象数据,传感器数据,监测设备GPS数据与视频信息的实时采集、处理、分析与显示,并且分段说明了该系统各部分的设计和实现的重点.该系统具有友好的人机界面,并且引入了流媒体技术、无线传感器技术等多种先进技术,能够最大程度地得到全面的表征油气田安全的数据与信息;该系统已经在研究实例龙岗油气田的实验中证明了具有安全生产方面的现实意义。     

基于TNPE的智能电网虚假数据注入攻击检测

为实现智能电网中虚假数据注入攻击的实时检测,提高电力系统运行的安全性,采用1种基于时序近邻保持嵌入的方法,对正常状态下采集到的电网历史量测数据建立离线模型,得到T2统计限,将实时数据通过模型获得的T2统计量与离线模型的统计限进行对比,若超过统计限,则说明存在虚假数据注入攻击.该方法在提取局部空间结构特征的基础上,可同时...