Olsen Spec Projects

[SilverJimmy]

Wow! That above is so true, especially in the normal automotive repair industry! I would sell a diagnostic platform to a typical tech who thought he knew everything and then he’d be pissed because it “told” him to change out XYZ but the car still wasn’t fixed! I tried to explain to them that a modern scanner was basically like a flashlight, it could only shine a light on what might be the problem, but they needed to be able to understand what was being spotlighted!
We won’t even go down the road of that one guy who told me that he knew what was wrong with a car when it pulled into his bay!

 

[Wiz02]

I'm really curious about your use of AI. Can you share any details without revealing proprietary information?

I was using AI to facilitate software development before retirement and it was helpful for fixing software syntax errors but you needed to know how to define the program requirements in detail to receive a satisfactory solution.

The other area that I used it for was wordsmithing emails and reviews which saved me a huge amount of time.

Helping techs learn the fundamentals of automotive systems diagnosis doesn't seem like an out of the box AI application. Did you have a custom solution developed?

I was and still am interested in AI / machine learning but it didn't make sense for me to change specialties so close to retirement.

 

[olsenmotorsports]

I'm really curious about your use of AI. Can you share any details without revealing proprietary information?

I was using AI to facilitate software development before retirement and it was helpful for fixing software syntax errors but you needed to know how to define the program requirements in detail to receive a satisfactory solution.

The other area that I used it for was wordsmithing emails and reviews which saved me a huge amount of time.

Helping techs learn the fundamentals of automotive systems diagnosis doesn't seem like an out of the box AI application. Did you have a custom solution developed?

I was and still am interested in AI / machine learning but it didn't make sense for me to change specialties so close to retirement.

Happy to - without giving away a ton, lets just say you can load your workshop manual into your chat GBT program and ask it anything. “Write me a test plan with wiring diagram on P0043 etc”. Spits it out in under 5 seconds and off you go.

 

[olsenmotorsports]

We won’t even go down the road of that one guy who told me that he knew what was wrong with a car when it pulled into his bay!

Love that attitude, to which I respond with “show me your data and lets walk through it”

 

[Wiz02]

Happy to - without giving away a ton, lets just say you can load your workshop manual into your chat GBT program and ask it anything. “Write me a test plan with wiring diagram on P0043 etc”. Spits it out in under 5 seconds and off you go.

Very cool way to use AI. Love it.

 

[olsenmotorsports]

Very cool way to use AI. Love it.

TBH we have not even scratched the surface on its capabilities yet. It is an incredibly useful tool that will change business forever IMHO and much needed. We aren’t the smartest people out there so much of how we use it is “hey I wonder if AI can help us”. Frustrating I know lol, but that’s about the limit of our capabilities

 

[Wiz02]

TBH we have not even scratched the surface on its capabilities yet. It is an incredibly useful tool that will change business forever IMHO and much needed. We aren’t the smartest people out there so much of how we use it is “hey I wonder if AI can help us”. Frustrating I know lol, but that’s about the limit of our capabilities

Instead of wondering, just ask AI. Nothing wrong with that at all, I think it should become second nature to ask, you just need to remember that the response is just as likely to be bs as valuable.

Here's what it returned when I asked about creating an AI engine to help technicians do diagnosis. Now this response is software developer oriented, but it has a lot of interesting points, especially on harvesting sensor data from diagnostic tools.

If I wasn't retired, I would love to work on a first phase of a project that goes something like this (later phases would integrate AI into the scanner) : The AI would ask the user to input a code or symptoms. For applicable issues, it then specifies where to connect diagnostic probes and how to set up the scanner.

Technician then compares actual readings to samples shown by AI and follows instructions on what to do based on the closest matching waveform.

Thinking about this a bit more, this doesn't really help a tech understand the root cause, it's just a more sophisticated version of: scanner reports code AXXXX and flowchart says replace component ZZZ

But having something like the above capabilities may still be very helpful since it seems like inexperienced technicians may not know what they are looking at, or even what to look at if they do actually hook up or want to hook up a diagnostic scanner.

FYI This was the AI response to my question about building a custom AI for automotive diagnostics :

Here’s a more detailed roadmap to help you get started:

1. Data Collection​

• Historical Data: Collect past diagnostic reports, repair logs, and service manuals. This data will help the AI learn common issues and their solutions.
• Sensor Data: Gather real-time data from vehicle sensors, including engine performance, emissions, and other critical parameters.

2. Data Preparation​

• Cleaning: Ensure the data is clean and free from errors. This involves removing duplicates, correcting inaccuracies, and standardizing formats.
• Labeling: Label the data with relevant tags, such as specific issues, symptoms, and corresponding fixes.

3. Choosing the Right AI Model​

• Machine Learning Models: Use models like decision trees, random forests, or neural networks to analyze the data and identify patterns.
• Natural Language Processing (NLP): Implement NLP to understand and process textual data from service manuals and diagnostic reports.

4. Training the Model​

• Supervised Learning: Train the model using labeled data. This involves feeding the model input data along with the correct output so it can learn the relationship between them.
• Validation and Testing: Split the data into training and testing sets to validate the model’s accuracy and performance.

5. Integration with Diagnostic Tools​

• OBD2 Scanners: Integrate the AI with OBD2 scanners to read and interpret diagnostic trouble codes (DTCs) in real-time.
• Diagnostic Software: Use platforms like ALLDATA DIY or MotorData to provide comprehensive diagnostic information and repair procedures.

6. Continuous Learning and Improvement​

• Feedback Loop: Implement a feedback loop where technicians can provide input on the AI’s recommendations, helping to refine and improve the model over time.
• Updates: Regularly update the AI with new data and repair techniques to keep it current with the latest automotive technologies.

Tools and Resources​

• Development Environment: Set up a robust development environment with tools like Python, TensorFlow, or PyTorch for building and training your models.
• Data Storage: Use databases like MongoDB or SQL for storing and managing your data.
• APIs and Integration: Develop APIs to integrate the AI engine with existing diagnostic tools and platforms.

Team and Expertise​

• Data Scientists: To handle data preparation, model training, and validation.
• Software Engineers: For developing the integration and deployment of the AI engine.
• Automotive Experts: To provide domain-specific knowledge and ensure the AI’s recommendations are accurate and practical.

By following this roadmap, you can create a powerful AI engine tailored to your specific needs.

 

[WoodsTruck]

Interesting for me to watch the AI capabilities.
I work for a large timber company in the R&D side, we are starting to get there, but still hard to wrap my head around AI growing trees. Probably something there, but we aren't there yet.

 

[Wiz02]

Interesting for me to watch the AI capabilities.
I work for a large timber company in the R&D side, we are starting to get there, but still hard to wrap my head around AI growing trees. Probably something there, but we aren't there yet.

While AI can't directly grow trees (at least not yet! ) if you teach it about your business, I'll bet that you will get some valuable insights into running it.

 

[olsenmotorsports]

Instead of wondering, just ask AI. Nothing wrong with that at all, I think it should become second nature to ask, you just need to remember that the response is just as likely to be bs as valuable.

Here's what it returned when I asked about creating an AI engine to help technicians do diagnosis. Now this response is software developer oriented, but it has a lot of interesting points, especially on harvesting sensor data from diagnostic tools.

If I wasn't retired, I would love to work on a first phase of a project that goes something like this (later phases would integrate AI into the scanner) : The AI would ask the user to input a code or symptoms. For applicable issues, it then specifies where to connect diagnostic probes and how to set up the scanner.

Technician then compares actual readings to samples shown by AI and follows instructions on what to do based on the closest matching waveform.

Thinking about this a bit more, this doesn't really help a tech understand the root cause, it's just a more sophisticated version of: scanner reports code AXXXX and flowchart says replace component ZZZ

But having something like the above capabilities may still be very helpful since it seems like inexperienced technicians may not know what they are looking at, or even what to look at if they do actually hook up or want to hook up a diagnostic scanner.

FYI This was the AI response to my question about building a custom AI for automotive diagnostics :

Here’s a more detailed roadmap to help you get started:

1. Data Collection​

• Historical Data: Collect past diagnostic reports, repair logs, and service manuals. This data will help the AI learn common issues and their solutions.
• Sensor Data: Gather real-time data from vehicle sensors, including engine performance, emissions, and other critical parameters.

2. Data Preparation​

• Cleaning: Ensure the data is clean and free from errors. This involves removing duplicates, correcting inaccuracies, and standardizing formats.
• Labeling: Label the data with relevant tags, such as specific issues, symptoms, and corresponding fixes.

3. Choosing the Right AI Model​

• Machine Learning Models: Use models like decision trees, random forests, or neural networks to analyze the data and identify patterns.
• Natural Language Processing (NLP): Implement NLP to understand and process textual data from service manuals and diagnostic reports.

4. Training the Model​

• Supervised Learning: Train the model using labeled data. This involves feeding the model input data along with the correct output so it can learn the relationship between them.
• Validation and Testing: Split the data into training and testing sets to validate the model’s accuracy and performance.

5. Integration with Diagnostic Tools​

• OBD2 Scanners: Integrate the AI with OBD2 scanners to read and interpret diagnostic trouble codes (DTCs) in real-time.
• Diagnostic Software: Use platforms like ALLDATA DIY or MotorData to provide comprehensive diagnostic information and repair procedures.

6. Continuous Learning and Improvement​

• Feedback Loop: Implement a feedback loop where technicians can provide input on the AI’s recommendations, helping to refine and improve the model over time.
• Updates: Regularly update the AI with new data and repair techniques to keep it current with the latest automotive technologies.

Tools and Resources​

• Development Environment: Set up a robust development environment with tools like Python, TensorFlow, or PyTorch for building and training your models.
• Data Storage: Use databases like MongoDB or SQL for storing and managing your data.
• APIs and Integration: Develop APIs to integrate the AI engine with existing diagnostic tools and platforms.

Team and Expertise​

• Data Scientists: To handle data preparation, model training, and validation.
• Software Engineers: For developing the integration and deployment of the AI engine.
• Automotive Experts: To provide domain-specific knowledge and ensure the AI’s recommendations are accurate and practical.

By following this roadmap, you can create a powerful AI engine tailored to your specific needs.

You hit the nail on the head. My experience has been that AI is always right when I provide the documentation for it to search/sort/analyze. It is when I just ask it to provide the answer that it usually fails in my space (I think it searches the forums )

I am currently searching for a program that I can load all of our stuff into but that doesn’t leave our server. Until that happens I just do not feel comfortable releasing information to the world. Ive got hundreds of thousands of dollars invested in information, hate for others to get it for free! (I know very selfish) I would love to build what you outlined above it would be an absolute game changer and would greatly increase efficiency allowing us to bring in more work.

Interesting for me to watch the AI capabilities.
I work for a large timber company in the R&D side, we are starting to get there, but still hard to wrap my head around AI growing trees. Probably something there, but we aren't there yet.

I love it. Hey at least you are trying. I feel like most aren’t yet. So you are way ahead of the curve.

 

[zanyad]

I am currently searching for a program that I can load all of our stuff into but that doesn’t leave our server.

From what I've heard recently, Microsoft Copilot may fit the bill. Just make sure to get an enterprise version, the free one definitely does leave your premises! I can wholeheartedly recommend our network services provider, send me a PM if you're interested!

 

[Wiz02]

My 2 cents is that AllData or Motors should be working on this AI engine project initially and then work with one of the scanner manufacturers to integrate it into their equipment.

Boy I would have enjoyed being the development manager for that project, as well as being hands on, but then again maybe not. It's time for the next generation to take over.

If you try to do this yourself, you are talking some significant development cost to develop a custom AI engine. Not to mention that unless you find a business analyst (BA) and software engineers who are into automotive diagnostics, you will have to be the subject matter expert.

Most likely your life will be even more chaotic as the BA who knows nothing about the subject pesters you constantly trying to write requirements that a clueless developer will try to implement.

Trust me when I say that the above scenario is not fun but it is definitely expensive. I think that this idea only makes sense for manufacturers that can amortize the development costs over the sale of many widgets.

But I would have loved to be part of it.

 

[olsenmotorsports]

From what I've heard recently, Microsoft Copilot may fit the bill. Just make sure to get an enterprise version, the free one definitely does leave your premises! I can wholeheartedly recommend our network services provider, send me a PM if you're interested!

That is currently what we use (yes we are enterprise). It is OK, just a little clunky to use, all of Microsoft is I find. Especially the new outlook (don’t get me started).

My 2 cents is that AllData or Motors should be working on this AI engine project initially and then work with one of the scanner manufacturers to integrate it into their equipment.

Boy I would have enjoyed being the development manager for that project, as well as being hands on, but then again maybe not. It's time for the next generation to take over.

If you try to do this yourself, you are talking some significant development cost to develop a custom AI engine. Not to mention that unless you find a business analyst (BA) and software engineers who are into automotive diagnostics, you will have to be the subject matter expert.

Most likely your life will be even more chaotic as the BA who knows nothing about the subject pesters you constantly trying to write requirements that a clueless developer will try to implement.

Trust me when I say that the above scenario is not fun but it is definitely expensive. I think that this idea only makes sense for manufacturers that can amortize the development costs over the sale of many widgets.

But I would have loved to be part of it.

I gotta be honest it is really hard to get mechanics to realize that A.I is the future and it is here. Really hard. This industry is so far stuck into the Stone Age in terms of systems and process’s (especially in the small independent space I am in) that it is infuriating to me.

I honestly do not know why or what it is. I think finances come into play for sure, but some of it is the “this is how we have always done it” mentality for sure. The above works of course, however there is this thing called innovation and the future. AI makes my brain spin, but if I don’t work on it now, get out in front of it, understand it, and integrate it, then someone else is. That puts them in front of me and I am not ok with that.

Speaking for myself, I have zero interest in creating the program etc, way above my pay grade and education level. I wish it wasn’t though!

 

[olsenmotorsports]

Some super nerdy stuff….



@MadeByMiller and I were chatting about some projects. So this is my first go around with 3d scanning (this is subcontracted out for now)



The old way we did it was rulers and tape measures and test fits. Man am I excited about this. We have our own line of intakes coming out in 2025. (All I can say) Here we are getting some fitted engine bay dimensions so we can maximize our room for heights.



Also we are in process of scanning an entire Porsche. This will be for future parts development. This is the center tunnel off a G body as we are also making shifters for release this year.

That is one of the beauties of having cars in various stages is if we need measurements we can do it pretty efficiently without taking anything apart.

Cannot wait to get started on developing some super trick Olsen Spec parts!

 

[Scuderia-F1]

Love to see that you’re back to lengthier videos on YouTube. Nothing better than workshop videos on a Friday night.

 

[olsenmotorsports]

Love to see that you’re back to lengthier videos on YouTube. Nothing better than workshop videos on a Friday night.

Thank you homie! We have some pretty big plans for media this year so hope we can live up to everyone’s standards!

 

[MadeByMiller]

Some super nerdy stuff….



@MadeByMiller and I were chatting about some projects. So this is my first go around with 3d scanning (this is subcontracted out for now)



The old way we did it was rulers and tape measures and test fits. Man am I excited about this. We have our own line of intakes coming out in 2025. (All I can say) Here we are getting some fitted engine bay dimensions so we can maximize our room for heights.



Also we are in process of scanning an entire Porsche. This will be for future parts development. This is the center tunnel off a G body as we are also making shifters for release this year.

That is one of the beauties of having cars in various stages is if we need measurements we can do it pretty efficiently without taking anything apart.

Cannot wait to get started on developing some super trick Olsen Spec parts!

Love to see it! I think you will remember this moment as the kickoff of a whole new way of approaching certain projects in the shop.

 

[Scuderia-F1]

Some super nerdy stuff….



@MadeByMiller and I were chatting about some projects. So this is my first go around with 3d scanning (this is subcontracted out for now)



The old way we did it was rulers and tape measures and test fits. Man am I excited about this. We have our own line of intakes coming out in 2025. (All I can say) Here we are getting some fitted engine bay dimensions so we can maximize our room for heights.



Also we are in process of scanning an entire Porsche. This will be for future parts development. This is the center tunnel off a G body as we are also making shifters for release this year.

That is one of the beauties of having cars in various stages is if we need measurements we can do it pretty efficiently without taking anything apart.

Cannot wait to get started on developing some super trick Olsen Spec parts!

This is such nice tech.
Recently I had a team over at my shop doing some reverse engineering on a 930 for a major player in the Porsche world (sadly can’t talk about that just yet. But to have these very special parts back in production will be awesome for everyone into cool 930 builds). The way this tech speeds up prototyping and reverse engineering is just amazing.

 

[olsenmotorsports]

We are closed this week for business and I am in Chicago this week for the holidays. Finally got around to the Bronco. I took this thing off the road in September and shipped it to CHI so I could build it. Haven’t touched it except to put the wheels on. Usually Try and spend my birthday week in the keys so plan is to get it all wrapped up and shipped there for the week to enjoy cruising around the keys in it.



Pulled it out of storage…. Love the new look with the wheels!



Pulled the front and rear bumpers off, running boards tow hooks, side rails. Pretty much everything bolted on to it. I am going to go Olsen Black on all that stuff - accent color will be the grey in the graphics. I was able to mix some paint that is a perfect match. After paint I will powdercoat over that in Casper Clear. Tow hooks, hitch receiver, mounts for running boards, and front tow hooks will all be that coastal grey. That will tie more contrast and flow into the exterior and give the Olsen Black a nice pop. The Olsen Black is a custom powder mix we made, it’s 40% clear, and has slight touches of purple and blue metallic. Very subtle and you have to look hard to see it, But it looks amazing on the Porsche trim.



Still have to take the spare tire carrier off as well as all the door handles and the exterior trim so clear bra can be applied to the whole car. Might do that today



I also got the seats out and shipped off to upholstery. Three tone with a dark sea blue Italian leather, a dark grey alcantara, and a jubilee stripe. Orange sticthing, and orange seat belts. That should tie the interior together better. Ford did a great job on the recaro’s making them comfy, but they fumbled at the 2 yard line on the styling aspect of the seat covers.



Seats out. Cannot wait to see these re done.

All that is left is to pull the center console and carpet out so we can install the sound deadnaming and lay in the wire looms for stereo. I was going to make pods for the rear speaker sections but TBH I just don’t think i will have the time to apply myself to do it. Just going to scrap that and make something work.

I will be in Florida most of January as the weather is nice there but I would really like to find some time before now and middle of March to devote 30-40 hours to get this project done (most of that is the stereo system)

 

[olsenmotorsports]

Question as I know there are a lot of smart people here.

I am trying to speed up the measuring process for engines, pistons, cylinders, crankshafts etc. Currently we are doing it by hand with micrometers etc (all Starret) but it is time consuming. I am looking to automate that process:

• The ability to log measurements and tie them to the work order
• Something that is super repeatable, easy to use and set up
• Accuracy within .001mm is all that is needed.
• Less than $5g (probably not realistic)

I have been looking at air gauges which seem to be the most economical of the stuff I am finding.

Any insight anyone can share?

 

[WhoWhatNow]

Question as I know there are a lot of smart people here.

I am trying to speed up the measuring process for engines, pistons, cylinders, crankshafts etc. Currently we are doing it by hand with micrometers etc (all Starret) but it is time consuming. I am looking to automate that process:

• The ability to log measurements and tie them to the work order
• Something that is super repeatable, easy to use and set up
• Accuracy within .001mm is all that is needed.
• Less than $5g (probably not realistic)

I have been looking at air gauges which seem to be the most economical of the stuff I am finding.

Any insight anyone can share?

You are looking for a CMM.

Coordinate Measuring Machines | Mitutoyo America Corporation

Mitutoyo manufactures top-quality coordinate measuring machines (CMM) that lead the industry in precision. Find the perfect solution for you.

www.mitutoyo.com

I believe Zeiss makes optical CMMs. I can’t speak to price but I don’t think they will be under $5K. I have seen them come up at machine shop auctions from time to time.

 

[scooby074]

Question as I know there are a lot of smart people here.

I am trying to speed up the measuring process for engines, pistons, cylinders, crankshafts etc. Currently we are doing it by hand with micrometers etc (all Starret) but it is time consuming. I am looking to automate that process:

• The ability to log measurements and tie them to the work order
• Something that is super repeatable, easy to use and set up
• Accuracy within .001mm is all that is needed.
• Less than $5g (probably not realistic)

I have been looking at air gauges which seem to be the most economical of the stuff I am finding.

Any insight anyone can share?

You can have hand held measurement tools that are connected to and upload data to PC. This can be entered into a a custom file to store this data. Attaching it to a work order should be no problem, heck I imagine there is existing software that will do all this off the shelf.

I dont remember exactly what models my old workplace used, but I think they were Mitutoyo. QC wasnt my dept. Check out what Mitsutoyo has to offer on their website as a starting point.

 

[Scuderia-F1]

Like mentioned above, Mitutoyo has what you’re looking for.
QMP Racing also has some very very neat fixtures etc if you wanna go next level with the way you are taking the measurements.

 

[olsenmotorsports]

. I can’t speak to price but I don’t think they will be under $5K.

My weekly hunt down the rabbit hole of “things I didn’t know I needed for $25g”

 

[olsenmotorsports]

You can have hand held measurement tools that are connected to and upload data to PC. This can be entered into a a custom file to store this data. Attaching it to a work order should be no problem, heck I imagine there is existing software that will do all this off the shelf.

I dont remember exactly what models my old workplace used, but I think they were Mitutoyo. QC wasnt my dept. Check out what Mitsutoyo has to offer on their website as a starting point.

Appreciate the insight sir!

Like mentioned above, Mitutoyo has what you’re looking for.
QMP Racing also has some very very neat fixtures etc if you wanna go next level with the way you are taking the measurements.

Those fixtures are pretty neat looking! Have any experience with them?

 

[danski0224]

My weekly hunt down the rabbit hole of “things I didn’t know I needed for $25g”

I wouldn't be surprised if you are dumping that much or more into your Bronco, which won't make you any money...

Not that there is anything wrong with that.

I'd like to have a spare 25g or so to dump into my 79 Ford truck.

 

[Scuderia-F1]

Appreciate the insight sir!

Those fixtures are pretty neat looking! Have any experience with them?

In my opinion they’re essential if you build engines, otherwise you just assemble them.

 

[BORING HOP YARD]

You guys sound like you're talking about SPC outputs on you measuring tools, operator pushes a button and sends the data to the computer. We used what was called "data trucks" that were small handheld devices to collect the date at the machine or in assembly documenting torques being done.
I don't see any type of CMM, FARO arm or laser scanning tooling working for this function.
My recommendation is to have two sets of measuring tools, the ones being used and a set to hold if you find one out of spec so you shut down your production. If we had a requirement of + or - .001 our certified tool had to be certified to + or - .0005.

 

[930dreamer]

So I know a guy that still wants an aircooled 911/930. What's the most practical way to find one, would prefer from a dry desert state.

 

[olsenmotorsports]

You guys sound like you're talking about SPC outputs on you measuring tools, operator pushes a button and sends the data to the computer. We used what was called "data trucks" that were small handheld devices to collect the date at the machine or in assembly documenting torques being done.
I don't see any type of CMM, FARO arm or laser scanning tooling working for this function.
My recommendation is to have two sets of measuring tools, the ones being used and a set to hold if you find one out of spec so you shut down your production. If we had a requirement of + or - .001 our certified tool had to be certified to + or - .0005.

I am actually looking to speed up the process. Currently we are measuring “old school” with micrometers etc. Nothing wrong with that, but I would like to accomplish two things - eliminate human error, or at least mitigate it, and become more efficient.

I was researching and came across air gauges (new to me) which seemed to fit a budget I was comfortable as well as fitting the tolerances we need to fit in.

Air Gauge Measuring Device

Im sure optical/laser/3d would be able to accomplish it, just probably out of scope of what I want to spend.

There is around 4 hours in labor built in to measuring an engine paid to a tech, and it takes all of that to do it properly. I would like to get that down to under an hour to make them more efficient, thus putting more money in their pocket.

Not sure if this is even feasible but it is something I am looking into!

 

[olsenmotorsports]

I wouldn't be surprised if you are dumping that much or more into your Bronco, which won't make you any money...

Not that there is anything wrong with that.

I'd like to have a spare 25g or so to dump into my 79 Ford truck.

Agree, but that is all personal projects as well as myself funding it.

Funny thing is, my personal vehicle collection actually does make me money, it clears my head, helps me unplug, thus making me more productive and creative as a result!

The BS that comes firing at me every single day has to be channeled into something - otherwise I would be untenable to deal with

 

[olsenmotorsports]

So I know a guy that still wants an aircooled 911/930. What's the most practical way to find one, would prefer from a dry desert state.

I would recommend using a broker - stay off the online sites etc. My staff will be happy to make some recommendations to steer you in the right direction!

 

[SilverJimmy]

From my experience there is a small amount of human miss accuracy when using hand held measuring tools. Every person will get a slightly different reading when using even the same tool measuring the same part. It might be negligible, but it does happen and it can add up. If you can eliminate that maybe the results will be worth the effort and expense.

 

[scooby074]

Appreciate the insight sir!

I did a little quick reading on Mits site and found this. It might help point you in the right direction https://www.mitutoyo.com/webfoo/wp-content/uploads/Data-Management-Systems_2275.pdf

 

[BORING HOP YARD]

You might consider a time study to determine the areas that could be improved in the current measurement process.
We had a couple of terms "value" or "non-value" added. The goal is to understand the waste in the process, make changes to eliminate the waste.
If there is a concern about how accurate the current process is, you could do a Gauge R/R study "repeatability reproducibility".
This would show you the capability of the measurement process between operators using the exact same tools.
Repeatability is about measuring equipment variations and reproducibility is about operator variation.

We purchased some air gauges, and they worked great, the part we didn't like was the dimensional range the gauges could measure.
If you were checking one size bore no big whoop, 10 different bores may require 10 separate gauges. The inspection area had a small footprint on the factory floor and the last thing we needed was more tool to store.
This was about 10 years ago air gauges could have improved.

Some of our measuring tools had to stay at a certain temperature in an air-conditioned room, like our CMMs.
Hope this helps.

 

[inphx]

When it comes to AI for troubleshooting and working with private datasets, I’m testing two approaches. One is running LLaMA locally on my hardware with a Retrieval-Augmented Generation (RAG) setup. The other is similar but hosted on Microsoft’s Azure private cloud SaaS, (ChatGPT with a RAG) which they attest keeps your data secure. Some colleagues at larger companies had their legal teams review Microsoft’s service agreement, especially with GDPR in mind, and they decided it was safe to use.

With LLaMA i have experimented with Cadillac shop manuals for a personal project with the goal of an ordered checklist for an engine out job that threads the manuals referred to sections and pulls in torque/cautions/required tools inline to a personally generated serialized to do guide.

I keep hearing about the need for private, on-premises data models, but the cost of the hardware—like NVIDIA GPUs and the memory needed for ElasticSearch—can be pretty high for just the hardware. My background is in aerospace but the domains I work in include health & safety, first responders, manufacturing, logistics and legal. I've also been thinking about how to combine AI with military-grade IETM (Interactive Electronic Technical Manual) systems. Adding blockchain could allow trusted collaborator entities around the world to refine troubleshooting and securely enrich or update the data. Blockchain would also help on intermittent connections scenarios like subs or trusted collaborators.

One of the Musk podcasts had him thinking out loud on a scenario that all the parked idle Teslas could be offering their on board compute as a service much like the idea your Tesla can someday be part of a robo taxi fleet. If the on premises “appliance” server was provided as a lease with support and preconfigured tools - I wonder if any client would be ok with idle time being used by the provider… (being compensated for power impact to be fair).

Fun times we are living in. ALL THE BEST for 2025! Very much appreciate the tech details in this thread.

 

[Wiz02]

When it comes to AI for troubleshooting and working with private datasets, I’m testing two approaches. One is running LLaMA locally on my hardware with a Retrieval-Augmented Generation (RAG) setup. The other is similar but hosted on Microsoft’s Azure private cloud SaaS, (ChatGPT with a RAG) which they attest keeps your data secure. Some colleagues at larger companies had their legal teams review Microsoft’s service agreement, especially with GDPR in mind, and they decided it was safe to use.

With LLaMA i have experimented with Cadillac shop manuals for a personal project with the goal of an ordered checklist for an engine out job that threads the manuals referred to sections and pulls in torque/cautions/required tools inline to a personally generated serialized to do guide.

I keep hearing about the need for private, on-premises data models, but the cost of the hardware—like NVIDIA GPUs and the memory needed for ElasticSearch—can be pretty high for just the hardware. My background is in aerospace but the domains I work in include health & safety, first responders, manufacturing, logistics and legal. I've also been thinking about how to combine AI with military-grade IETM (Interactive Electronic Technical Manual) systems. Adding blockchain could allow trusted collaborator entities around the world to refine troubleshooting and securely enrich or update the data. Blockchain would also help on intermittent connections scenarios like subs or trusted collaborators.

One of the Musk podcasts had him thinking out loud on a scenario that all the parked idle Teslas could be offering their on board compute as a service much like the idea your Tesla can someday be part of a robo taxi fleet. If the on premises “appliance” server was provided as a lease with support and preconfigured tools - I wonder if any client would be ok with idle time being used by the provider… (being compensated for power impact to be fair).

Fun times we are living in. ALL THE BEST for 2025! Very much appreciate the tech details in this thread.

Interesting ideas. I hope that one of they work out for you. I really like the idea of having an organized checklist with torque specs related to the asswbly's fasteners.

I hate when you have to jump all around because you need to go to another section of the manual to remove an assembly and then return to continue on the main task.

One other feature that factory service manuals (FSM) don't include that I think would help people doing complex service procedures infrequently, is a detailed description of exactly how to uninstall something. For example, show the locations of the Christmas tree fasteners that hold a door card and where to pry or the location of hidden fasteners.

YouTube has been a huge help with this, but the quality level of information varies wildly and you waste a lot of time until you find credible information.

On the computing front, when cloud services were starting to grow in popularity a few years ago, I was not huge fan of hosting a firm's intellectual property in the cloud.

These days, nearly every company regardless of size as well as the public sector, including the military, uses cloud based services and hosts proprietary data in the cloud.

Trying to duplicate the computational power, reliability and security of a data center in house is a tall task as I'm sure you know. These days the consensus is that it's cheaper and more secure to go cloud based, but I still believe that on premise still makes sense for some applications, but not CPU intensive ones.

 

[olsenmotorsports]

You might consider a time study to determine the areas that could be improved in the current measurement process.
We had a couple of terms "value" or "non-value" added. The goal is to understand the waste in the process, make changes to eliminate the waste.
If there is a concern about how accurate the current process is, you could do a Gauge R/R study "repeatability reproducibility".
This would show you the capability of the measurement process between operators using the exact same tools.
Repeatability is about measuring equipment variations and reproducibility is about operator variation.

We purchased some air gauges, and they worked great, the part we didn't like was the dimensional range the gauges could measure.
If you were checking one size bore no big whoop, 10 different bores may require 10 separate gauges. The inspection area had a small footprint on the factory floor and the last thing we needed was more tool to store.
This was about 10 years ago air gauges could have improved.

Some of our measuring tools had to stay at a certain temperature in an air-conditioned room, like our CMMs.
Hope this helps.

Exactly the feedback for what I was looking for thank you!

For bore sizes we are measuring 46-102mm so I am sure we will need a few different probes which I am ok with kitting up for.

Grateful for the feedback. I am thinking Air gauge will be the fastest learning curve and also allow us to efficiently measure and track.

The regular old school stuff works great, but to @SilverJimmy ‘s point there is that different operator different measurement.

This came up recently as we had cylinders fail post QC as the tear down guy measured them within spec and the assembler measured them out of round, delaying us a few days while we waited for parts/customer approval for new jugs and pistons. Nothing costs more money than putting the brakes on a project and waiting on parts.

I am also going to be making some fixtures to help us measure straighter and more accurately.

Right now we are all measure by hand and notes by hand. It is garbage efficiency lol.

 

[olsenmotorsports]

With LLaMA i have experimented with Cadillac shop manuals for a personal project with the goal of an ordered checklist for an engine out job that threads the manuals referred to sections and pulls in torque/cautions/required tools inline to a personally generated serialized to do guide.

We are doing much the same, had great success in sorting information faster. Really a time saver.

What is the Caddy project?

 

[inphx]

No comparison to your art work... very "ghetto leno" ... and now all my washers are backwards