That makes sense, and I think you are describing the real product problem.
Capturing data is the easy part. If the owner or technician has to manually dig through five minutes of millisecond-level logs, the product has already failed.
The device would be at the ECM harness, not at the OBD port, so I agree that data retrieval and event marking need to be thought through carefully.
The way I am thinking about the architecture is:
The recorder itself should not depend on a phone, app, Bluetooth, Wi-Fi, or cloud connection to capture the event. It should always keep a local rolling buffer and lock the event locally.
A button, phone app, or small cabin device would only act as an event marker. If the driver feels a stumble and presses the button 10–30 seconds later, the pre-buffer has to already contain the useful data.
For data retrieval, the practical options would be a sealed service USB lead, Wi-Fi download, or a phone/cabin companion device. I would not expect the owner to remove the ECM-side module or work with raw files directly.
The cloud or AI side would be for interpretation, not for capturing the event. The truck may have no connection when the issue happens, so the evidence has to be saved locally first. After that, cloud processing could help decode the data, compare it against baselines, and generate a readable report.
For the first version, I would keep the automatic triggers conservative and objective:
driver event marker
bus-off
error passive
voltage drop / brownout
device reset
FIFO or queue overflow
a normally periodic message disappearing
side-to-side communication mismatch, if the topology supports that
For “learning normal,” I agree with your point, but I would not want to overclaim it as automatic root-cause diagnosis at first.
A realistic first step would be learned baseline comparison for that specific vehicle and operating condition. For example, a value would only be compared against similar conditions:
RPM range
load / MAP
throttle position
gear / vehicle speed
coolant and oil temperature
battery voltage
AFM/DFM state, if decoded and validated
Then the report could flag things like:
this periodic message disappeared compared with its normal timing
this value deviated from this vehicle’s normal range under similar conditions
the same abnormal pattern repeated after the same type of event
the anomaly occurred together with voltage, oil-pressure, misfire, or communication changes
But I would still call that “abnormal pattern detected,” not “replace this part,” unless there is enough validated repair data behind it.
So the intended product would not be “here is a huge log.” It would need to be an event package:
what triggered the capture
how much pre/post data was preserved
what changed before and after the event
whether the device itself reset, overflowed, or saw a bus error
selected graphs around the event
raw data only as supporting evidence
From your perspective, what would make this kind of report useful instead of just another datalog?
For example:
What are the top 5 parameters or events you would want highlighted first?
Would you trust a learned baseline for that specific vehicle, or would you prefer fixed thresholds?
How much false-positive flagging would be acceptable before you stopped looking at the reports?
What would a one-page report need to show for an independent shop to take it seriously?
For misfire, AFM/DFM, oil pressure, or U-code complaints, what would you want the tool to flag automatically?
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now