Time-Series Prediction of Long-Term Sustainability of Grounds Improved by Chemical Grouting

In the field of geotechnical engineering, the problems of liquefaction and land subsidence are of major concern. In order to mitigate or prevent damage from liquefaction, the chemical injection method is actively used as one of the countermeasures for ground improvement. However, a complete understanding of the long-term sustainability of improved grounds is still unavailable due to a lack of knowledge of the influencing parameters. Thus, the chances of chemical injection accidents cannot be ruled out. In this study, the compressive strength of improved grounds by the granulated blast furnace slag (GBFS), one of the grouting materials used in the chemical injection method, was evaluated and used for a time-series prediction of long-term sustainability. The objective of this study was to evaluate the accuracy and validity of the prediction method by comparing the prediction results with the test results. The study was conducted for three different models, namely, the autoregressive integrated moving average (ARIMA) model, the state-space representation (SSR) model, and the machine learning predictive (MLP) model. The MLP model produced the most reliable results for the prediction of long-term data when the input information was sufficient. However, when the input data were scarce, the SSR model produced more reliable results overall. Meanwhile, the ARIMA model generated the highest degree of errors, although it produced the best results compared to the other models depending on the criteria. It is advised that studies should be continued in order to identify the parameters that can affect the long-term sustainability of improved grounds and to simulate various other models to determine the best model to be used in all situations. However, this study can be used as a reference for the selection of the best prediction model for similar patterned input data, in which remarkable changes are observed only at the beginning and become negligible at the end.