Using Large Language Models in Manufacturing | UReason Webinar

In this presentation, we will explore how LLMs, combined with the insights from our Control Valve App, can enhance decision-making processes in manufacturing. LLMs represent a cutting-edge class of artificial intelligence models designed to understand and generate human language. These models are notable for their immense scale, built upon deep neural networks with parameters ranging from hundreds of millions to tens of billions. Pre-trained on vast amounts of text data, LLMs excel in understanding linguistic nuances and performing a variety of language tasks.

1. Webinar: Large Language Models

2. Large Language Models Webinar June 26, 2024 Presenter: Jules Oudmans

3. Large Language Models? Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 3 A large language model (LLM) is a type of artificial intelligence (AI) algorithm that uses deep learning techniques and massively large data sets to understand, summarize, generate and predict new content. Source: Sylphai

4. UReason & LLMs Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 4 At UReason we experiment with LLMs to determine how we use them in our products: APM/APM Studio, the Control Valve and Pump/Motor App Our customers have: Large volumes of text resources (manuals, maintenance activities, inspection reports et cetera) Time-series data. Combining these data sources to support our users with additional advisories, recommendations and predictions is key! Hi Precious .. Long time no see, gosh we are growing old together. But this will be the last time we see each other I’m off on pension Hey Old Man .. I’d love to talk back to you but I can’t, my owners have never really cared about all the data I generate just about my functioning when they noticed it was off 

5. General LLM Usage Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 5

6. Prompt & Response - Support Example Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 6 Part of a longer response, response is very general and not specific to the type of actuator (as the model has not been trained on this)

7. Our Usage – Blending in Timeseries Data Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 7

8. Prompt with Time Series Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 8 Context of reply (technical) Prevent it from starting a dialogue Time series data for a control valve Response we are looking for + identify trends

9. Prompt with Time Series Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 9 Explaining the data format and providing examples Please provide your answer in the following template: QUERY: Write a summary of the query here. DESCRIPTION: Write a description of the health trend in detail. COMPARISON: Compare Valve B to Valve B and C. Explain which trend it follows more closely. DETAILS: Provide extensive analysis and reasoning.

10. Response on Time Series Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 10

11. Combining LLMs with CVA Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 11 Which valves will have issues like V001 in the next 3 months?

12. Wrap-up Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 12 LLMs provide great value to support users on technical questions LLMs can be combined with time-series data to provide predictions (requires specific processing) We can run LLMs on-premise (close to your data) or in cloud

13. Wrap-up Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 13 LLMs will enable decision automation and autonomous operations Source: Gartner

14. Download our LLM Findings! Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 14

15. UReason Asset Performance Management Software Copyright © 2024 UReason. All rights reserved. This presentation or any portion thereof may not be reproduced or used in any manner whatsoever without the express written permission of the author or rights holder 15 On Device / Edge / On-Premise & Cloud As Suite and Specific Asset Apps 20+ Years in Business Monitoring and optimization of components, assets and processes with data Industry Know-How Industry, Maintenance, Data, Data Science, OT and IT Low Code Development Platform APM Studio Asset Apps by UReason (Valve, Pump, Motor)

16. Q&A Webinar

17. Thank You The Netherlands UReason Boomgaardsstraat 32 3012 XD Rotterdam The Netherlands United Kingdom UReason – United Kingdom Regus Winnersh Triangle, Building 220, Wharfedale Road, Wokingham, RG41 5TP, United Kingdom Estonia UReason – Estonia Lift99 Telliskivi 60a/5, B-building, 10412 Tallinn, Estonia www.ureason.com Rotterdam Wokingham Tallinn