Trusted Deep Learning For Multi-Domain Engineering Systems. The efficient design and operation of engineering systems require trusted predictive model

多领域工程系统的可信深度学习。

• 批准号: 2651000
• 金额: --
• 依托单位: University of Sheffield
• 依托单位国家: 英国
• 项目类别: Studentship
• 财政年份: 2022
• 资助国家: 英国
• 起止时间: 2022 至 无数据
• 项目状态: 未结题
• 来源: https://gtr.ukri.org/projects?ref=studentship-2651000
• 关键词: Trusted; Deep; Learning; Multi; Domain

Trusted Deep Learning For Multi-Domain Engineering Systems. The efficient design and operation of engineering systems require trusted predictive models using knowledge sources from different physical domains. These domains (e..g. thermal, vibration, electrical, etc.) are traditionally siloed specialisms within industry with their own methods and language. This project addresses the challenges in producing unifying models that will be used to better holistically understand engineering system behaviour and health, while accommodating the realities of imperfect training data sets. The project will build upon the state-of-the-art in machine learning systems to create interconnected modular models that can be interpreted by experts (i.e. that are explicable) capturing both known physics and complex (e.g. emergent from domain coupling) behaviours captured only in real system data.

面向多领域工程系统的可信深度学习。工程系统的有效设计和运行需要使用来自不同物理域的知识源的可信预测模型。这些域(例如，热、振动、电气等)是行业内传统上孤立的专业，有自己的方法和语言。该项目解决了在产生统一模型方面的挑战，这些模型将用于更好地全面了解工程系统的行为和健康，同时适应不完美的训练数据集的现实。该项目将建立在最先进的机器学习系统的基础上，创建可由专家解释的相互连接的模块模型，以捕获仅在真实系统数据中捕获的已知物理和复杂(例如，从域耦合中出现的)行为。