During industrial production activities, industrial products serve as critical resources whose performance is subject to various external factors and usage conditions. To ensure uninterrupted production processes and to guarantee the safety of the production personnel, a real-time analysis of the industrial product quality and subsequent decision making are essential. Conventional detection methods have inherent limitations in meeting the real-time demands of processing large volumes of data and achieving high response speeds. For instance, the regular inspection and maintenance of cars can be time-consuming and labor-intensive if performed manually. Furthermore, monitoring the damage situation of bearings in real time through a manual inspection may lead to delays and may hinder production efficiency. Therefore, this paper presents online machine-learning-based methods to address these two practical problems and simulates them on various datasets to meet the requirements of efficiency and speed. Prior to being fed into the network for training, the data undergo identity parsing to transform them into easily identifiable streaming data. The training process demonstrates that online machine learning ensures timely model updates as small batches of data are sent to the network. The test results indicate that the online learning method exhibits highly stable and effective performance, optimizing the training process.

Transfer Learning (TL) has shown potential for body sensor network (BSN) applications. However, TL performance is influenced by various factors such as the BSN environment, user-specific physiological signs or sensor modalities, leading to negative transfer (NT) effects. Limited research has defined NT from perspectives such as data quality, domain transferability, and transfer components, but the cognition of NT in multimodal BSNs remains insufficient. In this paper, we explore the correlation between transfer modes and NT, proposing a knowledge transfer framework composed of common domain adaptation constraints to reveal the negative transfer effects in over-constrained environments. This effect can explain why strongly constrained transfer algorithms are effective but not necessarily optimal in terms of performance. In the three typical BSN scenarios of emotion recognition, fall detection, and daily activity recognition, we demonstrate the NT in over-constrained environments and its potential influencing factors using different constraint combinations and based on different multi-modal body sensor fusion architectures. Extensive experiments focus on discussing the impact of category-condition constraints, domain similarity, and fusion architectures on the NT effect in BSNs. This work can provide a new perspective for the design of multimodal BSN knowledge transfer scheme.

Warp knitting machines operate under high-speed and continuous working conditions, where minor abnormalities in yarn tension or vibration may quickly evolve into severe defects. Conventional vision-based inspection methods usually detect faults only after visible defects appear, resulting in substantial latency and limited capability for early warning and predictive maintenance. To address this issue, we propose KnitLLM, a multimodal semantic learning framework for early fault diagnosis in warp knitting machines, which integrates sensor sequences and fabric images for proactive anomaly detection. KnitLLM first employs a vector-quantized variational autoencoder to map heterogeneous temporal signals and image features into a unified discrete semantic space, enabling aligned fault primitive representation across modalities. A Transformer-based masked fault modeling strategy is then introduced to learn long-range dependencies and structural priors of the knitting process by reconstructing masked process states. In addition, a diffusion model is used to augment rare fault samples, alleviating class imbalance and improving recognition performance for minority fault categories. Theoretical analysis from the perspectives of manifold representation, quantization error, and attention structure further supports the semantic abstraction and information preservation capabilities of the proposed framework. Experiments on a real-world industrial dataset containing 10,000 h of warp knitting production data demonstrate that KnitLLM consistently outperforms competitive baselines across multiple metrics, achieving an F1-score of 0.961 and a mean detection latency of 124 ms. The results indicate that the proposed method can identify early faults before visible defects emerge, offering a practical solution for intelligent maintenance and fault warning in high-speed textile manufacturing.