Field study on unconfined compressive strength and drilling data of DSM columns: An ANN application

Abstract

Numerous limitations in transportation networks, such as the right of way, lack of proper base ground, especially in offshores, or far distances of borrow sites and the corresponding cut and fill costs, have forced construction projects to consider soil improvement methods seriously. Hence, the application of various soil treatment methods has been progressively increased, especially in airport runways. The deep soil mixing (DSM) method is an efficient ground improvement procedure using in-situ soils for treatment plans. In order to ensure the quality of the treated soil, the condition of the DSM elements should be evaluated using continuous coring procedures and unconfined compressive strength (UCS) tests on only 2-4% of all executed columns, based on the recommendations of the FHWA. Therefore, there are significant uncertainties and a lack of information regarding the UCS statues of other about 96% of the untested DSM elements in all projects. In this regard, Artificial Intelligence (AI) and methods based on the artificial neuron network (ANN) can be considered to reduce these limitations and deficiencies. In the current study, the Multilayer Perceptron (MLP) as an ANN-based method was implemented to predict the correlation algorithm between UCS test results and several drilling rig machine parameters such as drilling, lifting, and rotary speeds, pull-down, rotary and hoist pressures, and the mixture flow and pressure. The results of over 150 UCS tests and corresponding machine data of a DSM soil improvement field project in Qeshm island were considered to train and test the MLP algorithm. Finally, the tested algorithm was implemented to predict the UCS results of over 1000 executed DSM columns. In this case, instead of conducting a probability analysis to find out a DSM column UCS for using in design procedure, the simulated strength of each column by the ANN method can be considered.

Keywords: Soil improvement, Deep soil mixing (DSM) method; Field investigation; Unconfined compressive strength (UCS), Artificial neuron network (ANN), Multilayer Perceptron (MLP).