Multi-response-model confirmation measurement method based on probability box frame

Abstract

A multi-response-model confirmation measurement method based on a probability box frame comprises the steps that corresponding simulating calculation models are established according to relevant theories and input conditions of concrete problems; input parameters of the established corresponding simulating calculation models are divided into random uncertainty, cognition uncertainty and mixture uncertainty, and the parameters are evaluated to determine specific distribution; M*N double-layer sampling is conducted on the models to obtain multi-dimensional samples of M sets of models; mahalanobis distance conversion is conducted on each set of samples to obtain mahalanobis distance samples; the mahalanobis distance sample values are countered; an area measurement interval of two probabilityboxes is calculated. By adopting a probability box method, the simultaneously existing problem of random uncertainty and cognition uncertainty can be effectively described, existing probability modelsand interval number and evidence structures can be directly converted into the forms of probability boxes, uncertainty description conforms to project customs and is easily accepted and used by project workers.

Description

technical field [0001] The invention relates to a measurement method, in particular to a multi-response model confirmation measurement method based on a probability box framework, which is suitable for multi-input multi-output model confirmation measurement. Background technique [0002] In the face of increasingly large and complex engineering systems, it is difficult and expensive to conduct physical experiments on them. In engineering research, researchers try to use increasingly mature computer technology for modeling and simulation calculations, and replace complex and expensive real physical experiments by calculating the built models. However, when establishing a calculation model, the model is usually simplified due to the existing technology and calculation efficiency, or due to the cognitive limitations of the modeler, there is a certain deviation between the built model and the actual model. At this time, the accuracy and credibility of the calculation model will...

AI Technical Summary

Problems solved by technology

However, existing probability box methods address mostly single-response model validation problems

Images