Characteristics estimation of natural fibre reinforced plastic composites using deep multi-layer perceptron (MLP) technique Articles uri icon

authors

  • Sathish, T.
  • Sunagar, Prashant
  • Singh, Vikash
  • Boopathi, Sampath
  • sathyamurthy, Ravishankar
  • Al Enizi, Abdullah M.
  • PANDIT, BIDHAN
  • Gupta, Manish
  • Sehgal, Satbir S.

publication date

  • October 2023

volume

  • 337

International Standard Serial Number (ISSN)

  • 0045-6535

Electronic International Standard Serial Number (EISSN)

  • 1879-1298

abstract

  • Polymer Matrix Composite (PMC/Plastic Composite) often referred to as Plastic Composite with Natural fibre reinforcement has a huge interest in industries to manufacture components for various applications including medical, transportation, sports equipment etc. In the universe, different types of natural fibres are available which can be used for the reinforcement in PMC/Plastic Composite. So, the selection of appropriate fibre for the PMC/Plastic Composite/Plastic composite is a challenging task, but it can be done using an effective metaheuristic or optimization techniques. But in this type of optimal reinforcement fibre or matrix material selection, the optimization is formulated based on any one of the parameters of the composition. Hence to analyse the various parameter of any PMC/Plastic Composite/Plastic Composite without real manufacturing, a machine learning technique is recommended. The conventional simple or single-layer machine learning techniques were not sufficient to emulate the exact real-time performance of the PMC/Plastic Composite. Thus, a deep multi-layer perceptron (Deep MLP) algorithm is proposed to analyse the various parameter of PMC/Plastic Composite with natural fibre reinforcement. In the proposed technique the MLP is modified by including around 50 hidden layers to enhance its performance. In every hidden layer, the basis function is evaluated and subsequently, the sigmodal function-based activation is calculated. The proposed Deep MLP is utilized to evaluate the various parameters of PMC/Plastic Composite Tensile Strength, Tensile Modulus, Flexural Yield Strength, Flexural Yield Modulus, Young's Modulus, Elastic Modulus and Density. Then the obtained parameter is compared with the actual value and the performance of the proposed Deep MLP is evaluated based on the accuracy, precision, and recall. The proposed Deep MLP attained 87.2%, 87.18%, and 87.22% of accuracy, precision, and recall. Ultimately the proposed system proves that the proposed Deep MLP can perform better for the prediction of various parameters of PMC/Plastic Composite with natural fibre reinforcement.

keywords

  • deep learning; deep mlp; multi-layer perceptron; natural fibre reinforcement; parameter prediction; plastic composite