Artificial Neural Network and Nonlinear Models for Gelling Time and Maximum Curing Temperature Rise in Polymer Grouts

Abstract

In this study, the effects of initial temperature, catalyst content, and activator content on the gelling time and maximum curing temperature of two polymeric grouts were investigated. Because the grouts are used in various environmental conditions, the grouts were investigated at three different initial temperatures (4.4, 15.7, and 26.7°C, or 40, 60, and 80°F). The catalyst and activator contents varied from 0.5% to 3% of the total weight of grout mix. Gelling times for the polymer grouts were measured, and the curing temperatures for the mixtures were monitored during and after the gelling process. The gelling time reduced with increased catalyst and activator contents and increased initial temperature for both the polymer grouts. One grout exhibited shorter gelling time and lower temperature rise during the gelling process than the other grout. Both artificial neural network (ANN) and nonlinear relationship (NLR) models were used to predict the observed grout behaviors. For each grout, a total of 48 tests were performed and data from 32 tests (2/3 of the total data) were used for training and calibrating the two models. The data from the 16 remaining tests were used to verify the model predictions. Various ANN architectures with numbers of hidden layers and neurons were studied to determine the optimum network architecture for the current study. The optimum network required two hidden layers with four nodes in each layer. The NLR model and the ANN model predicted the gelling time and maximum curing temperature rises very well, and the coefficient of determination (R2) was in the range of 0.91 to 0.97.