Jet grouting induces a complex interaction between soil and injected fluids, and thus the properties of columns are dictated by a combination of the two systems. Aiming to improve the efficiency of projects and optimize execution, past research has focused on the prediction of the column properties understanding the mechanisms underlaying treatment execution. For the complexity of phenomena and the uncertain determination of soil properties, the question can be only partially addressed on the theoretical level, being important answers left to the empiricism of field trials, i.e. full scale experiments carried out to test specific jet grouting solutions on specific sites. The present paper reports the results of a field experiment whose peculiarity consists in being conceived to investigate the role of technology on a wider spectrum. Single and double fluid injection systems with various parameters have been simultaneously performed on a subsoil characterized by in situ tests. Columns have then been discovered to measure their diameter and samples of cemented material have been cored and subjected to uniaxial compressive tests. Results are herein summarized and compared with literature solutions to point out strength and deficiencies of currently adopted conceptual models.

In the paper, an indirect method for the identification of the final shape of the freshly executed jet-grouted column is developed. The method relies on the backward analysis of the temperatures measured inside the column, along the trace of the injecting pipe. Temperature changes in the column are caused by the hydration process of the cementitious grout. 2D axisymmetric unsteady heat conduction initial-boundary value problem is solved for finding the column shape which fits best the reference temperature measurements. The model of the column is solved using the finite element method. The search is performed using the global evolutionary optimization algorithm called differential evolution. It is shown that the proposed method can provide an accurate prediction of the column shape if only the model reflects the physical reality well. The advantage over previous results is that the cylindrical shape of the column does not have to be assumed anymore, and the full profile of the column along its length can be accurately identified.