ARTIFICIAL INTELLIGENCE IN PROCESS PLANT SAFETY

这项提案的主要目标是研究和论证 基于人工智能(AI)的新方法 化工过程危险检测、预防和控制的设计 控制系统。这样的系统对于 改善化工厂的职业安全，因为 现代加工厂的复杂性。工业统计数据显示 即使重大灾难和化学物质造成的灾难 工厂故障很少发生，轻微事故很常见， 在日常生活中发生，导致许多职业 伤病，每年给社会造成数十亿美元的损失 年。拟建工程以防治为目的。 如此频繁的、日复一日的、行业内的意外事件。 过去在故障诊断系统中的方法不能正确地 包括人类专家的推理策略和经验 因此不足以有效地纠正故障- 射击。我们建议通过设计系统来改善这一状况 这将更好地对问题解决过程进行建模 通过领域知识的适当表示 通过使用因果模型和推理从一开始 原理，类似于人类专家，通过利用一些 人工智能的最新进展。我们提出了一个 帮助开发专家系统的方法，该专家系统 独立于流程、推理透明、弹性强 在不可预见的故障组合下，并能够诊断 断层种类繁多。该系统的领域知识是 基于流程的故障和因果模型库 设备以及设备之间的物理互连 设备单元和过程状态之间的因果关系 变量。推理策略使用基于模型的推理来 分析植物的行为。我们描述了一个原型专家 系统，称为MODEX，基于我们的方法。系统 在原型化学品的测试用例上成功执行 加工厂，看起来很有希望。然而，在一次成功的 将这一方法转移到行业可以通过以下方式启动 提案中概述的研究问题的数量需要 通过试验更大的原型化学品来解决 植物。

The major goal of this proposal is to research and demonstrate new approaches based on Artificial Intelligence (AI) towards the design of chemical process hazard detection, prevention, and control systems. Such systems are extremely important for improving the occupational safety of chemical plants owing to the complexity of modern process plants. Industrial statistics show that even though major catastrophies and disasters from chemical plant failures are infrequent, minor accidents are very common, occurring on a day to day basis, resulting in many occupational injuries and illnesses, costing the society billions of dollars every year. The proposed project is aimed at the prevention and control of such frequent, day to day, accidental events in the industry. Past approaches in fault diagnostic systems did not properly include the human expert's reasoning strategies and experience and hence were not adequate in efficient and correct trouble- shooting. We propose to improve this status by designing systems that would have better modeling of the problem-solving process through an appropriate representation of the domain knowledge through the use of causal modeling and reasoning from first principles, similar to human experts, by exploiting some of the recent advances in Artificial Intelligence. We propose a methodology that aids the development of expert systems which are process-independent, transparent in their reasoning, resilient under unforeseen fault combinations, and capable of diagnosing a wide diversity of faults. The domain knowledge of the system is based on a library of fault and causal models of process equipments as well as on the physical interconnections between equipment units and causal relationships among process state variables. The inference strategy uses model-based reasoning for analyzing the plant behavior. We describe a prototype expert system, called MODEX, based on our methodology. The system has performed successfully on test cases of prototypical chemical process plants and looks promising. However, before a successful transfer of this methodology to the industry can be initiated a number of research issues outlined in the proposal need to be resolved by experimenting with larger prototypical chemical plants.