Scholarlink

<p>This study explores predictive modeling of mechanical properties tensile strength, flexural strength, impact strength, and hardness of natural fiber and filler cashew nutshell waste, sugarcane waste, and polyethylene terephthalate (PET) waste was used as fillers composite materials based on advanced machine learning algorithms. The experiment composition weigth percentages (0%, 5%, 10%, and 15%) were obtained through the literature and intermediate and longer compositions (1%–16%) were approximated using Artificial Neural Network (ANN) and Support Vector Regression (SVR) models. The performance of every algorithm was compared based on statistical measures such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and the coefficient of determination (R<sup>2</sup>). The ANN model exhibited better prediction performance with R<sup>2</sup> values greater than 0.99 in every property, with the lowest error rates, representing high reliability in interpolation as well as extrapolation. SVR also worked satisfactorily, albeit with marginally increased deviations in calculated values at some composition ranges. The work establishes machine learning models specifically ANN as an effective means of simulating composite materials’ mechanical behavior, and an effective method of material design optimization that can be done with less experimental labor.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A comparative analysis on usage of ANN and SVR algorithms for predicting the mechanical properties of natural fiber-based composites using experimental data

  • 期刊论文

  • 发表日期：

    2025年8月12日

查看全文
  • R. Alagulakshmi,
  • R. Ramalakshmi,
  • Arumugaprabu Veerasimman,
  • Geetha palani

摘要

This study explores predictive modeling of mechanical properties tensile strength, flexural strength, impact strength, and hardness of natural fiber and filler cashew nutshell waste, sugarcane waste, and polyethylene terephthalate (PET) waste was used as fillers composite materials based on advanced machine learning algorithms. The experiment composition weigth percentages (0%, 5%, 10%, and 15%) were obtained through the literature and intermediate and longer compositions (1%–16%) were approximated using Artificial Neural Network (ANN) and Support Vector Regression (SVR) models. The performance of every algorithm was compared based on statistical measures such as Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and the coefficient of determination (R2). The ANN model exhibited better prediction performance with R2 values greater than 0.99 in every property, with the lowest error rates, representing high reliability in interpolation as well as extrapolation. SVR also worked satisfactorily, albeit with marginally increased deviations in calculated values at some composition ranges. The work establishes machine learning models specifically ANN as an effective means of simulating composite materials’ mechanical behavior, and an effective method of material design optimization that can be done with less experimental labor.