AI-Analyst: Next Generation Advanced Pattern Recognition for Operations & Maintenance Supporting Delivery of a Low Carbon Future

AI-Analyst：下一代高级操作模式识别

• 批准号: 89639
• 金额: $ 41.27万
• 依托单位: COGNITIVE.BUSINESS LTD
• 依托单位国家: 英国
• 项目类别: Collaborative R&D
• 财政年份: 2021
• 资助国家: 英国
• 起止时间: 2021 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=89639
• 关键词: AI; Analyst; Next; Generation; Advanced

Considered by many to be the holy grail of predictive maintenance, transfer learning (TL) is the ability to identify failure symptoms (ISO-13379) from one asset and apply them automatically to another. Applied across the thousands of connected plant items in the Industrial Internet of Things (IIoT), it could unleash the sector's potential adding $14.2tn to the global economy by 2030 \[Accenture\].Breakthroughs in deep learning (DL) solving Big-Data problems, such as accurate image recognition, might provide the impression that DL would enable asset failure predictions in much the same way. However asset failure data is scarce, every asset has unique data signatures, and therefore IIoT is not Big-Data \[Uniper, 2017\].This is an industrial research programme building upon a successful novel proof-of-concept technology. The AI-Analyst provides automatic modelling, early-fault detection, and diagnosis using TL, practical for O&M requirements; delivering a genuinely unique offering which can be readily commercialised and exported globally to all IIoT connected assets

被许多人认为是预测性维护的圣杯，迁移学习（TL）是从一个资产中识别故障症状（ISO-13379）并将其自动应用于另一个资产的能力。应用于工业物联网（IIoT）中数千个连接的工厂项目，到2030年，它可以释放该行业的潜力，为全球经济增加14.2万亿美元。深度学习（DL）解决大数据问题的突破性进展，例如准确的图像识别，可能会给人留下这样的印象，即DL将以大致相同的方式实现资产故障预测。然而，资产故障数据是稀缺的，每个资产都有独特的数据签名，因此IIoT不是大数据[Uniper，2017]。这是一个基于成功的新型概念验证技术的工业研究计划。AI分析师使用TL提供自动建模、早期故障检测和诊断，可满足运维要求;提供真正独特的产品，可随时商业化并出口到全球所有IIoT连接资产