Lab Lessons Learned from Student Testing, Series 2
Road Quality Determination Using Vehicle Data
Last month we featured an application note by Purdue University Professor, Doug Adams, sharing a number of educational experiences his students have learned "the hard way."
Projects are carried out each spring by groups of students enrolled in a class called ME597A ‘‘Practical Experiences in Vibration’’ at Purdue University in the School of Mechanical Engineering. As the second installment of a three-part series, the experiments and results of a student team’s project are presented covering the analysis of road conditions. Road quality is determined with spectral analysis using an automotive suspension system as a mobile operating data acquisition system.
This course and others like it at institutions across the country recognize that experimental vibrations and acoustics are best learned if students are required to work in groups to conceptualize, design, conduct, and analyze their own experiments. Multiple offerings of this course at Purdue University have demonstrated that students retain more knowledge of measurement systems, signal processing, frequency response estimation, modal analysis, and other topics in experimental dynamics if they experience the frustrations and revelations of testing firsthand. Students sometimes find testing tedious; however, they always come away from the course with a solid understanding of the basics in structural dynamic testing: What is modal impact testing? Howcan it be used to obtain information about the free vibration response of a structure? What signal processing issues can cause problems in the analysis of data?
In this experimental vibrations course, the instructor serves primarily as a ‘‘measurement coach’’ by providing students with quick tutorials and suggestions at various points throughout the semester. By working together, students usually troubleshoot and solve their own measurement challenges. In fact, different groups often consult for one another during the semester leading to a highly collaborative learning environment.
In addition to benefiting students, this type of experiential and cooperative learning approach also benefits industry. In fact, the course at Purdue University is sponsored by notable industry partners (including Arvin Meritor, General Motors, Lord Corporation, The Modal Shop, PCB Piezotronics, among others) who have provided test structures, measurement instrumentation, and expertise in kind. Surveys of these sponsors from both industry and federal laboratories with an interest in structural dynamics indicate that by allowing students to conceptualize and design their own experiments, open-ended lab experiences provide students with the training necessary to hit the ground running when they enter the workplace. For example, students are exposed to the latest sensing, data acquisition, and data analysis (software) technologies during the semester-long course at Purdue University. VXI Technologies 1432 data acquisition systems with multireference impact testing and X-Modal interfaces, which were created by the University of Cincinnati Structural Dynamics Research Laboratory, are used by every student in the course. This exposure to multichannel dynamic measurement systems helps demystify experimental techniques so the student can quickly learn whichever system they are expected to use in industry. Surveys and hiring histories associated with programs, where hands-on, open-ended labs are offered, demonstrate that employers are more likely to hire undergraduate and graduate students who have participated in such programs.
ROAD QUALITY DETERMINATION IN TIPPECANOE COUNTY, INDIANA
Road quality is often evaluated using subjective ratings assigned by an inspector from the highway department. Cracking in the pavement can lead to problems with vehicle handling, uncomfortable ride dynamics and motion sickness, tire damage, or other potential driving hazards. The purpose of this project was to understand and quantify road quality using more objective features from experimental data. It was thought that a vehicle could first be instrumented with ICP accelerometers on its spindles and in other locations to collect data from which road quality features could be extracted. The vehicle could be driven on different roadways with known conditions to extract information relating to road quality. Differences in vibration levels and/or the bandwidth of the excitations provided by these ‘‘baseline’’ roadways would be the indicators of quality. However, students would face a challenge in interpreting the data they collected because the forced response of a vehicle contains information relating not only to the roadway profile forcing function but also to the vehicle dynamics.
Lesson Learned #1
Because the objective of this project was to characterize the roadway forcing function, students were required to develop an understanding of the differences between free- and forceresponse behavior. In a sense, the forced response of a vehicle is the product (or convolution) of the free response of the suspension and the roadway forcing characteristics. Furthermore, the forced response of a vehicle contains information related to the suspension system, tires, steering mechanisms, and so forth, in addition to information related to the road profile. In order to interpret the forced response, students would need to first develop a good understanding of the free-response characteristics of the vehicle. By conducting a series of tests starting with tests in which the roadway forcing characteristics are specified and ending with tests in which the roadway forces are unknown, students would be able to extract information related to road quality on those unknown road surfaces.
At first, the students struggled with the overwhelming task of setting up such a test involving multiple channels of data using accelerometers and data acquisition systems, which would need to survive sometimes harsh vehicle ride conditions. For example, students discussed the types of sensors they should use (e.g., ICP or DC capacitive, sensitivity, single- axis or triaxial accelerometers), the locations at which to place sensors (e.g., spindle, strut, chassis), and the required data acquisition parameters (e.g., frequency range, sample rate, filter settings) among other things. After researching the automotive literature, students discovered that SAE had already established standards (SAE J1490) for setting up and conducting such vehicle experiments.
Lesson Learned #2
When students search the literature to identify resources that may assist them in their projects, references to prior work that is either identical or directly related to their projects is always found. One instructor lesson learned in this project was that references to such standards should be made available to students early in the project to reduce the time students spend reinventing approaches that have already been developed by others. By reducing the time spent by students on the front end of their experiments, instructors can ensure that students are free to spend more time with analyzing and interpreting their data. Because one of the goals of this particular course is to have students develop their own experimental planning skills, the instructor does not usually want to provide students with ready-made plans for their tests. However, standards from SAE and other organizations involved in experimental mechanics are invaluable resources to instructors because they are rigorous test outlines, which still require students to plan and execute their experiments.
The students used this SAE standard as a starting point for their test plan. However, they soon learned that standards do not always apply strictly to tests that are not directly discussed by the standard. For example, the standard describes ride vibrations as oscillations of the vehicle in the frequency range from 1 to 25 Hz. Although the 1- to 25-Hz frequency range would contain the bulk of ride vibrations, students conducted additional surveys of the literature on vehicle suspension and passenger comfort to determine if this frequency range was sufficient to consider in their data acquisition and analysis. Students learned that suspensions are nonlinear leading to harmonic distortion in the response, which would contain energy in higher frequency ranges. Students determined that the harshness in a vehicle’s ride at these higher frequencies could possibly be indicative of roadway surface quality; therefore, they extended their frequency range of interest to 100 Hz.
Lesson Learned #3
Students learned that by conducting a rigorous literature survey, they could assemble a test and data analysis plan that was tailored to their measurement problem. It is rare that one reference or standard contains all the information required to solve an experimental measurement problem of interest.
As stated above, the students determined that they could use vehicle response sensors as a means of assessing road quality. Students also determined that in order to characterize the roadway forcing characteristics, which determine quality, the vehicle dynamics would first need to be studied using known roadway excitations. Therefore, a controlled durability course at the Bosch Proving Grounds in New Carlisle, IN (see Fig. 1) was used to conduct the first set of tests, which aimed to characterize the vehicle response given an excitation spectrum from the known roadway profile. The durability course included a cobblestone road (shown in Fig. 1), inverted chatter bumps, and an undulating road. The controlled environment allowed certain vehicle free-response characteristics to be accentuated, which resulted in a better understanding of the forced response behavior observed later when response data were acquired on actual roads.
The vehicle selected was a 2002 Saturn L200. It was instrumented in the same manner for both the proving ground and the public road quality tests. Six ICP accelerometers were mounted in the following locations: one triaxial accelerometer (PCB Piezotronics model T356B18, 1 V/g) at the base of the driver’s side strut and five single-axis accelerometers (PCB Piezotronics model 333B32, 100 mV/g), one each at the top of the driver’s side strut, the passenger side strut, the driver’s side seat track, the passenger side seat track, and in the trunk. These locations were selected to determine if true road quality could be quantitatively assessed from different points on the vehicle despite the suspension action, which filters input spectra from the road surface.
Lesson Learned #4
As students proceeded with the test execution, they realized that data analysis in the sensor locations furthest from the road surface, such as those on the passenger seat track and in the trunk, would be most difficult because these sensors were likely to contain more dynamics of the vehicle chassis, body, and other subsystems. Such noncollocated sensors always run the risk of introducing dynamics between the point of interest (roadway) and the measurement location. However, students felt that the triaxial spindle accelerometer would be an informative indicator of the roadway excitation forces given the location of this sensor directly above the tire.
Fig. 1: Cobblestone road surface at proving ground track
The accelerometer on the base of the strut at the spindle was mounted first with super glue and then wrapped tightly with electrical tape. It was important to secure the sensor because of the rough roadway conditions anticipated in the proving ground and public road tests. The cables were routed from the base of the strut to the back of the wheel well, tie strapped to the existing cables near the back of the wheel well, and finally routed through a small hole into the engine compartment area. Figure 2a shows the positioning of the triaxial accelerometer on the spindle. As mentioned above, this sensor was essential for assessing road quality because it provided multidimensional information below the suspension where road inputs were most observable. A laptop-based, portable data acquisition unit was situated in the back seat to record data as shown in Figure 2b.
Fig. 2: (a) Driver’s side spindle base triaxial accelerometer and (b) data acquisition system in the back seat
As stated above, different known road surfaces at the Bosch Proving Grounds were used to characterize the vehicle free response. Figure 3 shows the acceleration frequency spectra for the vehicle tested on (a) cobblestone and (b) undulating road surfaces. The low-frequency body modes (pitch and roll near 2–4 Hz) are more evident in the undulating road surface data (b), whereas the higher frequency suspension modes such as tire hop near 15 Hz are more evident in the cobblestone road surface (a). It is evident that the cobblestone road contains higher harmonic frequencies of the fundamental cobblestone period and the undulating road contains primarily lower frequencies.
Lesson Learned #5
This lesson learned regarding the different frequency content for various road surfaces was an important reminder to students of why they were conducting these tests on a known set of road surfaces in the first place. The undulating road would be characteristic of one public road the students would later test. The cobblestone road was very different from any public road that would be tested; however, the cobblestone pattern vividly illustrated how higher frequencies could be encountered on normal road surfaces. The pattern of the cobblestone road reminded the students of the square waves they had seen earlier in the semester in a Fourier series homework assignment. Students felt that it was invaluable to experience the nature of frequency analysis firsthand while riding in the vehicle on the proving ground.
The filtering action of the suspension system was also evident in Figure 3b for the undulating road surface. Note the lack of any significant spindle response in the frequency range below 5 Hz. Students realized that it was important for the tires to remain on the road to provide sufficient handling performance on such a rough roadway. In contrast, the trunk response (yellow) and passenger seat track response (blue) were very high in the range below 5 Hz due to the low-frequency pitch and roll body resonant modes.
Lesson Learned #6
A direct observation of vibration isolation illustrated in Figure 3 was a valuable lesson learned for the entire class. The automobile’s suspension system effectively filtered the large vibrations measured at the spindle base from the driver. Although other real-world examples of vibration isolation in machine tools and even automotive power trains had been discussed during the semester, students were much more convinced of how important vibration isolation can be when they experienced the difficult driving conditions encountered on the cobblestone roadway.
When analyzing the data in Figure 3a, students were surprised to observe that the vertical acceleration on the spindle (highest amplitude red curve) dropped below the spindle fore-aft acceleration (blue curve higher than other curves in 40–50 Hz range). The same phenomenon was observed in Figure 3b for the undulating road surface. Students determined that the stiffness in the fore-aft direction must be higher than in the vertical direction and that there might also be nonlinearities acting to introduce these higher frequency response components. Regardless of the source of the fore-aft spindle response, students were encouraged by this as it suggested that the decision to use a triaxial accelerometer on the spindle was fortuitous. When the different scales are taken into account in Figure 3a, b, it is evident that the fore-aft response for the cobblestone road is nearly a factor of five times the fore-aft response for the undulating road. This difference in the fore-aft response for the roughness of the road suggests that this measurement would be helpful in characterizing road quality when public roads are tested.
Fig. 3: Acceleration spectra at eight locations for (a) cobblestone road surface showing excitation of higher frequency suspension
Fig. 3: Acceleration spectra at eight locations for (b) undulating road surface showing lower frequency body/chassis modes
Lesson Learned #7
The importance of the vertical and fore-aft response in the spindle acceleration for characterizing the nature of the roadway demonstrated the importance of sensor selection to students. The triaxial accelerometer was invaluable in this test because it provided multidimensional response data at the spindle leading to more insight into the nature of the road roughness (undulating vs. cobblestone pattern).
Having characterized the free-response characteristics of the vehicle on the proving ground roadways, a second set of tests was conducted to identify the quality of several roads in Tippecanoe County, IN. The vehicle was instrumented at the same eight locations. The data were analyzed and road quality was determined using acceleration amplitudes at different frequencies. A total of five data sets of 60 s each were taken for each road in the second set of tests. Care was taken to maintain a constant vehicle velocity for the trials on each road.
Lesson Learned #8
When conducting these public road quality tests, students learned that it is sometimes difficult to control all test parameters. In this case, the vehicle speed was very important to control because it governed the frequency bandwidth of excitation provided by the roadway. If the speed could not be controlledandrecorded as constant, the roadway excitation would have been time varying and conclusions regarding road quality from road to road would have been difficult to formulate.
In order to provide a realistic measure of what a typical driver would experience, the vehicle speed was held constant near the speed limit for each road. The vehicle speed was 35 mph for US 52 and Lindberg Road, 55 mph for SR 26, and 75 mph for I-65. Figure 4 shows the frequency domain response results for each of the four roads for four sensor degrees of freedom on the spindle (a, vertical; b, fore-aft; and d, lateral) and driver seat track (c). Note that although US 52 and SR 26 both have high vibration levels in the vertical direction at the spindle, US 52 has far higher fore-aft accelerations indicating poorer road quality especially related to handling. This conclusion was drawn based on the test data acquired for the cobblestone roadway at the proving ground. This result also demonstrates that subjective driver ratings at the driver seat (c) are inappropriate for assessing road quality due to the suspension action. Closer inspection of the vertical spindle (a) and driver seat (c) accelerations also reveals that Lindberg Road possesses the most severe low-frequency undulations below 5 Hz. This result was not surprising because Lindberg Road spans a boggy area, which has caused sinking at various spots of the roadway leading to undulations similar to those tested at the proving ground.
Fig. 4: Acceleration frequency spectra for four different roads tested (colors) with four different channels (a) spindle vertical, (b) spindle fore-aft, (c) driver seat, and (d) spindle lateral
It was determined that the road quality of several roads in Tippecanoe County, IN, could be quantitatively assessed using vehicle response data acquired at the spindle base in three directions. In order to interpret data acquired when the vehicle was driving on public roads, two roadways at a proving ground (cobblestone and undulating) were used to excite the vehicle with known forcing characteristics containing low- and highfrequency bandwidths. By controlling the excitation spectrum in this manner, the free-response characteristics of the vehicle suspension could be distinguished from the influence of the roadway forcing function. This information was then used to interpret data acquired on the four public roadways. It was determined that the spindle fore-aft acceleration gave a clear indication of the road roughness, whereas the spindle vertical acceleration indicated the waviness of the road.
Students learned lessons related to the differences between the free and the forced response of a vehicle in order to identify the nature of the roadway excitation characteristics. One important instructor lesson learned was the need to provide students with test standards from SAE or other organizations when such standards are available to avoid delays in testing, reducing time available for data interpretation. Students also learned valuable lessons about the positioning of sensors and the challenges involved in analyzing sensor data far removed from the locations where forces of interest are applied. Other lessons learned included the differences between low- and high-frequency excitations, the impact of vibration isolation on the performance of mechanical systems like automotive suspensions, the importance of choosing proper sensors with adequate degrees of freedom for characterizing multidimensional responses, and general issues associated with experimentation and the control of test parameters.