Fuel Economy, Hypermiling, EcoModding News and Forum - EcoModder.com - View Single Post

bwilson4web · 06-07-2019, 07:15 PM

Hi,

Standard Range Plus Model 3 (version 19.16.2)
SOC ~75%, 15 mi (24 km) warm-up
3711 (empty weight) + 250 (driver) + 15 (EVSE, patch kit) ~= 3976 lbs (1,807 kg)
77F (25C)
asphalt road surface
wind 0 mph, 9:09 PM (21:09) CST

The acceleration data will be used to calculate the velocity. Then I'll add the EPA roll-down drag coefficients to calculate the total force on the car and eventually the vehicle HP (kW). This will be for both chill and standard modes.
. . .
Those 'down spikes' are the end of the maximum acceleration runs. I targeted 80 mph as my ending speed because the car sounds an excessive speed alarm at 90 mph. Yes, I could have disabled the speed alarm but earlier GPS measured runs showed 80 mph was adequate to our goals.

I'm using a Gulf Coast Data Concepts, Human Activity Monitor:

Code:
;Title, Gulf Coast Data Concepts, LLC, X16-MPU-ham, ADXL345, MPU-9250
;Version, 1191, Build date, Nov 15 2016, SN:CCDC3016B4874F1
;Start_time, 2019-06-05, 21:09:12.423
;Temperature, -999.00, deg C, Vbat, 4174, mv
;MPU SR, 200,Hz, Accel sens, 4096,counts/g, Gyro sens, 16,counts/dps, Mag SR, 10,Hz, Mag sens, 1666,counts/mT
;Deadband, 50, counts
;DeadbandTimeout, 0,sec
;Time, Ax, Ay, Az, Gx, Gy, Gz, Mx, My, Mz
0.004486,-170,1156,-4192,5,-54,-1
0.018432,96,1210,-3916,-21,-20,-9,-311,-383,-506
0.038421,-396,1242,-4240,-20,-34,0
0.058410,-370,1302,-4166,-25,-13,-8
0.078399,680,1450,-4722,-17,-10,-7,-304,-371,-506
0.098419,-384,1138,-4572,-18,-18,0
0.118408,-584,1142,-4106,-8,1,-7
0.138397,124,978,-3978,7,15,-8
0.158386,250,1096,-4122,16,1,-5
0.178406,-640,1222,-4154,17,-12,3,-296,-387,-514
Time - seconds from start of data file. Use the header to get the 'real time' clock and date.
Ax - side to side acceleration
Ay - front to rear acceleration
Az - top to bottom, gravity, which we use to scale 1 G
Gx - rotation about the door axis
Gy - rotation about the front to rear axis
Gz - rotation around a vertical axis
Mx - magnetic field along door axis
My - magnetic field along front to rear axis
Mz - magnetic field along the vertical axis
MEMS accelerometer data are noisy so I used a 7 element Gaussian filter (0.063, 0.250, 0.375, 0.250, 0.063) to do a weighted average. Unlike a linear average, this preserves the local peaks while significantly reducing the noise.

Each data file has 16,000 data samples covering about 5 minutes which puts a significant load on the OpenSource spreadsheet. Somewhat arbitrary, I used 500 counts, 500/4096 ~= 12.2% G, to trim the non-acceleration elements reducing the samples to 2,185 which was easily handled by the spreadsheet.

Perhaps more details than you were expecting, I was looking for an opportunity to share where the data came from.
. . .
So this is what the velocity looks like:

Remember, my car has a heavy driver, the EVSE, and a tire patch kit.
. . .
There are two parts:
Inertal HP - this is how much is used to move the mass
Drag HP - this is how much energy is lost due to rolling and aerodynamic drag
Since we have the mass, we can calculate the HP that accelerated the car and its contents to a given speed. Then add the drag HP to get the actual HP that was applied at the wheels.
. . .
Both Chill and Standard mode acceleration metrics:

HP - the inertial power needed to accelerate the car and contents
drag - the power needed to handle rolling, transmission, and aerodynamic drag
Total HP - the total power needed at the drive wheels

. . .
I'm using a recording, MEMS, accelerometer to capture the data and CAT (truck stop) weight for the mass. Then I used physics to calculate the speed and HP. I also used the EPA roll-down coefficients to calculate the drag power.

You are welcome to my data in the zip file. It isn't a well documented spreadsheet as I use it for my curiosity. If you have questions, I'll be happy to explain what is going on.

I could clean up the spreadsheet if there is interest in a more 'formal' presentation:

improved noise filter - I used a 5-element Gaussian filter but could increase it to 7-9 to further reduce the noise. I won't use a linear average because it suppresses the peaks even though it is more commonly used. Still, I could code the spreadsheet to let each method be used.
standard units - I mix both SAE and metric metrics and calculations which ever is easiest. I could change all of the metrics and calculations to metric and then at the end convert the results to SAE.
table parameters - in some calculations it was easier to hardcode the functions instead of referencing a well documented table. By using tables, different models and cars can be instrumented and values used to get useful results. Just I'm not in that business ... yet.

Bob Wilson

06-07-2019, 07:15 PM	#1 (permalink)
bwilson4web Engineering first Join Date: Mar 2009 Location: Huntsville, AL Posts: 843 17 i3-REx - '14 BMW i3-REx Last 3: 45.67 mpg (US) Blue Bob's - '19 Tesla Std Rng Plus Thanks: 94 Thanked 248 Times in 157 Posts	Some Standard Range Plus Model 3 metrics Hi, Standard Range Plus Model 3 (version 19.16.2) SOC ~75%, 15 mi (24 km) warm-up 3711 (empty weight) + 250 (driver) + 15 (EVSE, patch kit) ~= 3976 lbs (1,807 kg) 77F (25C) asphalt road surface wind 0 mph, 9:09 PM (21:09) CST The acceleration data will be used to calculate the velocity. Then I'll add the EPA roll-down drag coefficients to calculate the total force on the car and eventually the vehicle HP (kW). This will be for both chill and standard modes. . . . Those 'down spikes' are the end of the maximum acceleration runs. I targeted 80 mph as my ending speed because the car sounds an excessive speed alarm at 90 mph. Yes, I could have disabled the speed alarm but earlier GPS measured runs showed 80 mph was adequate to our goals. I'm using a Gulf Coast Data Concepts, Human Activity Monitor: Code: ;Title, Gulf Coast Data Concepts, LLC, X16-MPU-ham, ADXL345, MPU-9250 ;Version, 1191, Build date, Nov 15 2016, SN:CCDC3016B4874F1 ;Start_time, 2019-06-05, 21:09:12.423 ;Temperature, -999.00, deg C, Vbat, 4174, mv ;MPU SR, 200,Hz, Accel sens, 4096,counts/g, Gyro sens, 16,counts/dps, Mag SR, 10,Hz, Mag sens, 1666,counts/mT ;Deadband, 50, counts ;DeadbandTimeout, 0,sec ;Time, Ax, Ay, Az, Gx, Gy, Gz, Mx, My, Mz 0.004486,-170,1156,-4192,5,-54,-1 0.018432,96,1210,-3916,-21,-20,-9,-311,-383,-506 0.038421,-396,1242,-4240,-20,-34,0 0.058410,-370,1302,-4166,-25,-13,-8 0.078399,680,1450,-4722,-17,-10,-7,-304,-371,-506 0.098419,-384,1138,-4572,-18,-18,0 0.118408,-584,1142,-4106,-8,1,-7 0.138397,124,978,-3978,7,15,-8 0.158386,250,1096,-4122,16,1,-5 0.178406,-640,1222,-4154,17,-12,3,-296,-387,-514 Time - seconds from start of data file. Use the header to get the 'real time' clock and date. Ax - side to side acceleration Ay - front to rear acceleration Az - top to bottom, gravity, which we use to scale 1 G Gx - rotation about the door axis Gy - rotation about the front to rear axis Gz - rotation around a vertical axis Mx - magnetic field along door axis My - magnetic field along front to rear axis Mz - magnetic field along the vertical axis MEMS accelerometer data are noisy so I used a 7 element Gaussian filter (0.063, 0.250, 0.375, 0.250, 0.063) to do a weighted average. Unlike a linear average, this preserves the local peaks while significantly reducing the noise. Each data file has 16,000 data samples covering about 5 minutes which puts a significant load on the OpenSource spreadsheet. Somewhat arbitrary, I used 500 counts, 500/4096 ~= 12.2% G, to trim the non-acceleration elements reducing the samples to 2,185 which was easily handled by the spreadsheet. Perhaps more details than you were expecting, I was looking for an opportunity to share where the data came from. . . . So this is what the velocity looks like: Remember, my car has a heavy driver, the EVSE, and a tire patch kit. . . . There are two parts: Inertal HP - this is how much is used to move the mass Drag HP - this is how much energy is lost due to rolling and aerodynamic drag Since we have the mass, we can calculate the HP that accelerated the car and its contents to a given speed. Then add the drag HP to get the actual HP that was applied at the wheels. . . . Both Chill and Standard mode acceleration metrics: HP - the inertial power needed to accelerate the car and contents drag - the power needed to handle rolling, transmission, and aerodynamic drag Total HP - the total power needed at the drive wheels . . . I'm using a recording, MEMS, accelerometer to capture the data and CAT (truck stop) weight for the mass. Then I used physics to calculate the speed and HP. I also used the EPA roll-down coefficients to calculate the drag power. You are welcome to my data in the zip file. It isn't a well documented spreadsheet as I use it for my curiosity. If you have questions, I'll be happy to explain what is going on. I could clean up the spreadsheet if there is interest in a more 'formal' presentation: improved noise filter - I used a 5-element Gaussian filter but could increase it to 7-9 to further reduce the noise. I won't use a linear average because it suppresses the peaks even though it is more commonly used. Still, I could code the spreadsheet to let each method be used. standard units - I mix both SAE and metric metrics and calculations which ever is easiest. I could change all of the metrics and calculations to metric and then at the end convert the results to SAE. table parameters - in some calculations it was easier to hardcode the functions instead of referencing a well documented table. By using tables, different models and cars can be instrumented and values used to get useful results. Just I'm not in that business ... yet. Bob Wilson __________________ 2019 Tesla Model 3 Std. Range Plus - 215 mi EV 2017 BMW i3-REx - 106 mi EV, 88 mi mid-grade Retired engineer, Huntsville, AL Last edited by bwilson4web; 06-07-2019 at 11:11 PM..