Audio-Based Detection of Liquid Content in Concrete Mixtures

Noise generated during concrete mixing constitutes one of the primary environmental pollutants in the construction industry. However, the noise itself could be used as a potential indicator for concrete quality control. An intelligent method was proposed in this paper for detection of liquid content in concrete mixtures based on acoustic analysis of the noise. Through in-depth investigation of the noise, comprehensive features were extracted from multi-domains under different status of liquid content. These features, which were fused to form a discriminative feature set, reveal the correlation between acoustic characteristics of the noise and liquid content. A LightGBM-based classification model optimized by Bayesian hyperparameter tuning was developed to serve as an improved detector of liquid content. Experimental results demonstrated that the proposed method had better performance compared with traditional methods.