Linxuan Zhang

Lithium-ion batteries are the primary power source for new energy equipment, such as electrochemical energy storage systems and electric vehicles. Accurate state of health (SOH) estimation and remaining useful life (RUL) prediction are essential for ensuring safe operation and reducing lifecycle operation and maintenance costs. Since existing studies are largely based on ideal laboratory data, the proposed algorithms often exhibit limited generalizability and engineering implementation difficulties in real-world scenarios. To address this gap, this paper provides a comprehensive review of the latest research progress with field data as the core focus. First, it defines the scope, acquisition characteristics, and preprocessing strategies of in-service battery data and quantitatively evaluates the applicability of representative domestic and international open-source and industrial proprietary datasets. Second, it clarifies the quantitative standards, intrinsic relationships, and application-specific requirements for SOH and RUL. The performance, applicability, and limitations of model-based, data-driven, and hybrid approaches are comprehensively analyzed based on real-world datasets, with emphasis on their underlying principles, recent technical advances, and engineering potential. Finally, it summarizes challenges, including multi-factor coupled degradation, battery pack consistency assessment, and small-sample transfer learning, and it outlines future directions, including multimodal data fusion, digital twin modeling, and edge intelligence collaboration. Unlike previous reviews that primarily focus on laboratory data, this review establishes an engineering-oriented analytical framework centered on real-world operational data. The findings indicate that hybrid approaches present a highly promising qualitative trend for balancing prediction accuracy and model interpretability, offering valuable guidance for bridging the gap between laboratory research and practical industrial applications.

The skylight process represents a pivotal element of railway maintenance work; however, it encompasses substantial safety risks. The prompt identification of trespassing behaviors during the skylight process is crucial for maintaining railway stability and ensuring worker safety. Existing trespassing detection research is mainly based on railway operation scenes, with no prior studies focusing on trespassing detection in the context of the skylight process in railway maintenance. Addressing these challenges, this study proposes an algorithm for detecting trespassing by railway workers during the skylight process, utilizing the You Only Look Once instance segmentation model trained on the manually annotated data set with augmentation. This trespassing detection algorithm includes innovative elements, such as the horizontal line detection method, customized width ratio, and a customized rule for identifying the railway under maintenance. The proposed algorithm is designed to aid safety personnel at skylight process sites by providing enhanced supervision and effectively detecting instances of worker trespassing. With a detection precision exceeding 97% in the experimental results, this algorithm proved its efficiency in the railway skylight process, underscoring its extensive potential for diverse trespassing detection applications.

The correct use of personal protective equipment (PPE) is crucial for mitigating occupational hazards and ensuring the safety of workers. The standards for PPE usage can vary depending on the nature of the job and the individual’s role. Even within the same role, there may be multiple correct ways to wear PPE. The conventional approach to PPE compliance detection typically treats it as a classification problem, which often requires the reconstruction of the dataset and the retraining of the model, leading to limited scalability. This paper introduces a method that merges object detection and pose estimation techniques. This method is structured around three key components including a worker detection module, a part segmentation module, and a template matching module. By acquiring a small set of standard images and making adjustments to the standard template library, the method is able to adapt to various PPE requirements. Experimental results in two phases with two types of PPE indicate high accuracy in detecting worker PPE compliance. The method’s ability to swiftly and accurately identify compliance makes it well suited for real-world applications, and its capability to adjust to different PPE standards with minimal additional workload highlights its robust scalability.

In this chapter, we aim to advance our understanding of how education can harness students’ capabilities to collectively deal with and act on major challenges, such as the ecological crisis. We examine students’ agency formation in an environmental action...

Accurately predicting the probabilities of user feedback, such as clicks and conversions, is critical for advertisement ranking and bidding. However, there often exist unwanted mismatches between predicted probabilities and true likelihoods due to the rapid shift of data distributions and intrinsic model biases. Calibration aims to address this issue by post-processing model predictions, and field-aware calibration can adjust model output on different feature field values to satisfy fine-grained advertising demands. Unfortunately, the observed samples corresponding to certain field values can be seriously limited to make confident calibrations, which may yield bias amplification and online disturbance. In this paper, we propose a confidence-aware multi-field calibration method, which adaptively adjusts the calibration intensity based on confidence levels derived from sample statistics. It also utilizes multiple fields for joint model calibration according to their importance to mitigate the impact of data sparsity on a single field. Extensive offline and online experiments show the superiority of our method in boosting advertising performance and reducing prediction deviations.

Lithium-ion batteries, essential components in new energy vehicles and energy storage stations, play a crucial role in health-status investigation and ensuring safe operation. To address challenges such as limited estimation accuracy and a weak generalization ability in conventional battery state of health (SOH) estimation methods, this study presents an integrated approach for SOH estimation that incorporates multiple health indicators and utilizes the multi-scale convolutional neural network (MSCNN) model. Initially, the aging characteristics of the battery are comprehensively analyzed, and then the health indicators are extracted from the charging data, including the temperature, time, current, voltage, etc., and the statistical transformation is performed. Subsequently, Pearson’s method is employed to analyze the correlation between these health indicators and identify those with strong correlations. A regression-prediction model based on the MSCNN model is then developed for estimating battery SOH. Finally, validation using a publicly available lithium-ion battery dataset demonstrates that, under similar operating conditions, the mean absolute error (MAE) for SOH estimation is less than 0.67%, the mean absolute percentage error (MAPE) is less than 0.37%, and the root mean square error (RMSE) is less than 0.74%. The MSCNN has good generalization for datasets with different working conditions.