Scholarlink

It is becoming popular to replace destructive laboratory tests with related to non-destructive testing (NDT) techniques. Unfortunately, not all material properties can be determined accurately enough in this way. Supporting non-destructive methods with artificial intelligence (AI) offers potential for groundbreaking development in this area. A common approach in the construction industry is to determine compressive strength using, for example, the Schmidt Hammer, ultrasonic wave velocity (UPV), composite composition information, and machine learning (ML). The determination of pull-off strength can be approached similarly. In this work, presented ML model can be used to predict the pull-off strength of resin coatings containing granite powder and flax fibers. To obtain satisfactory results, selected ML algorithms were analyzed on a database consisting of 140 sets of values of parameters containing information about the composition of the resin coating. Metrics indicating high performance (R = 0.885; RMSE = 0.138 MPa; MAPE = 3.72%) were achieved by a model based on the random forest (RF) algorithm containing 160 trees with a depth of 10 nodes. A comparison of the predicted fb pull-off strength to that determined by in situ testing has been developed. The results obtained suggest that the use of AI to determine the fb of resin coatings is promising alternative.

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

Comparison of Models Predicting the Tensile Strength of Epoxy Resin Floors Modified with Granite Powder and Flax Fibers

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Mateusz Moj,
  • Łukasz Kampa,
  • Sławomir Czarnecki

摘要

It is becoming popular to replace destructive laboratory tests with related to non-destructive testing (NDT) techniques. Unfortunately, not all material properties can be determined accurately enough in this way. Supporting non-destructive methods with artificial intelligence (AI) offers potential for groundbreaking development in this area. A common approach in the construction industry is to determine compressive strength using, for example, the Schmidt Hammer, ultrasonic wave velocity (UPV), composite composition information, and machine learning (ML). The determination of pull-off strength can be approached similarly. In this work, presented ML model can be used to predict the pull-off strength of resin coatings containing granite powder and flax fibers. To obtain satisfactory results, selected ML algorithms were analyzed on a database consisting of 140 sets of values of parameters containing information about the composition of the resin coating. Metrics indicating high performance (R = 0.885; RMSE = 0.138 MPa; MAPE = 3.72%) were achieved by a model based on the random forest (RF) algorithm containing 160 trees with a depth of 10 nodes. A comparison of the predicted fb pull-off strength to that determined by in situ testing has been developed. The results obtained suggest that the use of AI to determine the fb of resin coatings is promising alternative.