Intelligent compaction quality evaluation based on multi-domain analysis and artificial neural network

Abstract

Intelligent compaction technology monitors the compaction quality of pavement materials in real-time by observing the sensor signals installed on the roller. At present, the existing intelligent compaction detection methods cannot always accurately evaluate the comprehensive compaction quality of soil from shallow layer to deep layer. In addition, the time domain features of the signal and the frequency domain features of the power spectrum have never been considered as an observation values to evaluate the soil compaction quality. In this paper, we combine these signal features with an artificial neural network to accurately evaluate the overall soil compaction quality. First, to achieve the minimum amount of data required by an artificial neural network, each original signal is split into one hundred pieces. Then the time-domain features and frequency-domain features of the signal fragment in the power spectrum is calculated and the correlation between these characteristics and the comprehensive compaction quality is analyzed. Finally, the eight features with the best correlation are used as the input of an artificial neural network to build a nonlinear model. The test results show that the nonlinear model can accurately classify the overall compaction quality of soil into three categories: under compaction, the best compaction, and over compaction.