Practical Intelligent Compaction with Modest Accelerometers: Real Time Soil Stiffness via Roller Accelerometer Vibration Analysis
Kwame Edjah, Laith Sadik, Sara Khoshnevisan, Mehdi Norouzi, Kaveh Bastani, Lei Wang, Nayyar SiddikiTraditional compaction assessment methods for soils and geomaterials offer only localized point measurements, making continuous spatial assessment inefficient and time-consuming. Intelligent compaction (IC) technologies have been introduced to address this limitation; however, many IC indicators still lack reliable standardized correlations with actual soil stiffness across different compaction conditions. To address this gap, this study introduces a novel real time ensemble modeling approach that integrates vibration data from roller-mounted accelerometers with in situ lightweight deflectometer measurements. This approach uses frequency domain analysis by introducing total harmonic distortion with subharmonics alongside the established compaction control value to capture harmonic and subharmonic response components that may be missed by conventional IC indicators. Because soil–roller interaction is nonlinear across compaction states, separate linear regression models are developed for empirically classified low- and high-compaction zones. The framework is developed and validated using No. 53 aggregate field data from two active highway construction sites in Indiana, US, with validation coefficient of determination values from 0.67–0.78. The combined data follow the nonlinear trend expected from soil mechanics, providing preliminary support that the framework captures a physically meaningful response rather than only a site-specific pattern. However, each compaction zone is represented by data from a single site; therefore, the observed response may partly reflect site conditions and roller dynamics in addition to compaction state. Therefore, the regression coefficients should be interpreted as material-, site-, and equipment-specific. Cross-site replication within each compaction zone and recalibration for other geomaterials, rollers, and sensor configurations remain necessary before broader application.