The Study On Optimal Selection Of Foundation Treatment Programs For Buildings In Collapsible Loess Region

With the rapid development of China’s construction industry, it has become aninevitable tendency to build various buildings in collapsible loess region. Selecting abetter foundation treatment program plays a vital role for an engineering project. Theselection of foundation treatment program of buildings in collapsible loess region is abroad and comprehensive work which involves a lot of uncertain information, affectedby treatment effect and other factors, such as the construction period, construction costand the environment problems. It has become an essential and urgent puzzle how toselect foundation treatment programs of planned buildings optimally taking thesefactors into consideration, which has strong research significance.The selection of foundation treatment programs is not only a technical issue, butalso a decision-making one which contains multiple targets. As a result, the optimalselection of foundation treatment programs in collapsible loess region needscomprehensive assessment which includes two steps: the selection of evaluation indexand setting weights. So this article includes the following achievements:（1） Introduce the elementary knowledge of collapsible loess and severalfoundation treatment methods which have been widely used. Combined with theprinciple of evaluation index selection, this paper determines the evaluation indexeswhich influence the optimal selection of foundation treatment programs in collapsibleloess regions.（2） Elaborate on the basic theory of Artificial Neural Network and thefundamental principle of traditional BP Neural Network. And analyze the feasibility that BP neural network model is applied for solving the problem of selecting foundationtreatment programs. This article applies for BP neural network to calculate the weightsof each index. In view of the shortcomings of both traditional BP algorithm andimproved algorithms, transfer function, error function and the BP algorithm have beenimproved in this paper based on them and then deduced the BP algorithm in detail.（3） By MATLAB language and the Neural network toolboxes, this paper buildsthe improved BP network model and the traditional BP network model respectively.Collect30sets of successful foundation treatment program data from the new area ofTongchuan and regard them as training samples and testing samples. Then train and testthe two models mentioned above. Learn from the training result and testing result bycomparing with each other that the improved BP network model is superior to thetraditional BP network model at convergence speed and accuracy.（4） Use the improved BP network model which has been trained to predictfoundation treatment programs of two buildings in planning. By designing them indetail, then put the foundation treatment programs into practice. When accomplishingthe construction of foundation treatment, test and monitor bearing capacity of compositefoundation, bearing capacity of a single pile, whether the collapsibility of loess has beeneliminated, sedimentation of the buildings and so on. From the testing results, the effectof these two foundation treatment programs meets the requirements of the buildings’normal use and shows good result.