Self-supervised learning (SSL) has significantly advanced speaker verification, especially in scenarios with limited labeled data. This paper introduces Energy-based Confidence-Aware Distillation (EBCA-DINO), an SSL enhancement for speaker verification that integrates Energy-Based Models (EBMs) into the DINO (Distillation with No Labels) framework. EBMs use energy scores to assess data complexity and uncertainty, guiding label-free self-distillation. The adaptive temperature scaling tailors the learning process to data characteristics, allowing the teacher model to dynamically adjust the student model’s focus based on sample difficulty. This energy-aware distillation optimizes speaker verification performance. Experimental results demonstrate that EBCA-DINO improves speaker verification with relative performance gains of 4.3%, 4.9%, and 8.7% on the Vox1-O, E, and H test trials, respectively.

In industry, many systems exhibit load-sharing characteristics. In a load-sharing system, failure of an asset, in addition to affect system reliability, increases the workloads of remaining surviving assets and so their failure rates. When managing such the assets in a system, it is important for decision makers to ensure overall performance of the system, by determining redundancy of assets and a preventive maintenance plan with consideration of load sharing and uncertain environmental conditions. This article proposes an approach for synthetically optimizing redundancy design and age-based preventive maintenance for a load-sharing system with identical assets. A two-stage stochastic programming model with recourse is established, which incorporates risk-aversion preference of decision makers. A decomposition algorithm is developed to solve the joint optimization model, incorporating analytical properties of system failure rate functions and models. A comparative study with deterministic optimization and robust optimization is conducted to demonstrate the advantages of the proposed risk-averse stochastic programming approach. Finally, a numerical study on an effluent treatment system is conducted to analyze the optimal redundancy design and maintenance plan and practical insights.

Consider a system consisting of multistate components that perform the same function, and each component occupies a location in the system. The deterioration processes of components differ due to different workloads, usage rates, or environmental stresses that are associated with the locations. This article proposes a condition-based maintenance policy, in which the component reallocation (CR) with distinct assignments of components to locations and the preventive replacement of system are dynamically implemented based on the system state. A Markov decision process (MDP) is formulated to optimize the proposed condition-based multi-CR maintenance policy by determining the actions for each system state that minimize the expected long-run system maintenance cost. In the current studies on the MDP for maintenance optimization, the actions mainly include component replacement, imperfect repair, and system replacement. In this article, including CRs of distinct assignments as actions and considering multiple times of CRs increase the action space of the MDP significantly. An enumeration-based value iteration algorithm and a genetic-algorithm-based value iteration algorithm are proposed. Numerical experiments on$k$-out-of-$n$:G systems and Monte Carlo simulation tests show the effectiveness of CRs on reducing the system maintenance cost and extending system lifetime and provide structural insights on the optimal maintenance policy.