Study On Calibration Of CAE Model And Prediction Based On Gaussian Process

In the modern engineering design, computer aided engineering analysis model plays an increasingly important role.However, in view of the actual problem, the CAE model is not easy to establish high degree of simulation.On the one hand, due to the complexity of the actual problem,it is difficult to use CAE simulation model based on single physical mechanism to simulate the actual problem.On the other hand, due to the practical problems inherent randomness and uncertainty in the observations, which leads to the difficulty in simulation and prediction of practical problems based on CAE model with deterministic input parameters.In view of this, in recent years,people began to consider the mechanism model and actual observation data together to simulate the complex practical problems.Gauss model integrated model and data is an effective method for simulation and prediction of the system.In this paper, we use the Gauss model to combine the traditional CAE model with the actual observation data, and give a new and efficient method of CAE analysis.The research shows that the Gauss model can achieve a small amount of data modeling, parameter uncertainty calibration and predict correction.Firstly, the basic principle of the Gauss model is introduced.Gauss model is a model based on Bayesian theory and statistical theory.Gauss model belongs to the non parametric model, which has the advantages of good flexibility, high accuracy and accurate quantification of uncertainty. It is suitable for processing high dimensional data and nonlinear complex problems.In the Gauss model prediction principle at the same time, this paper discusses the influence of selection of covariance function of the Gauss model, and verify the mean function to select the desired effect on Gauss model is very little. Accordingly, the follow-up work in the establishment of the Gauss model, focusing on how to select the appropriate covariance function need.Secondly, this paper constructs a simple example of the mechanical analysis of the cantilever beam to discuss the basic idea and feasibility of the application of Gauss model in structural analysis.In this example, the establishment of the Gauss model using the observation data of Timoshenko beam model and the Bernoulli beam mode to analysis cantilever beam.In order to achieve the purpose of using the Bernoulli beam model to simulate and predict the height of the model.Again, the paper discusses the basic process and key points of the application of Gauss process in general CAE analysis.This paper takes a plastic tensile plate with a circular hole as an example, study the establishment of the Gauss model using the experimental data and the traditional general CAE model, through the method of uncertainty parameters and predicted using the CAE model to replace the traditional.The research shows that the Gauss model is used to replace the traditional CAE model to carry out high precision simulation and prediction.Finally, using the Gauss model to solve the actual problem of the Santiago National Laboratory released the 2014V&V challenge. The whole process of the complex problem is solved by using the Gauss model are presented, including how to establish the Gauss model,how to fix the uncertain parameters and predict the output response, and how the Gauss model to predict the failure probability of the container and other details.