Incorporating distributed fibre optic sensing and machine learning into dry deep mixing-PPT Presentation

PPT Presentation

Pdf Summary

The presentation explores how distributed fibre optic strain (DFOS) sensing and machine learning (ML) could improve dry deep mixing (DDM), addressing two key challenges: measuring field performance and predicting the properties of DDM-treated soil.<br /><br />Laboratory and field work investigated DFOS monitoring. Laboratory tests revealed a complex relationship between strain in the fibre and strain in the surrounding soil, while field installation proved challenging. Field measurements also tracked strain over time and depth.<br /><br />For predicting DDM soil strength, the study compared random forest and gradient boosting models using training and test data with 10-fold cross-validation. Both ML approaches outperformed traditional strength-prediction methods, even without hyperparameter tuning.<br /><br />The findings suggest that combining high-resolution DFOS measurements with ML could support better assessment and prediction of DDM performance, helping optimize future projects, reduce natural-resource use, and lower costs. Further work is ongoing.

Keywords

distributed fibre optic strain

DFOS

machine learning

dry deep mixing

DDM

soil strength prediction

random forest

gradient boosting

strain monitoring

field performance