Human-centric AI for engineering design changes in the nuclear industry

作为核电厂整个寿命周期运行的一部分，由于故障、陈旧和效率提高等因素，核电厂的子部件和系统需要更换或升级。更换的设计和实施是一个耗时和人工密集的过程，需要从各种来源以及以前的业务经验中获取信息。一系列人工智能技术的集成可以导致设计过程的自动化和增强，从而显著提高效率，并消除成本高昂的非最佳重新设计。这将需要研究集中在以下方面的结合：基于知识的系统，基于案例的推理和以人为中心的人工智能。这将需要新的方法来正式的知识捕获和表示，沿着基于案例的推理技术，以支持保留和重用的关键设计决策和经验教训。这将包括开发一个通用本体，用于在安全标准、变更请求日志、设计文件和台站图纸等若干系统和资源之间交换和融合信息。以人为中心的人工智能专注于人类操作员如何与人工智能算法进行有效交互，这将是博士学位的关键要素。在核工业中，在这种受监管的环境中做出的任何决定都需要透明度和可重复性。因此，向用户提供正确的交互、解释和理由至关重要。这项研究将创建一个新的人工智能技术框架，解决这些问题，并提供设计建议，最终可以构建成一个自动化的决策支持系统。

As part of the through-life operation of nuclear power plants, sub-components and systems of the plant need replacing or upgrading by a number of drivers such as failures, obsolescence and efficiency improvements. The design and implementation of replacements is a time consuming and manually intensive process which requires information to be drawn from a wide variety of sources, as well as from previous operational experience. The integration of range of AI techniques could lead to automation and augmentation of the design process to deliver significant efficiency improvements and remove costly non-optimal redesigns. This will require research focused on the combination of: knowledge-based systems, case based reasoning and human-centric AI. This will require novel approaches to formal knowledge capture and representation, along with case-based reasoning techniques, to support the retention and re-use of key design decisions and lessons learned. This will include the development of a generic ontology for exchange and fusion of information between a number of systems and resources, such as safety standards, change request logs, design documents and station drawings. Human-centric AI is focused on how the human operator interacts effectively with the artificial intelligence algorithms, and this will be a key element of the PhD. Within the nuclear industry where there is a significant requirement for transparency and auditability of any decisions made in this regulated environment. Therefore, the correct interaction, explanations and justifications being made available to the user is critical. This research will create a novel framework of AI techniques which tackle the issues, and deliver design advice and can be ultimately built into an automated decision support system.