Abstract
The conventional approach of designing of SPR driven photonic crystal fiber-based sensor yields accurate results, it is time consuming especially in the situation where we deal with wide range of wavelengths and refractive indices. The utilization of machine learning (ML) driven optimization techniques can address such limitations and allow us to faster estimation of simulated results. This work covers the designing aspects of dual core PCF biosensor composed of hybrid layers of gold and graphene for rapid and quick estimation of biosensor performance without having time consuming repetitive full wave iterative simulations executed in the COMSOL Multiphysics environment. For successful implementation of ML approach, first simulated data set containing confinement loss (CL), analyte refractive indices and wavelength values of the designed structure is utilized to train and predict the CL data under two ML approaches. (1) The data driven combined stacked ensemble model consisted of RFR, GPR, SVM, and (2) an artificial neural network (ANN) model. The predicted data obtained from various ML models shows that the performance of the stacked ensemble model is better over ANN having performance metric of MAPE, RMSE and R2 values 0.792 %, 0.149013 and 0.9999 respectively. Secondly, the stacked ensemble model is further used to predict the real and imaginary parts of the effective refractive indices, Re(nₑff) and Im(nₑff) respectively of the design under analyte of RI variation 1.30 to 1.39 over the wavelength range 650–1700 nm. The predicted Im(nₑff) values were further employed in an inverse formulation to reconstruct the confinement loss spectra using the analytical relation between CL and Im(nₑff) values. This data set is useful in the prediction of maximum CL value and its associated resonant wavelength corresponding to each analyte RI value yielding maximum wavelength sensitivity of 30,000 nm/RIU. Thus, the ML approach can be utilized to predict the simulated results containing an effective RI, resonant wavelength and sensitivity. The proposed idea consisted of combined stacked ensembled ML model may be useful in the development of various photonic biosensors with reduced computational efforts and time.
Biography
Prof. Suneet Kumar Awasthi is a science post graduate from University of Lucknow,Uttar Pradesh, India. After completing his post-graduation, he joined the Photonics Research Group of Prof Usha Malaviya at Department of Physics, University of Lucknow. He did his doctoral thesis on “Optical Devices Based on Photonic Crystals”. He joined JIIT (Deemed to be University) as an Assistant Professor in Department of Physics and Material Science and Engineering, JIIT-128 Noida in January 2010. Prior to join JIIT-128, he has joined Amity University Uttar Pradesh as a Lecturer in Amity School of Engineering and Technology (November 2007 – July 2008). He worked as a Senior Lecturer in Physics Department, Amity University Rajasthan, Jaipur (July 2008 – January 2010). He also served Department of Physics and Department of Computer Science University of Lucknow as a guest faculty (2005-2007). Currently, he has more than seventeen years of post-Ph.D. research and teaching experience. Prof Awasthi is life member of Indian Science Congress Association Kolkata and Indian Association of Physics Teachers. He is editorial board member of scientific reports. Prof. Awasthi is a potential reviewer of various international and national journals of repute in the field of optical engineering, photonics, bio-sensors,plasma physics etc.
