Many small automotive companies and testing agencies are trying to understand road roughness to enable them to test vehicle component fatigue efficiently with a limited budget. No one wants to test to the absolute worst conditions, but many companies need to test to an 80th to 95th percentile road roughness.
To do this, how much and what type of data is needed to understand the statistical distribution of roughness? In a perfect world, this would consist of acceleration in six degrees of freedom. The perfect world would quantify the differences due to tire sizes, tire pressures, weight distributions, and single- frame systems vs. coupled systems for each area of the vehicle.
While this study only addresses the vertical acceleration on a passenger car at one tire pressure, the intent is to provide a method of evaluating road roughness in a quantitative format and an idea of how much data is necessary. This quantitative data could be used to determine automotive proving-grounds procedures and understand acceleration factors and/or simulation files for laboratory shake tests.
Because real-world roads are continually changing, interstate roads in different Midwest states in different conditions of repair, such as under construction, newly paved, and needing repair, were chosen. Interstates I-94, I-90, I-80, I-74, I-65, I-57, I-55, and I-35 in Illinois (809 miles), Indiana (197 miles), Minnesota (213 miles), Wisconsin (333 miles), and Michigan (268 miles) were used to collect data for a total of more than 1,800 miles.
Since the goals of this project were to understand the basic statistics of road roughness and provide a testing approach to obtain and analyze data, a single-channel vertical accelerometer was installed at the base of the driver seat. Operator perception can have a very significant influence on how the person drives, so a location was selected that would provide direct feedback.
All testing was done at base vehicle weight with only the driver and no ballast. The fuel level varied from full to near empty, trying to keep equal time at different levels. Tire pressure was not adjusted. Vehicle speeds were consistent with the normal traffic flow.
Since this data can only be related to one specific car with one specific tire brand and size with data on roads that will be constantly changing, data also was obtained at an automotive proving grounds to provide some type of reference that will not significantly change. The Bosch Automotive Proving Grounds durability-course events were measured at different vehicle speeds with the same vehicle and test setup.
Test Equipment
A SoMat 2100 Field Computer was used to collect the data. A single channel of acceleration was recorded at 500 samples/second (S/s). A global positioning system (GPS) module recorded longitude, latitude, vehicle speed, and elevation at 1 S/s. SoMat EASE Software was used to verify, review, and analyze the data. Calibration and zeroing were done before each trip.
Review and Analysis Of the Data
A visual review of the data initially was completed to look for potential problems. A frequency analysis using an auto power spectrum was performed to understand the proper sampling speed. Data indicated the significant frequency input was less than 50 Hz for the accelerometer data. A sampling rate of 500 S/s was chosen. By using the mapping software, the data was segregated into categories of interstate, highway, country road, and city.
To present the data in a statistical distribution and for the identification of the different roads, an arbitrary 1,000-s window was used (when possible) for each segment analysis. This relates to almost 20 miles at normal interstate speeds. The distance for each segment was determined by integration of the speed signal. This distance was used to normalize each segment into damage per mile or counts per mile.
The primary analysis consisted of a rain-flow analysis of the vertical acceleration. A bin size of 0.2g was used.
A secondary analysis performed a pseudofatigue analysis of the vertical acceleration. Since there is an approximately linear relationship of strain and acceleration in this frequency range, a scale factor of 900 was used on the acceleration data, and a strain life fatigue analysis was performed to provide qualitative damage distribution. Although the particular slope of a failure line will not be appropriate for all components, this analysis allows all levels of acceleration to be combined into one number of relative damage or roughness for each segment.
Presentation of the Data
The initial presentation of data consists of a rain-flow analysis of the acceleration data to indicate, for the total interstate mileage recorded, how many times per mile bumps occur that create specific acceleration ranges (Figure 1). Or for the large accelerations, how many miles occur between bumps.
The next presentation of data aids in understanding road roughness. The data indicates that road roughness is a skewed instead of a normal statistical distribution. It does appear logical that there are many smooth roads and a much smaller number of roads under construction or in different levels of deterioration. The pseudofatigue damage distribution indicates the smooth grouping and different levels of roughness (Figure 2).
Practical Applications of Results
Unless you would intentionally choose only roads that have been newly repaved, only 100 or 200 miles of data are required to obtain a respectable representation of road roughness. Charting the data in accumulative 100-mile increments indicated no significant difference after the first 100 miles.
It appears there is no convergence to a value as miles increase, just a random change due to random roads. Best of all, there does not appear to be any drastic change at any number of miles. Plotting the severity of each segment from the chronological recording dates indicated rough road segments typically are sporadic and irregularly encountered.
From the chronological data, it appears that road repair is done in small segments. There are no extensive portions of either extremely good roads or extremely bad roads. This allows short data collections to have a more reasonable probability of being representative. When using the damage-per-mile data, the high-end road roughness should be 15 to 35 times the smooth road roughness.
Another way of evaluating road roughness is shown in Figure 3. Approximately 43% of the roads accumulate damage at a rate of 10% or less compared to the worst road segment.
An interesting observation is how the real-world data relates to the data from the automotive proving grounds. By using combinations of durability events and speeds over the events, different acceleration factors (the acceleration factor is the ratio of real-world miles to proving-grounds miles for the equivalent damage) can be obtained. Also, there can be adjustments to high-load, low-frequency occurrences.
For example, if the data consists of 1,800 miles of real-world roads and the maximum acceleration only occurs every 500 miles, there is a significant probability that in 100,000 miles there would be a higher acceleration. Also, there is a significant probability that if 1,800 miles of real roads were tested, some other roads would produce higher acceleration. This data indicates that automotive proving grounds can be easily used to obtain these higher loads. The major decision is how much higher in roughness to allow.
The automotive proving-grounds durability courses typically match the real-world conditions very well (Figure 4). The low-magnitude inputs typically are considered nondamaging or of very little significance. These low-magnitude inputs are much lower per projected equivalent real-world miles on a durability course.
Typical acceleration factors of the proving grounds to the real world are 20:1 to 100:1. As the acceleration factor increases, the acceleration-factor variation increases for the different areas of the vehicle. The radiator may have a 50:1 acceleration, the instruments a 60:1 acceleration, and the rear speakers a 45:1 acceleration factor.
About the Author
Joe Janas, a sales engineer at SoMat, has more than 30 years of experience in testing of construction, forestry, mining, and agricultural equipment; military vehicles; and automotive equipment including passenger cars, light trucks, motor homes, utility vehicles, and heavy trucks. Mr. Janas earned a master’s degree in engineering and previously worked for Clark Equipment, AlliedSignal, and Case. SoMat, 1322 E. 800 N, LaPorte, IN 46350, 219-778-9932, e-mail: [email protected]
For More Information
SoMat Field Stories
www.rsleads.com/209ee-209
Return to EE Home Page
Published by EE-Evaluation Engineering
All contents © 2002 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
September 2002