Research On Electromagnetic Situation Analysis Technology Based On Knowledge Transfer Method

As research on the electromagnetic situation and knowledge transfer deepens,people realize that combining the two can lead to more achievements in the electromagnetic field.In the fields of electromagnetic target threat estimation and aircraft intent recognition,knowledge transfer can be used to address problems such as Target Threat Estimation and intent recognition.Therefore,knowledge transfer plays an important role in promoting the development of electromagnetic target threat estimation and aircraft intent recognition,and has significant research significance.(1)A knowledge transfer-based electromagnetic target threat estimation algorithm(KT-IPSO-BP)is proposed.To address the issue of the random initialization of weight and bias in the BP network.To begin with,identify the configurable parameters of the BP neural network,such as learning rate,initialization of weights and biases,and so on.Next,by utilizing knowledge transfer to learn the IPSO optimization method,which enables to encode the adjustable parameters of the BP network into the positions of particles.with the position range being the tunable parameter range of the BP network.Finally,the optimal particle position is decoded into the tunable parameters of the BP network,and the parameters of the BP network are updated,thus creating the KT-IPSO-BP network model.This model has optimal network parameters.By using this model for threat estimation,the best estimation result can be obtained by minimizing the sum of squares of the difference between the predicted output of the sample and the actual output.(2)A threat estimation algorithm supporting human intervention in the loop is proposed.Based on the(1)model,an electromagnetic target threat estimation prediction model(MAMETE)is proposed by analyzing the attributes of the electromagnetic target.The MAMETE model quantifies each attribute of the electromagnetic target mathematically to obtain a quantitative value,and then assigns weights to each attribute to obtain the threat estimation value of the electromagnetic target by weighting the quantitative value.By weighting the MAMETE model and the KT-IPSO-BP model,the proportion of weights between the two models can be adjusted to obtain the final electromagnetic target threat value,enabling human intervention in the loop.The two models are integrated and a system interface is developed to facilitate the intuitive display of electromagnetic target threat estimation values.(3)A knowledge transfer based intent recognition(KT-TLANet)network model is proposed.Using Transformer and LSTM functional structure through knowledge transfer method.The model consists of four parts,including data attribute dimension expansion module,sequential characteristic extraction and encoding module,time sequence feature fusion module and feature decoding module.The model can better capture the dependencies among data and has a better understanding of multiple complex attributes,thus enabling more accurate prediction of target intent.