Prediction of seam strength of superimposed and lapped seams on plain woven cotton fabric using artificial neural network

The strength of the seam is one of the most important factors in determining the garment’s durability, structural soundness, and useful life. Conventional experimental-based assessment techniques are robust, but ineffective in cases where different sewing parameters interact non-linearly. This work is focused on the formulation and validation of an Artificial Neural Network (ANN) model for predicting seam strength associated with superimposed and lapped seams under different sewing conditions. These seam classes are commonly used in woven apparel products such as shirts, trousers, denim garments, sportswear, and workwear, where seam durability and structural integrity are important performance requirements. 60 samples of seam were prepared from plain cotton fabrics. Five sewing thread linear densities (21-59 tex), six stitch densities (5-11 SPI), and two seam classes were systematically used in combination. The seam strength according to ASTM D 1683-04 was tested with a calibrated Universal Testing Machine. A feed-forward ANN with one hidden layer was implemented in MATLAB and trained using the Levenberg-Marquardt backpropagation algorithm. The performance of the model was assessed using mean square error (MSE), root mean square error (RMSE), mean absolute percentage error (MAPE), and correlation coefficient. For training, validation, testing, and independent datasets, the ANN exhibited very good prediction capability with correlation coefficients greater than 0.99. Independent verification exhibited an RMSE of 2.38 N and a MAPE of 6.96%, indicating improved prediction accuracy compared to earlier ANN-based seam strength models with an approximately 15-30% lower prediction error. The model was successful in capturing the non-linear relationships between stitch density, thread linear density, and seam geometry. The developed model may assist garment manufacturers and textile engineers in rapid sewing parameter selection, reduction of trial-and-error, minimization of material wastage, and optimization of seam quality during apparel product development.