Document Type : Primary Research paper
Assistant Professor, Department of Computer Science and Engineering, Avanthi Institute of Engineering and Technology, Gunthapally, Abdullahpurmet Mandal-501512, Hyderabad, Telangana.
Associate Professor and Head, Department of Computer Science and Engineering,St Peter's Engineering College,R R District, 500100.
Assistant professor (Sl.G),Department of EEE,KPR Institute of Engineering and Technology,Coimbatore-641407.
Lecturer,Department of Computer Science ,Prince Saatam Bin Abdul Aziz University , Saudi Arabia, Wadi Aldwassir , 1191.
Professor, Department of Computer Science and Engineering, Dr.Samuel George Institute of Engineering & Technology, Markapur, Prakasam Dt, Andhra Pradesh, 523316.
Associate Professor, Department of Computer Science and Engineering, Velammal Institute of Technology, Velammal Gardens, Panchetti, Chennai – 601204.
A Deep Convolutional Neural Network (DCNN) based model for predicting the advancement of temporal field esteems in transient electrodynamics is proposed in this paper. In our model, the Recurrent Neural Network (RNN) fills in as the focal part, which learns portrayals of the succession of its info information in long haul spatial-temporal connections. Simulations of plane wave scattering from dispersed using finite difference time domain, perfect electric conducting objects, we build an encoder-recurrent-decoder architecture educated on the data. The trained network is shown to simulate a transient electrodynamics issue with a simulation time that is more than 17 times faster than conventional finite difference time domain solvers, as shown in this paper. It contains a supervised machine learning model for estimated electromagnetic fields in a cavity with an arbitrary distribution of electrical spatial permittivity. Our model is quite predictive and more than 10 times faster than simulations with similar finite differential frequencies, which indicates that, for example, optical reverse design techniques may be employed in the future. Optical devices need the use of fast and precise simulations, which are thus essential. This article proposes a deep learning method to speed up a simulator's performance in solving Maxwell frequency-domain equations. Since our model forecasts 2D slit array transmission by wavelength under certain conditions, it is pretty accurate and delivers results 160,000 times faster than those achieved by the simulator.