An inverse predictive model for the design of functional textiles

Abstract

Coated woven fabrics are used for large scale structures including airports and sports stadia. Manufacturers produce a range of fabrics from which a single fabric is selected by the structural engineer based on design criteria such as stiffness, weight, strength and formability. Designs must therefore utilise a fabric with properties which may not be optimal for that particular application. This thesis develops and tests a model that allows a bespoke coated woven fabric to be designed with specified mechanical properties such as tensile stiffness, Poison’s ratio and shear stiffness. A method is developed to ‘invert’ an existing predictive mechanistic ‘unit cell’ model using the derivatives of the equations defining the unit cell. The existing model is altered to enable the prediction of shear characteristics in addition to tensile properties by the inclusion of the coating using a finite element representation. The ‘inverse’ model is shown to accurately design a fabric for specific and attainable targets of Young’s modulus, Poisson’s ratio, and shear stiffness which have been derived using the predictive model for various fabric stress states. The effect of variability in fabric parameters on the tensile response of a fabric is considered using both Monte Carlo and FORM analysis. The sensitivity of the fabric response to biaxial loading is calculated using the direction cosines defined in the FORM methodology. The calculation of fabric sensitivity also enables a detailed investigation of the sensitivity of fabric stress-strain behaviour to variation in individual fabric parameters. A method is developed to design fabrics with mechanical properties which are robust to changes in manufacturing parameters by altering the geometry of the fabric. The model is validated by comparing the inverse model output to unit cell model input and also to biaxial test results. The inverse model shows excellent fidelity with results calculated using the unit cell model, but fails to adequately reproduce the actual fabric geometry when target stiffness values are based on biaxial test data. A method for the removal of yarns from fabrics and tensile testing of coated fabric yarn specimens is also developed. iii It is common practice to use a plane stress formulation to approximate the stress-strain response of a coated woven fabric. Comparison of the model output with biaxial test results necessitated the creation of a method for the calculation of fabric tensile stiffness at multiple stress states instead of a single set of elastic constants. This approach takes into account the complex nonlinear behaviour of architectural fabrics by considering the variation in stress-strain behaviour at different biaxial stress states. The final inverse model provides a novel tool for the design of coated woven fabric with prescribed mechanical responses at multiple stress states that is robust to variations in its constituent parameters, with scope for future application in textile architecture, medical textiles and industrial textiles.