FIGURE 8 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
Source publication
In the design of conventional microwave devices, the parameters need to be continuously optimized to meet the desired targets, and the whole process is time-consuming and laborious. As a surrogate model, machine learning is an effective optimization method. However, in the modeling process, the high-dimensional data processing and the complex nonli...
Contexts in source publication
Context 1
... widely used in multi-standard mobile/wireless systems [32]. The optimized second antenna structure is the MIMO antenna, the schematic diagram is shown in Fig. 8, the top layer is shown on the left and the bottom layer is shown on the right. The size of the antenna is 41mm×25mm×1.6mm, and the bandwidth of the antenna is widened by a wrenchshaped microstrip feed line, and a rectangular structure is introduced in the ground plane of the antenna to obtain a good port isolation. The design ...
Context 2
... realize the notch characteristics of the MIMO antenna, C-shaped branches are introduced on the antenna radiator, and by adjusting the size of the notch structure and its position are used to determine the frequency range to be suppressed. The proposed DBN-ELM model can accurately predict the size of the C-shaped dendrites (for L 7 , W 8 and d in Fig. 8). Similarly, 500 sets of data in Table 4 are prepared as training input samples, and the corresponding S-parameters are calculated by HFSS simulation software simulation as training output samples. The sample acquisition process took 25,620 seconds. The 500 sets of training data are substituted into the DBN-ELM model for training to ...
Similar publications
In this article, an optimized, highly efficient, wideband 1 × 2 Multiple-Input-Multiple-Output (MIMO) fractal antenna is designed for high-speed short-range communication, video rate imaging, medical imaging, and explosive detection in the terahertz frequency band. The suggested fractal THz MIMO antenna configuration is originated from basic rectan...
Citations
... ML has been widely applied, such as data mining, securities market analysis, natural language processing, computer vision, speech and handwriting recognition, search engines, strategic games, robot applications, biometric recognition, medical diagnosis, detection of credit card fraud, DNA sequence sequencing, as well as in the field of electromagnetics [1] [2]. There are many concern ML algorithms applied in antennas optimization domain, including support vector machine [3], Gaussian process [4], deep Gaussian process [5], student's T process [6], extreme learning machine [7], broad learning system [8], artificial neural network [8], deep neural network [9], convolutional neural network [10], etc. Reference [3] measured electronically steerable parasitic array radiator patterns, and then used the support vector machine training process to handle antenna-based DOA estimation. Reference [4] developed a modeling method named progressive Gaussian process, which test sample with the largest predictive variance was inputted into an electromagnetic solver to compute its results when a Gaussian process was trained, and then added it to the training set to train the Gaussian process progressively. ...
... In [6], student's T process was used as the surrogate model in Bayesian optimization algorithm, and the estimation strategy function was improved based on the posterior output, realizing the improved Bayesian optimization, finally used for the antennas optimization. In order to design UWB antennas, [7] proposed a deep learning model, and the deep belief network structure is determined by particle swarm algorithm, and then combined it with extreme learning machine. Aiming to speed up the antennas design by the easy-to-measure parameter variables achieving more accurate prediction of hard-to-measure quality variables, [8] proposed auto-context for the regression scenario, using the current broad learning system training results as prior knowledge, which was taken as the context information and then combined them with the original inputs becoming the new inputs for further training. ...
Machine learning (ML) is a scientific technology related to data learning that can help machines learn patterns from existing complex data to predict future behavioral outcomes and trends. In this paper, we use 16 different ML algorithms and 5 different hyper-parameters optimal methods to model a circularly polarized omnidirectional base station antenna. The mean absolute error is 8.431, the root mean squared error is 11.100, the coefficient of determination is 0.967, and the adjusted coefficient of determination is 0.944, which means evaluation metrics are excellent.
... Compared to FPDM, IPDM is more in line with the public's needs. It has a large number of application cases in antenna design [20][21][22][23][24][25]. These applications have similarities with the design of metasurfaces [19,[26][27][28][29][30]. ...
To solve the time-consuming and complex design problems, the deep learning method is used to realize the inverse predictive design of a transmission-type linear-to-circular polarization control metasurface (TLCPCM). Firstly, the target-generation neural net-work model (TGNNM) is constructed based on a fully connected neural network (FCNN). The model selects the critical features of the required electromagnetic perfor-mance as design targets, and maps low-dimensional design targets to high-dimensional electromagnetic performance. Secondly, taking the output data of the TGNNM as input data, an inverse-mapping neural network model (IMNNM) is constructed by a convolu-tional neural network (CNN). The prediction performance of the IMNNM is compared with two other inverse-mapping models. The research results show that the IMNNM out-performs the other two networks. Finally, combining TGNNM and IMNNM, four sets of TLCPCM structural parameters are predicted. The research results show that the elec-tromagnetic performances of the metasurface determined by the predicted structural parameters are generally consistent with the given design targets. On this basis, one ex-perimental sample is manufactured. The measurement results are consistent with the simulation results. The research results demonstrate the validity and feasibility of the inverse predictive design method proposed in this paper.
... Recently, DL/ML methods have been applied to designing MIMO antennas. A DL model was proposed [34] for the design of UWB antennas which used the particle swarm optimization (PSO) algorithm to build the structure of the deep belief network (DBN). Using their model, they were able to learn large amounts of features, approximate nonlinear relationships between design structure and performance, and model the scattering parameters for a two-element notch MIMO antenna. ...
Traditional optimization methods for antenna design are time-consuming, whereby full-wave electromagnetic (EM) simulations are performed repeatedly to find the optimum design parameters for acceptable performance. However, deep learning (DL) is an effective method for determining the optimum physical parameters based on their complex, nonlinear relationship with the design characteristics. This paper proposes a time and resource-efficient DL approach for designing MIMO antenna arrays. A wideband, eight-element MIMO antenna array for fifth-generation (5G) cellphones is designed to operate at n77/n78/n79 sub-6 GHz new radio (NR) bands and optimized using the proposed approach. The proposed method utilizes the feature reduction method to reduce the design space and generate an effective dataset. A novel dual-channel deep neural network (DC-DNN) is developed to predict the scattering parameters. A direct optimization approach is used by utilizing the defined figure of merit (FOM) for multiple design objectives instead of using an optimization algorithm with the proposed DL model. The optimized MIMO antenna is fabricated and tested for performance evaluation. The optimized design achieved −6 dB impendence bandwidth of 45.5 % (3.28–5.21 GHz), isolation < –12.5 dB, TARC < –6 dB, ECC < 0.1, MEG < −6 dB, and CCL < 0.35 bps/Hz.
... For the proper impedance matching indication, at the location of the feed, the VSWR is the important parameter. The ratio of the maximum voltage to the minimum voltage in the antenna de nes the VSWR as given in Eq. (12). ...
In many applications such as aerospace systems, satellites, mobile radar, and other wireless applications, the Microstrip patch antenna (MPA) plays an important role due to its properties like simplicity, lightweight, low cost of production, and compact structure. Low gain, narrow frequency bandwidth, and high return loss are the shortcomings in the existing microstrip patch antenna (MPA) design approaches. Moreover, the developed models of the antenna are hard to design and larger in size. The antenna's geometrical specifications should be optimized to address this problem. In the proposed approach Reptile Search Algorithm (RSA) based Multilayer perceptron (MLP) neural network is employed to design the H-shaped antenna for Ku-band applications. The multilayer perceptron neural network is employed to calculate the fitness value of the RSA. To train the MLP neural network, the dataset is generated using MATLAB software that contains values of the substrate height, dielectric constant, length of the patch, resonant frequency, width of the patch as the input parameters, and the values of gain and return loss as the output parameters. The performance of the optimally designed antenna with the proposed approach is evaluated in terms of the radiation pattern, return loss, Voltage Standing Wave Ratio (VSWR), gain, computation time, directivity, and convergence speed. The experimental and simulation results of the proposed approach show better performance with 8.89 dB gain, -33.06 dB return loss, and 1.07 VSWR.
... In [14], planar ultra-wideband antennas were designed by combining deep structures (the cascade of an extreme learning machine, a deep belief network, and a restricted Boltzmann machine) and a particle swarm optimization as a global optimizer. ...
In this communication, artificial neural networks are used to estimate the initial structure of a multiband planar antenna. The neural networks are trained on a set of selected normalized multiband antennas characterized by time-efficient modal analysis with limited accuracy. Using the Deep Learning Toolbox in Matlab, several types of neural networks have been created and trained on the sample planar multiband antennas. In the neural network learning process, suitable network types were selected for the design of these antennas. The trained networks, depending on the desired operating bands, will select the appropriate antenna geometry. This is further optimized using Newton’s method in HFSS. The use of the neural pre-design concept speeds up and simplifies the design of multiband planar antennas. The findings presented in this paper will be used to refine and accelerate the design of planar multiband antennas.
... Similarly, particle swarm based optimization have been extensively in the recent years for optimization of deep neural networks [6][7][8]. In [9] an extreme machine learning model have been employed for determining the UWB antenna parameters. Optimization algorithms such as simulated annealing algorithm, genetic algorithm (GA), ant colony optimization, differential evolution, grey wolf optimization, sine-cosine optimization, and many others can be used to optimize the design parameters of the antenna [10][11][12][13][14][15]. ...
Antenna design involves continuously optimizing antenna parameters to meet the desired requirements. Since the process is manual, laborious, and time-consuming, a surrogate model based on machine learning provides an effective solution. The conventional approach for selecting antenna parameters is mapped to a regression problem to predict the antenna performance in terms of S parameters. In this regard, a heuristic approach is employed using an optimized random forest model. The design parameters are obtained from an ultrawideband (UWB) antenna simulated using the high-frequency structure simulator (HFSS). The designed antenna is an embedded structure consisting of a circular monopole with a rectangle. The ground plane of the proposed antenna is reduced to realize the wider impedance bandwidth. The lowered ground plane will create a new current channel that affects the uniform current distribution and helps in achieving the wider impedance bandwidth. Initially, data were preprocessed, and feature extraction was performed using additive regression. Further, ten different regression models with optimized parameters are used to determine the best values for antenna design. The proposed method was evaluated by splitting the dataset into train and test data in the ratio of 60:40 and by employing a ten-fold cross-validation scheme. A correlation coefficient of 0.99 was obtained using the optimized random forest model.
... A Deep Belief Network-Extreme Learning Machine (DBN-ELM) model based on PSO (Particle Swarm Optimization) was proposed by the author [8]. Results demonstrate that the model can rapidly extract samples, minimize the complexity and computational cost imposed by repeated simulations in antenna design, and significantly increase the effectiveness of antenna design. ...
Machine learning (ML) will be heavily used in the future generation of wireless communication networks. The development of diverse communication-based applications is expected to boost coverage and spectrum efficiency in relation to conventional systems. ML may be employed to develop solutions in a wide range of domains, such as antennas. This article describes the design and optimization of a circular patch antenna. The optimization is done through ML algorithms. Six ML models, Decision Tree, Random Forest, XG-Boost Regression, K-Nearest Neighbour (KNN), Gradient Boosting Regression (GBR), and Light Gradient Boosting Regression (LGBR), were employed in this work to predict the antenna's return loss (S11). The findings show that all of these models work well, with KNN having the highest accuracy in predicting return loss of 98.5%. The antenna design & optimization process can be accelerated with the support of ML. These developments allow designers to push beyond the limits of antenna technology, optimize performance, and offer novel solutions for emerging applications such as 5G, 6G, IoT, and flexible wireless communication systems).
... The results show that the deep network is more accurate when modelling microwave devices than the traditional shallow neural network. Reference [6] proposes using a particle swarm algorithm (PSO) to * Corresponding authors: Chen Yang (yangchen6526@163.com). identify a deep trust network's model structure, followed by a deep model coupled with an extreme learning machine to get the best fractal antenna design. ...
... Step 5: PSOCF algorithm optimization DMLP. In the training process, particle optimization is carried out through PSOCF, and the optimal particle position corresponding to the final DMLP (node, p) is output, then the optimized DMLP model is obtained [6]. Set the particle interval; each node takes the value of [130, 1000]; p takes the value of [0.15, 0.8]; the average accuracy of cross-validation of test samples is selected as the fitness function; the number of PSOCF iterations is 20; both learning factors are 2.05; the number of particles is 5; the batch geometric is 128; the number of neural network iterations is 2000; the learning rate is 0.001; use the MSE loss function and Adam optimizer for model optimization. ...
... Nan et al. suggest a deep learning model that establishes the model's structure for developing ultra-wideband (UWB) antennas. They created Deep Belief Network and Extreme Learning Machine Surrogate Models (DBNELM) model, which has good accuracy, significantly improving antenna design efficiency and the ability to minimize root mean square error values [4]. Gaussian process, least absolute shrinkage and selection operator (LASSO) semi-supervised learning, ANNs techniques [5], single-output Gaussian process regression (SO-GPR), multioutput Gaussian process regression (MO-GPR), regression based learning algorithm is proposed in this research. ...
A smart antenna synthesis approach is described as automatically choosing the optimum antenna type and providing the best geometric characteristics under the demands of antenna performance. Different antenna performance characteristics are examined, and using decision tree classifier, the optimal antenna is suggested using an intelligent antenna selection model. Finally, the geometric characteristics of the antenna are given before the fuzzy inference system is developed by merging five primary learners to fully exploit the benefits of each type of learner. Rectangular patch antenna, pyramidal horn antenna, and helical antenna are the three types of antennas that are classified by a decision tree classifier, and the optimal antenna size parameters are determined using a fuzzy inference method. The performance of decision tree classifier measured using accuracy and FIS is measured using Mean Square Error (MSE) and MAPE. The system demonstrates excellent capability in parameter prediction with antenna categorization with a MAPE of less than 5.8% and accuracy over 99% achieved in our proposed method. The recommended methodology might be widely applied in actual smart antenna design.
... In Ref. [25,26], the reference-design-free constrained modelling method and self-adaptive GP modelling method were used, respectively, to reduce the cost and training dataset required for training the surrogate model. For the same purpose, the paper in Ref. [27] uses deep belief network and extreme learning machine. ...
This paper suggests an optimisation method to design multi‐band antenna using artificial neural networks. The proposed network, surrogate‐based model using auto‐encoder (SBM‐AE), is composed of two parts, ordinary neural network and auto‐encoder. First, the front neural network predicts the encoded antenna characteristics and then decodes the predicted data to obtain the antenna characteristics. After training the encoder to obtain a characteristic signature vector, the front neural network regresses only the characteristic signature vectors, reducing the complexity and number of parameters of the neural network. This not only reduces the training time but also significantly reduces the number of training data required. We confirm the effectiveness of the design method by designing a multi‐band antenna where the proposed SBM‐AE required 270 training data while the conventional neural network without auto‐encoder needed 1350 training data to achieve comparatively the same error rate.
