Abstract

The research described here aimed to model the water absorption behavior of reinforced polystyrene (PS) composites developed from powders of plantain peel (PPC) and bamboo fiber (BFC) using the Artificial Neural Network (ANN) model. The composites were developed by manual mixing and hand layup at room temperature (25 ± 2 °C) and cured by open molding at room temperature for 7 days. Water absorption tests were performed according to the ASTM standard method (D1037-99, ASTM, 1999). The water absorption was observed to increase with both filler loading and immersion time for both PPC and BFC. The coefficient of determination (R²) values >0.98 were achieved for training, validation, and testing for both composite types. The model results showed low root mean squared error values (<1 wt%), revealing that in the utilization of the model a high accuracy threshold was expected for the ANN predictions. Parity plots revealed that the models gave a good balance between over-predictions and under-predictions and the accuracy could be substantiated both at low and high water absorption prediction values. ANOVA revealed that the results were statistically significant at a significance level of p < 0.05.

摘要

此处描述的研究旨在使用人工神经网络 (ANN) 模型对由车前草皮 (PPC) 和竹纤维 (BFC) 粉末开发的增强聚苯乙烯 (PS) 复合材料的吸水行为进行建模。复合材料在室温 (25 ± 2 °C) 下通过手动混合和手工铺层开发，并在室温下通过开模固化 7 天。根据 ASTM 标准方法 (D1037-99, ASTM, 1999) 进行吸水率测试。观察到 PPC 和 BFC 的吸水率随着填料用量和浸泡时间的增加而增加。决定系数 ( R ²) 两种复合类型的训练、验证和测试都达到了 >0.98 的值。模型结果显示低均方根误差值 (<1 wt%)，表明在使用该模型时，对于 ANN 预测期望具有高精度阈值。奇偶图显示模型在高预测和低预测之间取得了良好的平衡，并且在低吸水率和高吸水率预测值下都可以证实准确性。方差分析显示结果在p  < 0.05的显着性水平上具有统计学意义。