How is AI being used in your Piping career? 3

[Mabooz]

Hello there,

I am wondering how different organization around the world are adapting to AI, especially in area of Piping Engineering or process plants in general.

I am interested to implement AI to Piping design in stages like FEED and Detail Design.

How is this possible? is AI being adapted in your company? if so, how?

lets gather ideas together to benefit each other

 

[swazimatt]

We aren't using it yet, but i did attend a seminar where they were using it for pipe CCTV inspections and also drones + AI for bridge inspections.
The red flags are then double checked by a human

I believe anything that is repetitive, has specific parameters to check against will be done by AI in the future

 

[Andrew ONeill]

Before you get too far into it, might want to check what "AI" actually means in the current hype cycle.
Broadly there are two things
1) "Machine Learning" - has been around forever, based around neural networks. From an end user point of view, feed the computer samples of things, and manually categorize them, and the computer can "learn" then apply that to other similar samples. In recent years it's gotten cheaper and more accessible, but ultimately does the same thing. The simplistic practical example is automatic tagging of images that the googles and facebooks of the world do. This is what @swazimaat is refering to with CCTV inspectiosn etc.

2) Large Language Models - (i.e. chat GPT) - is a statistical model that predicts what should come next based on an input. It is trained on whatever text you feed it (typically publicly available internet websites). Note the same type of technology is used for AI image generators. This is what is peaking on the current hype cycle.

... and that's about it.

There might be ways to apply these tools to piping design, but I can't fathom where these tools would help with that in a major way. Even with "chat bot" type help, I've found it easier to use google to get me in the right direction then open the relevant standard / guidebook directly.
Anyway that's my 2 cents.

Andrew O'Neill
Specialist Mechanical Engineer
Australia

 

[Mabooz]

Andrew,

AI in a sense of providing prompt to do certain task, we have seen how AI propmpt these days can create for you a whole software code or a whole website, or some other complicated things. and its been only couple years for CHATGPT, what will happen in 5 years? when the advancement in curve is exponential?

is it not possible in the future that I feed an AI a P&ID and ask it to review it based on certain codes and standards? or as ask it to draft P&ID?

 

[RVAmeche]

Its worth noting AI like ChatGPT is largely committing mass copyright fraud by just scanning everything and using what makes sense based on your prompt. I wouldn't be surprised if they start to get expensive to afford some sort of settlement with the major players.

I think it will be a long time before an AI does anything super productive in the piping world, like automatically navigate a pointcloud and create a preliminary routing and whatnot.

As of right now I see no way you could have it create a P&ID. Based on what? If you ask for a tank P&ID it may give you something generic, but that's nothing. Right now they are far too generic language based and not industry specific.

 

[Gator]

So, if a Large Language Model finds that the majority of resources scanned say that 2+2=7, the ChatGPT AI will report that 2+2=7?

 

[GBTorpenhow]

AI can play a significant role in process engineering FEED (Front End Engineering Design) studies in several ways:

1) Data Analysis and Pattern Recognition: AI algorithms can analyze historical data from similar projects to identify patterns and trends. This analysis can provide insights into optimal design parameters, potential risks, and areas for improvement.
2) Optimization: AI techniques such as machine learning and optimization algorithms can be used to optimize process parameters, equipment sizing, and layout design. These algorithms can consider various constraints and objectives to find the most efficient and cost-effective solutions.
3)Simulation and Modeling: AI-powered simulation tools can simulate the behavior of complex processes under different operating conditions. These simulations can help engineers understand the dynamic behavior of the system, identify bottlenecks, and evaluate the performance of alternative design options.
4) Predictive Maintenance: AI can be used to predict equipment failures and maintenance needs based on real-time sensor data and historical maintenance records. This proactive approach to maintenance can help minimize downtime and maximize the reliability of the process.
5) Risk Assessment and Mitigation: AI can analyze data from various sources, including historical project data, environmental factors, and regulatory requirements, to assess risks associated with the project. This analysis can help identify potential hazards and develop mitigation strategies to ensure the safety and compliance of the design.
6) Decision Support: AI systems can provide decision support to engineers by analyzing large amounts of data and providing recommendations based on predefined criteria. This can help streamline the decision-making process and ensure that design decisions are based on data-driven insights.

Overall, AI has the potential to revolutionize the process engineering FEED process by enabling faster, more informed decision-making, optimizing designs for efficiency and cost-effectiveness, and mitigating risks associated with complex engineering projects.

 

[Nutzman]

The Two companies I work for (International companies), have both blocked access to ChatGPT and Bing Copilot, based on the comment "we are reviewing the AI in terms of the companies requirements....". I have used both the above to provide engineering calculations. AI comes back with a list of formula. I provide some data and request AI to solve the problem, by applying the data into the formula. AI does the numbers and comes back with a negative (value) answer. Then concludes that "if the answer is a negative value, then the calculation is incorrect. I'm happy the companies have blocked AI. This stops staff from chasing solutions that could well be incorrect.

 

[RVAmeche]

At least right now, ChaptGPT and similar are NOT computational programs. They don't understand anything about math. They're just regurgitating things they've found that seem right but they're not Wolfram Alpha on steroids.

 

[1503-44]

The answer is 42

https://www.mindprison.cc/p/the-question-that-no-llm-can-answer

--Einstein gave the same test to students every year. When asked why he would do something like that, "Because the answers had changed."