Auto Enginuity vs MG1

Anytime you get a new instrument, the first steps should be to check the 'calibration.' A badly calibrated instrument has led many an engineer into a 'bad place.' So I was interested in seeing how well Auto Enginuity (AE) handles MG1 data.

MG1 Torque - USELESS!

This chart includes a test drive that includes numerous ICE off and similar conditions. But this is the distribution of AE MG1 (aka., generator) torque values:

When the shifter is put in "N", there is no torque applied to the engine but this chart shows what appears to be a linear, signed value whose units make no sense. The Power Split Device (PSD) provides 28% counter torque to the engine. A value of either 500 or -500 Nm would imply an engine torque of 500/.28 ~= 1,756 Nm. Now I like our 1.5 L engine but it is NOT providing this much torque!!

MG1 Current V - Possible

This distribution of current values is consistent with "energy recirculate mode" (the left) and "normal mode" (the right.) Energy recirculate mode has also been called 'heretical mode' although the definition is a little murky:

We still need a mapping of current to torque, NM, but the shape of the current graphs are consistent with the expected profile. However, I'm a little concerned about the maximum and minimum current values.

MG1 Current W - Promising!

This data looks much more consistent with the expected MG1 torque:

We have both the positive and negative signs but more importantly, these values are consistent with expected ICE torque. Of the three values, this is the one that most makes sense to map to true MG1 torque, and engine ICE torque.

Follow-Up

I need to do a 'hill climb' test recording ICE rpm, MAF flow (seems to have reasonable values) and 'MG1 Current W'. Then I can compare the values to the Graham Miniscanner values. FYI, I have a dual-headed, DCL so I may (if there is no conflict on controller IDs) to record both. If so, a conversion function should be possible.

Bob Wilson

 

Here is another plot showing the absolute value of the two MG1 currents vs ICE rpm:

Of the two, MG1 watt seems to be more consistent with the expected torque from the ICE. The MG1 current has too many outliers.

What I need to do is record traction battery, MG1 and MG2 current and watt. In a steady state, the sum should be the vehicle overhead current plus some energy loss percentage. Most of the time, MG1 should be opposite sign and identical magnitude to MG2 current in steady state operation.

BTW, the raw data record looks terribly inconsistent:

	Column 1	Column 2	Column 3	Column 4	Column 5	Column 6
0	Time	Engine (RPM)	MG1 rpm	MG1 Torque	MG1 (A)	MG1 W (A)
1	09:49:01.711	0	0	0	0	0
2	09:49:03.227	0	0	0	0	0
3	09:49:04.742	0	0	0	0	0
4	09:49:06.242	0	0	0	-0.49	0
5	09:49:08.180	0	0	0	0	0
6	09:49:09.680	0	-128	0	0	0
7	09:49:11.180	0	-128	-116	0	0
8	09:49:12.695	0	-256	-412	0	0
9	09:49:14.227	0	1152	500.01	-0.49	50.75
10	09:49:15.742	1696	2304	-372	5.37	-58.56
11	09:49:17.242	2016	2560	-188	-68.32	70.76
12	09:49:18.914	2336	2816	-496	-16.1	-75.64
13	09:49:20.445	2496	2688	-412	71.74	74.66
14	09:49:21.945	2432	2304	400	67.34	-58.56
15	09:49:23.445	2208	1664	-352	66.37	-39.04
16	09:49:24.945	2080	1536	296	50.75	-45.87
17	09:49:26.445	2144	1280	-260	62.95	-33.67
18	09:49:27.945	2144	1280	-252	60.02	63.44
19	09:49:29.445	2144	1152	-404	0	23.91
20	09:49:30.945	992	-2048	404	-64.42	6.34
21	09:49:32.445	1184	256	-96	0.98	51.73

Start of recorded data. Only the ICE rpm time-stamp is retained because the individual data time-stamps obscure data analysis.

The MG1 rpm units are mislabeled "(Nm)" in the label " Generator - MG1 Revolution (Nm) [-16384 - 16383]". The values look right but the units are wrong. I've edited the column text to make sense, as much as is possible.

We know MG1 spins up the ICE before it is started and we can see this 9:49:09 to 9:49:12. But where is the ICE rpm during the startup?

MG1 Torque has already been shown to be useless. It is as if a random sampling of a counter were being captured and reported. The distribution of values has more in common with a random number generator than usable engineering data. Note that MG1 Torque appears to span a byte, a normal 'atomic' value. In contrast, the two MG1 current and watt values are single bytes.

MG1 (A) and MG1 W (A) seem to have some association with expected values. The "0" value at 9:49:29 is something that seems to happen with Auto Enginuity. Data distribution suggests MG1 (A) has more outliers than MG1 W (A). Also, the sign of MG1 W (A) seems to flip somewhat randomly. The complete data distribution suggests sign flipping is a problem with both MG1 (A) and MG1 W (A).

One follow up analysis will be to compare ICE, MG1, and MG2 rpm against the well know ratios Graham Davies reported:

MG1 = (3.6 * ICE) - (2.6 * MG2)
​

This will clearly show how trustable the Auto Enginuity data is for engineering analysis.

Bob Wilson

 

I've had the good luck to start with a Graham Miniscanner and that unit's data is solid. But this unit continues to disappoint. Not every data element but enough that I find it hard to trust it. For example, I found today that MG2 rpm increases by 128, which means the low order, six bits are not being read or reported.

If you can, record MG2 rpm and see what the data looks like to you.

Bob Wilson

 

I think we're seeing the same thing. For engineering data, AE is weak. We can probably map out the useful data and I'll keep AE informed of what we see. It doesn't mean they'll fix it but they can't fix what they don't know about.

Bob Wilson