University Park, Pa. - Penn State researchers have developed the first neural network pattern recognition software - "a tool that learns" - to predict the stresses on pavement from truck radial-ply and truck bias-ply tires, the truck tires found most often on U.S. roads.
Dr. Moustafa El-Gindy, senior research associate and director of the Crash Safety Center at the Pennsylvania Transportation Institute, a Penn State intercollege research program, led the project. He says, "The new tool will provide engineers with a unique set of integrated and coordinated capabilities needed to design better tires and roads."
El-Gindy described the new tool in a paper, "Tire/Pavement Contact-Stress Model Based on an Artificial Neural Network", presented Tuesday, July 20, at the second annual Pennsylvania Transportation Conference at the University. His co-authors are Heidi L. Lewis, former Penn State student, and A. Scott Lewis, research associate at Penn State's Applied Research Laboratory.
"Recent measurements have shown that the vertical stress of a tire moving at slow speed on pavement is not uniformly distributed, creating an enormously complex problem for anyone who wants to describe, mathematically, what is happening to the road. It may even be impossible to develop ordinary mathematical models that take into account the non-uniform contact stress distribution", El-Gindy says.
"Our new artificial neural network-based model, however, can be trained to predict the complex stress distribution patterns under a tire, at any given load and inflation pressure and for any specific tire type," he says.
The researchers currently have four programs, two each for bias-ply and radial-ply tires. The programs predict the vertical stress distribution at the contact patch of each tire at constant inflation pressure and variable vertical loads and at constant vertical load and variable inflation pressure. All programs were developed on Matlab, are user friendly and can be run in a PC environment.
The programs were developed using actual, precisely measured, three-dimensional contact-stress distribution patterns obtained from low speed rolling tire tests conducted by the University of California in association with California Department of Transportation and the Nevada Automotive Test Center. The Federal Highway Agency supplied the data to the Penn State group.
The new software, called the "Neuro-Tire Patch Model," is built on artificial neural nets, which are mathematical tools with excellent capabilities in data storage and retrieval as well as pattern matching, recognition and classification. Inspired by the biological behavior of brain cells, neural nets can "learn" when given a large body of data on which to "train."
Using the FHA-supplied data and the neural net technique, the Penn State researchers "trained" the software using precisely measured three-dimensional contact-stress distribution patterns from low-speed rolling tire tests that included 50 sets of measurements. Having been trained on known data sets, the software has been shown to accurately predict outcomes for data sets on which it has not been trained. The user need only input one of the specified vertical loads and the required inflation pressure to obtain the contact patch stress distribution.
El-Gindy notes, "Inaccurate measurements, of course, may lead to poor predictions by the neural networks. However, the software has been trained on the best data sets available."
The research was supported, in part, by the Federal Highway Administration Turner Fairbank Highway Research Center. Bill Kenis, truck/pavement interaction program leader, represented FHA on the project.
The above post is reprinted from materials provided by Penn State. Note: Materials may be edited for content and length.
Cite This Page: