Search for: All records

Crystallization due to liquid → solid transformation is observed in many natural and engineering processes. Extant literature indicates that crystallization in supercooled liquids is initiated by precursory metastable phases or states, also called non-classical nucleation. For face-centered cubic (FCC) materials, latest experimental and computational studies suggest that metastable hexagonal-closed packed (HCP) structures facilitate equilibrium FCC formation. However, the underlying nucleation mechanism remains unclear. Here, we examine structural changes and energetic barriers associated with such a non-classical mechanism, by performing molecular dynamics (MD) simulations using pure Al, Al-0.5 at. %Cu, and Al-0.5 at. %Ni (all FCC-formers) and phenomenologically coupling MD results with phase-field (PF) modeling. Such a coupling involved initializing PF simulation domains and constructing Landau polynomials—consistent with MD observations. Unsupervised machine learning was utilized to capture nuclei structures from MD simulations, while neural networks helped in extracting equilibrium interfacial energies from PF modeling. Atomistic simulations showed that precursory nuclei are comprised of collection of metastable-HCP states with medium ranged ordering. The pockets of HCP states later transform to critical nuclei—containing an FCC core and an outer layer of HCP. PF modeling qualitatively replicated the precursory-to-critical nuclei transformation and showed that the energetic barriers between the precursory and critical nuclei are substantially smaller than predictions obtained from classical nucleation theory. Together, these observations permitted us to propose a holistic non-classical mechanism that links triangular motifs within Al-based supercooled liquids to the critical nuclei via in-liquid structural transformations. 

The increasing adoption of smart home devices has raised significant concerns regarding privacy, security, and vulnerability to cyber threats. This study addresses these challenges by presenting a federated learning framework enhanced with blockchain technology to detect intrusions in smart home environments. The proposed approach combines knowledge distillation and transfer learning to support heterogeneous IoT devices with varying computational capacities, ensuring efficient local training without compromising privacy. Blockchain technology is integrated to provide decentralized, tamper-resistant access control through Role-Based Access Control (RBAC), allowing only authenticated devices to participate in the federated learning process. This combination ensures data confidentiality, system integrity, and trust among devices. This framework’s performance was evaluated using the N-BaIoT dataset, showcasing its ability to detect anomalies caused by botnets such as Mirai and BASHLITE across diverse IoT devices. Results demonstrate significant improvements in intrusion detection accuracy, particularly for resource-constrained devices, while maintaining privacy and adaptability in dynamic smart home environments. These findings highlight the potential of this blockchain-enhanced federated learning system to offer a scalable, robust, and privacy-preserving solution for securing smart homes against evolving threats. 

The state-of-the-art determines the remaining useful lifetime (RUL) through a steady-state, fixed power cycling tests (PCT) without considering the impact of dynamically changing environmental conditions. It has resulted in considerable RUL prediction errors in the real world. However, the dynamic changing conditions (e.g., large temperature swings) may affect the degradation evolution of SiC MOSFET, which could eventually result in RUL changes. Thus, it must be integrated to make accurate predictions. To precisely understand the RUL variation complexity, the junction temperature (Tj) has been measured with a Negative Thermal Coefficient (NTC) thermistor, Temperature Sensitive Electrical Parameter (TSEP), and these profiles have been modeled through the thermal model RC foster network using Extended Kalman Filter (EKF). Then, the on-state resistance (Rds,on) variations and Degradation Acceleration Factor (DAF) under the dynamic environment conditions are integrated into a lifetime prediction model to accurately predict the RUL through the Long Short-Term Memory (LSTM) machine learning algorithm. 

A generalized logistic differential equation model of biomolecular memristors improves the tuning of hyperparameters of parallel-memristor physical reservoir computing systems by enablingex situtraining.