The Application Of Genetic Programming On Data Process For Deformation Monitoring

It is an important means to analyze and research the deformation observation data to master the working character of building, the state degree of building also can be judged according to the observation material. It is a key problem in deformation monitoring that how to make full use of holding information to judge the deformation of building objectively and truly. There are some shortages of traditional mathematical models in application. The Genetic Programming model was studied, the calculating method of this model was embedded analyzed and discussed, and the improved measures were advanced, then were used in the data processing for deformation monitoring. The main contents of this paper are as follows:1. Parameters impacted on the results of the GP model were embedded analyzed. According to the testing to different dams and projects, different probabilities of reproduction, crossover and mutation are obtained.2. The methods to create a parse tree were particularly studied, and the shortages of "the method of growth" and "the method of full" were analyzed. The equal gradient method was studied. The example indicates, the chromosomes created by this method are distributed symmetrically, and could cover with the space of questions better.3. The methods to enhance the capability of initial population were embedded improved. The present methods in common use were studied. Improved optimal and sorting method was advanced. Comparing with other methods, it's better to enhance the capability of initial population.4. The editing operator was particularly studied. On the base of the common editing rules, and the editing operator based on a child tree was discussed. And it was proved that this method could weaken the redundancy of a parse tree.5. The practicability of GP model was analyzed by some examples. Applied the GP model, the Regression model and the ANN model on the processing of dam observation data separately, the monitoring model for each point is obtained, and prediction and anlysis are done by these models. By comparing the results of different models with the observations, it is proved that the GP model has better precision and reliability, and possesses the higher practically applicable value.