The physical and mechanical properties of composite materials, such as strength and durability, are highly dependent on their microstructure and micro-viscoelastic properties. Characterizing these aspects is essential, as the microstructure dictates load distribution and damage resistance, while the micro-viscoelastic properties govern time-dependent behaviors like creep and stress relaxation. This characterization is crucial for optimizing material performance in industries like pharmaceutical and aerospace. In this study, the microstructural and micro-viscoelastic properties of pharmaceutical tablets (as composite materials) are correlated with their performance and quality. A systematic, quantitative correlation is established between the microstructural properties of oral solid dosages (OSDs), such as porosity and particle size distribution, and performance characteristics like dissolution rate and hardness. This is achieved through ultrasonic analysis, providing deeper insights into the formulation-process-performance relationships, ultimately improving product development and quality control. 

The U.S. pharmaceutical supply chain is vulnerable, and transitioning from batch to continuous manufacturing (CM), supported by the FDA, aims to enhance efficiency and security. My research focuses on addressing gaps in Real-Time Release (RTR) testing in CM to strengthen supply chain resilience. In this study, we are developing a Level I prototype, integrating a blockchain-supported cyber-backbone, ML modeling, and real-time sensor data of the Process Parameters (PPs) and Material Attributes (MAs) from the CM line. This autonomous experimental rig construct with the capacity to characterize physical mechanical, and microstructural properties of 200 pharmaceutical tablets per hour without human intervention ultrasonic non-destructive manner. Collaborative robot (UR3e), Industrial Gripper (EGK MB), BOA Spot Vision System, bowl feeder, line feeder, are integrated to ensure the automation.  

Additive Manufacturing (AM) processes are highly unstable, leading to variability in part quality even with consistent parameters, as highlighted in the 2023 AMSC Roadmap. Due to this uncertainty, timely anomaly detection and defect prediction are critical. My research aims to address this gap by focusing on real-time quality assurance and early defect detection in 3D-printed parts in an ultrasonic non-destructive manner. I developed in-situ monitoring systems using ultrasonic and laser-induced techniques to assess build integrity. Key contributions include introducing Phononic Crystal Artifacts (PCAs) for simulating build complexities and optimizing material performance through ultrasonic evaluation. I am currently working on integrating state-of-the-art machine learning algorithms (such as principal component analysis, decision trees, and deep neural networks) with ultrasonic non-destructive evaluation technologies and advanced data analytics to develop a robust autonomous testing system. This system, embedded in 3D printers, is designed to predict quality attributes and enable early defect detection in additive manufacturing and 3D printing processes.

The AMSC 2023 Roadmap highlights a significant research gap, referred to as Gap DE14: Designing to be Cleaned, which notes the absence of design guidelines for ensuring device cleanability after production. In my work on nanoparticle removal from 3D printed parts, I developed a Dry Laser Cleaning (DLC) method that uses laser-induced thermo-elastic waves for non-contact removal of particles, enhancing cleaning efficacy while preventing re-deposition risks. Additionally, I introduced a laser-induced plasma (LIP) shockwave technique, which generates shockwaves to remove particles from intricate geometries in additive manufactured components. Both methods are non-destructive and significantly improve cleaning efficiency, addressing post-production challenges and enhancing the reliability of 3D printed parts, especially for sensitive applications like aerospace and biomedical fields.

Inverse mathematical problems are challenging due to their ill-posed nature, where solutions may be non-unique, highly sensitive to data errors, or nonexistent. They are often nonlinear, computationally expensive, and require regularization techniques to manage instability. In my work, I developed a novel Machine Learning-based (Multi-Output Regression Neural Network) technique for the first time to solve this inverse mathematical problem for extracting micro-viscoelastic and micro-structural properties of compressed pharmaceutical oral solid dosage (OSD) forms directly from ultrasonic waveforms. Synthetic waveforms with a given set of micro-properties of virtual tablets are computationally generated to train, validate, and test the developed ML models for their effectiveness in the inverse problem of recovering specified micro-scale properties. 

Ear segmentation is critical in ear-based biometric systems as it ensures accurate feature extraction for reliable human recognition, especially in non-cooperative or contactless settings. I have applied the Novel deep-learning-based MASK-RCNN technique for ear recognition and segmentation in an autonomous manner by providing precise detection and instance segmentation, which is essential for improving the performance of ear recognition systems, particularly in complex or real-world scenarios. Its effectiveness in isolating the ear from challenging backgrounds enhances the accuracy of biometric identification. Additionally, I have demonstrated the effectiveness of the Deep Learning-based segmentation technique for post-mortem human iris segmentation.