The effect of mean stress is a significant factor in design for fatigue, especially under high cycle service conditions. The incorporation of mean stress effect in random loading fatigue problems using the frequency domain method is still a challenge. The problem is due to the fact that all cycle by cycle mean stress effects are aggregated during the Fourier transform process into a single zero frequency content. Artificial neural network (ANN) has great scope for non-linear generalization. This paper presents artificial neural network methods for including the effect of mean stress in the frequency domain approach for predicting fatigue damage. The materials considered in this work are metallic alloys. The results obtained present the ANN method as a viable approach to make fatigue damage predictions including the effect of mean stress. Greater resolution was obtained with the ANN method than with other available methods.

No abstract

a b s t r a c tThis paper presents artificial neural networks (ANN) and wavelet analysis as methods that can assist high resolution of multiple defects in close proximity in components. Without careful attention to analysis, multiple defects can be mis-interpreted as single defects and with the possibility of significantly underestimated sizes. The analysis in this work focussed on A-scan type ultrasonic signal. Amplitudes corresponding to the sizes of two defects as well as the phase shift parameter representing the distance between them were determined. The results obtained demonstrate very good correlation for sizes and distances respectively even in cases involving noisy signal data.

No abstract

Abstract. An inverse artificial neural network (ANN) assessment for locating defects in bars with or without notches is presented in the paper. Postulated void defects of 1mm x 1mm were introduced into bars that were impacted with an impulse step load; the resultant elastic waves propagate impinging on the defects. The resultant transient strain field was analyzed using the finite element method. Transient strain data was collected at nodal points or sensors locations on the boundary of the bars and used to train and assess ANNs. The paper demonstrates quantitatively, the effects of features such as the design of ANN, sensing parameters such as number of data collection points, and the effect of geometric features such as notches in the bars.