Data Fusion Reliability Evaluation Model Based On The Statistical Learning

With the rapid development of electronic devices,data processing and network technology,people’s requirements for product reliability are increasing day by day.Relying only on the processing of a single data source,the reliability assessment accuracy of complex new equipment may not meet the requirements.Multi-source data fusion technology has gradually become an effective method for information synthesis.This technology can increase the confidence of data and improve the accuracy of system reliability assessment,and has been widely used in industry and military.In order to solve the problem that the estimation accuracy is not high when only a single data source is used for reliability assessment,this paper aims to establish a reliability assessment model of data statistical fusion suitable for various situations.For large samples,kernel density estimation and regression analysis are used to determine the density function,and the weighting method is used to weight the density function to obtain a composite density function,and the inverse function of the distribution function is used to simulate and generate random samples.In the case of small samples,the Bootstrap method is used to estimate the distribution characteristic parameters,and data-level fusion is performed by parameter weighting.Multidimensional data are fused with the help of multidimensional kernel density estimation.And a reliability analysis model is established based on the fusion algorithm.Regarding the fusion of large sample semi-physical simulation test data and small sample field test data,the high-confidence field test data is used as the target domain,the semi-physical simulation test data is used as the source domain,and the principal component analysis method is used to extract the data respectively.feature,and embed the feature distribution into the Grassmann manifold space as two points on it,construct a geodesic flow core curve between the two points,and map the source and target domains to a certain point through the geodesic flow core curve.A public space to successfully realize data fusion.In order to prevent the field test data from being overwhelmed by the hardware-in-the-loop simulation test data during reliability assessment,expert scoring and the smote algorithm are used to expand the mapped field test data.The field test data is used to establish a reliability evaluation model.The method solves the problems that the reliable information of the small sample field test data is easily distorted,the distribution type is difficult to determine,and the distribution parameter estimation is difficult.The verification results of an example show that the fusion result of the algorithm proposed in this paper is closer to the real reliability of the product than the reliability evaluation using only a single data source,and the evaluation of the fusion of multi-source data information is more comprehensive and objective.