فهرست نشریات علمی دسترسی آزاد ایران

The present study utilized an artificial neural network (ANN) model to anticipate Barcol hardness, impact strength, and heat deflection temperature data for epoxy resin specimens with varying weight percentages of graphite additive exposed in different types of water. A feedforward backpropagation algorithm was used for predictive modeling with two input parameters: the weight percentage of the graphite additive (0, 5, 10, 15, and 25 wt.%) and the type of water used (dry specimen, potable water, distilled water, alkaline solution, and acidic solution). Experimental test data for mechanical properties were used to train the ANN model. The network was validated by comparing the predicted outputs with experimental data and by evaluating performance metrics. The results conclude that the ANN model is a practical and accurate approach for rapidly predicting mechanical performance and can be considered a substitute for traditional procedures used to characterize composite materials through experimental methods. Among the two input parameters, the weight percentage of the graphite additive was the most essential input parameter used to predict the mechanical properties of composites. Besides, the key findings of this work can also be a reference for the engineering practice of composite materials under mechanical and moisture environments.