Assessment of Concrete Slump Through Deep Learning: A Novel Approach for Enhanced Accuracy and Reliability

The assessment of concrete slump is a component in the evaluation of concrete quality. However, the availability of methods for real-time detection of concrete slump is conspicuously limited in current practice. This study introduces an innovative methodology for the real-time assessment of concrete slump, employing video perception algorithms coupled with a specially designed concrete retention bucket detection system. The mechanism involves the introduction of mixed concrete into a funnel-shaped apparatus, followed by the use of a camera synchronized with a sophisticated deep learning network to continuously monitor the slump of the concrete within the bucket. Empirical results from this investigation affirm the effectiveness of the proposed detection method in accurately classifying concrete slump, evidenced by a classification accuracy of 90.90%. The methodology introduced in this study pioneers the utilization of video stream data for the real-time monitoring of concrete slump variations, setting a new standard for classification accuracy and reliability in this field.