Neighborhood Graph Research Articles

Quantifying how TCR sequence influences cell fate commitment in chronic LCMV infection

Abstract Repetitive T cell receptor (TCR) engagement on CD8 T cells during chronic infection is a critical driver of the dysfunctional state termed exhaustion. These exhausted cells are characterized by additional heterogeneity including a memory-like progenitor cell (Tpro) that continuously gives rise to either non-cytolytic exhausted cells (Texh) or cytotoxic effector cells (Teff). While we and others have shown that TCR avidity influence cell fate commitment to Texh or Teff, it is not certain if TCR sequence features influence this fate choice. In this study, we identify how TCR V(D)J gene recombination influences T cell differentiation at the single-cell resolution. To investigate this, we performed single-cell resolution TCR sequencing (scTCR-seq) on GP33+ tetramer sorted cells within lymphocytic choriomeningitis virus (LCMV) Clone 13 model of chronic infection across 7 different mice. Utilizing TCRdist and Clonotype Neighbor Graph Analysis (CoNGA), we construct correlations between gene expression and TCR features. We find that both chain usage and cdr3 motifs correlate to effector or exhausted cell fate commitment across these different mice. Overall, our results suggest that TCR gene rearrangement is critical not only for antigen specificity, but also determining CD8 T cell phenotypic outcomes during chronic viral infection.

Characterizing unstructured data with the nearest neighbor permutation entropy.

Permutation entropy and its associated frameworks are remarkable examples of physics-inspired techniques adept at processing complex and extensive datasets. Despite substantial progress in developing and applying these tools, their use has been predominantly limited to structured datasets such as time series or images. Here, we introduce the k-nearest neighbor permutation entropy, an innovative extension of the permutation entropy tailored for unstructured data, irrespective of their spatial or temporal configuration and dimensionality. Our approach builds upon nearest neighbor graphs to establish neighborhood relations and uses random walks to extract ordinal patterns and their distribution, thereby defining the k-nearest neighbor permutation entropy. This tool not only adeptly identifies variations in patterns of unstructured data but also does so with a precision that significantly surpasses conventional measures such as spatial autocorrelation. Additionally, it provides a natural approach for incorporating amplitude information and time gaps when analyzing time series or images, thus significantly enhancing its noise resilience and predictive capabilities compared to the usual permutation entropy. Our research substantially expands the applicability of ordinal methods to more general data types, opening promising research avenues for extending the permutation entropy toolkit for unstructured data.

Persistent homology and topological statistics of hyperuniform point clouds

Hyperuniformity, the suppression of density fluctuations at large length scales, is observed across a wide variety of domains, from cosmology to condensed matter and biological systems. Although the standard definition of hyperuniformity only utilizes information at the largest scales, hyperuniform configurations have distinctive local characteristics. However, the influence of global hyperuniformity on local structure has remained largely unexplored; establishing this connection can help uncover long-range interaction mechanisms and detect hyperuniform traits in finite-size systems. Here, we study the topological properties of hyperuniform point clouds by characterizing their persistent homology and the statistics of local graph neighborhoods. We find that varying the structure factor results in configurations with systematically different topological properties. Moreover, these topological properties are conserved for subsets of hyperuniform point clouds, establishing a connection between finite-sized systems and idealized reference arrangements. Comparing distributions of local topological neighborhoods reveals that the hyperuniform arrangements lie along a primarily one-dimensional manifold reflecting an order-to-disorder transition via hyperuniform configurations. The results presented here complement existing characterizations of hyperuniform phases of matter, and they show how local topological features can be used to detect hyperuniformity in size-limited simulations and experiments. Published by the American Physical Society 2024

Non-parameter clustering algorithm based on chain propagation and natural neighbor

Clustering analysis is a powerful tool for discovering potential knowledge in datasets. However, numerous existing clustering algorithms suffer from heavy reliance on parameter settings and cannot cluster complex manifold data very well. Moreover, although non-parameter clustering algorithms aim to lower the usage threshold, their actual performance in clustering is often unsatisfactory. Addressing how to achieve similar clustering effects for numerous data points with only a few core data points during clustering is a valuable consideration. To alleviate these challenges, a non-parameter clustering algorithm is proposed and named Non-parameter Clustering Algorithm based on Chain Propagation and Natural Neighbor (NPCCPN) in this paper, by jointly using chain propagation and natural neighbor. Specifically, NPCCPN identifies core points through chain propagation and clusters them using the saturated neighborhood graph. This makes the core data extraction and clustering process efficient and non-parameter. Finally, the performance of the method is validated on 15 complex synthetic datasets and 10 real datasets from public UCI database. The experimental results show that the effectiveness and superiority of the proposed algorithm. Purity scores first. ACC scores second without parameters, but the score first algorithm needs to set the parameters.

Semi-supervised few-shot fault diagnosis driven by multi-head dynamic graph attention network under speed fluctuations

Although graph neural networks (GNNs) have achieved great success in the field of semi-supervised bearing fault diagnosis recently, most GNN-based studies still face the challenge of insufficient labeled samples. Meanwhile, the signals collected from bearings usually have a time offset due to the alternating speed in the actual working conditions, reducing the applicability of traditional GNN diagnosis methods relying on Euclidean distance. And incomplete use of unlabeled data and insufficient exploitation of neighborhood information further reduce the reliability of fault diagnosis models. To solve these problems, a semi-supervised few-shot fault diagnosis method driven by multi-head dynamic graph attention network (MHDGAT) is proposed in this study. First, dynamic time warping (DTW) similarity is introduced in the construction of K-nearest neighbor graph to solve the problem of local time offset caused by speed change. Second, some pseudo labels generated by smoothness assumption are injected into the raw training set, and more comprehensive information can be learned in the process of model training. Then, the presented MHDGAT model can mine rich fault characteristics from different scales by the multi-head attention and extract the most sensitive features due to the proposed dynamic graph mechanism adaptively. Finally, the fault categories can be obtained by a softmax classifier located at the end of the model. The proposed model has been validated in two case studies, and the experimental results indicate that MHDGAT exerts strong performance even under extremely low-label ratios.

M-step inverse fuzzy mixed competition graphs

The notion of competition graphs has a wide range of applications and has been utilized in designing various ecological problems and real-world competitions. Here, the idea of m-step inverse fuzzy mixed competition graphs has been introduced and some of its basic properties have been discussed. Also, some ideas associated with the concept of m-step inverse fuzzy mixed competition graphs, such as m-step inverse fuzzy neighborhood graph, inverse fuzzy mixed economic competition graph, and m-step inverse fuzzy mixed economic competition graph have been defined. Our motive was to focus on the real-life problems that can be properly addressed with graphs involving directed and undirected edges. In this case, the concept of an m-step fuzzy competition graph is not appropriate for solving these problems. However, the idea of m-step inverse fuzzy mixed competition graphs is more relevant to address them. An application of m-step inverse fuzzy mixed competition graphs has been identified as a suitable solution to a real-life problem.

Sublinear Time Approximation of the Cost of a Metric \({k}\)-Nearest Neighbor Graph

Sublinear Time Approximation of the Cost of a Metric \({k}\)-Nearest Neighbor Graph

Consensus Affinity Graph Learning via Structure Graph Fusion and Block Diagonal Representation for Multiview Clustering

Learning a robust affinity graph is fundamental to graph-based clustering methods. However, some existing affinity graph learning methods have encountered the following problems. First, the constructed affinity graphs cannot capture the intrinsic structure of data well. Second, when fusing all view-specific affinity graphs, most of them obtain a fusion graph by simply taking the average of multiple views, or directly learning a common graph from multiple views, without considering the discriminative property among diverse views. Third, the fusion graph does not maintain an explicit cluster structure. To alleviate these problems, the adaptive neighbor graph learning approach and the data self-expression approach are first integrated into a structure graph fusion framework to obtain a view-specific structure affinity graph to capture the local and global structures of data. Then, all the structural affinity graphs are weighted dynamically into a consensus affinity graph, which not only effectively incorporates the complementary affinity structure of important views but also has the capability of preserving the consensus affinity structure that is shared by all views. Finally, a k–block diagonal regularizer is introduced for the consensus affinity graph to encourage it to have an explicit cluster structure. An efficient optimization algorithm is developed to tackle the resultant optimization problem. Extensive experiments on benchmark datasets validate the superiority of the proposed method.

Datascape: exploring heterogeneous dataspace

Data science is a powerful field for gaining insights, comparing, and predicting behaviors from datasets. However, the diversity of methods and hypotheses needed to abstract a dataset exhibits a lack of genericity. Moreover, the shape of a dataset, which structures its contained information and uncertainties, is rarely considered. Inspired by state-of-the-art manifold learning and hull estimations algorithms, we propose a novel framework, the datascape, that leverages topology and graph theory to abstract heterogeneous datasets. Built upon the combination of a nearest neighbor graph, a set of convex hulls, and a metric distance that respects the shape of the data, the datascape allows exploration of the dataset’s underlying space. We show that the datascape can uncover underlying functions from simulated datasets, build predictive algorithms with performance close to state-of-the-art algorithms, and reveal insightful geodesic paths between points. It demonstrates versatility through ecological, medical, and simulated data use cases.

Structure-aware preserving projections with applications to medical image clustering

The application of dimensionality reduction (DR) in effectively handling high-dimensional data is becoming increasingly prominent. However, mainstream methods in projection learning exhibit a reliance on either global or local structures, resulting in the inability to effectively learn the discriminative projection. To address this issue, we propose a novel unsupervised DR algorithm, called Structure-Aware Preserving Projections (SAPP). Specifically, SAPP exploits a powerful combination of subspace clustering and graph construction to effectively capture the intrinsic structure of the data. It is able to find discriminative projections, which is mainly attributed to two key factors: (1) can capture the local structure of the data by constructing the nearest neighbor graph and introducing the grouping effect of the representation into projection learning; and (2) preserves the global structure and respects the complex relationship between data points by integrating least squares regression into the dimension reduction process. Moreover, SAPP proves to be a versatile framework that is easily extended to semi-supervised scenarios. Extensive experiments on medical image datasets confirm the effectiveness of our proposed method, showcasing superior clustering accuracy compared to state-of-the-art approaches.

Sybil Attacks Detection and Traceability Mechanism Based on Beacon Packets in Connected Automobile Vehicles.

Connected Automobile Vehicles (CAVs) enable cooperative driving and traffic management by sharing traffic information between them and other vehicles and infrastructures. However, malicious vehicles create Sybil vehicles by forging multiple identities and sharing false location information with CAVs, misleading their decisions and behaviors. The existing work on defending against Sybil attacks has almost exclusively focused on detecting Sybil vehicles, ignoring the traceability of malicious vehicles. As a result, they cannot fundamentally alleviate Sybil attacks. In this work, we focus on tracking the attack source of malicious vehicles by using a novel detection mechanism that relies on vehicle broadcast beacon packets. Firstly, the roadside units (RSUs) randomly instruct vehicles to perform customized key broadcasting and listening within communication range. This allows the vehicle to prove its physical presence by broadcasting. Then, RSU analyzes the beacon packets listened to by the vehicle and constructs a neighbor graph between the vehicles based on the customized particular fields in the beacon packets. Finally, the vehicle's credibility is determined by calculating the edge success probability of vehicles in the neighbor graph, ultimately achieving the detection of Sybil vehicles and tracing malicious vehicles. The experimental results demonstrate that our scheme achieves the real-time detection and tracking of Sybil vehicles, with precision and recall rates of 98.53% and 95.93%, respectively, solving the challenge of existing detection schemes failing to combat Sybil attacks from the root.

Orthogonal graph regularized non-negative matrix factorization under sparse constraints for clustering

The standard NMF algorithm is not suitable for sampling data from low-dimensional manifolds embedded in high-dimensional environmental spaces, as the geometric information hidden in feature manifolds and sample manifolds is rarely learned. In order to obtain better clustering performance based on NMF, manifold and orthogonal constraint, a new type of model named Orthogonal Graph regularized Non-negative Matrix Factorization model under Sparse Constraints (OGNMFSC) is proposed. Firstly, this type of model constructs a nearest neighbor graph to encode the geometric information of the data space, in order to obtain more discriminative ability by preserving the structure of the graph. Secondly, this type of model adds orthogonal constraints to achieve better local representation and significantly reduce the inconsistency between the original matrix and the basis vectors. Thirdly, by adding sparse constraints to obtain a sparser representation matrix, the clustering performance of the model can be improved. The main conclusion of this paper is that two effective algorithms have been generated to solve the model, which not only provides theoretical convergence proof for these two algorithms, but also demonstrates significant clustering performance in experiments compared to classical models such as K-means, PCA, NMF, Semi-NMF, NMFSC, ONMF, GNMF, NeNMF.

Structural Information Enhanced Graph Representation for Link Prediction

Link prediction is a fundamental task of graph machine learning, and Graph Neural Network (GNN) based methods have become the mainstream approach due to their good performance. However, the typical practice learns node representations through neighborhood aggregation, lacking awareness of the structural relationships between target nodes. Recently, some methods have attempted to address this issue by node labeling tricks. However, they still rely on the node-centric neighborhood message passing of GNNs, which we believe involves two limitations in terms of information perception and transmission for link prediction. First, it cannot perceive long-range structural information due to the restricted receptive fields. Second, there may be information loss of node-centric model on link-centric task. In addition, we empirically find that the neighbor node features could introduce noise for link prediction. To address these issues, we propose a structural information enhanced link prediction framework, which involves removing the neighbor node features while fitting neighborhood graph structures more focused through GNN. Furthermore, we introduce Binary Structural Transformer (BST) to encode the structural relationships between target nodes, complementing the deficiency of GNN. Our approach achieves remarkable results on multiple popular benchmarks, including ranking first on ogbl-ppa, ogbl-citation2 and Pubmed.

Hierarchical and Incremental Structural Entropy Minimization for Unsupervised Social Event Detection

As a trending approach for social event detection, graph neural network (GNN)-based methods enable a fusion of natural language semantics and the complex social network structural information, thus showing SOTA performance. However, GNN-based methods can miss useful message correlations. Moreover, they require manual labeling for training and predetermining the number of events for prediction. In this work, we address social event detection via graph structural entropy (SE) minimization. While keeping the merits of the GNN-based methods, the proposed framework, HISEvent, constructs more informative message graphs, is unsupervised, and does not require the number of events given a priori. Specifically, we incrementally explore the graph neighborhoods using 1-dimensional (1D) SE minimization to supplement the existing message graph with edges between semantically related messages. We then detect events from the message graph by hierarchically minimizing 2-dimensional (2D) SE. Our proposed 1D and 2D SE minimization algorithms are customized for social event detection and effectively tackle the efficiency problem of the existing SE minimization algorithms. Extensive experiments show that HISEvent consistently outperforms GNN-based methods and achieves the new SOTA for social event detection under both closed- and open-set settings while being efficient and robust.

Large-scale graph-machine-learning surrogate models for 3D-flowfield prediction in external aerodynamics

The article presents the application of inductive graph machine learning surrogate models for accurate and efficient prediction of 3D flow for industrial geometries, explicitly focusing here on external aerodynamics for a motorsport case. The final aim is to build a surrogate model that can provide quick predictions, bypassing in this way the unfeasible computational burden of traditional computational fluid dynamics (CFD) simulations. We investigate in this contribution the usage of graph neural networks, given their ability to smoothly deal with unstructured data, which is the typical context for industrial simulations. We integrate an efficient subgraph-sampling approach with our model, specifically tailored for large dataset training. REV-GNN is the chosen graph machine learning model, that stands out for its capacity to extract deeper insights from neighboring graph regions. Additionally, its unique feature lies in its reversible architecture, which allows keeping the memory usage constant while increasing the number of network layers. We tested the methodology by applying it to a parametric Navier–Stokes problem, where the parameters control the surface shape of the industrial artifact at hand, here a motorbike.

Abstract 6189: Deep-learning model trained on multiplex immunofluorescence-stained tissue samples predicts the survival of patients with non-small cell lung cancer better than PD-L1 TPS alone

Abstract In non-small cell lung cancer (NSCLC), the composition and structure of the tumor immune microenvironment (TME) critically impact patient outcomes, necessitating improved computational approaches to characterize the TME beyond traditional metrics like the PD-L1 tumor proportion score (TPS). Here, we developed an interpretable deep-learning model that robustly predicts patient overall survival (OS) and identifies unique PD-L1 enriched neighborhoods linked to patient prognosis. We utilized a dataset of 507 NSCLC patient biopsy samples prospectively acquired as part of the DFCI ImmunoPROFILE project. Each sample was stained for Cytokeratin as tumor marker, PD-L1, PD-1, CD8, and FOXP3 using a targeted multiplex immunofluorescence assay. Inner tumor regions of interest were automatically processed for spatially resolved identification and quantification of marker expression at the single-cell level. The dataset was divided into 380 patients for model development and 127 for testing. The deep-learning model, a graph neural network (GNN), was trained to predict OS based on local cell neighborhoods using a weakly supervised training paradigm. The GNN models neighborhoods as a graph, with cells and their corresponding marker expression representing nodes and edges connecting adjacent cells, for an average of 42 cells per neighborhood. Survival predictions were obtained by averaging predictions over all neighborhoods from one sample. The GNN model demonstrated robust performance with a concordance index (c-index) of 0.79 on the test data, surpassing traditional metrics such as TPS and immune marker density-based models (c-index: 0.50 and 0.74, respectively). Beyond assessing performance, we investigated the features learned by the GNN using k-means clustering in the model’s feature space. We identified three clusters that were notably enriched in PD-L1 positive tumor cells. The first cluster shows an especially “high TPS” phenotype, where the GNN predicted such neighborhoods to be generally poor for survival. The other two clusters were additionally enriched with PD-L1 positive immune cells (ICs) and associated with favorable survival predictions. Notably, one cluster was also enriched with other types of ICs (”PD-L1 IC mixed”), which the GNN predicted as especially favorable for survival compared to exclusive enrichment of PD-L1 ICs (”PD-L1 IC dominant”). A score summarizing these three PD-L1 expression patterns was significantly associated with OS (p-value: 0.01; HR: 0.72, 95% CI: 0.16-1.27), validating the features learned by the GNN. In sum, we developed a highly interpretable deep-learning model leveraging cellular neighborhood graphs, which effectively identifies expression patterns correlating with NSCLC prognosis, offering promise for developing personalized treatment strategies. Citation Format: Katharina Viktoria Hoebel, James R. Lindsay, Joao V. Alessi, Jason L. Weirather, Ian D. Dryg, Jennifer Altreuter, Mark M. Awad, Scott J. Rodig, William E. Lotter. Deep-learning model trained on multiplex immunofluorescence-stained tissue samples predicts the survival of patients with non-small cell lung cancer better than PD-L1 TPS alone [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 6189.

Abstract 2295: Circulating T cell receptor repertoire analysis improves cancer early detection

Abstract Background: Cancer screening by liquid biopsy promises to detect cancer early when it can be cured. However, current liquid biopsy approaches based on circulating tumor DNA (ctDNA) detect only a minority of early-stage cancers, because these tumors shed little ctDNA and plasma sample volume is limited. For liquid biopsy to realize its potential for the early detection of cancer, sensitivity for the earliest stage disease must be improved. Methods: We set out to improve sensitivity of liquid biopsy for cancer early detection by exploiting tumor recognition by T cells through sequencing of the circulating T cell receptor repertoire. Using gDNA extracted from blood buffy coats, we studied a cohort of 451 lung cancer patients (86% with stage I disease) and 525 subjects without cancer, enriched for older individuals with a history of smoking. We performed genomic DNA TCR beta chain sequencing to yield a median of 81,088 TCR clonotypes in cancer patients and 112,588 clonotypes in non-cancer controls. We built a TCR sequence similarity nearest neighbor graph to cluster >68M distinct TCR clonotypes into TCR repertoire functional units (RFUs). The TCR frequencies of RFUs were tested for association with cancer status and significantly associated RFUs were combined using a support vector machine model into a TCR RFU cancer score. This model was evaluated by 10-fold cross-validation (CV) and compared to a ctDNA panel of 237 mutation hotspots in 154 lung cancer driver genes (IDT™) and to a panel of 17 lung cancer related protein biomarkers (Olink®) in 87 subjects. Results: We identified 177 cancer-associated TCR RFUs at FDR≤0.10, including 83 RFUs that were enriched in cancer samples with a TCR count log2-fold change between 0.05 and 1.31, and 94 that were enriched in controls with a log2-fold change between 0.07 and 0.40. We observed 56 of the 83 cancer-enriched RFUs had decreasing TCR counts with increasing age across the cohort, coupled with higher counts in cancer patients relative to age-matched controls. The TCR RFU cancer score achieved a cross-validated ROC AUC of 0.69 for cancer status prediction, with the AUC for stage I cancers similar to that of stage II-IV (0.69 vs. 0.70). 44% of stage I cancers (test samples of each CV fold) were detected by the TCR model at a specificity of 80%. We saw a substantial gain in sensitivity for stage I cancer when TCR RFUs were added to ctDNA and proteins, with a ~15-20%-point increase seen at the 90% target specificity level typical for cancer screening tests. Likewise, sensitivity increased at the 99% specificity level used in multi-cancer early detection, with 35% of stage I lung cancers detected, which could not be achieved by any analyte alone. Conclusion: We demonstrate detection of a significant fraction of early lung cancer cases from blood by analyzing the circulating TCR repertoire and show that this signal is complementary to established analytes such as proteins and ctDNA. Citation Format: Yilong Li, Michelle Nahas, Dennis Stephens, Kate Froburg, Emma Hintz, Devin Champagne, Amaneet Lochab, Markus Brown, Jesse Abhyankar, Jasper Braun, Michelle Dinh, María Antonia Fortuño, María-del-Mar Ocon, Andrea Pasquier, Inés María Luque, Christopher W. Seder, Jeffrey A. Borgia, Luis Miguel Seijo, Luis M. Montuenga, Roman Yelensky. Circulating T cell receptor repertoire analysis improves cancer early detection [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2295.

Online course evaluation model based on graph auto-encoder

In the post-epidemic era, online learning has gained increasing attention due to the advancements in information and big data technology, leading to large-scale online course data with various student behaviors. Online data mining has become a popular and important way of extracting valuable insights from large amounts of data. However, previous online course analysis methods often focused on individual aspects of the data and neglected the correlation among the large-scale learning behavior data, which can lead to an incomplete understanding of the overall learning behavior and patterns within the online course. To solve the problems, this paper proposes an online course evaluation model based on a graph auto-encoder. In our method, the features of collected online course data are used to construct K-Nearest Neighbor(KNN) graphs to represent the association among the courses. Then the variational graph auto-encoder(VGAE) is introduced to learn the useful implicit features. Finally, we feed the learned implicit features into unsupervised and semi-supervised downstream tasks for online course evaluation, respectively. We conduct experiments on two datasets. In the clustering task, our method showed a more than tenfold increase in the Calinski-Harabasz index compared to unoptimized features, demonstrating significant structural distinction and group coherence. In the classification task, compared to traditional methods, our model exhibited an overall performance improvement of about 10%, indicating its effectiveness in handling complex network data.

Heterogeneous graph community detection method based on K-nearest neighbor graph neural network

Traditional community detection models either ignore the feature space information and require a large amount of domain knowledge to define the meta-paths manually, or fail to distinguish the importance of different meta-paths. To overcome these limitations, we propose a novel heterogeneous graph community detection method (called KGNN_HCD, heterogeneous graph Community Detection method based on K-nearest neighbor Graph Neural Network). Firstly, the similarity matrix is generated to construct the topological structure of K-nearest neighbor graph; secondly, the meta-path information matrix is generated using a meta-path transformation layer (Mp-Trans Layer) by adding weighted convolution; finally, a graph convolutional network (GCN) is used to learn high-quality node representation, and the k-means algorithm is adopted on node embeddings to detect the community structure. We perform extensive experiments and on three heterogeneous datasets, ACM, DBLP and IMDB, and we consider as competitors 11 community detection methods such as CP-GNN and GTN. The experimental results show that the proposed KGNN_HCD method improves 2.54% and 2.56% on the ACM dataset, 2.59% and 1.47% on the DBLP dataset, and 1.22% and 1.67% on the IMDB dataset for both NMI and ARI. Experiments findings suggest that the proposed KGNN_HCD method is reasonable and effective, and KGNN_HCD can be applied to complex network classification and clustering tasks.

Intelligent fault diagnosis of ultrasonic motors based on graph-regularized CNN-BiLSTM

The ultrasonic motor (USM) is peculiarly prone to failure due to continuous high-frequency friction-related power transfer, whose failure mechanisms are remarkably different from traditional induction motors. Intelligent fault diagnosis provides a way to alarm and avoid catastrophic losses proactively. However, previous studies using deep learning usually ignore the inherent geometric structure of the signal distribution. This paper proposes an intelligent multi-signal fault diagnosis framework for USMs to restore the linear or nonlinear manifold structure by preserving the internal structure by integrating graph regularization with deep neural networks. Firstly, the one-dimensional CNN to learn spatial correlations and BiLSTM to exploit temporal dependencies are coalesced to build the deep neural network. Then, an improved k-nearest neighbor graph is proposed to protect the geometric structure information and force the latent features to be more concentrated within their classes. Moreover, the layer in the deep architecture to integrate graph regularization is designed to reduce computation cost, and an adaptive decay strategy is considered to adjust the coefficient of graph regularized automatically. A two-stage training algorithm is developed by considering the time to calculate the graph regularization term. Finally, the proposed multi-signal fault diagnosis framework is validated using datasets from the fault injection experiment of similar USMs in China’s Yutu rover of Chang’e lunar probe. Experimental results show that the proposed method can effectively discriminate different fault types.