The Effect of Alkaline Treatment on Mechanical Performance of Natural Fibers-Reinforced Plaster: Part II Optimization Comparison between ANN and RSM Statistics

ABSTRACT

The present study is a continuation of a previously published by the authors. In Part I, mechanical data of Alkaline-treated natural fibers-reinforced plaster was examined using response surface methodology (RSM) statistics, while in this work (Part II), the data are analyzed using Artificial neural network (ANN) statistics. Many studies have focused on substituting synthetic fibers by natural fibers in plaster matrix. The present study reports ANN statistics data analysis on the mechanical properties of plaster and natural fibers. ANN and RSM methods are employed and their findings discussed. During this study, flexural properties of sodium hydroxide solution (NaOH)-treated natural fibers in plaster mortars were investigated. Experimental data effects were established by analysis of variance (ANOVA) method. Fiber length optimization and NaOH percentage treatment of fibers were also performed utilizing desirability function DF to obtain maximum flexural properties. Experimental results showed good agreement with those obtained statistically. The ANN and RSM models also have correlated highly to the experimental data. However, the ANN model proves more to be accurate.

摘要

本研究是作者先前发表的一篇论文的继续. 在第一部分中, 使用响应面法 (RSM) 统计分析了碱处理天然纤维增强石膏的力学数据, 而在本工作 (第二部分) 中, 使用人工神经网络 (ANN) 统计分析了数据. 许多研究集中在石膏基质中用天然纤维替代合成纤维. 本研究报告了石膏和天然纤维力学性能的人工神经网络统计数据分析. 采用了人工神经网络和RSM方法, 并对其结果进行了讨论. 在这项研究中, 研究了经氢氧化钠溶液 (NaOH) 处理的石膏砂浆中天然纤维的弯曲性能. 通过方差分析 (ANOVA) 方法建立实验数据效应. 利用期望函数DF优化纤维长度和NaOH百分比处理纤维, 以获得最大弯曲性能. 实验结果与统计结果吻合较好. ANN和RSM模型也与实验数据高度相关. 此外, 优化结果表明, RSM误差在0.79%到3%之间, 而ANN误差在0.10%到1.60%之间.

结果表明, 人工神经网络模型更为准确.

KEYWORDS:

• Natural fibers
• plaster
• artificial neural network
• flexural properties
• optimization

关键词:

• 木棉非织造布
• 天然纤维
• 灰泥
• 人工神经网络
• 弯曲性能
• 优化