Data analysis is a critical component to the engineering design cycle. Without data, one cannot make informed design decisions. Thanks to the infrastructure set up on the Baja electronics team, we were able to collect around 30 million data points over the course of a 4 hour race in Pennsylvania.
Background
The data analysis will focus on M23’s endurance race run at the Pennsylvania College of Technology in Williamsport, Pennsylvania. The data spans over the course of 5.2 hours, with the race itself being 4 hours.
The weather was sunny, with mild humidity.
During the course of the race, we pitted 4-5 times, with the last pitting being after multiple failures of the drivetrain, limiting the car to 1WD.
Breakdown of Variables
- cycle_start_us
The microsecond counter of the Teensy microcontroller. This variable is used as the main time base for all data analysis.
NOTE
The microsecond counter is stored in a long long, and will overflow at certain points in the log. This just means that the counter will wrap around to 0.
- engine_rpm
The RPM (rotations per minute) of the engine, before being filtered. This value should only range from around 1800 - 3800 (idle to max legal engine RPM).
- filtered_engine_rpm
The RPM of the engine, but passed through an IIR low pass filter. This value is used in the controls cycle.
- secondary_rpm
The RPM of the secondary sheave, the output of the ECVT.
- filtered_secondary_rpm
The RPM of the secondary sheave, filtered through a low pass IIR filter. This value is also used on the car, but exactly how it’s used shall not be disclosed.
- target_rpm
The current engine RPM value that the transmission’s PID controller is trying to target. In controls, this is know as a setpoint or reference!
- velocity_command
The speed that the Odrive is commanded to move at. Measured in rotations per second.
- velocity_estimate
The Odrive’s estimate of it’s current speed based on the motor’s encoder. Velocity estimate and velocity command values can diverge based on the tuning scheme of the Odrive and based on our velocity ramping coefficient.
- position_estimate
The position estimate is the current position of the Odrive, as measured by the encoder. Measurement unit is unknown…
- bus_voltage
The voltage the Odrive measures at the battery input terminals. For M23, nominal battery voltage was 48V.
- bus_current
The current flow the Odrive measures at the battery input terminals. Positive currents indicate battery consumption, while negative current indicate regenerative braking.
- iq_measured
The current flow in the phases of the BLDC motor. This value directly corresponds to the torque of the motor through a constant defined by the motor itself. This value is also the basis upon which the Odrive achieves torque control.
- iq_setpoint
The target value of the Odrive’s internal motor PID controller.
- inbound_limit_switch
A 0/1 value corresponding to whether or not the inbound limit switch has been pressed. The variable is 1 if the switch has been pressed. The inbound switch is the switch closest to the engine and indicates to the system that the maximum transmission ratio has been reached.
- outbound_limit_switch
A 0/1 value corresponding to whether or not the outbound limit switch has been pressed. The variable is 1 if the switch has been pressed. The outbound switch sits the farthest away from the engine, and tell the engine that the clutch position has been achieved. The clutch position means that the engine and transmission are decoupled.
- engage_limit_switch
A 0/1 value corresponding to whether or not the engage limit switch has been pressed. 1 if switch is pressed. The engage switch sits in between outbound and inbound and sits right at the edge of belt engagement.
- throttle
A value ranging from 0 to 1, indicating the percentage of throttle travel achieved. 1 indicates the throttle has been pressed all the way through its travel and 0 corresponds to the throttle not being pressed at all.
- brake
A value ranging from 0 to 1, indicating the percentage of brake travel achieved. 1 indicates the brakes are pressed all the way and 0 corresponds to the brakes not being pressed at all.
- d_throttle
Derivative of the throttle measurement
Data Cleaning
- cycle_start_us must be processed to provide continuation when the variable overflows and wraps around. A short Python script is provided to accomplish that. Please adjust for your use case:
adjustment = 0
max_seen = 0
new_values = []
for i in range(len(df)):
if i > 0 and df['cycle_start_us'][i] < df['cycle_start_us'][i-1]:
# Loop detected, add the last max_seen value to adjustment
adjustment += max_seen
max_seen = 0 # Reset max_seen for the new cycle
max_seen = max(max_seen, df['cycle_start_us'][i])
new_values.append(df['cycle_start_us'][i] + adjustment)
df['cycle_start_us'] = new_valuesAnalysis
Electrical Power Analysis
In this section, we explore the current data (bus_current) as well as the voltage data (bus_voltage) to gain insight into the power usage of the actuator. This section will provide statistics about power consumption (helping us justify the power of the motor), voltage sag (shows the health and capacity of the battery), and current draw (provides information about torque and allows us to spec the right size battery).
Bus Current Consumption Over Time
Current Consumption Distribution
- What distribution does the current consumption roughly follow, and what current do we need to support to capture 95% of current consumption cases?
Bus Voltage Consumption Over Time
Total Current Consumption Over Endurance
- integral
Total Voltage Sag Over Endurance
- (difference
Electrical Power Consumption Over Time
Total Electrical Power Consumption
- Watt Hours
Controls Performance Analysis
In this section, we want to gain insight into the software side of the car’s performance. This section will provide insight into the efficiency of the code, potential bottlenecks for controls, and hopefully give us direction for future controls improvements.
Velocity Command vs Time
Engine RPM vs Time
Filtered Engine RPM vs Engine RPM
- Correlation, data loss
Target RPM vs Engine RPM and Difference
- Perspective on tracking performance
Cycle Tick Consistency
- Consistency of controls system evaluation.
Actuator Mechanical Analysis
In this section, we explore the main mechanical characteristics of the actuator.
Velocity Command Over Time
Velocity Estimate Over Time
Velocity Command Vs. Velocity Estimate.
- Difference between the two.
- Provides insight into how well the inner motor control loop is tracking our commands. (filtered and unfiltered)
Position Estimate Over Time
Actuator Utilization
- (histogram of actuator position when outbound limit switch is not activated)
Limit Switch Repeatability
- Histogram of actuator position when inbound limit switch activated. Provides insight into accuracy of actuator position estimate and our motor’s stopping ability.
- Histogram of actuator position when outbound limit switch activated
Actuator Torque
- Torque applied over time. Can use electrical power from previous section as “total power”. Be careful of units, velocity estimate is in rotations/sec.
Vehicle Dynamic Analysis
In this section we explore data analysis that’s not necessarily linked to one part of the car, but to the performance of the car as a whole.
Wheel Speed Analysis
Car Speed Analysis
Engine RPM Analysis
- Engine RPM over time
- Histogram of Engine RPM
Miscellaneous Analysis
Throttle Position
- Vs time
- Histogram
Brake Position
- Brake vs Time
- Histogram of Brake
Derivative of throttle
- d_throttle vs Time
- Histogram of d_throttle