Rahul Kaushik

Scientific Officer | Raja Ramanna Centre for Advanced Technology| 2022 - Present | Indore, India

• Interferometer Lock Loss Prediction: Developed a predictive model using seismic channel time-series data, applying DMAT (Digital Matrix / Dynamic Matrix Analysis Tool) -based time-history learning and transfer-learning CNNs ResNet50/EfficientNet) to accurately predict lock-loss events.

• Development of Control System and Machine Learning Model for Optic Steering: Led the design and implementation of a precision OSEM-based control system, analyzed 10,000+ actuator-displacement data points and developed predictive ML models (KNN, Random Forest, AdaBoost) achieving ±100 micrometers positioning accuracy, enhancing system responsiveness and improving model prediction accuracy by 25%.

• Development of Single-Stage Suspension for End Station: Modeled single-stage suspension in Mathematica incorporating flexural corrections, thermoelastic damping, and dissipation dilution; validated CAD designs and experiments, achieving first six modes below 80 Hz and improving low-frequency vibration isolation by 15%.

• Development of Triple-Stage Suspension for Beam Splitter: Mathematically modeled three-stage suspension in Mathematica with transfer functions for low-frequency isolation; addressed OSEM actuation constraints, achieving first 18 modes below 60 Hz and enhancing seismic isolation by 20%, enabling control strategy development.

• Development of Lightweight Back-Pocketed Mirror using Diffusion Bonding: Designed CAD and FEA-optimized 500 mm mirror with mass-reducing pockets, reducing weight from 50 kg to 10.5 kg while limiting deflection to 163 nm (from 544 nm), improving rigidity by 70%, and establishing diffusion bonding parameters for manufacturable optical components.

• Development of Semi-Continuous Production Facility for Optical Glass: Led the design and thermal optimization of a platinum-based optical glass production system using CAD and transient/steady-state thermal analysis, ensuring optical homogeneity of 10^-4; optimized power input, insulation thickness and melt flow rate to maintain uniform temperature, prevent bubble entrapment and significantly improve energy efficiency and glass quality.

• Quality Prediction of Produced Optical Glass: Developed a machine-learning–based quality prediction framework for optical glass by correlating process parameters with optical homogeneity; applied Random Forest regression and PCA-based dimensionality reduction, achieving improved prediction accuracy and a 20% reduction in RMSE for reliable optical glass performance assessment.