Framework and models for an event-based early warning system design

基于事件的预警系统设计的框架和模型

• 批准号: RGPIN-2014-05812
• 负责人: Ahmed, Salim
• 金额: $ 1.46万
• 依托单位: Memorial University of Newfoundland
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2014
• 资助国家: 加拿大
• 起止时间: 2014-01-01 至 2015-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=567405
• 关键词: Framework; models; event; based; early

I have proposed a five-year research program that aims to develop an event-based approach to the design and management of warning systems used in process industries. The special features and novelty of the proposed approach are in the allocation of warnings to specific events, the use of risk as a basis for alarm annunciation and categorization, and the use of prediction for warnings. The challenges in the design of such a warning system are tied to identification of events, selection of their associated variable sets, and continuous estimation of the risk of corresponding events. Moreover, the use of prediction will require the development of identification schemes to estimate models with the ability to predict variables around safety limits. My experience with the development of novel system identification algorithms and our recent works in the field of alarm system design will be valuable tools to overcome the above challenges. My record of contributions, previous and current HQP training performance, and detailed plan for HQP training will ensure the success of the program. The outcome of the program will be a novel warning system design framework and models, corresponding intellectual properties, and a group of highly qualified personnel. This program will also help to establish the applicant as a leading contributor to this field of significant industrial importance. Warning systems for industrial operations, including alerts and both process and safety alarms, play a key role in monitoring process safety. These systems warn operators of abnormal operating conditions, and seek the attention of an operator when intervention is required. Timely action, based on real warnings, is needed to help enhance plant safety and minimize costs through the effective prevention, control and mitigation of abnormal situations. In industrial plants, however, misleading warnings are a regular occurrence. Moreover, when flooded with spurious alarms, the warning system may itself contribute to abnormal situations and to plant accidents. A root cause of the ineffectiveness of existing warning systems is the large number of warnings that result from the use of single-variable settings. The number of warnings in a plant needs to be significantly reduced. Warnings should be made informative, actionable and clear indicators of the risk associated with the current state of process/system variables. Plant operators should not need to monitor the trend of individual variables; instead, the risk profiles of a plant should be monitored, with warnings that have predictive capabilities. To achieve adoption of this novel approach, a paradigm shift in the overall design of the warning systems is needed. To overcome the problems associated with existing warning systems (use of single variables, alarm flooding), the proposed research program will develop an event-based approach to warning system design and management that will improve the ability of plant operators to identify potential problems, make timely decisions and take any necessary actions. By assigning warnings to events, the number of warnings will be significantly reduced, as will the number of false warnings. Incorporation of risk in the methodology for annunciation will make it possible to accommodate the importance of the process variables and the criticality of an abnormal situation. Thus it will be possible to categorize warnings using a well-defined metric. Finally, the use of prediction will provide operators with extra time to take necessary measures. The new warning system will be a valuable addition to the arsenal of plant operators in their management of abnormal situations, the prevention of plant accidents, and improve process safety.

我提出了一个为期五年的研究计划，旨在开发一种基于事件的方法来设计和管理流程工业中使用的警报系统。所提出方法的特殊性和新颖性在于将警告分配给特定事件、使用风险作为警报通告和分类的基础以及使用警告预测。设计这种预警系统的挑战与事件的识别、相关变量集的选择以及相应事件风险的连续估计有关。此外，预测的使用将需要开发识别方案来估计具有预测安全极限附近变量的能力的模型。我在开发新型系统识别算法方面的经验以及我们最近在报警系统设计领域的工作将成为克服上述挑战的宝贵工具。我的贡献记录、以前和当前的 HQP 培训表现以及 HQP 培训的详细计划将确保该计划的成功。该项目的成果将是一种新颖的预警系统设计框架和模型、相应的知识产权以及一批高素质人才。该计划还将帮助申请人成为这一具有重要工业重要性的领域的主要贡献者。工业操作的警告系统，包括警报以及过程和安全警报，在监控过程安全方面发挥着关键作用。这些系统会警告操作员异常的操作情况，并在需要干预时寻求操作员的注意。需要根据真实警报及时采取行动，通过有效预防、控制和缓解异常情况来帮助提高工厂安全并最大限度地降低成本。然而，在工厂中，误导性警告经常发生。此外，当充斥着虚假警报时，警报系统本身可能会导致异常情况和工厂事故。现有警告系统无效的根本原因是使用单变量设置导致大量警告。工厂中的警告数量需要大幅减少。警告应提供信息丰富、可操作且明确的与过程/系统变量当前状态相关的风险指标。工厂操作员不需要监控单个变量的趋势；相反，应该监控工厂的风险状况，并发出具有预测能力的警告。为了采用这种新颖的方法，需要对预警系统的整体设计进行范式转变。为了克服与现有预警系统相关的问题（使用单一变量、警报泛滥），拟议的研究计划将开发一种基于事件的预警系统设计和管理方法，这将提高工厂操作人员识别潜在问题、及时做出决策和采取任何必要行动的能力。通过向事件分配警告，警告数量将显着减少，错误警告的数量也将显着减少。将风险纳入通报方法将能够适应过程变量的重要性和异常情况的严重性。因此，可以使用明确定义的指标对警告进行分类。最后，预测的使用将为操作员提供额外的时间来采取必要的措施。新的警报系统将成为工厂操作员管理异常情况、预防工厂事故和提高过程安全性的重要补充。