back-propagation algorithm

Resilient modulus (Mr) of subgrade soils is considered as one of the most important factors for designing flexible pavements using empirical methods as well as mechanistic-empirical methods. The resilient modulus is commonly measured by a dynamic triaxial loading test, which is complex and expensive. In this research, back-propagation artificial neural network method has been employed to model the resilient modulus of clayey subgrade soils based on the results of the cone penetration test. The prediction of the resilient modulus of clayey subgrade soil can be possible through the developed neural network based on the parameters of the cone tip resistance (qc), sleeve friction (fs), moisture content (w), and dry density (γd). The results of the present study show that the coefficients of determination (R2) for training and testing sets are 0.9837 and 0.9757, respectively. According to the sensitivity analysis results, the moisture content is the least important parameter to predict the resilient modulus of clayey subgrade soils, while the importance of other parameters is almost the same. In this study, the effect of different parameters on the resilient modulus of clayey subgrade soil was evaluated using parametric analysis and it was found that with increasing the cone tip resistance (qc), the sleeve friction (fs) and the dry density (γd) and also with decreasing the moisture content (w) of soils, the resilient modulus of clayey subgrade soils increases.

Artifitial neural network (ANN) is an information processing system that is formed by a large number of simple processing elements, known as artificial nerves. It is formed by a number of nodes and weights connecting the nodes. Using the trained data, the designed ANN can be adjusted in an iterative procedure to determine optimal parameters of ANN. Then for an unknown input, we can compute corresponding output using the trained ANN. There are many methods for training the network and modifications of the weights. One of the most famous and simplest methods is a back-propagation algorithm that trains the network in two stages: Feed-forward and feed-backward. In the feed-forward process, the input parameters are moved to the output layer. In this stage, the output parameters the next stage is done In this study, a 3-layer perceptron neural network was used with 28 neurons in a hidden layer for modeling the eastern component (VE) and 27 neurons in a hidden layer for modeling the northern component (VN) velocity field of the earth's crust in Iran. The minimum relative error obtained from this evaluation for the eastern component was -3.57% and for the northern component was +0.16%: also the maximum relative error for the eastern component was +38.1 % and for the northern component was +95.3%. In this study, a polynomial of degree 5 with 18 coefficients was used to model the east and north components for the evaluation of artificial neural networks in estimating the velocity rate of geodetic points. A comparison of the relative error from the polynomial model and the relative error from the neural network illustrated the superiority of the neural model with respect to the polynomial model in this region.

Although CO2 injection is one of the most common methods in enhanced oil recovery, it could alter fluid properties of oil and cause some problems such as asphaltene precipitation. The maximum amount of asphaltene precipitation occurs near the fluid pressure and concentration saturation. According to the description of asphaltene deposition onset, the bubble point pressure has a very special importance in optimization of the miscible CO2 injection. The purpose of this research is to predict the onset of asphaltene and bubble point pressure of fluid reservoir using artificial intelligence developed models including a software simulator called “Intelligent Proxy Simulator (IPS)” based on structure artificial neural networks and “adaptive neural fuzzy inference system”, which is a combination of fuzzy logic and neural networks. To evaluate the predictions by artificial intelligence networks at the onset of deposition, a solid model using Winprop software was employed. Standing correlations were used for comparison of bubble point pressure. The results obtained using artificial intelligence models in prediction of the onset of asphaltene deposition and bubble point pressure during injection of CO2 were more accurate than those obtained from the thermodynamics Solid model and the Standing correlation respectively.