Brooks Robin, Thorpe Richard, Wilson John
Curvaceous Software Limited, P.O. Box 43, Gerrards Cross, Bucks SL98UX, UK.
J Hazard Mater. 2004 Nov 11;115(1-3):169-74. doi: 10.1016/j.jhazmat.2004.05.040.
A new mathematical treatment of alarms that considers them as multi-variable interactions between process variables has provided the first-ever method to calculate values for alarm limits. This has resulted in substantial reductions in false alarms and hence in alarm annunciation rates in field trials. It has also unified alarm management, process control and product quality control into a single mathematical framework so that operations improvement and hence economic benefits are obtained at the same time as increased process safety. Additionally, an algorithm has been developed that advises what changes should be made to Manipulable process variables to clear an alarm. The multi-variable Best Operating Zone at the heart of the method is derived from existing historical data using equation-free methods. It does not require a first-principles process model or an expensive series of process identification experiments. Integral with the method is a new format Process Operator Display that uses only existing variables to fully describe the multi-variable operating space. This combination of features makes it an affordable and maintainable solution for small plants and single items of equipment as well as for the largest plants. In many cases, it also provides the justification for the investments about to be made or already made in process historian systems. Field Trials have been and are being conducted at IneosChlor and Mallinckrodt Chemicals, both in the UK, of the new geometric process control (GPC) method for improving the quality of both process operations and product by providing Process Alarms and Alerts of much high quality than ever before. The paper describes the methods used, including a simple visual method for Alarm Rationalisation that quickly delivers large sets of Consistent Alarm Limits, and the extension to full Alert Management with highlights from the Field Trials to indicate the overall effectiveness of the method in practice.
一种将警报视为过程变量之间多变量相互作用的全新数学处理方法，提供了有史以来第一种计算警报限值的方法。这使得现场试验中的误报大幅减少，从而降低了警报通知率。它还将警报管理、过程控制和产品质量控制统一到一个单一的数学框架中，从而在提高过程安全性的同时实现了操作改进并获得了经济效益。此外，还开发了一种算法，可建议对可操作过程变量进行哪些更改以消除警报。该方法核心的多变量最佳操作区是使用无方程方法从现有历史数据中推导出来的。它不需要第一性原理过程模型或一系列昂贵的过程识别实验。与该方法集成的是一种新格式的过程操作员显示屏，它仅使用现有变量来全面描述多变量操作空间。这些特性的组合使其成为小型工厂和单件设备以及大型工厂都负担得起且易于维护的解决方案。在许多情况下，它还为即将进行或已经在过程历史记录系统中进行的投资提供了理由。英国的英力士氯碱公司和马林克罗德特化学品公司正在对新的几何过程控制（GPC）方法进行现场试验，该方法通过提供比以往任何时候都高质量的过程警报和提醒来提高过程操作和产品的质量。本文描述了所使用的方法，包括一种用于警报合理化的简单可视化方法，该方法可快速提供大量一致的警报限值，以及扩展到完整的警报管理，并重点介绍了现场试验的情况，以表明该方法在实际应用中的整体有效性。
