Road Structure Estimation (RSE)

Road profile is one of the most important factors affecting vehicle performance and characteristics, e.g. ride comfort, handling, fuel consumption, and tire wear. Information about the road profile is also important for road management to plan and follow up road conditions. This is especially valuable during winter season when the road conditions can change rapidly due to snowfall. The road profile is measured directly using a profilograph or a profilometer. To make measurements at larger distances, high speed profilometers are needed. However, high speed profilometers are expensive and must be handled by trained operators. The monitoring of the road network is therefore limited, especially during winter season when laser methods are impractical. Because of these limitations, indirect methods using low cost measurement systems that can be installed on a fleet of vehicles are of main interest. An attractive type of sensor is accelerometers that are inexpensive, rugged, and easy to mount. In addition, there are accelerometers already in vehicles, such as in active suspension systems. An alternative to use accelerometers is to use microphones. As often experienced, one can hear how the sound in the car compartment changes in level and frequency distribution when the car passes from for example a section with smooth asphalt to a section with more coarse asphalt.

ISO 13473-1 divides the road texture in 4 categories depending on the texture wavelength (distance between periodically repeated parts of the road profile):

• Microtexture   – texture wavelengths smaller than 0.5 mm
• Macrotexture  – texture wavelengths between 0.5 and 50 mm
• Megatexture   – texture wavelengths between 50 mm and 0.5 m
• Unevenness  – texture wavelengths between 0.5 and 50 m

One limitation of indirect methods based on accelerometer measurements is the limitation in frequency range, especially with regard to low frequencies. For example a wavelength of 100 m driving at 50 km/h gives an excitation frequency of 0.14 Hz and a wavelength of 10 cm at 110 km/h gives a frequency of 305 Hz. For high frequencies, e.g. the macrotexture of worn asphalt, the amount of vibrations that pass through to the hub and the suspension is limited because the tyre acts as a lowpass filter. For such short texture wavelengths microphones is an alternative to using accelerometers. As often experienced, one can hear how the sound in the car compartment changes in level and frequency distribution when the car passes from a section with smooth asphalt to a section with more coarse asphalt.

In a first study, it was shown that both interior and exterior measurements of tyre noise and axle acceleration can be used to distinguish fresh laid asphalt from worn. In a second study, axle acceleration measurements and an inverse quarter-car model was used to reconstruct a well-defined test track at 30 and 50 km/h. In a third study measurements was done together with Vectura using a high speed profilometer.

Study 1 – Classification of coarse and smooth asphalt

The audible changes in the sound inside the car compartment for different pavements are caused by a change in the tyre/road interaction. Intuitively, tyre noise increases with rougher road texture. However, low and high frequencies are affected differently by a change in the road texture amplitude. Low-frequency noise increases with increasing road texture amplitudes for wavelengths between 10-500 mm. On the contrary, the high frequencies noise content will decrease for increasing road texture amplitudes for wavelengths between 0.5-10mm.

The tyre/road interaction causes noise but also vibrations that are transferred through the wheel and suspension to the car body (structure borne noise). As well as the noise is influenced by the road texture the same is true for the vibrations. The macrotexture of the road and its amplitude should be possible to predict with either the emitted sound or the vibrations in for example the wheel hub. In this study a car was equipped with microphones and accelerometers to investigate how different road textures influences sound and vibrations. Texture wavelengths from 0.5 cm up to a few centimetres were analysed. For collecting data a car was equipped with accelerometers and microphones both on the outside and the inside of the car compartment, see figure below. The external microphone was mounted just in front of the right rear wheel and the exterior accelerometer was placed on the lower mounting point of the wheel suspension. Both the interior accelerometer and microphone was mounted in the car trunk.

In total 10 different road sections in the area of Luleå were measured at three different velocities (30, 50 and 70 km/h). Due to traffic and speed limits all speeds could not be measured for all sections. The 10 road sections shown in figure below consisted of 8 asphalt pavements ranging from freshly laid asphalt to worn asphalt in need to be replaced, one concrete pavement, and one concrete tile pavement. The asphalt road sections were classified in two groups (smooth and coarse) based on asperity height unevenness, asperity spacing and asperity radius. Road sections S1-S4 were classified as smooth and sections S5-S8 as coarse.

Based on the sum of the level of the octave bands 500 and 1k Hz a classification method was developed. The road sections were arranged by the sum of the octave bands for the different sensors and speeds. The results showed that there is information in both the sound and acceleration signal that can classify texture amplitudes for wavelength of 0.5 cm to a few centimetres. The asphalt road sections were successfully classified at 50 and 70 km/h based on octave bands levels.

Study 2 – Road profile reconstruction from vehicle acceleration measurements

This study describes an indirect method to estimate the road profile from vehicle acceleration measurements using an inverse quarter-car model. Quarter-car models are commonly used to assess the dynamic response of vehicles. A two-degree of freedom quarter-car model subjected to road excitation (zr) is illustrated in the figure to left. The tyre dynamics are described as a spring with no damping. The system can be redefined and solved by using the state space model. The inverse problem – to obtain the effective road elevation from acceleration measurement – was solved by inverting the state space model. The models were implemented in MATLAB and model parameters for the car were determined by measuring weight, static compression and analysing system resonances of recorded axle acceleration data. Due to nonlinearities in the suspension; different parameters for each speed were used. The frequency response of the quarter car model and its inverse is plotted in figure below.

Measurements were performed on the Frequency Sweep Excitation Test Track. The track was 100 m long with ribs; height 20 mm and width 40 mm. The acceleration at the rear left tyre was measured near the wheel axle. The reconstructed profile and the derived effective height are shown in figure below. Linear trends were removed from the reconstructed data. Tyre resonances (80-90 Hz), especially apparent at 50 km/h, are seen in the measured acceleration as in the reconstructed profile. The indirect method using axle acceleration has limitation in low frequencies due to sensor sensitivity. For high frequencies (above 80 Hz which correspond to 10-18 cm texture wavelength for speed range of 30-50 km/h) tyre resonances impair the result.

Study 3 – High speed profilometer measurements

Together with Vectura 5 road sections around Stockholm were measured using a high speed profilometer.

Projektledare: Roger Johnsson