Research On Data-driven Health Prediction Technology Of Radar

As a representative electronic system,radar is becoming more and more complex and integrated.Radar's characteristics of working continuously and reliably make it necessary that the users need to be alarmed to eliminate hidden troubles in time when its performance degraded.Therefore,it is necessary to implement health management for radar,improve the open health management system architecture of radar,establish a storage and excavator system and conduct research on health prediction technology applicable to radar.The main contents of the thesis and the conclusions are as follows:1.In view of the imperfect system architecture of radar health management system and the irregular development in engineering application,the overall architecture of an open radar health management system was summarized,including three levels:system level,subsystem level and module level.2.Since there is no effective design consideration for the collection,transmission,storage and processing of health information in the engineering application,in order to support the efficient execution of functions such as prediction,the data storage architecture in the health management system architecture was optimized and perfected.Distributed information collection,hierarchical processing and transmission,centralized storage and mining mechanisms for health information were established.3.The system-level,subsystem-level,module-level and parameter-level evaluation indicators of radar performance were combed,and some characteristics of radar performance parameter sequences were analyzed.For the prediction of radar performance parameter sequences,chaotic theory was used to analyze the predictability of the performance parameter sequences.It turns out that some of the monitoring sequences are predictable.4.For the problem that the statistical state of the performance parameter monitoring sequences change with time when the health status and working conditions of the radar system differ at different times,an adaptive prediction algorithm was proposed,and two types of models with different orders and parameters were used to model and predict.Simple and efficient models are chosen according to the proposed criteria.Multi-threaded parallel strategy was put forward in engineering applications.5.The structure of radar is complex,and the degradation of radar's performance is caused by unrelated failure factors,which makes it difficult for a single model to predict effectively.The idea of distinguishing failure factors and modeling them respectively was proposed.The linear and nonlinear models were employed to model different components.The optimal split model of failure factors was established and the polynomial coefficient autoregressive model was improved to model and predict different sequence components,which effectively reduces the modeling bias.6.The application of the prediction function as a plug-in in the open health management system is realized by using MATLAB and Qt hybrid programming method.