Stimulated by growing requirements for security and reliability in Cyber-Physical Systems (CPS), this paper proposes a federated intrusion detection framework integrating improved prototype learning and adjusted-logits Cross-Entropy loss to address the challenges of hybrid feature-label shifts. A dual-path federated learning approach (DP-FL) with prototype alignment and dynamic logits is proposed, featuring a two-branch architecture: At first, a dynamically log-weighted prototype aggregation mechanism employing dual adaptive factors is introduced, achieving a more balanced and informative global prototype. Building upon this, a dynamic logits adjustment mechanism is further designed to calibrate the decision boundaries of local client models by jointly considering both label frequency and prototype divergence, thereby strengthening the discriminative capability and generalization efficacy of the model. Finally, the validation of the proposed DP-FL framework’s effectiveness is conducted on the NSL-KDD and CICIDS2017 datasets. Experimental results show that the proposed DP-FL framework outperforms existing methods under scenarios involving hybrid feature-label distribution shifts.

Given the pervasive presence of feedback mechanisms and inertial loops inherent in industrial processes, increasing research efforts have focused on integrating latent feature dynamics into deep learning architectures to address the issue of strong dynamic autocorrelation in industrial processes. However, the complex temporal dependencies present in latent features and the limitations imposed by traditional alternating training methods have already constrained the performance of such models. In this paper, we propose a latent probabilistic dynamics embedding supervised deep networks for soft sensing. In detail, a probabilistic dynamic model with graph-guiding based on the past and current latent features is constructed in the latent space of the supervised deep networks to capture the complex dependencies between the latent sequences. The article introduces a new approach involving a probability-distribution-based predictive regularization term for latent features. By jointly training the model, the network parameters are optimized to ensure the overall convergence of the network. An improved variational graph recurrent neural network with the inclusion of randomness into the high-level latent space was proposed to model the latent dynamics for supervised deep networks and additional graph structure information to help analyze temporal dependencies. Finally, the proposed methods are implemented on two real industrial cases to demonstrate their effectiveness and superiority. Comparative experiments and ablation experiments are designed to illustrate the higher prediction accuracy and effectiveness of the proposed method.

Nonstationarity in industrial processes, guided by factors, such as equipment aging and changing upstream load demands, inherently exhibits heterogeneous characteristics. This complex overlay of homogeneous stationarity poses great difficulty in process monitoring and analysis. Therefore, this study presents a new model (Hs- ${\mathrm {H}}_{\mathrm {n}}$ ) that peels the homogeneous and heterogeneous nonstationarity, which has four components: a differentiated learning network (DL-Net), a peeling network (Pe-Net), an adaptive reweighting network (AR-Net), and a global decoder network. DL-Net obtains the differentiated representation by leveraging a new differentiated learning approach to unique inputs, which is based on the cognitive understanding and derivation of functional specialization and content learning during network training. The aim is to maximize functional diversity and minimize content overlap. Furthermore, Pe-Net extracts the stationarity and nonstationarity (S-N) components from each differentiated scale, formulated as an encoder-decoder-encoder architecture with an integrated identity subtraction skip connection. A min-max S-N constraint regulates the peeling process and controls the extracted content. AR-Net additionally refines homogeneous stationarity across each scale and reweights the individual components to adaptively adjust their contributions. Last, reweighted components are fused and input into the global decoder to facilitate unsupervised learning. Experimental results on three processes demonstrate the effectiveness of Hs-Hn.

Due to the scarcity of fault samples in real-world industrial scenarios, few-shot fault diagnosis (FSFD) has attracted increasing attention, as it is vital for ensuring industrial safety and engineering reliability. However, existing metrics-based meta learning approaches generally calculate fault similarity relying on global pairwise, which insufficiently exploit fault information and thus limit discrimination among fault types. To address the insensitivity to buried fault information in complete processes, we propose a novel framework called the Local-Global Consistency Relation Network (LGCRN). The framework captures more comprehensive fault feature information through two components: the global branch, which focuses on macro-information involving all join process units, and the local branch, which emphasizes finer granularity from specific units. Additionally, a local-global alignment loss is utilized to optimize LGCRN, where self-cross relation improves the discriminability and stability of fault features by maintaining consistency between local and global branches. The self-cross relation comprises self-sample relation, ensuring consistency within the same sample at different granularities, and cross-sample relation, maintaining consistency across samples within the same class. Experiments on two benchmarks and a real semiconductor process demonstrate the feasibility and superiority of the proposed method, confirming its effectiveness in addressing limited-sample industrial fault diagnosis tasks and enabling timely maintenance crucial for operational safety in real-world scenarios.

Since the low labeled rate, especially the extremely low labeling rate with only one label data for each class, will greatly limit the performance of traditional fault diagnosis algorithms, this paper proposes a pseudo-label guided rotating machinery fault diagnosis method. Firstly, the vibration signal is converted into power spectral density (PSD) data and the structural similarity (SSIM) index is used to measure the relationship between PSD, which reasonably characterizes the structural relationship between the data. On this basis, more reliable pseudo labels are selected according to the adaptive threshold and assigned to the unlabeled data, which increases the number of available labeled samples and eliminates the influence of noisy labels. Then, in the pre-training stage, the weights based on pseudo-labels are improved by contrastive learning and embedded into the stacked autoencoder (ICSAE), enabling the model to learn deep features that are easier to classify. During the fine-tuning stage, labeled data and pseudo-label data with different weights are used to help improve the accuracy of model fault diagnosis. Finally, the laboratory bearing dataset verify the excellent performance of ICSAE in extremely low labeled rate scenarios.

Since the poor quality of signals and redundant features may reduce the accuracy of fault diagnosis of rotating machinery components, an ensemble targeted stacked denoising autoencoders (ETSDAE) method is proposed for machinery running under harsh environment. At first, a targeted denoising strategy is designed for ensemble models to remove noise from time-domain and frequency-domain data in the encode stage. The multidomian data makes up for the limitations of single-domain data while the denoising strategy improves the anti-noise ability of ETSDAE. On the other hand, two indexes based on mutual information are designed into cost function as a soft constraint to learn min-redundant deep features that have max-relevance to the class targets, thus it can get rid of the dependence on additional feature selection procedure. On this basis, integrated deep features are directly input to single-hidden layer feedforward neural network to realize fault diagnosis. Finally, the effectiveness of the proposed method are verified by rolling bearing test rig and industrial reciprocating pump. The results show that ETSDAE has excellent performance in fault diagnosis, especially in terms of anti-noise and feature learning.

In deep-learning-based process monitoring, obtaining an effective feature representation is a critical step in constructing a reliable deep-learning monitoring model. Conventional deep-learning methods like stacked auto-encoders (SAEs) capture feature representation by minimizing the data reconstruction errors, which lack the expression of essential information and ultimately lead to degradation of the monitoring performance. To solve this problem, variational discriminative SAE (VDSAE) is proposed in this article. First, a variational generative discriminative structure is designed to obtain a reliable prelearned discriminator. Based on this new variational discriminator, the authenticity of the reconstructed data is evaluated as an important criterion for feature learning. Then, an SAE incorporating the prelearned discriminator is trained by both minimizing the reconstruction error and maximizing the data authenticity. In this way, the prelearned discriminator makes the network effectively capture the essential expression of the reconstructed data. The proposed approach enables SAE to learn a better feature representation owing to the excellent reconstruction performance. Finally, the feature representation and fault detection performance of VDSAE are verified in two cases. The results show that the average fault detection rates (FDRs) of the multiphase flow facility and the waste-water treatment process (WWTP) can be improved to 72% and 97%, respectively, compared with the other fault detection methods.

Due to the complex static, dynamic, and large-scale characteristics for modern industrial processes, in this article, we propose a double-layer distributed monitoring approach based on multiblock slow feature analysis and multiblock independent component analysis. To this end, the processed dataset is divided into the static and dynamic blocks on the basis of the sequential information of each variable in the first layer. Considering the correlations between the variables in the large-scale processes, the sequential correlation matrices in two blocks are calculated, which serves as the second-layer block division rule. Then, the static and dynamic blocks are further divided into several static and dynamic subblocks in which the variables in each subblock are strongly correlated and in the same state. The slow feature analysis and independent component analysis monitoring models are, respectively, generated for the dynamic and static subblocks. Finally, the monitoring results in each subblock are integrated by Bayesian inference to get the final statistics. The average fault detection rate of the proposed method for the Tennessee Eastman process is 0.842, while those of the other traditional methods are lower than 0.75, which shows the advantages of the proposed method.