Research On Model Validation Method For Small Sample And Service-oriented Tools

Simulation model validation,one of the key issues in simulation filed,is an important means to ensure that simulation systems can be used to replace the real systems.In actual engineering applications,we can obtain a small sample of real-system output with the limited of factors such as test conditions,cycles,and expenses,which has brought difficulties to the validation of complex simulation system.Therefore,it has practical significance to study the model validation problem with small sample.In addition,with the development of model validation theory and service-oriented technology,model validation has emerged with some new demands.Thus,it is necessary to develop a service-oriented network model validation platform that supports reusable,scalable,and collaboration.Furthermore,the platform not only supports users to obtain model validation services on demand,but also supports users to publish their own service functions on the platform,which can realize the sharing of model validation services.Firstly,a validation framework for small sample data models based on statistical analysis is presented.If the sample obeys non-normal distribution,validation is performed using a non-parametric test method.If the sample obeys normal distribution,the appropriate parameter test method for two populations is selected according to the observed sample capacity.For observed sample with small size which less than or equal to 10,a method based on Bootstrap and Bayes parameter estimation is proposed.Bootstrap method is used to expand the observed sample to get the resamples.Then the hyper-parameters of Bayes prior distribution are estimated by regenerating samples.Next,Bayes method is used to estimate the mean and variance of observed samples.At last,the parameter test for single population is used to validate the model.Moreover,an improved Bootstrap method is proposed to solve the problem that the traditional Bootstrap method has a limited range of regenerating samples and is easy to deviate from the real distribution.The numerical examples show that the improved Bootstrap method is correct and the validity of the proposed model verification method for small sample.Secondly,the key technologies of service resource description and service composition based on SOA model validation are studied.In order to improve the retrieval efficiency of service resources and reduce the redundancy of resource storage,a service resource description method based on metadata and XML is proposed,which is used to describe the basic attributes and functional attributes.Furthermore,in order to reduce the coupling between various functional modules,realize the reusable of algorithm and provide a good user interface,a service composition scheme based on graph component is proposed.Also,in order to choose the combination service,a service selection mechanism based on the model validation service quality is adopted.At last,a specific example is used to verify the feasibility of the service resource description method and service composition scheme.Finally,a service-oriented simulation model validation tool is designed and developed.First of all,the requirement analysis of the tool is given,the architecture,function structure and database of the tool are designed.Then,the user management,service management,project management and model validation module are designed in detail.Developing service-oriented model validation tool functionality using the Qt Creator environment.The tool is applied to verify a simulation model and test the effectiveness and practicability of the service-oriented model validation tool.