Effect of Process and Material Parameters on the Permeabilities of Reinforcement Mats: Experimentations and Machine Learning Techniques

The main objective of this work was to evaluate the effect of processing and material parameters on the reinforcement mat permeability through mould-filling experiments and to model the reinforcement mat permeability as a function of porosity, mat layers, test-fluid viscosity and injection pressure using machine learning (ML) techniques. Two experimental methods based on electrical sensors and visualization techniques were employed to measure the permeability through temporal flow front tracking. The fibre wetting analysis was performed using contact angle measurements to analyse the test fluid saturation at the reinforcement mats and its effect on mat permeability. Artificial neural network (ANN) and the adaptive neuro-fuzzy inference system (ANFIS) ML models were adopted to model effective permeability as a function of four input parameters using the experimental data. From the results, the order of permeability was obtained between 8 × 10^–10 to 8 × 10^–9 m² for chopped strand glass-fibre mat, 8.8 × 10^–10 to 8 × 10^–9 m² for jute fibre mat, 8.9 × 10^–10 to 8.5 × 10^–9 m² for woven roving glass-fibre mat, and 8.9 × 10^–10 to 1 × 10^–8 m² for hemp fibre mat. From the fibre wetting analysis, it was found that the mat permeability decreases with the increase in the test fluid–fibre surface wetting time. From the modelling analysis, it was found that the adopted ANN and ANFIS techniques predicted permeability values qualitatively and quantitatively with R² values of 0.967 and 0.975, respectively. From the statistical analysis, ANFIS has shown an efficient correlation with the experimental permeability as a function of input key parameters than the ANN approach.

Graphical Abstract