1. Attachments are working again! Check out this thread for more details and to report any other bugs.

ICE, MG1, and MG2 rpm vs Graham Ratio

Discussion in 'Generation 1 Prius Discussion' started by bwilson4web, Apr 28, 2010.

  1. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,375
    15,513
    0
    Location:
    Huntsville AL
    Vehicle:
    2018 Tesla Model 3
    Model:
    Prime Plus
    As part of the Auto Enginuity testing, using "ISO-9141-2" and "KWP2000 Forced ECM Init" achieved the longest recorded file, ~35 minutes. Data recorded during this test:

    • Engine Speed (RPM) [0 - 8000]
    • Generator - MG1 Revolution (Nm) [-16384 - 16383]
    • Motor - MG2 Revolution (RPM) [-16384 - 16383]
    • Motor Current W (A) [-750.00 - 750.01]
    The first three data items are the elements of the famous Graham Power Split Device ratios:
    MG1 = 3.6 * ICE - 2.6 * MG2 (source)

    MG1 - Generator MG1 rpm
    ICE - Engine Speed rpm
    MG2 - Motor MG2 rpm
    ​
    So I loaded the data into a spreadsheet realizing there were would be some data sampling errors and plotted the Auto Enginuity MG1 reported value against the calculated value using the ICE and MG2 rpm only it didn't work.

    I noticed that when the car was not moving, MG2 = 0, the numbers were off by a factor of two. Then when MG2 was rotating, the shape of the curves was similar but offset. So I modified the calculation to easily change the MG2 multiplier and derived this formula:
    MG1 = (3.6 * ICE - 5.52 * MG2) / 2

    MG1 - Generator MG1 rpm
    ICE - Engine Speed rpm
    MG2 - Motor MG2 rpm
    ​
    This is the plot of the adjusted formula:
    [​IMG]

    There are two unknowns in a single formula so I can't really quantify the errors. However, I have a shorter file recording that includes MG2 and vehicle speed to see how MG1 vs speed looks compared to the Graham ratio:
    MG2 = 59.1 * MPH
    ​
    I had recorded in another file MG2 rpm and vehicle MPH and they are off by a factor of ~2.1 . . . Of course here are two different values in a single, separate formula. But this one clearly shows the problem of internal consistency of rpm values.

    It is entirely possible that Graham's formulas and data from the Graham Miniscanner are off. However, years of experience with and a substantial body of work indicates this is not the case. Occams Razor says Auto Enginuity data values need calibration or independent validation (aka. customer testing) before use.

    It is too soon to fault-isolate the ProLine or control software. The ambiguity group is too large. One consistent pattern has been byte sized numbers seem to pass the reasonableness test. It looks like data elements larger than a byte need to be validated.

    Bob Wilson

    ps. This is just a quick posting. If we use the 2.1 factor for MG2 error, we may be able to estimate the MG1 error. If simple correction constants can be formulated, we should preserve the Graham ratio and then be able to use these two data values.
     
    #1 bwilson4web, Apr 28, 2010
  2. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,375
    15,513
    0
    Location:
    Huntsville AL
    Vehicle:
    2018 Tesla Model 3
    Model:
    Prime Plus
    <head slap>

    There are times in the data when the one of the variables goes to zero:

    • stopped and engine running -> MG2 = 0
    • engine off and moving -> ICE = 0
    Tossing out the transients, the ratios look like:

    • ICE / MG1 ~= 1.73 (1)
    • MG1 / MG2 ~= -2.88 (2)
    Now we're in a position to compare the Auto Enginuity numbers versus the geometry derived Graham numbers:
    MG1 = ICE / 1.73 (1)
    MG1 = -2.88 * MG2 (2)

    2 * MG1 = (ICE / 1.73) + (-2.88 * MG2)
    2 * MG1 = ( ICE + ( -4.98 * MG2 ) ) / 1.73
    3.46 * MG1 = ICE + ( -4.98 * MG2 )

    MG1 = ( ICE + (-4.98 * MG2) ) / 3.46
    ​
    This is pretty different from the Graham, geometry model.
    MG1 = 3.6 * ICE - 2.6 * MG2
    ​
    Something must be wrong. Of course it is only fair to capture these three numbers, ICE, MG1 and MG2, and do the same analysis. It will take a day to build the adapter.

    Bob Wilson
     
    #2 bwilson4web, Apr 28, 2010
    1 person likes this.
  3. usbseawolf2000

    usbseawolf2000 HSD PhD

    Joined:
    Sep 22, 2004
    14,487
    2,996
    0
    Location:
    Fort Lee, NJ
    Vehicle:
    2012 Prius Plug-in
    Model:
    Plug-in Base
    Bob, are the data from NHW11 or ZVW30? I am guessing NHW11 because you had to initiate in KWP2000 mode.
     
    #3 usbseawolf2000, Apr 28, 2010
  4. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,375
    15,513
    0
    Location:
    Huntsville AL
    Vehicle:
    2018 Tesla Model 3
    Model:
    Prime Plus
    Correct. We have an NHW11 and a ZVW30 and as you've noticed, setting up the AE to work reliably with this model has been interesting.

    I suspect it works much better with the NHW20 but I don't have one to test. But by sharing my methodology, folks are welcome to check the values with the NHW20.

    I strongly suspect we are seeing the effects of a Layer 1 issue, getting the 10.5 baud, variable length responses. When there is a subtle timing problem, the errors tend to congregate at the end of larger packets. Yes, I know, 'no body has network timing problems today' . . . I've been there and have the bloody T-shirt. <grins>

    Bob Wilson
     
    #4 bwilson4web, Apr 28, 2010
  5. usbseawolf2000

    usbseawolf2000 HSD PhD

    Joined:
    Sep 22, 2004
    14,487
    2,996
    0
    Location:
    Fort Lee, NJ
    Vehicle:
    2012 Prius Plug-in
    Model:
    Plug-in Base
    When MG2 = 0 and ICE is warming up (~1,280 rpm), MG1 should be ~4,600 rpm. If you are seeing by a factor of two, are you seeing ~9,200 rpm? That can not be possible because NHW11 MG1 max is 6,500 rpm.
     
    #5 usbseawolf2000, Apr 29, 2010
  6. bwilson4web

    bwilson4web BMW i3 and Model 3

    Joined:
    Nov 25, 2005
    27,375
    15,513
    0
    Location:
    Huntsville AL
    Vehicle:
    2018 Tesla Model 3
    Model:
    Prime Plus


    Good comment!

    What I initially did was 'curve fitting', which meant I parameterized the Graham formula so I could change different factors. The divide by two was an initial hack to bring things into the region. Then I tweaked the MG2 multiplier until the curves fit.

    Here is some of the data:
    Column 1 Column 2 Column 3 Column 4 Column 5 Column 6
    0 time (mm:ss) Engine rpm MG1 rpm MG2 rpm MG1 calc rpm err
    1 12:51 1184 2048 0 2131 -83
    2 12:52 1184 2048 0 2131 -83
    3 12:53 1184 2176 0 2131 45
    4 12:54 1184 2048 0 2131 -83
    5 12:55 1184 2048 0 2131 -83
    6 12:57 1184 2048 0 2131 -83
    7 12:58 1248 2048 0 2246 -198
    8 12:59 160 0 0 288 -288
    9 13:00 0 0 0 0 0
    10 13:01 0 -128 0 0 -128
    11 13:03 0 -128 0 0 -128
    12 13:04 0 -256 0 0 -256
    13 13:05 0 -384 0 0 -384
    14 13:06 0 -384 128 -353 -31
    15 13:07 0 -384 0 0 -384
    16 13:09 0 -384 128 -353 -31
    17 13:10 0 1152 128 -353 1505
    18 13:11 1728 2560 256 2404 156
    19 13:12 2144 3072 384 2799 273
    20 13:13 2464 3200 384 3375 -175
    21 13:15 2592 3200 512 3252 -52
    22 13:16 2688 3200 640 3072 128
    First four columns from AE and last are calculated error.

    Bob Wilson
     
    #6 bwilson4web, Apr 30, 2010
  7. usbseawolf2000

    usbseawolf2000 HSD PhD

    Joined:
    Sep 22, 2004
    14,487
    2,996
    0
    Location:
    Fort Lee, NJ
    Vehicle:
    2012 Prius Plug-in
    Model:
    Plug-in Base
    Ok, so AE is reporting half the value we were expecting.
     
    #7 usbseawolf2000, Apr 30, 2010
  8. 2009Prius

    2009Prius A Wimpy DIYer

    Joined:
    Mar 25, 2009
    2,705
    510
    63
    Location:
    USA
    Vehicle:
    2009 Prius
    I suppose not many people use the feature so AutoEnginuity didn't even bother to double check their software. I almost decided to buy it but didn't in the end after learning from the manufacturer that it can talk to only one ECU in the same session - useless if I want to record data from multiple ECUs all together. I ended up writing my own program and here is what I got for the RPM numbers: (essentially verified the linear relationship 3.6 ICE = MG1 + 2.6 MG2)
    http://priuschat.com/forums/gen-ii-...un-early-problem-detection-2.html#post1224867
     
    #8 2009Prius, Dec 9, 2010
    1 person likes this.
  9. w2co

    w2co Member

    Joined:
    Dec 11, 2009
    366
    81
    0
    Location:
    Longmont, CO.
    Vehicle:
    2006 Prius
    Model:
    N/A
    Interesting indeed, I also have both AE and scangaugeII and know that some of the recorded data may be invalid from AE but the question still stands - is it good at reading dtc when a real one occurs. That is what the average person would use it for. Sure one can take all the data they want when things are running good and try to understand it all, even build up a baseline (normal) but the acid test is mentioned above, we know it can clear the codes but how good does it correctly detect them is the question in my mind. I have not had any dtc's in either the gen1 or gen2 since I've owned AE so can't say on this yet, but it does work much better on the gen2, the can bus works much better and less glitches, but I have not really had the time to look at the stuff you've been collecting. I also know in both gen1&2 AE does a nice job at anything battery ECU related. I consider this probably the highest probability of a failure in a gen1 with the hv battery aging now.
    But still good!
     
    #9 w2co, Dec 14, 2010
(You must log in or sign up to post here.)