The pitfalls inherent in the indiscriminate application of artificial neural networks to numerical modeling problems are illustrated. An example is used of an apparently successful (but ultimately unsuccessful) attempt at training a neural network constitutive model for computing the resilient modulus of gravels as a function of stress state and various material properties. Issues such as the quantity and quality of data needed to successfully train a neural network are explored, and the importance of an independent test set to verify network performance is examined.
Get full access to this article
View all access options for this article.
References
1.
HicksR. G., and MonismithC. L.. Factors Influencing the Resilient Properties of Granular Materials. In Highway Research Record 345, HRB, National Research Council, Washington D.C., 1971, pp. 15–31.
2.
MayR. W., and WitczakM. W.. Effective Granular Modulus To Model Pavement Responses. In Transportation Research Record 810, TRB, National Research Council, Washington D.C., 1981, pp. 1–9.
3.
UzanJ.Characterization of Granular Materials. In Transportation Research Record 1022, TRB, National Research Council, Washington D.C., 1985, pp. 52–59.
4.
WitczakM. W., and UzanJ.. The Universal Airport Pavement Design System, Report I of V: Granular Material Characterization.University of Maryland, Department of Civil Engineering, College Park, 1988.
5.
ItaniS. Y.Behavior of Base Materials Containing Large-Sized Particles. Ph.D. thesis. Georgia Institute of Technology, Atlanta, Sept. 1990.
6.
FelibertiM.Critical Evaluation of Parameters Affecting Resilient Modulus Tests on Subgrades. M.S. thesis. University of Texas at El Paso, Nov. 1991.
7.
PezoR. F.A General Method of Reporting Resilient Modulus Tests of Soils—A Pavement Engineer's Point of View. Presented at 72nd Annual Meeting of the Transportation Research Board, Washington D.C., 1993.
8.
BarksdaleR. D., and ItaniS. Y.. Influence of Aggregate Shape on Base Behavior. In Transportation Research Record 1227, TRB, National Research Council, Washington, D.C., 1989, pp. 173–182.
9.
GhaboussiJ., GarrettJ. H., and WuX.Knowledge-Based Modeling of Material Behavior with Neural Networks. Journal of Engineering Mechanics, Vol. 117, No. 1, 1991, pp. 132–153.
10.
EllisG. W., YaoC., and ZhaoR.. (1992). Neural Network Modeling of the Mechanical Behavior of Sand. Proc., 9th Conference on Engineering Mechanics, College Station, Tex., 1992, pp. 421–424.
11.
PidapartiR. M., and PalakalM. J.. Material Model for Composites Using Neural Networks. AIAA Journal, Vol. 31, No. 8, 1993, pp. 1533–1535.
12.
PenumaduD.Strain Rate Effects in Pressuremeter Testing and Neural Network Approach for Soil Modeling. Ph.D. thesis. Georgia Institute of Technology, Atlanta, 1993.
13.
AlbaJ. L.Laboratory Determination of Resilient Modulus of Granular Materials for Flexible Pavement Design. Ph.D. thesis. Georgia Institute of Technology, Atlanta, 1993.
14.
Resilient Modulus of Unbound Granular Base/Subbase Materials and Subgrade Soils. In SHRP-LTPP Laboratory Material Testing Guide.Strategic Highway Research Program, 1993.
15.
RumelhartD. E., HintonG. E., and WilliamsR. J.. Learning Internal Representation by Error Propagation. In Parallel Distributed Processing: Explorations in the Microstructures of Cognition (RumelhartD. E., and McClellandJ. L., eds.), MIT Press, Cambridge, Mass., 1986, pp. 318–362.
16.
CybenkoG.Approximation by Superposition of a Sigmoidal Function. Mathematics of Control, Signals, and Systems, Vol. 2, 1989, pp. 303–314.
17.
HornikK., StinchcombeM., and WhiteH.. Multilayer Feedforward Networks Are Universal Approximators. Neural Networks, Vol. 2, 1989, pp. 359–366.
18.
FreemanJ. A., and SkapuraD. M.. Neural Networks: Algorithms, Applications, and Programming Techniques.Addison-Wesley, Reading, Mass., 1991, p. 104.
