The distribution discrepancy metrics are the core foundation of achieving domain confusion. Therefore, they mainly determine the performance of deep transfer diagnosis models. However, their effectiveness relies on the stability of data local distributions, making them unsuitable for cross-domain machine diagnosis tasks under continuous time-varying conditions. Hence, a new integrated-dispersion manifold distance (IDMD) is proposed to enhance the discrepancy representation capability in dynamic data structures. The maximum entropy-based local distribution (MELD) selection mechanism is designed to represent the global distribution information of time-varying monitoring signals adaptively. Furthermore, the ensemble Grassmann manifold geodesic (EGMG) measurement is constructed to characterize the intrinsic distribution discrepancy information due to complex nonlinear structures of high-dimensional data. The proposed IDMD distribution discrepancy metric is validated against two fault transfer diagnosis experiments under time-varying conditions, including laboratory planetary gearboxes and actual wind turbine bearings. The experimental results demonstrate its effectiveness and advantage over the existing advanced methods.

As a proactive maintenance approach, remaining useful life (RUL) prediction plays a key role in smart operation and maintenance of industrial systems. To enhance the interpretability of deep neural network, and to measure the uncertainty of complex systems in the degradation process, an RUL prediction approach based on interpretable serialized variational autoencoder with drift-diffusion stochastic equation (ISVAE-DDSE) is proposed. Specifically, considering a dynamic sequential modeling method, this article proposes a generative deep learning approach to ensure that the model effectively captures the distribution characteristics of degradation data. On this basis, from the perspective of probabilistic deep generative network, this article derives a new type of generative loss function with the aid of the Bayesian theory. Furthermore, this article proposes an interpretable latent variable construction pattern based on DDSE, which integrates the dynamic representation of states, and rate of state change. In this sense, the network model can understand, and predict the evolutionary behavior of complex systems over time. Moreover, a Gaussian distribution network is designed to evaluate the RUL prediction’s uncertainty. This article demonstrates the advantages of the ISVAE-DDSE using a NASA aircraft turbofan engine dataset.

This paper presents a data-driven framework for integrating encryption transmission and attack detection in cyber-physical systems (CPS) with nonlinear physical plants. The main focus of this research is to use deep neural networks to realize the coprime factorization (CF) of nonlinear systems. The definition of the CF guides the network training and designing process, and the model’s topology is designed in the state-space form, which improves the interpretability of the data-driven CF. Based on the CF-aided neural networks, an encrypted transmission module is designed that projects information related to system dynamics into a perpendicular data space, which complements existing encryption methods from a control theory perspective. Subsequently, an anomaly detector are designed using the same CF pairs. This detector not only provides high-accuracy detection of attacks but also distinguishes between attacks and faults, thereby reducing the false positive rate and enhancing the reliability of the attack detection. The proposed method has been validated in a real CPS using a mecanum-wheeled vehicle as the physical plant, demonstrating its effectiveness and applicability.

Aiming at the current challenges in converting electrical parameters to indicator diagrams, a subspace-aided indicator diagram estimation approach is proposed to establish a data-driven mapping model from electrical to force parameters, which helps avoid the need for analyzing the mechanism model of tower-type pumping systems. Specifically, the lifting technique is adopted based on the subspace method to construct the space of the electrical parameter signals, addressing the correspondence between input and output signals, while preventing the loss of effective information. Then, a recursive indicator diagram estimation approach is proposed, utilizing the updating/downdating of the Cholesky decomposition to enable online updating of the data-driven mapping model. In addition, for tower-type pumping systems operating under multiple conditions, a gap metric indicator is developed as a test statistic to determine the switching of operating conditions. The effectiveness of the proposed methods is verified through experimental measurements from tower-type pumping systems in actual oil wells.

Source-free domain adaptation (SFDA) has emerged as a promising and practical approach to achieve better cross-domain fault diagnosis in privacy-preserving scenarios, yet face challenges in identifying target-private faults within open-set (OS) scenarios. To address this limitation, a theoretical generalization bound error is employed to analyze the root causes, which primarily stem from domain shift and OS scenarios. Guided by this theoretical foundation, a novel SFDA-OS approach is proposed to integrate target adaptation process and OS separation using entropy-based confidence index and corresponding confidence sets. It incorporates a comprehensive loss function for adaptation, combining pseudolabel learning, clustering, and uncertainty-aware updating for high-confidence samples, alongside additional clustering for low-confidence samples. Extensive experimental results validate the effectiveness of the proposed method, demonstrating its capability to provide a potential, practical, and privacy-compliant solution for deployable fault diagnosis in actual engineering systems where unknown faults emerge and source data access is restricted.

This article designs a data-driven security defense and monitoring approach that involves encrypted transmission and attack detection to defend cyber-physical systems (CPS) against stealthy attacks. The approach starts by using subspace theory to achieve data-driven coprime factorization of the closed-loop CPS. Based on this, the physical dynamics are encrypted from a control perspective to minimize the risk of information leakage and hinder the creation of stealthy attacks. Meanwhile, the security monitoring approach is designed using the same offline-learned coprime factorization. This approach is effective in detecting and distinguishing between cyber-physical attacks and machine-induced faults, which enables effective maintenance measures to be taken for different anomalies. The proposed encrypted transmission and security monitoring approach provides a comprehensive defense against nonstealthy and stealthy attacks. The effectiveness of our work is illustrated through a numerical example and experimental results on a Mecanum-wheeled vehicle platform.

This article focuses on utilizing process data to detect faults in nonlinear systems. To accomplish this, stable image/kernel representation is learned for nonlinear systems using deep neural networks, which serve as the basis for residual generators and fault detection. First, the closed-loop image representation of nonlinear systems is identified using gate recurrent units and fully connected neural networks. The involved network topology is designed to learn the nonlinear mapping in the form of linear time-varying state space, allowing the extension of existing linear methods to nonlinear systems. Then, with the identified image representation, the data-driven realization of kernel representation is derived. Finally, the residual generator is developed utilizing the system's kernel representation to enable precise fault detection in nonlinear systems. The effectiveness of our study is demonstrated through a numerical benchmark study and an actual experiment on a real Mecanum-wheeled vehicle platform.

This article designs a data-driven unsupervised defense scheme for nonlinear systems by proposing a machine learning approach called gate recurrent unit-based modified denoising and stable image representation-aided autoencoders. The proposed scheme decomposes original data into two subspaces through orthogonal projection. For secure transmission, information related to the system's dynamics, which is in the image space of the controlled system, is hidden through filtering, whereas only the dynamic-independent information is plaintext for transmission, which supplements the cryptographic encryption methods from a control perspective. Moreover, attack detection for nonstealthy and stealthy attacks is achieved simultaneously under the same framework. A case study is conducted for validation on the a hardware-in-the-loop platform with a mecanum-wheeled vehicle. The comparative experiments with well-known unsupervised data-driven methods show the high detection accuracy of the proposed defense scheme for nonstealthy and stealthy attacks and the excellent encryption capability.

Accurate remaining useful life (RUL) prediction of lithium-ion batteries is critical for energy supply systems. In conventional data-driven RUL prediction approaches, the battery's degradation mechanism is difficult into incorporate in the RUL prediction. Furthermore, there are notable limitations in reflecting the significance of different time instances, and the uncertainty in the degradation process. Consequently, a novel data-model interactive RUL prediction approach based on particle filter-temporal attention mechanism-bidirectional gated recurrent unit (PF-BiGRU-TSAM) is proposed. Specifically, BiGRU-TSAM is trained offline through historical data, which assigns corresponding significance to battery capacities at different time instances. Moreover, regarding the interactive data-model for the online prediction phase based on PF-BiGRU-TSAM, the advantages of data-driven and model-based approaches are integrated, which accomplishes the purpose of modifying each other. The proposed PF-BiGRU-TSAM approach is validated with a real-world battery dataset. Experimental results demonstrate the proposed approach is better than some published approaches. Taking the 50th operational cycle of the four batteries B0005, B0006, B0007, and B0018 in the dataset as an instance, the absolute errors of the proposed PF-BiGRU-TSAM are 0, 1, 3, 3, respectively, which represents the proposed approach has an excellent performance.

Accurate bearing fault diagnosis is of great significance of the safety and reliability of rotary mechanical system. In practice, the sample proportion between faulty data and healthy data in rotating mechanical system is imbalanced. Furthermore, there are commonalities between the bearing fault detection, classification, and identification tasks. Based on these observations, this article proposes a novel integrated multitasking intelligent bearing fault diagnosis scheme with the aid of representation learning under imbalanced sample condition, which realizes bearing fault detection, classification, and unknown fault identification. Specifically, in the unsupervised condition, a bearing fault detection approach based on modified denoising autoencoder (DAE) with self-attention mechanism for bottleneck layer (MDAE-SAMB) is proposed in the integrated scheme, which only uses the healthy data for training. The self-attention mechanism is introduced into the neurons in the bottleneck layer, which can assign different weights to the neurons in the bottleneck layer. Moreover, the transfer learning based on representation learning is proposed for few-shot fault classification. Only a few fault samples are used for offline training, and high-accuracy online bearing fault classification is achieved. Finally, according to the known fault data, the unknown bearing faults can be effectively identified. A bearing dataset generated by rotor dynamics experiment rig (RDER) and a public bearing dataset demonstrates the applicability of the proposed integrated fault diagnosis scheme.

Knee segmentation and landmark localization from 3D MRI are two significant tasks for diagnosis and treatment of knee diseases. With the development of deep learning, Convolutional Neural Network (CNN) based methods have become the mainstream. However, the existing CNN methods are mostly single-task methods. Due to the complex structure of bone, cartilage and ligament in the knee, it is challenging to complete the segmentation or landmark localization alone. And establishing independent models for all tasks will bring difficulties for surgeon's clinical using. In this paper, a Spatial Dependence Multi-task Transformer (SDMT) network is proposed for 3D knee MRI segmentation and landmark localization. We use a shared encoder for feature extraction, then SDMT utilizes the spatial dependence of segmentation results and landmark position to mutually promote the two tasks. Specifically, SDMT adds spatial encoding to the features, and a task hybrided multi-head attention mechanism is designed, in which the attention heads are divided into the inter-task attention head and the intra-task attention head. The two attention head deal with the spatial dependence between two tasks and correlation within the single task, respectively. Finally, we design a dynamic weight multi-task loss function to balance the training process of two task. The proposed method is validated on our 3D knee MRI multi-task datasets. Dice can reach 83.91% in the segmentation task, and MRE can reach 2.12 mm in the landmark localization task, it is competitive and superior over other state-of-the-art single-task methods.

Cup anemometers are widely used instruments for wind turbines to measure wind speed in wind farm. Aimed to reduce the adverse impact on wind energy resource estimation, this paper proposes a data-driven fault diagnosis approach for assessing the anemometer health status. Auto-associative netural network (AANN) is developed to reconstruct the anemometer measurement data after data pre-processing, and residual analysis is performed between the anemometer measurement data and the AANN reconstruction data. In addition, the quantitative indicators that can reflect the health status of the anemometer gained from residuals are obtained through the K-Means clustering algorithm, based on which the faulty anemometers in the wind farm can be identified. The approach can provide guidance for the production and operation of the wind farm.