Drivetrain measurements

[bwilson4web]

I have another, substantially larger file, that includes climbing a 525 ft (160m) hill at 55 mph (true.) It includes the drive to the hill:

Some quick observations:

• 30 mph indicated - there is almost no power transfered via the electrical path
• 48 mph indicated (50 mph GPS true) - the car appears to run on flat at fairly steady-state, long durations.
• 52 mph indicated (55 mph GPS true) climb 8% grade - the car switches from "energy recirculate" to normal power flow. My eyes suggest the ratio of electrical path to mechanical path does not change that much.

I am happy to share the data from this longer run but it exceeds the quotas allowed in Ecomodder posting. The normalized data is a substantially huge, 252K, file. Post or send me a request for the data and we'll see what can be done. Alternatively, they are at the YahooGroup, 'Prius_Technical_Stuff" in a folder 'Graham Miniscanner Tools.'

I did take the hill climb data and generate this efficiency chart:

Again, higher efficiencies, 85-99%, are found at higher power, normal power split operation. Lower efficiencies, 75-95%, are found at lower power, "energy recirculate" mode. But this is just the electrical power path that takes passes only a fraction of the total ICE power.

Here is a set of data at 50 mph GPS on relatively flat land:


I am not very interested in further analysis of this particular large data file versus working on an improved, data collection system (my GPS receiver arrived and works great.) By integrating GPS and Graham Miniscanner data, we can get a total energy profile with good ephemeris. This can lead to more precise measurements that are likely to improve NHW11 operation. It also brings us closer to tank-to-pavement efficiency, what the owner/driver has to pay 'real money' to achieve.

I have some data from 2007/10/08 from Nashville TN to Huntsville AL:

Notice what happens to MG1/MG2 temperature whenever the speed exceeds 65 mph. Heat, higher temperature, is the signature of lost efficiency. Based upon this data, the maximum, NHW11 Prius speed should be 65 mph.

We can see that the ratio of power that is split between the two paths is a function of load and speed. Furthermore, the normal flow is running at ~80% on the fraction of power that takes the MG1-to-MG2 path, regardless of speed and power. The "energy recirculate" mode is in the 60% range but it also lets the engine turn much slower and in an improved thermodynamic range. The ratio of electrical to ICE power is also a function of vehicle speed and power. Although useful for engineering insights, the driver needs total vehicle efficiency, not just the transaxle efficiency at any single operating point.

Bob Wilson

[bwilson4web]

On Saturday, July 30, we had scattered showers that cooled things down and suppressed traffic to support more extensive benchmarks that include:

• GPS - provides location, velocity, and altitude to calculate the watt change in potential and kinetic energy.
• Graham miniscanner data - in addition to MG1/MG2 rpm and torque, I've replaced traction battery with mass air flow for fuel consumption and use a 2nd degree polynomial for the traction battery voltage as a function of current to estimate traction battery watts.

The following graph shows the car stopped for on coming traffic and then a turn from Whitesburg Drive on to Martin road on to the gate at Redstone:

The significant events:

• Standing start - the traction battery draw to start the ICE and initial ICE acceleration. This uses "normal" transaxle power flow where MG1 works as a generator and MG2 works as just a motor. The thin black line is the watts needed to change the kinetic and potential energy.
• Electric constant speed - there is a period of nearly 30 seconds where the car velocity, 35 mph (GPS) was sustained.
• Traffic slow down - merging traffic required slowing the car. Notice the kinetic and potential energy Watts closely agrees with the MG2 power generation. At slow speeds, there should be close agreement as there are little aerodynamic drag effects.
• Acceleration to 50 mph - the kinetic/potential energy Watts were nearly constant but as the speed built up, aerodynamic effects increased the difference between the power and kinetic/potential energy effects. Notice there was battery charging going on in no small part due to the early 30 seconds of traction battery operation.
• Sustained 50 mph with brief adjustment - shows the rate of fuel consumption to sustain 50 mph was about half of that used to reach 50 mph over 30-40 second acceleration.
• Deceleration - shows the engine braking that ends as the car falls below the 42 mph threshold that requires engine spinning. The difference between MG2 and kinetic/potential energy was larger at higher speeds because of the aerodynamic drag. As the velocity reduced, MG2 generated power more closely approached the watts changing kinetic and potential energy.
• Final braking - from 10 mph and slower, the mechanical brakes are absorbing the remaining kinetic energy change and these are not measured.

To calculate efficiency from the power split device when in "normal" mode, I've added 443 W vehicle electrical overhead (not directly measured,) traction battery charging power (estimated from 3d degree polynomial) to the MG2 output divided by MG1 electrical input. Similar relationships are used for "energy recirculate mode. After adjusting the X-axis time scales to be roughly aligned, here is the driving profile and below it the efficiency chart:


The big takeaway is:

• "normal" - has higher efficiency at higher power levels
• "energy recirculate" - has a little lower efficiency at lower power levels as it forces the engine into more fuel efficiency rpm ranges.

Analysis of this data has been a technical challenge as the GPS data has to be aligned with the Graham miniscanner data. To make sensible charts, we have to approximate from values sampled at different times and this often leads to 'outliers.' Right now, I'm using excel since it is easy to test these algorithms. But I have more data collected and to reduce it I'll have to write a Perl routine to normalize the data and remove outliers. This is not a big deal but it does take time.

Bob Wilson