Jump to content
Private Messaging is activated - check "How to" on how to disable it ×

Is anyone using AI to help with Calypso? Or inspections in general? If so, how?


---
 Share

Recommended Posts

---

We have recently been using AI software to generate bubbled prints to match numbered CMM programs and it has saved a lot of time, I also us AI to help program batch programs to manage data. I am hoping to find a software that checks the PDF report and checks it to a print to find mistakes with nominals/tolerances, I have used chatGPT for this with poor results but there is definitely potential. 

Anyone else started to incorporate artificial intelligence?

Link to comment
Share on other sites

---

I have little experience in using Ai programs, so how exactly did you instruct an AI to compare a Calypso program to a print? How does it know which dimension is which? (I'm thinking of several identical dimensions scattered across the print)

Edited
Link to comment
Share on other sites

---

Please sign in to view this quote.

If you are interested in converting a PDF to plain text file because it is easier to parse, look into XpdfReader in conjunction with your choice of scripting.

Link to comment
Share on other sites

---

Please sign in to view this quote.

I feed ChatGPT the PDF result from piweb, and I feed it the PDF print. I haven't had the chance to really get it hammered out to the point where it works well, but I have taught it to recognize nominals on the print itself and nominals on the report and compare them. Big issue is that it skips and doesn't recognize all of them from the print, it also really struggles with tolerances since they are ±. DWC's idea of text data might help that on the report but probably not the print itself.

Link to comment
Share on other sites

---

Something like analyzing the CAD model to gather all the geometry (size/form/position/orientation) and sending all that data to automatically assign the correct probes, strategy, clearance, filtering, etc.

Link to comment
Share on other sites

---

Is nobody worried about privacy issues with putting customer drawings/models/programs intro a 3rd party companies hands?

 

Chat GPT and the like are useful, but most of us can't share a drawing on this relatively private forum, giving it to OpenAI or Elon doesn't sound like a smart move either.

Link to comment
Share on other sites

---

Please sign in to view this quote.

I definitely would not use anything like ChatGPT for just this reason... but, I believe that there are private/offline options available.

Link to comment
Share on other sites

---

Please sign in to view this quote.

Please sign in to view this username.

 Not being smart here, but shouldn't the programmer check this when proving out the program? After that, unless a rev change, it shouldn't change right? 

 

But speaking to your point on automating tasks to check for mistakes, I have a script that looks at calibration reports (PDF) and converts each file to a plain text file with the same name as the PDF, except now as a txt, then looks at the text files to see if certain phrases are missing. If the phrase is missing, it copies the original PDF with the txt file name to a separate folder for easy locating. I can check hundreds of certificates in seconds for missing information. The script utilizes XpdfReader (free) pdftotext.exe.

 

# Define the folder containing PDF files
$folderPath = "C:\PATH TO YOUR PDF FILES"  # Change this to your folder path
$outputTextFolderPath = "C:\PATH TO PLAIN TEXT FILES"  # Change this to your desired output folder
$missingFilesFolderPath = "C:\PATH TO MISSING PHRASE FOLDER"  # Path for missing certs folder

# Path to the pdftotext executable
$pdftotextPath = "C:\Program Files\Glyph & Cog\XpdfReader-win64\bin64\pdftotext.exe"

# Ensure the output text directory exists
if (-Not (Test-Path -Path $outputTextFolderPath)) {
    New-Item -ItemType Directory -Path $outputTextFolderPath | Out-Null
}

# Ensure the folder for missing certs exists
if (-Not (Test-Path -Path $missingFilesFolderPath)) {
    New-Item -ItemType Directory -Path $missingFilesFolderPath | Out-Null
}

# Get all PDF files in the specified folder
$pdfFiles = Get-ChildItem -Path $folderPath -Filter "*.pdf"

# Initialize an array to hold the filenames that do not contain the specified phrase
$missingPhraseFiles = @()

# Define the phrase to search for
$searchPhrase = "YOUR PHRASE HERE"  # Replace with the actual phrase

# Loop through each PDF file and convert it to text
foreach ($pdf in $pdfFiles) {
    # Define the output text file path
    $outputFilePath = Join-Path -Path $outputTextFolderPath -ChildPath ($pdf.BaseName + ".txt")

    # Check if the output file already exists
    if (Test-Path -Path $outputFilePath) {
        Write-Host "Output file already exists: $outputFilePath"
        continue
    }

    # Execute pdftotext command
    & $pdftotextPath -layout "$($pdf.FullName)" "$outputFilePath"

    # Check if the conversion was successful
    if ($?) {
        Write-Host "Converted: $($pdf.FullName) to $outputFilePath"
        
        # Read text content from the output text file
        $foundPhrase = $false
        Get-Content -Path $outputFilePath | ForEach-Object {
            if ($_ -like "*$searchPhrase*") {
                $foundPhrase = $true
                return $true  # Exit the loop early
            }
        }

        # If the phrase is not found, copy the PDF file to the missing folder
        if (-not $foundPhrase) {
            Copy-Item -Path $pdf.FullName -Destination $missingFilesFolderPath
            $missingPhraseFiles += [PSCustomObject]@{ FileName = $pdf.BaseName }
        }
    } else {
        Write-Host "Failed to create text file for: $($pdf.FullName)"
    }
}

# Display the files missing the specified phrase
if ($missingPhraseFiles.Count -gt 0) {
    Write-Host "Files missing the phrase '$searchPhrase':"
    $missingPhraseFiles | ForEach-Object { Write-Host $_.FileName }
} else {
    Write-Host "All files contain the phrase '$searchPhrase'."
}

Write-Host "Conversion and extraction complete."

 

 

 

Link to comment
Share on other sites

---

I asked out of pure technical curiosity.

Well, you may be able to run a lesser AI model locally. Not feasible with ChatGPT of course, but there are (and will be) others out there. Depends on how much of a pimped-up computer one is willing to provide.

Link to comment
Share on other sites

---

Please sign in to view this quote.

I agree it is a concern, I don't work in a security clearance environment, but IP security is always a concern. For the sake of debate I would argue that few people would question sending models/prints over MS outlook or Gmail, ChatGPT just doesn't have that reputation yet(maybe it never will).

 

I haven't looked into any local AI programs but that might be the safer route. 

Link to comment
Share on other sites

---

Well, recently I was playing around with an image enhancement AI which I installed locally. The download was around 15 Gig and took forever. The installation was easy because it was all handled by the Pinokio browser, but without it it would have been a geek's nightmare. Unfortunately I wasn't able to process any image yet, because the PC I installed it on has no gfx card and without one calculations take hours if not days. What I'm trying to say: Right now we're still far away from a convenient and affordable way of running AIs locally.

Link to comment
Share on other sites

 Share

×
×
  • Create New...