In this last part of the introduction to data analysis, the sensor checks and the system memory will be discussed. There are important procedures to ensure that the data acquisition system are working well. This article proposes a brief review about these procedures.

Sensor checks

There are some checks that should be made with respect to sensors. These are the installation procedures, the amount of spare parts and the logging conditions. In terms of installation, it should be verified if the sensor housing is too exposed to vibrations, interference, rubbing, angles and temperatures. Usually sensors are designed to deal with some vibrations, but their fixtures must be proper selected. The sensor wiring should be well fixed to avoid rubbing and interference. In terms of temperatures, sensors are very sensitive to it. Normally, suppliers have a data sheet with the proper correction according to the temperature. The track activities are very intense, thus it is important to have spare sensors available and already calibrated. For instance, the velocity sensors are the most vulnerable.

Sensor stop setup

The steering sensor, when mounted on the steering box, must not be used as the stop for the system. The steering lock is usually placed on the rack stop. The reason is that the sensor can not be used as a stop, or rather it will break very easily. In the same manner, the track rod must not use the rim as a stop for obvious reasons. The throttle signal check should be done at zero and full throttle. However, the main detail is the position of the pedal stop. The first point is, the sensor can not be used as the physical stop, because it is quite weak, thus it will fade very fast. The second point is the position of the physical stop. If it is positioned before the final stroke of the pedal, the sensor will read less than 100%, thus the engine will not deliver full power. On the other hand, if the pedal stop is positioned after the end of the pedal stroke (101%), the sensor will be damaged. Hence, the solution is to displace the pedal stop at about 98% of the stroke and correct this on the math channel. In this case, the correction adds the difference from 100% when the gas pedal is at 98% for some seconds.

Speed sensor compensation

The speed sensor is another one that requires some compensations. In this case, it is to account the longitudinal slip, as seen on Figure 1. The main point is that the tire circumference changes according to the applied loads. This affects the rolling resistance. Since essentially the vehicle speed is conversion of the wheel rotation speed, this requires thee wheel radius. The loaded radius is a function of the applied load, camber angle and inflation pressure. To compensate for the variable radius it should be defined the loaded radius as a function of the applied load, then a testing factor can be obtained.

Vertical loads influence

The effect of the loads comes from Gx and Gy. Hence, it is possible to estimate the four wheel load by those two accelerations. A math channel can be built in order to account the wheel loads. In addition, with the data from the wind tunnel, it is possible to estimate the aerodynamic effects as a function of speed. Summarizing, from the longitudinal and the lateral G it can be accounted the load transfer and from the aerodynamic effects. Hence, it is possible to predict the vertical load on each tire. In the garage, it is possible to test different vertical loads which enables measuring the loaded radius. This test is repeated for 2 or 3 different inflation pressures to measure the centrifuge. Hence, a map with pressure, load and radius can be built. Finally, the loaded radius is applied by the equations below:

• S = (u-Rw)/u ; If u > Rw
• S = (Rw-u)/Rw ; If u < Rw

These describe the longitudinal slip, which can be applied in order to compensate the signal of the velocity when under accelerating and braking maneuvers.

Preliminary sensor checks

The preliminary checks start with the engine shut down. All the sensors are verified prior to the first start and these should be ok at this conditions. For instance, at the shop it is possible to check the accelerometer operation. In this case, the sensor is rotated in order to verify if it measures 1 G in all directions. After this, the engine is turned on, but holds at idle. This condition is useful to check the sensor operation under a some noise and significant vibrations. Then the engine is kept at low revs and the sensors are checked again. The same check is performed at high engine revs. The objective is always the same, to check the sensor operation under noise and vibration. Another important verification is radio interference on the sensor signal. Since the radio is very common in the motorsport field, it is important to check if there is any level of interference. In this case, the main point is checking if the sensor insulation is working properly. Regarding the sensor operation under wet conditions, this is an important check for connector insulation. This is the reason why it is important to check the sensor installation and its operation under engine working. These are the tests of the sensor with the engine running at idle, low and high revs.

System memory

The system memory is an important characteristic that affects the number of channels, the logging frequency and the analogic to digital converter (A/D), which is the number of the bits. These aspects are relative to the system capacity. For instance, considering a basic toolkit with a maximum logging capacity of 400 data/s, it is possible to define a suitable frequency, for each sensor. First, the time should be logged. If one sensor is logged at 100 Hz, time should be logged at 100 Hz. Distance is an important parameter, but it is calculated instead of logged. It is ran an outing, then the data generated is submitted to a Fourier analysis in order to verify the frequency content in each signal.

For instance, the throttle, the steering, the speed and the longitudinal acceleration could have about 3, 2.5, 2 and 1.5 Hz, respectively. In this case, one can notice that the frequencies are quite low. Additionally, between these the throttle and the steer are the widest ones. The reason is, because these are driver inputs. Since the car responds to the lower frequencies, it is sort of feedback from the outputs. The car response has some delay, thus this has even lower frequencies with respect to the inputs. Based on this information, a suggested logging frequency for the throttle could be 10 Hz. The reason is that to spot the variation of a channel and lately perform its derivative, the logging frequency should be the double of the one found on the Fourier analysis.

Considering now the decision about the longitudinal G logging. This parameter exhibits many some spikes. These is a consequence of raw data which are normally filtered. Generally, after filtering there is not significant difference between 50 Hz and 5 Hz data. The longitudinal G plot also allow to verify the gear shifts. Hence, it is required a high frequency to spot those shifts. However, this also highlights local bumps and patches. Another interesting point about raw and filtered data, is that this last one is able to indicate the braking maneuver. This means that the brake maneuver is a low frequency one. On the other hand, the filtered data usually hides the gear shifts. The frequency content of the longitudinal acceleration varies according to the car. The reason is, because each car has different features and these affects the frequency of them.

The frequency content is obtained applying the Fourier transform as seen on Figure 3. In this case it was applied on the lateral acceleration data (G_y). This technique allows to go into the frequency domain, then select the Fourier low pass filter and re-build the signal in the time domain. This is done by the reverse Fourier transform. For instance, a signal logged at 200 Hz. After applying the low pass filter it is possible to reconstruct G_y signal. One can notice that the higher the band the higher the oscillations on the signal. The lower the band, the lower the oscillations. The trade-off is the understanding of the meaning of the oscillations of each channel. Then, the low pass filter should be proper selected. Hence, the Fourier analysis indicates the frequency for a proper sampling rate. The procedure is having an outing for verification (installation outing), the Fourier analysis is ran and then the sampling is selected. This should consider the relevance of the signal for the vehicle behavior analysis and the allowance of the signal derivation. For instance, G_x is usually logged 50 Hz based on the Fourier analysis.

Yaw angle versus heading angle

The accelerometer is a little box with some wires. Hence, the lateral G sensor is fixed to the car, thus it is not normal to the trajectory. This means that even a simple measurement of the lateral G is not precise by definition due to the yaw angle. In this point there are two important definitions, the yaw and the heading angle. The difference between these is that, the first has the car as reference, while the heading angle is the angle between the fixed reference on the car and the fixed reference on the system (ground). The heading angle is important if the wind is too strong, which is not an usual condition in racing. Hence, yaw is more important for vehicle dynamics. The yaw angle is for the car what the slip angle is for the tires with respect to performance.

Banking curves

Another important situation with respect to the lateral G sensor is the banking curve found in oval tracks. With banking and slope, the lateral G sensor measures the banking angle, because it is fixed on the car. Hence, if there is a banking or a slope, it will measure only these. In the case of a 90° banking curve, the lateral G sensor measures the weight of the car since it is fully vertical. Hence, it is necessary to adjust the reading of the sensor with the banking angle. Actually, it is possible to correct the sensor signal for banking curve, installation and roll angles. It is also possible to correct the signal banking and roll.

Roll angle measure

To measure the roll angle it is necessary another sensor. However, there are some procedures to estimate the roll angle with a good precision. Considering a basic data acquisition toolkit and that the car is the garage. It is necessary to measure the roll stiffness of the car. In the garage, the car should be pulled at sideways on the center of gravity point. The amount of applied force is divided by the vehicle mass, thus the lateral G that would be obtained at the track is measured. Similarly, the roll angle is measured.

Banking angle measure

The banking angle is usually given by the race track owner. It is usually available the banking elevation as a function of the distance. For instance, it is possible to program a channel that, from Gy signal, it identifies a corner. Then it is possible to create the banking by the lap distance. This last one can be obtained with the car speed and the elapsed time. Then it is possible to compensate banking according to the distance.

Uphill and downhill

These considerations for banking are similarly applied for slopes. In this case there are two possible situations, uphill and downhill. Instead of roll, the car exhibits pitch. The procedures to correct the sensor measurement is the same done for banking angles. In an uphill, the longitudinal G sensor is measuring more acceleration, while in the downhill it is measuring less acceleration. This occurs, because the sensor is sensitive to weight.

Sensor quality

As mentioned in the previous articles, there six basic sensors in usual data acquisition systems. These are the driver input and velocity sensor are rather clean. This means that those signals used to not have signal interference. The level of disturbance is low. In terms of the car outputs, these are RPM, G_x and G_y and these are noisy signals. The engine rotation is usually noisy due to the engine vibrations. In the case of the longitudinal and lateral G, the noise comes from the fact that this sensor is quite sensitive to small changes. In addition, when it is logged the second third decimal place of the accelerations, this means a thousands of G. In the case of speed, 280, 280.5 and 281 km/h is rather comparable, but not so noisy as G accelerations.

References

• This article was written based on the lecture notes written during the Applied Vehicle Dynamics lectures attended at Dallara Accademy;
• Segers. J. Analisys Tequiniques for Racecar Data Acquisition, 1° Edição. Warrendale, PA. SAE International. 2008.