Rotating mechanical systems (RMSs) are widely applied in various industrial fields. Intelligent fault diagnosis technology plays a significant role in improving the reliability and safety of industrial equipment. A new algorithm based on improved multiscale fuzzy entropy and support vector machine (IMFE-SVM) is proposed for the automatic diagnosis of various fault types in elevator rotating mechanical systems. First, the empirical mode decomposition (EMD) method is utilized to construct a decomposition model of the vibration data for the extraction of relevant parameters related to the fault feature. Secondly, the improved multiscale fuzzy entropy (IMFE) model is employed, where the scale factor of the multiscale fuzzy entropy (MFE) is extended to multiple subsequences to resolve the problem of insufficient coarse granularity in the traditional MFE. Subsequently, linear discriminant analysis (LDA) is applied to reduce the dimensionality of the extracted features in order to overcome the problem of feature redundancy. Finally, a support vector machine (SVM) model is utilized to construct the optimal hyperplane for the diagnosis of fault types. Experimental results indicate that the proposed method outperforms other state-of-the-art methods in the fault diagnosis of elevator systems.

Rotating machinery plays a crucial role in intelligent manufacturing systems enabled by the industrial Internet of Things (IIoT), where operational stability is essential for reliability and safety. As key mechanical components, rolling bearings are susceptible to faults during prolonged operation, highlighting the need for timely and accurate fault diagnosis. Conventional diagnostic methods rely on manual feature extraction, which limits pattern recognition, multimodal fusion, and generalization under multicondition and few-shot scenarios. Pre-trained language models have shown potential in overcoming these limitations by automatically capturing complex patterns, integrating multimodal features, and leveraging prior knowledge. To leverage the transferable knowledge of pre-trained language models for time-series fault diagnosis, a two-stage multimodal framework, BERTFD, is proposed for intelligent bearing fault diagnosis. A pre-trained bidirectional encoder representation from the transformers (BERT) model, along with learnable multimodal sequence tokens (LMSTs), is used to extract and fuse multimodal features efficiently. A time–frequency semantic fusion module (TFSFM) is designed to enhance semantic representations and produce a unified diagnostic embedding. The two-stage training strategy combines self-supervised pre-training with few-shot fine-tuning, thereby reducing dependence on labeled data and enhancing model adaptability. Experimental results demonstrate that, compared with state-of-the-art (SOTA) methods, BERTFD achieves superior fault diagnosis accuracy and generalization performance in IIoT-enabled multicondition and few-shot scenarios.

This work considers the $H\_{\infty }$ fuzzy control issue of 2-D fuzzy systems susceptible to hybrid cyberattacks due to the unreliable communication networks modeled by the Markov chain, in which the hybrid cyberattacks include deception attacks and denial-of-service attacks. Meanwhile, considering the concealment characteristics of cyberattacks, it may be impossible to know which type of attack the system is suffered. Therefore, a unified hidden Markov model is used in this work, in which the hidden mode represents the real attack mode, the observed mode is the estimated value of the attack mode and it can be directly utilized by the designed security controller. By utilizing the attack mode-dependent Lyapunov function, the asymptotically mean square stability of the 2-D resultant system can be ensured and a set of sufficient conditions can be obtained. Finally, two illustrative examples are provided to demonstrate the applicability of the obtained results.

Deep learning has made remarkable advances in the identification of crop pests and diseases. However, it requires a significant quantity of labeled data, w