Application of Soft Computing Techniques and Multiple Regression Models for CBR prediction of Soils


The techniques of soft computing technique such as Artificial Neutral Network (ANN) have improved the predicting capability and have actually discovered application in Geotechnical engineering. The aim of this research is to utilize the soft computing technique and Multiple Regression Models (MLR) for forecasting the California bearing ratio ) CBR( of soil from its index properties. The indicator of CBR for soil could be predicted from various soils characterizing parameters with the assist of MLR and ANN methods. The data base that collected from the laboratory by conducting tests on 86 soil samples that gathered from different projects in Basrah districts. Data gained from the experimental result were used in the regression models and soft computing techniques by using artificial neural network. The liquid limit, plastic index , modified compaction test and the CBR test have been determined. In this work, different ANN and MLR models were formulated with the different collection of inputs to be able to recognize their significance in the prediction of CBR. The strengths of the models that were developed been examined in terms of regression coefficient (R2), relative error (RE%) and mean square error (MSE) values. From the results of this paper, it absolutely was noticed that all the proposed ANN models perform better than that of MLR model. In a specific ANN model with all input parameters reveals better outcomes than other ANN models.


: MLR, OMC, LL, PI, CBR, ANN, Prediction, Soils