Friday, March 13, 1998
4:00 - 5:00 pm
Although laboratory road simulators have reduced testing time, two obstacles to improved process automation remain. The first is that because simulation control parameters are specimen dependent, new data must be acquired for each specimen to ensure the most accurate test. A new data acquisition can add 6-8 weeks and $200,000 to the testing process. Additionally, drivable prototypes may be costly or entirely unavailable. For one very important class of durability road simulators, this problem is acute. These road simulators are referred to as spindle-coupled simulators, because the simulator attaches at the vehicle, spindle eliminating the tire from the road simulator. Prior work at Ford and GM has focused on physically including the missing tire dynamics as part of simulation process in the desire to move from vehicle specimen dependent parameters (spindle force or acceleration) to a more specimen independent control parameter such as the road profile.
The second impediment to improved test automation is that an open-loop iterative control process is used. This process is time consuming, requiring 2-4 weeks of a skilled engineer's time. The iterative control process is very time-consuming and does not compensate for changes in the test vehicle's parameters due to design evolution or suspension degradation as the durability test proceeds. Techniques for control of road simulators have been documented in Craig and in Cryer, et.al. The current industry-standard approaches to control all rely on the use of simple, experimentally determined models of the simulator and vehicle dynamics. These models are used to predict the control inputs necessary to achieve the desired responses. Although the desired vehicle response channels are known apriori, an open-loop iterative control process is used.
This talk will demonstrate two major process improvements. The first is a method to reduce the need for acquiring new data for the spindle-coupled simulation process. This will be accomplished by using a model of the dynamics that includes physical elements such as the actuators and simulator/vehicle dynamics as well as a model of the tire dynamics. Experimental results of the effectiveness of the model will be shown. The second improvement to testing automation is to demonstrate an improved control process. The approach taken will be to use an H-infinity approach that allows for frequency dependent weightings to represent sensor noise Wn, and model uncertainty, Wu. Performance will be specified by the weighting We. The structure of the controller, shown in Figure 2, will be investigated to determine the appropriate combination of feedback, K, feedforward, F, and preview control terms. Simulation and experimental results will be shown.
© UofM College of Engineering Control Research Group -