Research On Geotechnical Engineering And Project Network Planning Based On Computational Intelligence Methods

Computational Intelligence, which includes Neural Network, EvolutionaryComputation, etc., is developing rapidly and used widely in civil engineering in recentyears. In this paper, Genetic Algorithm, Feedforward Neural Network, andSelf-organizing Mapping Neural Network are studied, and their applications infields of geotechnical engineering and project network planning are carried out. Themain content of the paper is as follows:1.The advances and basic theories of GA and ANN are discussed, and theirapplications in geotechnical engineering and project network planning are reviewed inthe meantime.2. The SOM-BP compound neural network model is presented to evaluate slopestability. Firstly, by applying SOM neural network to cluster self-organizingly theinput patterns, the input information of neural network is improved greatly. Then byapplying BP neural network to train the patterns treated by SOM network, thenon-linear associations between the inputs and their corresponding outputs areobtained. Experimental results show that the SOM-BP neural network model forevaluation of slope stability is superior to BP network model.3. Considering that clustering results are sensitive to the original cluster centerand depend on the order of input patterns in the ordinary dynamic clustering method,a new dynamic clustering method for evaluation of slope stability is developed basedon genetic algorithm. By incorporating features of the problems discussed, thecorresponding genetic operators such as selection strategy, crossover operator, andmutation operator are designed to promote global search.4. An evolutionary neural network method for displacement back analysis onsupporting structures of deep foundation pits is proposed by combining geneticalgorithm and neural network. The patterns data is used to train the neural network tofind out the non-linear relationship between deep foundation pits mechanicalparameters and displacement of its supporting structure, then genetic algorithm isadopted to search the optimal mechanical parameters in their global ranges. 5. GA is applied to solve two resource optimization problems in constructionproject network: resource-constrained allocation and unlimited resource leveling.Based on the theorem of genetic algorithm, an integral encoding mechanism isadopted to represent chromosomes which has higher efficiency compared to binaryencoding mechanism;a new repairing operator aiming to repair unfeasiblechromosomes caused by crossover operation is proposed;and a new mutationoperator is designed to make the mutation operation executed in feasible ranges.A genetic algorithm combined with quadratic template is present to solve theconstruction time-cost trade-off problems. A quadratic template is introduced todescribe the nonlinearity relationship between time and direct cost, and then the GAwith the self-adapted mutation operator is used to search optimal time as well as theminimal cost.