Scholarlink

Statistical regression as well as machine learning models for predicting compressive strength of conventional concrete reinforced with steel reinforcement, whilst making use of nondestructive techniques (NDT) and correction factors, are abundantly available for correlating the NDT data (rebound hammer number) with compressive strength. However, these generalized models are not suitable for concrete with glass fiber reinforced polymer bars (GFRPC). With the growth in usage of GFRPC, there is an ardent need to develop new empirical relationships to determine its compressive strength for the commonly used concrete grades (30, 35, 40 MPa) from rebound number data. The experimental program described in this paper develops a new perceptron of a machine learning nonlinear technique for this purpose. Data obtained from experimental work were analyzed through the MATLAB learner classification application over 35 models. The support vector mechanism (SVM) performed well in deriving the compressive strength of GFRPC from rebound number data. The equation of functional margin for training the data was derived using Beta and bias values extracted from the SVM classifier. After classification, the refined data of rebound hammer readings were reintroduced into the regression problem and an improved correlation equation was then developed to ascertain the precise values of compressive strength of GFRPC. The associated errors have reduced considerably.

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

Machine Learning Driven SVM Algorithms to Evaluate Compressive Strength of Concrete Reinforced with Glass Fiber Reinforced Polymer Bars Using Non-destructive Technique

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Bharti Tekwani,
  • Archana Bohra Gupta

摘要

Statistical regression as well as machine learning models for predicting compressive strength of conventional concrete reinforced with steel reinforcement, whilst making use of nondestructive techniques (NDT) and correction factors, are abundantly available for correlating the NDT data (rebound hammer number) with compressive strength. However, these generalized models are not suitable for concrete with glass fiber reinforced polymer bars (GFRPC). With the growth in usage of GFRPC, there is an ardent need to develop new empirical relationships to determine its compressive strength for the commonly used concrete grades (30, 35, 40 MPa) from rebound number data. The experimental program described in this paper develops a new perceptron of a machine learning nonlinear technique for this purpose. Data obtained from experimental work were analyzed through the MATLAB learner classification application over 35 models. The support vector mechanism (SVM) performed well in deriving the compressive strength of GFRPC from rebound number data. The equation of functional margin for training the data was derived using Beta and bias values extracted from the SVM classifier. After classification, the refined data of rebound hammer readings were reintroduced into the regression problem and an improved correlation equation was then developed to ascertain the precise values of compressive strength of GFRPC. The associated errors have reduced considerably.