Predicting the tensile strength of single wool fibers using artificial neural network and multiple linear regression models based on acoustic emission

The acoustic emission (AE) technique is widely used at the present time for almost any kind of material characterization. The main aim of the present study was to predict the tensile strength of wool by using artificial neural networks and multiple linear regression analysis based on AE detection. With this aim, a number of single wool fibers were stretched to fracture and the signals at break were recorded by the AE technique. The energy, amplitude, duration, number of hits, average rectified value and root mean square value were used as input parameters to predict the strength of the wool. A feed-forward neural network with a backpropagation (BP) algorithm was successfully trained and tested using the measured data. The same input parameters were used by multiple stepwise regression models for the estimation of wool strength. The coefficients of determination of the BP neural network and stepwise regression indicate that there is a strong correlation between the measured and predicted strength of wool with an acceptable error value. The comparative analysis of the two modeling techniques shows that the neural network performs better than the stepwise regression models. Meanwhile, the relative importance of the input parameters was determined by using rank analysis. The prediction models established in the present work can be applied to AE studies of fiber bundles or fiber-reinforced composite materials.

Automated summary

This study successfully predicted the tensile strength of wool using the AE technique combined with artificial neural networks and multiple linear regression analysis, with a focus on determining the relative importance of input parameters through rank analysis. The results indicate that the neural network model outperformed the stepwise regression model in predicting wool strength.