Exploring Artificial Intelligence

Published On: June 20, 2024By Categories: Drilling, Groundwater & Wells

How smart machines work and why we use them.

By Thom Hanna, PG

As executive director of the Durango Research Institute (DRI), I thought it would be good to have an assistant besides the lab (Figure 1), so I looked to ChatGBT for help.

Figure 1. Lab at the Durango Research Institute.

Artificial intelligence, or AI for short, is like having a smart friend who can help us with all sorts of things. But how does it work, and why do we want to use it?

Imagine AI as a super smart robot. It learns things just like we do, but it uses computer programming instead of a brain. These programs are like recipes telling the robot how to think and make decisions.

So how does AI learn? One way is machine learning, which is similar to teaching a baby how to recognize shapes and colors. Show the AI a lot of examples, and it will learn from them. If you wanted the AI to recognize cats, you’d show it several pictures of cats and tell it, “These are cats.” Over time, the AI will recognize cats all by itself.

Another cool thing about AI is that it can handle a lot of information quickly. This is because it works with data. The data can be anything from pictures and videos to numbers and words. The more data the AI has, the smarter it becomes.

So, why do we want to use AI? Well, imagine if you had a helper who never got tired and never made mistakes. That’s what AI can be! It can help with all sorts of tasks like answering questions, organizing information, and even driving cars.

AI has helped doctors find out what’s wrong with patients faster and more accurately. By looking at lots of data, AI can spot patterns that humans might miss. This means doctors are treating people better and helping them stay healthy.

Figure 2. Well Wizard logo.

Moreover, AI can make our lives easier by doing things for us. For instance, it can help us find the fastest route to school or suggest movies we might like to watch. It’s literally like having a personal assistant.

AI is that smart friend who can help us with all sorts of things. It learns from examples, handles lots of information, and can make our lives easier and better. So, whether it’s answering questions, helping doctors, or just making our daily tasks simpler, AI is here to make the world a smarter and more helpful place.

So when I went to use AI, one of the first things ChatGBT wanted me to do is create a logo and a name for my new assistant—so Well Wizard it is.

From there I was communicating with ChatGBT on setting a methodology to select a filter pack and slot size. This was done by asking questions and receiving responses. After reading a response, I would ask more questions and provide more information and correct mistakes.

There was a number of iterations that took place before I got to the dialog that follows. I chose the topic of selecting a filter pack as it was the subject of the last Groundwater & Wells column in the Water Well Journal April 2024 issue.

Below we pick up the progress after about one hour of clarifying and educating the Well Wizard.

Thom:

I am not aware of empirical methods for selecting slot size; the purpose of the screen is to keep the filter pack in place.

Well Wizard:

Figure 3. Official logo of the DRI.

You’re absolutely correct, and I appreciate the clarification. The primary purpose of the well screen’s slot size is indeed to retain the filter pack, preventing it from entering the well while allowing water to pass through. The selection of slot size is more directly related to the size
of the filter pack rather than empirical methods that are often more relevant for other aspects of well design. Let’s refine the methodology with a focus on this purpose:

Step 1: Analyze the Formation Grain Size Distribution

  • Grain Size Analysis: Conduct a grain size analysis of the aquifer material to understand the distribution of particle sizes.
  • Determine Key Sizes: Calculate the D10, D50, and D70 of the formation material, which will inform the selection of the filter pack.

Step 2: Select the Filter Pack

  • Choose Filter Pack Material: Select a filter pack material that is coarser than the formation material. The typical rule is to ensure the D10 of the filter pack is about four to five times the D10 of the formation material although this can be adjusted based on the 70%
    retained rule and specific site conditions.
  • Determine the Multiplier: The multiplier (three to seven times the D70 of the formation) helps in selecting a filter pack size that balances filtration efficiency with hydraulic conductivity. The specific multiplier used depends on the formation’s characteristics and design objectives.

Step 3: Determine Filter Pack Thickness

  • General Thickness Guidelines: A common guideline is to have a thickness of 1 to 2 feet beyond the borehole diameter, though this can vary based on local geological conditions and well design.
  • Site-Specific Adjustments: Adjust the thickness as needed based on the aquifer’s characteristics, ensuring effective filtration without significantly impeding water flow.

Step 4: Determine the Slot Size

  • Slot Size to Retain Filter Pack: The slot size should be chosen to effectively retain the filter pack material. A common approach is to select a slot size that will retain 90% to 95% of the filter pack grains.
  • Alignment with Filter Pack Size: Ensure that the slot size is appropriately matched with the selected filter pack size, facilitating effective retention, and preventing the loss of filter material into the well.

Step 5: Install the Filter Pack

  • Even Distribution: Carefully distribute the filter pack around the well screen to prevent channeling and ensure uniform filtration.
  • Avoid Mixing: Prevent the filter pack from mixing with the formation material during installation.

Step 6: Develop the Well

  • Pumping to Remove Fines: Develop the well by pumping to remove fine particles, thereby stabilizing the interface around the well screen.
  • Clarity Assessment: Monitor the water clarity to determine when the development process has been effective and the filter zone is stable.

Step 7: Ongoing Monitoring and Maintenance

  • Periodic Inspections: Regularly check the well for signs of filter pack migration or sand production, which could indicate issues with the screen or filter pack.
  • Maintenance Plan: Establish a proactive maintenance plan to address potential issues, ensuring the well’s operational longevity and efficiency.

By focusing on the primary function of the well screen and the interplay between the screen and the filter pack, this methodology aims to optimize the well’s design for effective long-term performance.

Thom:

If the filter pack is 1 to 2 feet, it will be very difficult to develop the well. I think not more than 5 inches thick is better.

Well Wizard:

Thank you for pointing that out. The thickness of the filter pack is indeed a critical factor, and it must be optimized to balance effective filtration while ensuring that the well can be adequately developed and maintained. A filter pack that is too thick can indeed make well development challenging and may not be necessary, depending on the specific site conditions. Let’s adjust the methodology with a more suitable filter pack thickness:

Step 1: Analyze the Formation Grain Size Distribution

  • Grain Size Analysis: Conduct a grain size analysis of the aquifer material.
  • Determine Key Percentile Sizes: Identify the D10, D50, and D70 of the formation material to inform the filter pack selection.

Step 2: Select the Filter Pack

  • Filter Pack Material Selection: Choose a filter pack that is coarser than the formation material, typically ensuring the D10 of the filter pack is about four to five times the D10 of the formation material.
  • Multiplier Determination: Use a multiplier of three to seven times the D70 of the formation material to select the filter pack size, adjusting based on the specific geological characteristics and objectives.

Step 3: Determine Filter Pack Thickness

  • Optimal Thickness: Aim for a filter pack thickness of up to 5 inches, which is sufficient to ensure effective filtration while facilitating well development and maintenance.
  • Site-Specific Consideration: The exact thickness should be determined based on the local geological conditions, well diameter, and the anticipated ease of well development.

Step 4: Determine the Slot Size

  • Slot Size Selection: Choose a slot size that will retain the filter pack material, typically aiming to retain 90% to 95% of the filter pack grains.
  • Ensure Compatibility: The slot size should be matched with the filter pack size to ensure effective retention while allowing for water passage.

Step 5: Install the Filter Pack

  • Ensure Even Distribution: Carefully distribute the filter pack around the well screen to ensure uniformity and prevent channeling.
  • Prevent Contamination: During installation, take measures to prevent the filter pack from mixing with the native formation material.

Step 6: Develop the Well

  • Effective Development: Develop the well by pumping to remove fines and stabilize the filter zone, ensuring the well is clear of sediment and functioning efficiently.
  • Monitor Development: Assess the development process by monitoring water clarity and sand production, adjusting the development strategy as needed.

Step 7: Ongoing Monitoring and Maintenance

  • Regular Inspections: Monitor the well for any signs of filter pack migration, sand production, or changes in water quality and yield.
  • Proactive Maintenance: Maintain a schedule for regular well maintenance to address any issues and ensure the well’s long-term performance and sustainability.

By adjusting the filter pack thickness to a more practical measure, this revised methodology aims to ensure that the well is effectively filtered, easily developed, and maintainable over its operational lifespan.

Conclusion

AI is limited by what is available on the internet, but as you can see, the Well Wizard learned from me as I kept asking questions. It took me about 90 minutes to get to the revised methodology with numerous interactions.

It is important to have some knowledge of the subject if you are going to have AI take a technical look at things to ensure it doesn’t give information not related to what you are trying to accomplish.

AI hallucinations occur when artificial intelligence systems generate incorrect, misleading, or entirely fabricated information. These errors can result from several factors.

Firstly, data limitations mean that AI models trained on large datasets might still lack coverage for every scenario or contain errors. When asked about less-represented topics, the AI might produce plausible but incorrect answers.

Secondly, AI models generate text based on patterns learned during training, and if a question is complex or ambiguous, the AI might fit an incorrect pattern to it. The generative nature of models like GPT-4, which rely on word sequence probabilities, can also lead to factually inaccurate sentences.

Overfitting and underfitting are additional issues. Overfitting happens when a model learns specific details too well, leading to inaccuracies with new data, while underfitting occurs when the model fails to capture important patterns. Current AI models process and generate text based on patterns rather than true understanding, leading to confident but incorrect information.

Examples of AI hallucinations include fabricated references, incorrect facts, and misinterpreted questions. To mitigate these issues, strategies such as cross-checking information, using high-quality training data, continuously refining models, and involving human review are essential. Understanding AI hallucinations is crucial for developing reliable AI systems and for users to critically evaluate AI-generated
content.

There is more to AI than I understand in this, my first try at it. It only knows what it knows or can find online or in databases. The water well industry is one where much of the well design work and studies were completed in the 1950s-1980s, so there are large gaps in the literature and information that AI can find. However, it is a great tool for writing computer codes, accomplishing small tasks, and quickly gathering information on a subject. I believe it will evolve over time.

Try it out. It might be a help for some of your work and tasks around your office, but watch out for the hallucinations.

And I do think I will not let ChatGPT redesign the logo for the DRI (Figure 3). As I said, all great research institutes need a good lab!

Get the Top-Selling Reference Book in the Groundwater Industry
Groundwater & Wells, Third Edition, is the top-selling reference book year after year. This tremendous resource should be on your desk if you are serious about the groundwater industry.

It features 20 years of advancements in technology and knowledge gained in the field. Included are new borehole geophysical techniques, information on underbalanced and horizontal drilling, new approaches to well development, as well as details on ASR, groundwater monitoring and remediation, prepack screens for water interactive well design and well maintenance programs for water well contractors and design engineers.

Click here to order it, call (800) 551-7379, fax (614) 898-7786, or email customerservice@ngwa.org.


Thomas M. Hanna, PG, is a technical director of water well products/hydrogeologist for Johnson Screens where he works in areas of well design, development, and well rehabilitation. He is a registered professional geologist in Arizona, Kentucky, and Wyoming and has worked for several groundwater consulting firms. Hanna can be reached at thom.hanna@johnsonscreens.com.

Tags:

Read the Current Issue

you might also like