الفهرس 
IntroductionOnly 14 pages are availabe for public view
 |  | from | 62 |  |  |
| | | from | 62 | | |
Abstract
As the study of fatigue failure of composite materials needs a large number of experiments as well as long time, so there is a need for new computational technique to expand the spectrum of the results and to save time. The present work represents a new technique to predict the fatigue life of Woven-Roving Glass Fiber Reinforced Polyester (GFRP) under combined bending and torsional moments, with different fluctuating stresses. Fatigue experimental tests were conducted on two fiber orientations, [0o,90o]2 and [±45o]2 thin walled tubular specimens with different ratio of the flexural stress (A) to the torsional shear stress (B), these ratios were A/B= 2, 1. And to study the effect of mean stress of fatigue behavior. Specimens were fatigue tested at different stress ratio (R=𝑚𝑖𝑛.𝑠𝑡𝑟𝑒𝑠𝑠 𝑚𝑎𝑥.𝑠𝑡𝑟𝑒𝑠𝑠 ), R= -1, -0.75, -0.5, -0.25, 0 for [0o,90o]2 specimens with A/B= 2 and R= -1, -0.75, -0.5, 0. For [±45o]2 specimens with A/B= 1. Three neural network structures, feed-forward (FFNN), generalized regression (GRNN) and radial basis (RBNN), are applied, trained and tested. The groups of data considered are the maximum stress, the shear stress, the stress ratio and A/B ratio. On the other hand, more accurate prediction method is obtained by using a useful expert system which is designed to aid the designer to decide whether his suggested data for the composite structure is suitable or not. In this expert system, a neural network is designed to consider the design data as input and to get the fatigue life cycles (N) as output. The results showed that the experimental data matching to the predictive results with a very simple and acceptable error ratio. Also, the feed-forward neural network shows better results than that given by the generalized regression and radial basis network. The Artificial neural network system will be helping the designer with a100% correct conclusions about his decision of the combination of the proposed data.