scholarly journals Modeling of the 5G-Band Patch Antennas Using ANNs under the Uncertainty of the Geometrical Design Parameters Associated with the Manufacturing Process

Algorithms ◽  
2021 ◽  
Vol 15 (1) ◽  
pp. 7
Author(s):  
Piotr Górniak
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Reflection Coefficient ◽  
Antenna Arrays ◽  
Patch Antenna ◽  
Manufacturing Process ◽  
Design Parameters ◽  
Patch Antennas ◽  
Testing Data ◽  
Artificial Neural

In the paper, the author deals with modeling the stochastic behavior of ordinary patch antennas in terms of the mean and standard deviation of their reflection coefficient |S11| under the geometrical uncertainty associated with their manufacturing process. The Artificial Neural Network is used to model the stochastic reflection coefficient of the antennas. The Polynomial Chaos Expansion and FDTD computations are used to obtain the training and testing data for the Artificial Neural Network. For the first time, the author uses his analytical transformations to reduce the required number of highly time-consuming FDTD simulations for a given set of nominal values of the design parameters of the ordinary patch antenna. An analysis is performed for n257 and n258 frequency bands (24.5–28.7 GHz). The probability distributions of the design parameters are extracted from the measurement results obtained for a series of manufactured patch antenna arrays for three different frequencies in the C, X, and Ka bands. Patch antennas are chosen as the subject of the scientific analysis in this paper because of the popularity of the patch antennas in the scientific literature concerning antennas, as well as because of a simple form of these antennas that is reflected in the time required for computation of training and testing data for the Artificial Neural Network.

An artificial neural network-based critical values for multiple hypothesis testing: data-snooping case

Survey Review ◽  
2021 ◽  
pp. 1-16
Author(s):  
Vinicius Francisco Rofatto ◽  
Marcelo Tomio Matsuoka ◽  
Ivandro Klein ◽  
Maria Luísa Silva Bonimani ◽  
Bruno Póvoa Rodrigues ◽  
...  
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Hypothesis Testing ◽  
Multiple Hypothesis Testing ◽  
Critical Values ◽  
Data Snooping ◽  
Multiple Hypothesis ◽  
Testing Data ◽  
Artificial Neural

scholarly journals A critical evaluation of the methods for the determination of required volumes for detention tank

2018 ◽  
Vol 45 ◽  
pp. 00088 ◽  
Author(s):  
Mariusz Starzec
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Catchment Area ◽  
Critical Evaluation ◽  
Design Parameters ◽  
Method Of Calculation ◽  
Rainfall Duration ◽  
Detention Tank ◽  
Artificial Neural

Simplified methods allow a straightforward and quick determination of parameters of interest. A simplified method of calculation to be used must provide sufficiently accurate simulation results. This paper presents the results of tests completed to evaluate the effects of the parameters which describe a sewer catchment area and network on the value of Tp, a parameter applied in the Dziopak method [18]. The results of 2997 hydrodynamic simulations allowed to formulate an artificial neural network the application of which enabled the determination of the value of Tp dependent on the design parameters of a sewer catchment area and network. The artificial neural network had a very low error R2 = 0.9972 between the expected and determined values of Tp. The completed tests indicated a relationship by which an increase of the rainfall duration, a parameter used in the dimensioning of detention tank, is concomitant to an increase in the value of Tp. The calculations made so far included an assumption that the Tp value is constant irrespective of the design rainfall duration for the dimensioning of detention tank; this assumption has led to gross calculation errors. The paper also provides proof that the inclusion of these relationships allows a more precise determination of the service volume required for a multi-chamber detention tank.

CFD-based blade shape optimization of MGT-70(3)axial flow compressor

2019 ◽  
Vol 30 (6) ◽  
pp. 3307-3321 ◽  
Author(s):  
Mohammad Reza Pakatchian ◽  
Hossein Saeidi ◽  
Alireza Ziamolki
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Shape Optimization ◽  
Fluid Dynamic ◽  
Axial Compressor ◽  
Axial Flow ◽  
Design Parameters ◽  
Content Type ◽  
Wide Range ◽  
Artificial Neural

Purpose This study aims at enhancing the performance of a 16-stage axial compressor and improving the operating stability. The adopted approaches for upgrading the compressor are artificial neural network, optimization algorithms and computational fluid dynamics. Design/methodology/approach The process starts with developing several data sets for certain 2D sections by means of training several artificial neural networks (ANNs) as surrogate models. Afterward, the trained ANNs are applied to the 3D shape optimization along with parametrization of the blade stacking line. Specifying the significant design parameters, a wide range of geometrical variations are considered by implementation of appropriate number of design variables. The optimized shapes are analyzed by applying computational fluid dynamic to obtain the best geometry. Findings 3D optimal results show improvements, especially in the case of decreasing or elimination of near walls corner separations. In addition, in comparison with the base geometry, numerical optimization shows an increase of 1.15 per cent in total isentropic efficiency in the first four stages, which results in 0.6 per cent improvement for the whole compressor, even while keeping the rest of the stages unchanged. To evaluate the numerical results, experimental data are compared with obtained data from simulation. Based on the results, the highest absolute relative deviation between experimental and numerical static pressure is approximately 7.5 per cent. Originality/value The blades geometry of an axial compressor used in a heavy-duty gas turbine is optimized by applying artificial neural network, and the results are compared with the base geometry numerically and experimentally.

Reliability Assessment of Wafer Level Package using Artificial Neural Network Regression Model

2019 ◽  
Vol 35 (6) ◽  
pp. 829-837 ◽  
Author(s):  
P. H. Chou ◽  
K.N. Chiang ◽  
Steven Y. Liang
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Regression Model ◽  
Electronic Packaging ◽  
Design Parameters ◽  
Simulation Technology ◽  
Wafer Level ◽  
Structure Reliability ◽  
Artificial Neural ◽  
Packaging Structure

ABSTRACTFor electronic packaging structure, there are many design parameters that will affect its reliability performance, using experimental way to obtain the reliability result will take a considerable amount of time. Therefore, how to shorten the design time becomes a critical issue for new electronic packaging structure development. This research will combine artificial intelligence (AI) and simulation technology to assess the long-term reliability of wafer level packaging (WLP). A simulation technology using finite element method (FEM) with appropriate mechanics theories has been validated by multiple experiments will replace the experiment to create reliability results for different WLP structures. After a big WLP structure-reliability database created, this study will apply artificial neural network (ANN) theory to analyze this database and obtains a regression model for structure-reliability relationship of WLP. Once the regression model is established and validated, the WLP geometry, such as pad size, die and buffer layer thickness, and solder volume, etc. can be simply entered, and then the WLP reliability results can be immediately obtained through the ANN regression model.

Concrete strength prediction using artificial neural network and genetic programming

2018 ◽  
Vol 9 (3) ◽  
pp. 75
Author(s):  
Preeti Kulkarni ◽  
Shreenivas N. Londhe
Keyword(s):  
Neural Network ◽  
Artificial Neural Network ◽  
Genetic Programming ◽  
Construction Material ◽  
Concrete Strength ◽  
Maximum Size ◽  
Design Parameters ◽  
Ann Model ◽  
Influencing Parameters ◽  
Artificial Neural

Concrete is a highly complex composite construction material and modeling using computing tools to predict concrete strength is a difficult task. In this work an effort is made to predict compressive strength of concrete after 28 days of curing, using Artificial Neural Network (ANN) and Genetic programming (GP). The data for analysis mainly consists of mix design parameters of concrete, coefficient of soft sand and maximum size of aggregates as input parameters. ANN yields trained weights and biases as the final model which sometime may impediment in its application at operational level. GP on other hand yields an equation as its output making its plausible tool for operational use. Comparison of the prediction results displays the result the model accuracy of both ANN and GP as satisfactory, giving GP a working advantage owing to its output in an equation form. A knowledge extraction technique used with the weights and biases of ANN model to understand the most influencing parameters to predict the 28 day strength of concrete, promises to prove ANN as grey box rather than a black box. GP models, in form of explicit equations, show the influencing parameters with reference to the presence of the relevant parameters in the equations.

Predictive modeling of geometric shapes of different objects using image processing and an artificial neural network

2016 ◽  
Vol 231 (6) ◽  
pp. 1206-1216 ◽  
Author(s):  
Mustafa Ayyıldız ◽  
Kerim Çetinkaya
Keyword(s):  
Neural Network ◽  
Image Processing ◽  
Artificial Neural Network ◽  
Conjugate Gradient ◽  
Back Propagation ◽  
Neural Network Modeling ◽  
Back Propagation Algorithm ◽  
Geometric Shapes ◽  
Testing Data ◽  
Artificial Neural

In this study, an artificial neural network model was developed to predict the geometric shapes of different objects using image processing. These objects with various sizes and shapes (circle, square, triangle, and rectangle) were used for the experimental process. In order to extract the features of these geometric shapes, morphological features, including the area, perimeter, compactness, elongation, rectangularity, and roundness, were applied. For the artificial neural network modeling, the standard back-propagation algorithm was found to be the optimum choice for training the model. In the building of the network structure, five different learning algorithms were used: the Levenberg–Marquardt, the quasi-Newton back propagation, the scaled conjugate gradient, the resilient back propagation, and the conjugate gradient back propagation. The best result was obtained by 6-5-1 network architectures with single hidden layers for the geometric shapes. After artificial neural network training, the correlation coefficients ( R2) of the geometric shape values for training and testing data were very close to 1. Similarly, the root-mean-square error and mean error percentage values for the training and testing data were less than 0.9% and 0.004%, respectively. These results demonstrated that the artificial neural network is an admissible model for the estimation of geometric shapes using image processing.

Sign in / Sign up

Export Citation Format

Share Document