Monitoring engine health with OBD datalogging

Status
Not open for further replies.
Joined
Oct 20, 2005
Messages
11,956
Location
PA
I want to try to use OBD datalogging to catch problems with aging injectors/sensors/etc. as they develop, before they're bad enough to trigger a CEL. The idea is to randomly datalog my drives to build up a history, and then watch for deviations from the baseline.

Anyone have feedback on good parameters to measure? I’ll be using Torque Pro with a generic OBD Bluetooth dongle.

Fuel trims and cat temp are two fairly obvious examples. Another might be ignition timing. I'd imagine any of those would have to be compared against IATs, throttle, measured load, etc. But that bridge can be crossed later.

I can’t be the first to think of this. Anyone heard of others doing it? What do tuners monitor when they do remote tunes?

Any thoughts welcome.
 
My question would be, what are you going to do with all that data? I mean, sample engine temp at once/second, and that's sixty floats per minute, times however long you drive. Times however many trips over time. Then you need some sort of record of what type of driving. Track day or errand running.

Need some sort of tool to open the datalogs and mine out the data, one which can open multiple datalogs at once, plotting trends from different runs. You could copy&paste into Excel, I suppose.
 
Originally Posted By: supton
My question would be, what are you going to do with all that data? I mean, sample engine temp at once/second, and that's sixty floats per minute, times however long you drive. Times however many trips over time. Then you need some sort of record of what type of driving. Track day or errand running.

Need some sort of tool to open the datalogs and mine out the data, one which can open multiple datalogs at once, plotting trends from different runs. You could copy&paste into Excel, I suppose.

Good call. Have a few thoughts on that; may address it later. For now, I'm mainly trying to figure out how feasible it'd be in theory.
 
What's with the comments about having too much time on your hands? Data logging is done in the background and the data is easy to digest, so long as it can be copied to an excel file.

I have no experience with BT dongles and phone logging, but I used to datalog frequently using Cobb's AccessTunerRace software and a cheap netbook, which I hear is no longer being offered to AccessPort owners.

I don't have any logs in front of me, so, from memory, I believe I logged the following (either all at once or several at a time, depending on my objective):

Requested torque
Coolant temp
Intake air temp
Boost
Boost error
Feedback knock correction
Fine learning knock correction
Dynamic (Ignition) advance multiplier
Long-term fuel trim
Short-term fuel trim
Injector duty cycle
Engine load
EGT
AFR 1 (upstream)


Ignition timing will have too many variables and it will be tough to find a pattern there. Requested torque, AFR, knock and fuel trims (and boost, in my case) were logged all the time. IDC to see if I was going above 100%. Others, as desired.

I believe sample rates were 1/s but it might have been 10/s - can't recall the resolution.

Use conditional formatting in excel to quickly highlight any anamolous or desired data. I would highlight low boost in light green, high boost in dark green; knock in red; anamolous fuel trims (I used +\- 4 for fuel trim highlighting), etc.
 
Last edited:
I found the data that is usually captured in the $06 section of OBDII analysis to be very helpful, especially the misfire information.

What I found was that I had a small number of misfires across cylinders 1, 3, 5 (Bank 1)...It was small enough so as not to trigger a CEL (less than 10 per cylinder), but greater than zero, which is what was shown for cylinders 2, 4, 6 (Bank 2).

As a result of that data, I changed the pre-cat O2 sensor on that side, and ran valve cleaner through the air intake to see if that would help...Too soon to tell, but the number it shows for the last 10 cycles keeps coming down, so I am cautiously optimistic it helped.

Long story, short...It helped me identify and address an issue, before it became a bigger problem, so I would recommend it to others.
 
I'm a data guy and very fluent in Excel. What you are proposing to do is time consuming and difficult unless you understand vehicle data well. I was troubleshooting a CEL and suspected an O2 sensor. Teaching myself how those sensors work took a little time. Then trying to compare my data graphs to known good graphs and bad graphs I found on the internet was not easy. It's like looking at a cardiogram. You need to know what to look for. If it's not bad enough to throw a code, it's going to be subtle. A trained doctor can look at a cardiogram and easily tell normal from abnormal. A trained person can do the same with a cardiogram or vehicle data. You just have to know what to look at.

I suggest you do it. If nothing else, you'll learn a lot about vehicle data and data analysis. And plotting graphs in Excel.
 
Originally Posted By: gathermewool
What's with the comments about having too much time on your hands?


Because it's true. What will the data be useful for? To pay extra for more fancy fluids or parts before they're needed? Seems like a worse outcome. Operational observations and fuel economy changes are much less effort and strike right at the hear of the issue which is whether the vehicle is operating acceptably in its purpose as a vehicle.

Quote:
Data logging is done in the background


I just snapped my fingers and don't have data logging set up and done in the background. Therefore, it takes time.

Quote:
and the data is easy to digest, so long as it can be copied to an excel file.


Takes time, and takes time. And more time.

Quote:
I have no experience with BT dongles and phone logging, but I used to datalog frequently using Cobb's AccessTunerRace software and a cheap netbook, which I hear is no longer being offered to AccessPort owners.


So you're suggesting that you would have to... spend time to do this?

It's one thing to use data to tune a custom engine/exhaust/etc setup for peak performance, to win a race or even out of vanity.

It's another thing and fair use of time if it's considered a curiosity, a hobby, but as far as practical applications go, mostly a waste of time for stock daily drivers.

When something needs done, it will become apparent and then pull the OBDII codes. It would be different if there weren't a computer constantly compensating for several variables.
 
My STI engine ran just fine, and my boost numbers are right on with a large turbo inlet tear. I heard a faint squeal (like letting the air out of a balloon) that was barely perceptible at first, but slowly got worse. I knew to actively start looking for an issue when my fuel trims started getting wonky. The tear was post-MAF, pre-turbo, so it was sucking in extra unmetered air.

//

Again, I don't know how easily datalogs can be ported from phone apps, but if it's as easy as copying-and-pasting from a .csv file to .xls and creating some conditional formatting, then Leo99 is way off in his assessment in time spent. Once I got my conditional formatting set, it takes literally single-digit minutes to pour over hours of data!

I don't care about most of the run time that looks great. I care about that time I boosted a little more in 6th than I should and saw some knock as a result or specific trends in fuel trims that could indicate leaks, or a million other things that could indicate anything from sub-par fuel (bad fuel can be indicated in increased knock values, well above ghost knock and normal tip-in throttle response during too-aggressive driving for the gear you're in; all of which is imperceptible audibly) to major issues that should be quickly trouble-shot. Seriously, I would have tanks where I'd see nothing but the occasional ghost -1.4 degrees of feedback knock, but my learned knock would be 0 for the cell and DAM would be 1.0 (max); then, the next time, I'd see -2+ retard at increased load levels.

This isn't rocket surgery, folks. Once you get a system down (which is really fun to figure out and learn about,) you can see things that are out-of-whack very quickly. This isn't like staring at your coolant temp idiot gage for the first drive to see if you burped the radiator well enough or checking your O2 sensor response by watching the volt-meter cycle up and down (i.e., very little to be gleaned via these methods) - it CAN BE pretty darned informational.

//

I really hope that these apps show enough data, at a high enough resolution (1 to 10 samples per second) or else I've pretty much put my foot in my mouth, since I'd then be comparing apples (Cobb AP and AccessTunerRace software) to some Oranges (some rough Apple app.)
27.gif


//

In the end, and to emphasize an earlier point: this kind of stuff IS INTERESTING AND FUN. This is the kind of thing we're on a freaking automotive forum to discuss, for crying out loud! Data from a logger will be a heckuva lot more valuable than trying to figure out what that 2 ppm Si in your UOA means!
24.gif
36.gif
 
Last edited:
Thanks, guys.
cheers3.gif


gathermewool and ae7456t, seems like you understand where I'm coming from 100%. What you describe doing with these data is exactly the kind of stuff I was hoping to achieve.

Leo99, point taken about time and expertise needed, and I agree with your last line; even if this fails, I'll have learned something.

CT8, I'd definitely use MPG if my driving didn't vary so widely. However, I am still monitoring it.

I've also had some feedback from other people so hopefully I'll be pulling something together soon.
 
This sort of goes against the electronic monitoring you are interested in but IMO will give better results in getting the job you want done. If you don't already know how to use one for diagnostics read up on it, you will be amazed at the abilities of one of these gauges.
It will indicate when the valves are getting dirty on DI engines for example in an instant with no scan tool interpretation required, everything effects engine vacuum. One of my all time favorite diagnostic tools.

https://www.amazon.com/Auto-Meter-2337-Autogage-Vacuum/dp/B00062YVT8

Edit: I should mention it is also the perfect tool for diagnosing failing cats due to plugging.
 
Last edited:
Originally Posted By: d00df00d
Originally Posted By: supton
My question would be, what are you going to do with all that data? I mean, sample engine temp at once/second, and that's sixty floats per minute, times however long you drive. Times however many trips over time. Then you need some sort of record of what type of driving. Track day or errand running.

Need some sort of tool to open the datalogs and mine out the data, one which can open multiple datalogs at once, plotting trends from different runs. You could copy&paste into Excel, I suppose.

Good call. Have a few thoughts on that; may address it later. For now, I'm mainly trying to figure out how feasible it'd be in theory.


You could keep all logs saving the file with name.
Also you could append them into one big worksheet with an additional column for date when file captured.

Like that you can work a bit more on historical data/trends/what's a good average/exceptions......
 
Status
Not open for further replies.
Back
Top