Rapid detection of crushed stone aggregate void fraction based on image data and crushed stone aggregate grading and amounts
To address the challenges associated with traditional methods for detecting void ratio in crushed aggregates,such as destructive testing,lack of representativity,manual operation errors,time consumption and discontinuity,this study proposed a novel method for continuous,non-destructive,rapid,and accurate detection of void ratio in crushed aggregates.Imaging equipment was employed to capture surface images of compacted crushed aggregates,and image processing techniques were utilized to extract surface image data of the aggregates.The extracted image information was analyzed to establish an image index system that reflected the surface characteristics of the crushed aggregates.The extracted image indices,along with aggregate quantity and gradation information,were used as input data for predicting the void ratio of the crushed aggregates using a Random Forest model.An indoor compaction simulation test scheme was designed,and 65 sets of void ratio measurements of crushed aggregates with varying quantities and gradations were collected to construct a dataset for training the machine learning model.The results showed that the proposed method achieved an average absolute percentage error of 1.69%,an average absolute error of 0.589,a root mean square error of 0.914,and a correlation coefficient of 0.974 on the test set,with a prediction accuracy of 98.31%.This method enabled continuous,non-destructive,rapid,and accurate real-time detection of void ratio in crushed aggregates.
NETL
NSTL
万方数据
