Table 4 - uploaded by Olivera Pronić-Rančić
Content may be subject to copyright.
Final ANN models for the switch electrical and mechanical characteristics with the number of training and test samples
Source publication
![Fig. 6 Actuation voltage calculated by using the ANN model [42]](profile/Olivera-Pronic-Rancic/publication/292682082/figure/fig2/AS:667601835421705@1536180089449/Actuation-voltage-calculated-by-using-the-ANN-model-42_Q320.jpg)
![Table 6 RF MEMS switch modelling results: PI V [48] s L [µm] res f...](profile/Olivera-Pronic-Rancic/publication/292682082/figure/tbl2/AS:667601835405325@1536180089478/RF-MEMS-switch-modelling-results-PI-V-48-s-L-m-res-f-GHz-PI-V-V-inv-f-L-m_Q320.jpg)
The increased growth of the applications of RF MEMS switches in modern communication systems has created an increased need for their accurate and efficient models. Artificial neural networks have appeared as a fast and efficient modelling tool providing similar accuracy as standard commercial simulation packages. This paper gives an overview of the...
Context in source publication
Context 1
... to acquire more training data in these critical parts of the input space, the developed direct neural models were used for generating more training samples. The ANNs showing the best performance for each model are listed in Table 4, together with the number of training samples. To illustrate the accuracy of the inverse modelling, in Fig. 9 a comparison of the determined L f and its target value is plotted in the form of scatter plots. ...
Similar publications
RF MEMS switches have been efficiently exploited in various applications in communication systems. As the dimensions of the switch bridge influence the switch behaviour, during the design of a switch it is necessary to perform inverse modeling, i.e. to determine the bridge dimensions to ensure the desired switch characteristics, such as the resonan...
Citations
... In contrast to other algorithms like the generic algorithm, APLAC optimization, gravitational search optimization, Taguchi method, discrete simulated annealing, particle swarm optimization, fuzzy, statistics, and heuristic algorithms, ANN can process huge data (Chapman et al. 1993;Shalaby and Saitou 2004;Fasihuddin et al. 2019;Kula and Lup 2017;Shahraki and Zahiri 2013). Such neural network models are utilized to optimize the physical, RF and electrical characteristics of RF MEMS switch (Ć irić et al. 2017Marinković et al. 2016;Marinkovi et al. 2015;Marinkovic et al. 2016;Chawla and Khanna 2012a;Mafinejad et al. 2017;Filipovic 2005a, 2005b;Suganthi et al. 2012;Joslin Percy et al. 2022). In (Ć irić et al. 2017Marinković et al. 2016;Marinkovi et al. 2015;Marinkovic et al. 2016;Chawla and Khanna 2012a;Mafinejad et al. 2017;Filipovic 2005a, 2005b;Suganthi et al. 2012;Joslin Percy et al. 2022), direct and inverse ANN models are used to design different switch structures. ...
... Such neural network models are utilized to optimize the physical, RF and electrical characteristics of RF MEMS switch (Ć irić et al. 2017Marinković et al. 2016;Marinkovi et al. 2015;Marinkovic et al. 2016;Chawla and Khanna 2012a;Mafinejad et al. 2017;Filipovic 2005a, 2005b;Suganthi et al. 2012;Joslin Percy et al. 2022). In (Ć irić et al. 2017Marinković et al. 2016;Marinkovi et al. 2015;Marinkovic et al. 2016;Chawla and Khanna 2012a;Mafinejad et al. 2017;Filipovic 2005a, 2005b;Suganthi et al. 2012;Joslin Percy et al. 2022), direct and inverse ANN models are used to design different switch structures. ...
In designing and developing high-speed digital circuits it is important to develop a switch on a transmission line having higher phase velocity, low insertion loss and high isolation. An electro-statically actuated Radio Frequency Micro Electro Mechanical System (RF MEMS) switch realized on Elevated Coplanar Waveguide (ECPW) platform is fabricated and tested for RF characteristics. The introduced MEMS switch is capable of accomplishing high speed since it does not include a transistor in its structure. In this study, a 20 GHz ECPW based RF MEMS switch is designed, fabricated and characterized. Comparing the ECPW switch to normal CPW switches, the pull-in voltage is substantially lower and it is about 2–2.8 V. This decreased pull-in lowers the time required to change between states and reveals the usage of these switches for high-speed digital circuitry. At 20 GHz, measured RF performance reveals an insertion loss of − 2.7 dB and an isolation of − 5 dB. Parametric study and optimization need to be carried out to improve the isolation of the proposed switch. This paper gives an overview of the applicability of Artificial Neural Networks (ANN) in modelling a better ECPW-based RF MEMS switch. ANN Models are used to determine the required switch dimensions to obtain better radio frequency characteristics. ANN approach has been used to overcome time-consuming optimization methods for determining the radio frequency parameters in EM simulators like HFSS. In this work, the proposed ANN model reduces the switch’s design time by 98.9862% compared to EM simulator HFSS, which takes approximately 39 min and 27 s.
... It takes a long computational time as it involves a 3D structure design with a wide range of parameters, thin layers, and vias [19][20][21][22][23].To reduce the computational time involved in 3D EM simulation, recently Artificial Neural Networks (ANN) which are non-linear flexible models have been successfully used to predict unknown input-output relationships [24][25][26][27][28].It can handle a large amount of data compared to other algorithms such as generic algorithm, discrete simulated annealing, gravitational search optimization technique, particle swarm optimization, Taguchi method, APLAC optimization routines, fuzzy, statistics, heuristic algorithms [29][30][31][32][33][34]. ANNs have been used to model RF MEMS switch physical dimensions and radiofrequency electrical characteristics [35][36][37][38][39][40][41][42][43][44]. The direct and inverse approaches of neural network models with feedforward and back-propagation technique, most commonly the Levenberg-Marquardt algorithm is utilized to model membranes of different switch structures [35][36][37][38][39][40][41][42][43][44]. ...
... ANNs have been used to model RF MEMS switch physical dimensions and radiofrequency electrical characteristics [35][36][37][38][39][40][41][42][43][44]. The direct and inverse approaches of neural network models with feedforward and back-propagation technique, most commonly the Levenberg-Marquardt algorithm is utilized to model membranes of different switch structures [35][36][37][38][39][40][41][42][43][44]. ...
Radio Frequency Micro Electro Mechanical System (RF MEMS) switches are rapidly evolving due to the demand for low cost, high performance, and compact communication systems. An electrothermally actuated bistable lateral MEMS switch for redundancy applications has been already fabricated and tested for mechanical characteristics. In this paper to make this switch as a suitable candidate for RF applications,sidewall metallization using gold is proposed.Modelling of sidewall metallization in Bistable Lateral RF MEMS switch using Cascade Feedforward Scaled Conjugate Gradient (CFSCG) Artificial Neural Network approach is reported. Using an inverse approach of ANN, the desired length of sidewall coating required for better RF performance in terms of return loss and insertion loss is predicted. Time-consuming optimization procedures to determine the sidewall coating in customized EM simulators such as HFSS have been overcome using the proposed CFSCG approach. In this paper, the proposed CFSCG approach reduces the design time of the switch with sidewall coating by 99.4048% compared to the conventional EM simulator which takes approximately 14 h. Validation is done by comparing the obtained results with the HFSS model and good agreement is obtained. The simulation of the proposed switch results in an insertion loss less than -1 dB in the frequency range between 1-10 GHz with gold sidewall metallization of bistable lateral RF MEMS switch.
... The membrane critical stress analysis helps to reduce the buckling effect. The critical stress primarily depends on the membrane material and dimensions, it can be expressed as [28], [24] Dragonfly algorithm (DA) Switch directions Considering multiple parameters is not possible [25] ANN Radio frequency Complexity is high [26] Topology optimization Electrical Less accuracy [27] ANN Radio frequency Complexity is high The spring constant (K) is the major performance deciding factor. Maintaining low spring constant helps to reduce pull-in voltage and the switching time. ...
In this paper, a target application-based design approach for RF MEMS switches using artificial neural networks (ANN) is presented. ANN approach is used to decide the dimensions of the RF MEMS switch for the targeted application. The ANN approach design and analysis is done using MATLAB. The results obtained by ANN method are validated using the FEM tool simulation, which show that the developed models have good accuracy over the range of switch dimension values for the intended application. The switch dimensions are extracted for L, S, C, X, Ku, K, and Ka-band applications using ANN. Eventually, a novel RF MEMS switch based on the ANN approach for C-band applications is proposed. The required pull-in voltage will increase because of perforation, and this problem is handled by adding extra weight to the membrane with pillars and slabs. Analyzed load distribution in membrane with perforation. FEM design is implemented using solid mechanics and electrostatic-based Multi-physics in COMSOL environment. Different studies, i.e., stationary, time-dependent, and frequency domain are performed to extract electrical, mechanical, and Radio Frequency parameters of the proposed design. The switch designed for C-band applications offering an actuation voltage of 7.4 V, Insertion loss of − 0.2 dB, and Isolation loss of − 57 dB.
... Therefore during the design of a switch, it is necessary to determine the bridge dimensions to ensure the desired switch characteristics, such as resonant frequency, i.e. to perform the inverse modeling of the switch. The authors of this work proposed earlier a black-box inverse modeling of the RF MEMS capacitive switches where the bridge lateral dimensions were determined for given electrical or mechanical switches [19][20][21][22][23][24][25]. In this work, the neural based inverse modeling approach is extended in a way that the novel approach provides not only determination of the bridge dimensions but also the values of the corresponding lumped model elements, resulting in a lumped element model ready to be used for further simulations of the circuits containing the considered switch. ...
... The proposed approach is a hybrid approach combining neural modeling with a lumped element equivalent circuit. In other words, it is a combination of the black-box neural inverse modeling approach [19][20][21] and a modification of the scalable lumped element model proposed in [18]. Schematic diagram of proposed model is shown in Fig. 2. The aim of the first ANN (ANN 1) is to determine the length of the fingered part L f for the desired resonant frequency [19,20,22]. ...
... In other words, it is a combination of the black-box neural inverse modeling approach [19][20][21] and a modification of the scalable lumped element model proposed in [18]. Schematic diagram of proposed model is shown in Fig. 2. The aim of the first ANN (ANN 1) is to determine the length of the fingered part L f for the desired resonant frequency [19,20,22]. ...
RF MEMS switches have been efficiently exploited in various applications in communication systems. As the dimensions of the switch bridge influence the switch behaviour, during the design of a switch it is necessary to perform inverse modeling, i.e. to determine the bridge dimensions to ensure the desired switch characteristics, such as the resonant frequency. In this paper a novel inverse modeling approach based on combination of artificial neural networks and a lumped element circuit model has been considered. This approach allows determination of the bridge fingered part length for the given resonant frequency and the bridge solid part length, generating at the same time values of the elements of the switch lumped element model. Validity of the model is demonstrated by appropriate numerical examples.
... ANNs can be used in various applications in the field of microwaves, due to their learning and generalization abilities [3], [5][6][7][8][9][10][11]. Using them for modeling in the RF and microwaves field is beneficial, because it allows users to accurately reproduce S-parameters of the device, without knowing its equivalent circuit model. ...
... A generalization ability of an ANN is the ability to generate a correct response even for the input values that were not part of the initial training set. Due to their generalization ability, ANNs are often used in the field of RF and microwaves for developing device models from the measured data [3,[5][6][7][8][9][10][11]. ...
In this paper, S-parameters of SMS 7630 zero-bias Schottky diode are modeled with artificial neural network (ANN). A single network is trained with data obtained from measurements using vector network analyzer. Frequency band of those measurements is 0.5 – 5 GHz, with 0.01 GHz step, and input power is ranging from -25 dBm to 5 dBm, in 5 dBm increment. Learning and generalization capability of the developed ANN were tested, and results turned up to be satisfactory and useful, especially considering limited number of different input powers for network training.
... The field of electrical engineering has proved to be a very suitable for the ANN applications [26][27][28][29][30][31][32][33][34][35]. By learning the dependence between two datasets, ANNs have the capability to approximate any nonlinear function, whereby the knowledge about the physical characteristics of the problem to be modeled is not needed [36]. ...
The noise wave model has appeared as a very appropriate model for the purpose of transistor noise modeling at microwave frequencies. The transistor noise wave model parameters are usually extracted from the measured transistor noise parameters by using time-consuming optimization procedures in microwave circuit simulators. Therefore, three different Computer-Aided Design methods that enable more efficient automatic determination of these parameters in the case of high electron-mobility transistors were developed. All of these extraction methods are based on different noise de-embedding procedures, which are described in detail within this paper. In order to validate the presented extraction methods, they were applied for the noise modeling of a specific GaAs high electron-mobility transistor. Finally, the obtained results were used for the comparative analysis of the presented extraction approaches in terms of accuracy, complexity and effectiveness.
... Also Shalaby et al. [21] optimized the Actuation voltage of RF-MEMS direct contact cantilever switch. In 2016, Marankovic et al. [22] used the Artificial Neural Networks (ANN) [17] to model a RF MEMS Shunt Capacitive Switch and Emami and Sei [23] used the PSO method to optimize a NEMS Switch for optimal switching time but they lack of flexure hinges and perforation holes unlike the reported model. Additionally, the algorithms used in previous works have a tendency of getting stuck in local optima [17] and many improvements to these algorithms have been proposed in the state of art. ...
... As both T d C dx 22 and T dC dx are nonlinear functions of x, we need to approximate to have a closed form solution for critical distance In this regard, we apply Taylor series expansion around 0 x , and approximate it up to second order terms. ...
... Similarly, using (9), we can express T d C dx 22 in the form of ...
... Artificial neural networks (ANNs) have been applied for modeling RF and microwave devices [4]. Recently they have found applications in modeling of RF MEMS switches, providing the models which are able to determine the switch electrical and/or mechanical characteristics in a very short time [5]- [12]. In that way the time needed for the simulation and optimization of circuits containing the considered switches can be significantly reduced. ...
... This paper is devoted to modeling of the mechanical characteristics of the RF MEMS capacitive switches, i.e., to modeling of their actuation voltage versus the switch dimensions. Previously, the multilayer perceptron (MLP) ANNs have been applied for this purpose [8], [12]. The aim of this paper is to investigate the performances of the neural model based on the radial basis function (RBF) ANN for development of the switch actuation voltage model, which relates the dimensions of the switch with the actuation voltage. ...
... The aim of this paper is to investigate the performances of the neural model based on the radial basis function (RBF) ANN for development of the switch actuation voltage model, which relates the dimensions of the switch with the actuation voltage. The model is developed and verified for a particular RF MEMS capacitive switch for which different MLP ANN models have been developed [7]- [12]. Further, the results are compared with the corresponding results obtained by the earlier developed model based on the MLP ANNs. ...
Artificial neural networks (ANNs) have been exploited as an efficient tool in modeling of many electronic devices, among them RF MEMS devices. The design of RF MEMS devices requires determination of their electrical and mechanical characteristics according to the application requirements. ANNs have been proposed to be used for modeling RF MEMS devices and can be used further as an alternative and efficient simulation and optimization tool replacing time consuming simulations in standard electrical and mechanical simulators. The aim of this paper is to investigate possibilities of the radial basis function (RBF) ANNs to be applied for modeling of mechanical characteristics of RF MEMS capacitive switches, relating the switch geometry parameters and the actuation voltage. The achieved results obtained by the developed RBF neural model are compared with the results from the earlier developed multilayer perceptron (MLP) neural model. Moreover, effectiveness and accuracy of these two ANN models are analysed. The results confirm the efficiency of the both modelling approaches.
This communication discusses the role of nonlinear algorithms in training the neural network, which predicts the optimized RF MEMS switch dimensions. A dedicated dataset, i.e., DrTLN-RF-MEMS-SWITCH-DATASET-v1, was created by considering the most appropriate input and output variable suitable to predict the cantilever dimensions, crab leg and serpentine structure-based RF MEMS switches. The distinct artificial neural networks (ANN) performance is analysed using various training methods. The hardware implementation possible neural network algorithms, i.e., Fitting and Cascade Feed Forward Network, are considered for learning and prediction. The ANN algorithm's performance in predicting and optimizing RF MEMS switch is analysed using nonlinear training methods like Levenberg–Marquardt (LM) and Scaled Conjugate Gradient (SCG). The cascaded forward network with LM training combination offers the best performance compared with other varieties. A comprehensive study is performed using neural networks and finite element simulation results. The study revealed that the error percentage is below 15.08% for most of the parameters.
