Evaluation of Physical Properties and Time-Series Prediction of Long-Term Durability for Improved Ground by Ultrafine Particle Grouting Materials

In Japan, lots of cities are suffering from land subsidence problems, resulting from liquefaction due to its soft ground constitution. Among the various preventive and/or mitigation measures practiced for combating it, the chemical injection method is commonly selected for it with the intention of increasing the strength of the ground. However, the limitation of this technology lies in the scarcity of data on the probable influence and effectiveness of the sprayed chemicals over the long-term period into the surrounding grounds. It is necessary to study in detail to ensure that incidents like chemical accidents do not occur in the future due to the failure in strength over the long-term period. Therefore, the objective of this study will be the evaluation of ultrafine particle grouting materials properties in terms of durability over the long-term period by using the prediction method. The results generated by these prediction methods are verified by comparing them with the test results. In this study, the prediction is made by the Autoregressive integrated moving average (ARIMA) model, machine learning predictive model (MLPM), and state-space representation model (SSRM). For the lower input data amount, the results generated by the ARIMA model can be of higher precision comparatively, but the margin of error is still high. Thus, in order to conclude on the best model, a higher degree of research is required to identify all other influencing parameters and comparative studies of other methods.