Antenna Optimazition Design Based On Artificial Intelligence

With the development of wireless communication technology,antenna design is becoming more and more complex,but manual design and optimization of antennas have problems such as complex process and low efficiency.Antennas,as a key component of wireless communication systems,have long been a focus for researchers.Antenna is a device that transmits and receives electromagnetic waves in a wireless communication system,and its performance directly affects the quality of the entire communication system.Therefore,it is necessary not only to propose a variety of new antennas that meet the design specifications,but also to propose antenna optimization schemes that can be quick and accurate.This dissertation proposes two artificial intelligence antenna optimization schemes.By training the autoencoder surrogate model to fit the complex nonlinear relationship between the antenna physical parameters and S11,it is possible to avoid frequent calls to electromagnetic simulation software during antenna simulation optimization.In this dissertation,heuristic algorithms such as neural networks and particle swarm optimization algorithms are used in the optimization design of antennas,and the autoencoder surrogate model proposed in this dissertation is used to achieve rapid simulation and optimization of antennas.The main work of this dissertation is as follows:(1)This dissertation proposes an antenna optimization design scheme based on the autoencoder surrogate model and the fully connected neural network.Firstly,the scheme models the optimized antenna HFSS,and uses the Python-HFSS joint simulation program to randomly perturb the physical parameters of the antenna to automatically collect data.Among them,the dataset collecting antenna physical parameters and antenna S11 real virtual curve is used to train the autoencoder surrogate model,and the dataset collecting antenna S11 real virtual curve and antenna physical parameter perturbation values is used to train fully connected neural networks.In this scheme,the physical parameters of the antenna are updated according to the prediction results of the fully connected neural network until the antenna S11 curve meets the optimization design index.In this dissertation,the antenna optimization design scheme is used for the optimization design of an S-band“Beidou” satellite receiving antenna,and the experimental results show that the return loss of the optimized antenna at 2.492 GHz is less than-20 d B,and the bandwidth of the antenna at-10 d B is 90 MHz,which is in line with the set optimization index,and the optimization result is compared with the HFSS calculation result,which proves the effectiveness and accuracy of the proposed method.(2)This dissertation proposes an antenna optimization design scheme based on convolutional autoencoder surrogate model and particle swarm algorithm.This scheme improves the above autoencoder surrogate model,adds a convolutional layer network to the encoder,increases the feature extraction capability of the model,and shortens the training time of the model.The convolutional autoencoder surrogate model is combined with the particle swarm optimization algorithm to achieve automatic optimization of antenna physical parameters by giving the objective function.In this dissertation,the scheme is used for the optimal design of coaxial feeder microstrip antenna,and the experimental results show that the return loss of the optimized antenna at 2.45 GHz is less than-30 d B,and the optimization results are compared with the HFSS calculation results to prove the effectiveness and reliability of the optimized antenna design scheme.