No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Construction and Evolution of Fault Diagnosis Knowledge Graph in Industrial Process
Abstract
The steel industry production line is complicated and contains a substantial amount of equipment, leading to serious problems in fault diagnosis such as wrong inspection strategies, fault location, maintenance, and so on. To realize accurate and efficient equipment fault diagnosis, this paper proposes a steel production line equipment fault diagnosis knowledge graph (SPLEFD-KG) based on a novel relation-oriented model with global context information for jointly extracting overlapping relations and entities (ROMGCJE). A low-level self-learning SPLEFD-KG is first constructed using the triples extracted by ROMGCJE. However, this low-level SPLEFD-KG is incomplete, and only contains sparse paths for fault reasoning. To overcome this problem, a reinforcement learning framework is applied to mine hidden semantic knowledge to complete the missing relation. Besides, the graph neural networks are introduced to compute the embedding vector of new entities outside of the SPLEFD-KG for continuously completing missing entities. Finally, the low-level self-learning SPLEFD-KG evolves to one high-level SPLEFD-KG, which can provide information-rich and accurate fault-related knowledge. Extensive experiments conducted on the steel production line equipment failure dataset indicate that the novel SPLEFD-KG significantly improves fault diagnosis results and provides effective maintenance programs.
... In recent years, KGs have attracted much attention, and various approaches have emerged [12], by constructing a graph-based data organization structure [13] that describes the various concepts, entities, and relationships between entities of a device; provides better organization, management, and understanding of large amounts of information [14]; supports efficient data retrieval; and handles complex and varied associative representations [15] to form a KG for the troubleshooting of RM equipment. Fault knowledge is organized and stored as ternary groups to achieve semantic representation of faults, which is conducive to the mining of diagnostic knowledge [16][17][18], solves the problem of the accuracy of personnel diagnosis and decision-making, improves the intelligence of equipment FD, and reduces the occurrence of issues such as unplanned shutdowns of steel production line equipment. Li et al. [19] proposed an alarm KG based on graph neural networks to guide fault localization using network alarm knowledge. ...
... This study covers KG building in its KG-based malfunction diagnostics of RM [16]: (1) KG construction; (2) data acquisition and process; (3) feature match and inference; (4) FD and repair. ...
In the process of rotating machinery fault diagnosis (RMFD), the lack of feature conditions leads to the problem of low accuracy of traditional rule-based reasoning methods FD. This paper proposed a knowledge graph (KG)-driven device FD method and applied it to RMFD. First, we proposed a multi-level KG construction method to get multi-source data based on each level and analyzed the levels that affected the system state. A single-level KG was constructed through data features, and a multilevel KG with a stereostructure was built using a multi-source data fusion model as data support for FD. Second, we proposed an approach based on multilevel KG and Bayes theory to detect the system state and located the source of faults by combining the KG reasoning based on relational paths, then used the relationships between the structures of rotating mechanical equipment for fault cause reasoning and used the KG as a knowledge base for a reason using machine learning. Finally, the proposed method was validated using a steelworks motor as an example and compared with other ways, such as rule-based FD. The results show that under the condition of missing input features, the accuracy of the proposed method reaches 91.1%, which is significantly higher than other methods and effectively solves the problem of low diagnostic accuracy.
... Although KG technology has not been previously applied in railway alignment design, many KGs, such as WikiData (Vrandečić & Krötzsch, 2014), Freebase (Bollacker, Evans, Paritosh, Sturge, & Taylor, 2008) and YAGO (Hoffart, Suchanek, Berberich, & Weikum, 2013), have been constructed in the big data field. KG has been successfully extended in fields such as medical diagnosis (Han, Pan, et al., 2022;Wang, Luo, He, & Shao, 2023), mechanical fault diagnosis (Han, Wang, Wang, & Chen, 2022), enterprise efficient supply chain management (Deng, Chen, Huang, Chen, & Cheng, 2023), geoscience studies (Liu et al., 2023), surveying and remote-sensing technology (Hao et al., 2021) by the virtue of relation expression and complex relation reasoning. ...
... With KG as a core enabler, ) designed a reasoning approach using an equipment-based schema while (Mo et al. 2023) integrated digital twins and AI algorithms for dynamic reconfiguration of production systems to support decision-making based on user needs. In the context of diagnostics, (Xia et al. 2023) presented a link prediction approach to predict and explain potential faults in maintenance tasks while (Han et al. 2022) used a self-learning relation-oriented KG model to extract overlapping relations and entities for fault diagnostics recommendations. Alternative reasoning methods include case-based reasoning (Nakhjiri, Salamó, and Sànchezmarrè 2020), and graph-based causal linguistic patterns recognition for root cause analysis (Xu and Dang 2023). ...
The increasing complexity of industrial systems demands more effective and intelligent maintenance approaches to address manufacturing defects arising from faults in multiple asset modules. Traditional digital twin (DT) systems, however, face limitations in interoperability, knowledge sharing, and causal inference. As such, cognitive digital twins (CDTs) can add value by managing a collaborative web of interconnected systems, facilitating advanced cross-domain analysis and dynamic context considerations. This paper introduces a CDT system that leverages industrial knowledge graphs (iKGs) to support maintenance planning and operations. By employing a design structure matrix (DSM) to model dependencies and relationships, a semantic translation approach maps the knowledge into a graph-based representation for reasoning and analysis. An automatic solution generation mechanism, utilising graph sequencing with Louvain and PageRank algorithms, derives feasible solutions, which can be validated via simulation to minimise production disruption impacts. The CDT system can also identify potential disruptions in new product designs, thus enabling preventive actions to be taken. A case study featuring a print production manufacturing line illustrates the CDT system’s capabilities in causal inference and solution explainability. The study concludes with a discussion of limitations and future directions, providing valuable guidelines for manufacturers aiming to enhance reactive and predictive maintenance strategies.
... For instance, Bai et al created a knowledge graph that simultaneously extracts the system's entities and interrelationships to mine relationships between processes [18]. Similarly, Han et al developed a self-learning adaptive network based on a knowledge graph, enabling the knowledge fusion of human-cyber-physical systems [19]. Nevertheless, they are capable of successfully expressing a system's prior knowledge, these methods have certain limitations. ...
Each chemical process industry system possesses unique process knowledge, which serves as a representation of the system's state. As graph-theory based methods are capable of embedding process knowledge, they have become increasingly crucial in the field of process industry diagnosis. The fault representation ability of the diagnosis model is directly associated with the quality of the graph. Unfortunately, simple fully connected graphs fail to strengthen the internal connections within the same process but weaken the interactive connections between different processes. Moreover, each node in the graph is considered equally important, making it impossible to prioritize crucial system monitoring indicators. To address the above shortcomings, this paper presents a spatial weighted graph-driven fault diagnosis method of complex process industry considering technological process flow. Initially, the physical space sensor layout of the technological process flow is mapped into the spatial graph structure, where each sensor is regarded as a node and these nodes are connected by the k nearest neighbor algorithm. Subsequently, according to the mechanism knowledge, the sensors in the process are divided into different importance categories and weight coefficients are assigned to their nodes. The similarities between these weighted nodes are calculated, and the resulting edge information are used to construct the spatial weighted graphs. Finally, the spatial weighted graphs are input to a graph convolutional network, facilitating fault representation learning for fault diagnosis of complex process industry. Validation experiments are conducted using public industrial datasets, and the results demonstrate that the proposed method can effectively integrate the process knowledge to improve the fault diagnosis accuracy of the model.
Multi-hop question answering (QA) over knowledge graph (KG) poses significant challenges in the context of industrial processes, due to the intricate semantics of natural language queries and the obscure relationships among triples in the KG. This paper presents a multi-hop fault diagnosis QA framework based on KG reasoning and decision, called KG-MQA. It consists of three modules: 1) The embedding module is designed to learn representations of queried questions, and encode entities and relations within the input fault diagnosis KG. 2) The KG reasoning module is proposed to learn to reason candidate answers based on query entities and context. 3) The answer decision module is applied to obtain final answer by leveraging the attention mechanism to compute the attention scores between the input question and candidate answers. Extensive experiments have been conducted on the dataset from a hot rolling line, and our KG-MQA achieves an average accuracy of 95.7%, surpassing other state-of-the-art models. Ablation studies demonstrate the efficiency of KG-MQA in helping users obtain equipment fault-related information from massive data through the professional knowledge structure. It facilitates reasoning during data queries to efficiently provide multiple rounds of QA for equipment fault diagnosis in hot rolling lines.
The knowledge graph is widely used in industrial fields due to its structural characteristics. In order to reduce the cost of wrong decision-making, it is more important for the industrial knowledge graph to guarantee the quality and comprehensiveness of knowledge. In order to obtain the trustworthiness information of triples in the graph, this paper proposed a model to evaluate triple trustworthiness for industrial knowledge graphs(TT-IKG) and obtains the final score of triples by fusing the output of three sub-modules that investigate the local confidence of triples, the confidence of schema-matching and the confidence of global paths. Experiments on the real industrial knowledge graph of enterprises verified the effectiveness of the model, and the experimental results on the error detection and graph completion tasks of the knowledge graph show that the effect is better than that of other models.
The safety of batteries has become a major obstacle to the promotion and application of electric vehicles, and the use of cloud-based vehicle practical big data to summarize the fault knowledge of batteries to improve product quality and reduce maintenance costs has attracted widespread attention from academia and industrial communities. In this paper, a method is proposed to construct the battery fault knowledge graph which supports online knowledge query and fault inference. Reliability models for battery undervoltage, inconsistency, and capacity loss are built based on cloud data, and are deployed and continuously updated in the cloud platform to accommodate the migration of the models to different battery products. A bidirectional long short-term memory (Bi-LSTM) neural network was established for knowledge extraction of fault logs, and the results were imported into Neo4j to form a battery fault knowledge graph. Finally, a fault knowledge online query front-end interface was built to conduct inference tests on battery faults of a manufacturer, which proves the feasibility and effectiveness of the proposed method.
In complex power systems, when power equipment fails, multiple concurrent failures usually occur instead of a single failure. Concurrent failures are so common and hidden in complex systems that diagnosis requires not only analysis of failure characteristics, but also correlation between failures. Therefore, in this paper, a concurrent fault diagnosis method is proposed for power equipment based on graph neural networks and knowledge graphs. First, an electrical equipment failure knowledge map is created based on operational and maintenance records to emphasize the relevance of the failed equipment or component. Next, a lightweight graph neural network model is built to detect concurrent faults in the graph data. Finally, a city’s transformer concurrent fault is taken as an example for simulation and validation. Simulation results show that the accuracy and acquisition rate of graph neural network mining in Knowledge Graph is superior to traditional algorithms such as convolutional neural networks, which can achieve the effectiveness and robustness of concurrent fault mining.
Extracting entities and relations from unstructured text is an essential task in the field of information extraction. Existing work mainly pipeline extraction and joint decoding methods. However, these methods are unable to extract overlapping entities and relations, and ignore the task correlation between entity and relation extraction. In this paper, we first introduce the BERT pre-training model to model the text more finely. Then, we decompose the extraction into relation extraction and entity recognition. Relation extraction is transformed into a relation classification task. Entity recognition is transformed into a sequence labeling task. The recognition entity includes a head entity and a tail entity. We evaluate the model on the New York Times (NYT) and WebNLG datasets. Compared with most existing models, excellent results have been obtained. Experimental results show that our model can fully capture the semantic interdependence between the two tasks of entity and relation extraction, reduce the interference of unrelated entity pairs, and effectively solve the problem of entity overlap.
Relation extraction aims to identify semantic relations between entities in text. In recent years, this task has been extended to the joint extraction of entities and relations, which requires the simultaneous identification of entities and their relations from sentences. However, existing methods, limited by the existing tagging scheme, fail to identify more complex entities, which in turn limits the performance of the joint extraction task. This article presents a joint extraction model for entities and relations called MLRA-LSTM-CRF that uses multi-label tagging and relational alignment to transform this task into a multi-label tag recognition problem. The proposed model first tags the entities and their relations according to the multi-label tagging scheme and then uses a joint entity and relation extraction module with a multi-layer attention mechanism to extract the triplets in the sentence. Finally, the relational alignment module is used to align the predicted relation classification results. Experimental results on the New York Times and Wiki-KBP datasets indicate that MLRA-LSTM-CRF is significantly better than that of several state-of-the-art models and baseline.
A knowledge graph is a collection of triples, often represented in the form of “subject,” “relation,” “object.” The task of knowledge graph completion (KGC) is to automatically predict missing links by reasoning over the information already present in the knowledge graph. Recent popularization of graph neural networks has also been spread to KGC. Typical techniques like SACN achieve dramatic achievements and beat previous state-of-the-art. However, those models still lack the ability to distinguish different local structures within a graph, which leads to the over smoothing problem. In this work, we propose SD-GAT, a graph attention network with a structure-distinguishable neighborhood aggregation scheme, which models the injective function to aggregate information from the neighborhood. The model is constituted of two modules. The encoder is a graph attention network that improved with our neighborhood aggregation scheme, which could be applied for a more distinct representation of entities and relations. The decoder is a convolutional neural network using \(3\times 3\) convolution filters. Our empirical research provides an effective solution to increase the discriminative power of graph attention networks, and we show significant improvement of the proposed SD-GAT compared to the state-of-the-art methods on standard FB15K-237 and WN18RR datasets.
Human knowledge provides a formal understanding of the world. Knowledge graphs that represent structural relations between entities have become an increasingly popular research direction toward cognition and human-level intelligence. In this survey, we provide a comprehensive review of the knowledge graph covering overall research topics about: 1) knowledge graph representation learning; 2) knowledge acquisition and completion; 3) temporal knowledge graph; and 4) knowledge-aware applications and summarize recent breakthroughs and perspective directions to facilitate future research. We propose a full-view categorization and new taxonomies on these topics. Knowledge graph embedding is organized from four aspects of representation space, scoring function, encoding models, and auxiliary information. For knowledge acquisition, especially knowledge graph completion, embedding methods, path inference, and logical rule reasoning are reviewed. We further explore several emerging topics, including metarelational learning, commonsense reasoning, and temporal knowledge graphs. To facilitate future research on knowledge graphs, we also provide a curated collection of data sets and open-source libraries on different tasks. In the end, we have a thorough outlook on several promising research directions.
A relation tuple consists of two entities and the relation between them, and often such tuples are found in unstructured text. There may be multiple relation tuples present in a text and they may share one or both entities among them. Extracting such relation tuples from a sentence is a difficult task and sharing of entities or overlapping entities among the tuples makes it more challenging. Most prior work adopted a pipeline approach where entities were identified first followed by finding the relations among them, thus missing the interaction among the relation tuples in a sentence. In this paper, we propose two approaches to use encoder-decoder architecture for jointly extracting entities and relations. In the first approach, we propose a representation scheme for relation tuples which enables the decoder to generate one word at a time like machine translation models and still finds all the tuples present in a sentence with full entity names of different length and with overlapping entities. Next, we propose a pointer network-based decoding approach where an entire tuple is generated at every time step. Experiments on the publicly available New York Times corpus show that our proposed approaches outperform previous work and achieve significantly higher F1 scores.
Link prediction for knowledge graphs is the task of predicting missing relationships between entities. Previous work on link prediction has focused on shallow, fast models which can scale to large knowledge graphs. However, these models learn less expressive features than deep, multi-layer models - which potentially limits performance. In this work we introduce ConvE, a multi-layer convolutional network model for link prediction, and report state-of-the-art results for several established datasets. We also show that the model is highly parameter efficient, yielding the same performance as DistMult and R-GCN with 8x and 17x fewer parameters. Analysis of our model suggests that it is particularly effective at modelling nodes with high indegree - which are common in highly-connected, complex knowledge graphs such as Freebase and YAGO3. In addition, it has been noted that the WN18 and FB15k datasets suffer from test set leakage, due to inverse relations from the training set being present in the test set - however, the extent of this issue has so far not been quantified. We find this problem to be severe: a simple rule-based model can achieve state-of-the-art results on both WN18 and FB15k. To ensure that models are evaluated on datasets where simply exploiting inverse relations cannot yield competitive results, we investigate and validate several commonly used datasets - deriving robust variants where necessary. We then perform experiments on these robust datasets for our own and several previously proposed models, and find that ConvE achieves state-of-the-art Mean Reciprocal Rank across all datasets.
In this work, we introduce a convolutional neural network model, ConvE, for the task of link prediction where we apply 2D convolution directly on embeddings, thus inducing spatial structure in embedding space. To scale to large knowledge graphs and prevent overfitting due to over-parametrization, previous work seeks to reduce parameters by performing simple transformations in embedding space. We take inspiration from computer vision, where convolution is able to learn multiple layers of non-linear features while reducing the number of parameters through weight sharing. Applied naively, convolutional models for link prediction are computationally costly. However, by predicting all links simultaneously we improve test time performance by more than 300x on FB15k. We report state-of-the-art results for numerous previously introduced link prediction benchmarks, but concerns have been raised regarding the quality of the well-established FB15k and WN18 datasets. Previous work noted that these datasets contain many reversible triples, but the severity of this issue was not quantified. To investigate this, we design a simple model that uses a single rule which reverses relations and achieves state-of-the-art results. We introduce WN18RR, a subset of WN18 which was constructed the same way as the previously proposed FB15k-237, to alleviate this problem and report results for our own and previously proposed models for all datasets. Analysis of our convolutional model suggests that it is particularly good at modelling nodes with high indegree and nodes with high PageRank and that 2D convolution applied on embeddings seems to induce contrasting pixel-level structures.
State-of-the-art models for semantic segmentation are based on adaptations of convolutional networks that had originally been designed for image classification. However, dense prediction problems such as semantic segmentation are structurally different from image classification. In this work, we develop a new convolutional network module that is specifically designed for dense prediction. The presented module uses dilated convolutions to systematically aggregate multi-scale contextual information without losing resolution. The architecture is based on the fact that dilated convolutions support exponential expansion of the receptive field without loss of resolution or coverage. We show that the presented context module increases the accuracy of state-of-the-art semantic segmentation systems. In addition, we examine the adaptation of image classification networks to dense prediction and show that simplifying the adapted network can increase accuracy.
Recently, knowledge graph embedding, which projects symbolic entities and
relations into continuous vector space, has become a new, hot topic in
artificial intelligence. This paper addresses a new issue of \textbf{multiple
relation semantics} that a relation may have multiple meanings revealed by the
entity pairs associated with the corresponding triples, and proposes a novel
Gaussian mixture model for embedding, \textbf{TransG}. The new model can
discover latent semantics for a relation and leverage a mixture of relation
component vectors for embedding a fact triple. To the best of our knowledge,
this is the first generative model for knowledge graph embedding, which is able
to deal with multiple relation semantics. Extensive experiments show that the
proposed model achieves substantial improvements against the state-of-the-art
baselines.
We deal with embedding a large scale knowledge graph composed of entities and relations into a continuous vector space. TransE is a promising method proposed recently, which is very efficient while achieving state-of-the-art predictive performance. We discuss some mapping properties of relations which should be considered in embedding, such as reflexive, one-to-many, many-to-one, and many-to-many. We note that TransE does not do well in dealing with these properties. Some complex models are capable of preserving these mapping properties but sacrifice efficiency in the process. To make a good trade-off between model capacity and efficiency, in this paper we propose TransH which models a relation as a hyperplane together with a translation operation on it. In this way, we can well preserve the above mapping properties of relations with almost the same model complexity of TransE. Additionally, as a practical knowledge graph is often far from completed, how to construct negative examples to reduce false negative labels in training is very important. Utilizing the one-to-many/many-to-one mapping property of a relation, we propose a simple trick to reduce the possibility of false negative labeling. We conduct extensive experiments on link prediction, triplet classification and fact extraction on benchmark datasets like WordNet and Freebase. Experiments show TransH delivers significant improvements over TransE on predictive accuracy with comparable capability to scale up.
Knowledge graphs (KGs) provide a wealth of prior knowledge for the research on social networks. Cross-lingual entity alignment aims at integrating complementary KGs from different languages and thus benefits various knowledge-driven social network studies. Recent entity alignment methods often take an embedding-based approach to model the entity and relation embedding of KGs. However, these studies mostly focus on the information of the entity itself and its structural features but ignore the influence of multiple types of data in KGs. In this paper, we propose a new embedding-based framework named multiview highway graph convolutional network (MHGCN), which considers the entity alignment from the views of entity semantic, relation semantic, and entity attribute. To learn the structural features of an entity, the MHGCN employs a highway graph convolutional network (GCN) for entity embedding in each view. In addition, the MHGCN weights and fuses the multiple views according to the importance of the embedding from each view to obtain a better entity embedding. The alignment entities are identified based on the similarity of entity embeddings. The experimental results show that the MHGCN consistently outperforms the state-of-the-art alignment methods. The research also will benefit knowledge fusion through cross-lingual KG entity alignment.
Industrial Internet of Things (IIoT) systems connect a plethora of smart devices, such as sensors, actuators, and controllers, to enable efficient industrial productions in manners observable and controllable by human beings. Plain model-based and data-driven diagnosis approaches can be used for fault detection and isolation of specific IIoT components. However, the physical models, signal patterns, and machine learning algorithms need to be carefully designed to describe system faults. Besides, the ever-increasing level of connectivity among devices can induce exponential complexity. Knowledge-based fault diagnosis approaches improve interoperability via ontologies so that high-level reasoning and inquiry response can be provided to nonexpert users. Therefore, knowledge-based fault diagnosis approaches are preferred over plain model-based and data-driven diagnosis approaches in recent IIoT systems. In the context of IIoT systems, this work reviews the recent progress on the construction of knowledge bases via ontologies and deductive/inductive reasoning for knowledge-based fault diagnosis. Besides, general inductive reasoning methods are discussed to shed light on their successful applications in knowledge-based fault diagnosis for IIoT systems. Following the trend of large-system decentralization, future fault diagnosis also requires decentralized implementations. Therefore, we conclude this survey by discussing several interesting open problems for decentralized knowledge-based fault diagnosis for IIoT systems.
Named entity recognition (NER) isa preliminary task in natural language processing (NLP). Recognizing Chinese named entities from unstructured texts is challenging due to the lack of word boundaries. Even if performing Chinese Word Segmentation (CWS) could help to determine word boundaries, it is still difficult to determine which words should be clustered together for entity identification, since entities are often composed of multiple-segmented words. As dependency relationships between segmented words could help to determine entity boundaries, it is crucial to employ information related to syntactic dependency relationships to improve NER performance. In this paper, we propose a novel NER model to learn information about syntactic dependency graphs with graph neural networks, and merge learned information into the classic Bidirectional Long Short-Term Memory (BiLSTM) - Conditional Random Field (CRF) NER scheme. In addition, we extract various kinds of task-specific hidden information from multiple CWS and part-of-speech (POS) tagging tasks, to further improve the NER model. We finally leverage multiple self-attention components to integrate multiple kinds of extracted information for named entity identification. Experimental results on three public benchmark datasets show that our model outperforms the state-of-the-art baselines in most scenarios.
Existing methods in relation extraction have leveraged the lexical features in the word sequence and the syntactic features in the parse tree. Though effective, the lexical features extracted from the successive word sequence may introduce some noise that has little or no meaningful content. Meanwhile, the syntactic features are usually encoded via graph convolutional networks which have restricted receptive field. In addition, the relation between lexical and syntactic features in the representation space has been largely neglected. To address the above limitations, we propose a multi-scale representation and metric learning framework to exploit the feature and relation hierarchy for RE tasks. Methodologically, we
build a lexical and syntactic feature and relation hierarchy
in text data. Technically, we first develop
a multi-scale convolutional neural network
to aggregate the non-successive lexical patterns in the word sequence. We also design
a multi-scale graph convolutional network
to increase the receptive field via the coarsened syntactic graph. Moreover, we present
a multi-scale metric learning
paradigm to exploit both the feature-level relation between lexical and syntactic features and the sample-level relation between instances with the same or different classes. Extensive experiments on three public datasets for two RE tasks prove that our model achieves a new state-of-the-art performance.
With the development of sensors, more and more multimodal data are accumulated, especially in biomedical and bioinformatics fields. Therefore, multimodal data analysis becomes very important and urgent. In this study, we combine multi-kernel learning and transfer learning, and propose a feature-level multi-modality fusion model with insufficient training samples. To be specific, we firstly extend kernel Ridge regression to its multi-kernel version under the
l
<sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">p</sub>
-norm constraint to explore complementary patterns contained in multimodal data. Then we use marginal probability distribution adaption to minimize the distribution differences between the source domain and the target domain to solve the problem of insufficient training samples. Based on epilepsy EEG data provided by the University of Bonn, we construct 12 multi-modality & transfer scenarios to evaluate our model. Experimental results show that compared with baselines, our model performs better on most scenarios.
Fuzzy Petri nets (FPNs) are a promising modeling tool for knowledge representation and reasoning of rule-based expert systems. However, there exist limitations in representing ambiguous knowledge and performing approximate inference in traditional FPNs. Additionally, knowledge parameters are usually provided by few experts in existing FPN methods. In response to these issues, a new version of FPNs, named spherical linguistic Petri nets (SLPNs), is introduced in this paper for knowledge representation and reasoning in the large group context. To this end, spherical linguistic sets are applied to capture imprecise knowledge and represent the uncertainty of experts judgements. Furthermore, a large group knowledge acquisition approach is developed to determine knowledge parameters. A bidirectional inference algorithm is developed for implementing the reasoning process and identifying the root causes of an appointed event. Finally, the efficacy and superiority of our developed SLPNs are illustrated by a realistic example regarding stampede risk level assessment in a high-speed railway station.
Recently, cross-domain fault diagnosis based on transfer learning methods has been extensively explored and well-addressed when class-balance data with supervision information is available. However, data under machine faulty states are mostly difficult to collect, there is a huge divide between current transfer learning methods based on implicit class-balance data and real industrial applications. In this paper, we propose a class-imbalance adversarial transfer learning network with input being imbalanced data to learn domain-invariant and knowledge. Within this framework, class-imbalance learning is embedded into the adversarial training process to learn class-separate diagnostic knowledge with imbalanced data, double-level adversarial transfer learning including marginal and conditional distribution adaptations is conducted to learn domain-invariant knowledge. Extensive experiments on a planetary gearbox rig with imbalanced data verify the effectiveness and generalization of the proposed method and show it is superior performance over contrastive transfer learning methods. Moreover, the proposed method relaxes the underlying assumption that the success of current transfer learning regimes is rooted in class-balance data and extends the application of the transfer learning method for real-industrial cross-domain diagnosis tasks.
Knowledge inference using knowledge graphs is one of the most recent research trends for fault diagnosis. As complex Industrial Internet of Things (IIoT) systems evolve to incorporate distributed subsystems, each subsystem may have local measurements interpreted by a local Knowledge Base (KB). In this work, we propose a distributed fault diagnosis framework with distributed path-based reasoning. In contrast to centralized reasoning that requires merging of the distributed KBs, the proposed framework employs a centralized coordinator to transfer the model among decentralized agents and allows an agent to handover training paths for others to continue training by filling in missing links using their local KBs. Our work is the very first distributed knowledge inference framework to reason among distributed fault diagnosis systems to tackle the KB isolation, scalability and flexibility issues of centralized fault diagnosis systems. To validate our framework with IIoT datasets, we build a fault knowledge graph from the Tennessee Eastman (TE) process simulations. Evaluations show that our framework outperforms the non-cooperative distributed reasoning method in terms of prediction accuracy. With the TE process dataset, the mean reciprocal rank is improved from 0.1612 to 0.19 with transferred models, and further to 0.206 with hand-over queries between agents.
Chemical-induced disease (CID) relation extraction from biomedical articles plays an important role in disease treatment and drug development. Existing methods are insufficient for capturing complete document level semantic information due to ignoring semantic information of entities in different sentences. In this work, we proposed an effective document-level relation extraction model to automatically extract intra-/inter-sentential CID relations from articles. Firstly, our model employed BERT to generate contextual semantic representations of the title, abstract and shortest dependency paths (SDPs). Secondly, to enhance the semantic representation of the whole document, cross attention with self-attention (named cross2self-attention) between abstract, title and SDPs was proposed to learn the mutual semantic information. Thirdly, to distinguish the importance of the target entity in different sentences, the Gaussian probability distribution was utilized to compute the weights of the co-occurrence sentence and its adjacent entity sentences. More complete semantic information of the target entity is collected from all entities occurring in the document via our presented document-level R-BERT (DocR-BERT). Finally, the related representations were concatenated and fed into the softmax function to extract CIDs. We evaluated the model on the CDR corpus provided by BioCreative V. The proposed model without external resources is superior in performance as compared with other state-of-the-art models (our model achieves 53.5%, 70%, and 63.7% of the F1-score on inter-/intra-sentential and overall CDR dataset). The experimental results indicate that cross2self-attention, the Gaussian probability distribution and DocR-BERT can effectively improve the CID extraction performance. Furthermore, the mutual semantic information learned by the cross self-attention from abstract towards title can significantly influence the extraction performance of document-level biomedical relation extraction tasks.
With the rapid development and widespread application of Knowledge graphs (KGs) in many artificial intelligence tasks, a large number of efforts have been made to refine them and increase their quality. Knowedge graph embedding (KGE) has become one of the main refinement tasks, which aims to predict missing facts based on existing ones in KGs. However, there are still mainly two difficult unresolved challenges: (i) how to leverage the local structural features of entities and the potential soft logical rules to learn more expressive embedding of entites and relations; and (ii) how to combine these two learning processes into one unified model. To conquer these problems, we propose a novel KGE model named JSSKGE, which can
J
ointly learn the local
S
tructural features of entities and
S
oft logical rules. First, we employ graph attention networks which are specially designed for graph-structured data to aggregate the local structural information of nodes. Then, we utilize soft logical rules implicated in KGs as an expert to further rectify the embeddings of entities and relations. By jointly learning, we can obtain more informative embeddings to predict new facts. With experiments on four commonly used datasets, the JSSKGE obtains better performance than state-of-the-art approaches.
We propose methods for extracting triples from Wikipedia’s HTML tables using a reference knowledge graph. Our methods use a distant-supervision approach to find existing triples in the knowledge graph for pairs of entities on the same row of a table, postulating the corresponding relation for pairs of entities from other rows in the corresponding columns, thus extracting novel candidate triples. Binary classifiers are applied on these candidates to detect correct triples and thus increase the precision of the output triples. We extend this approach with a preliminary step where we first group and merge similar tables, thereafter applying extraction on the larger merged tables. More specifically, we propose an
observed schema
for individual tables, which is used to group and merge tables. We compare the precision and number of triples extracted with and without table merging, where we show that with merging, we can extract a larger number of triples at a similar precision. Ultimately, from the tables of English Wikipedia, we extract 5.9 million novel and unique triples for Wikidata at an estimated precision of 0.718.
The incompleteness of knowledge graphs triggers considerable research interest in relation prediction. As the key to predicting relations among entities, many efforts have been devoted to learning the embeddings of entities and relations by incorporating a variety of neighbors' information which includes not only the information from direct outgoing and incoming neighbors but also the ones from the indirect neighbors on the multihop paths. However, previous models usually consider entity paths of limited length or ignore sequential information of the paths. Either simplification will make the model lack a global understanding of knowledge graphs and may result in the loss of important and indispensable information. In this article, we propose a novel global graph attention embedding network (GGAE) for relation prediction by combining global information from both direct neighbors and multihop neighbors. Concretely, given a knowledge graph, we first introduce the path construction algorithms to obtain meaningful paths, then design path modeling methods to capture the potential long-distance sequential information in the multihop paths, final propose an entity graph attention and a relation graph attention mechanisms to obtain entity embeddings and relation embeddings. Moreover, an entity graph attention mechanism is proposed to calculate the entity embeddings by aggregating direct incoming and outgoing neighbors from: 1) an original knowledge graph with the original entity and relation embeddings and 2) a new knowledge graph constructed by the paths whose embeddings are updated by path modeling methods. for each relation, we construct a new graph with related entities and present a relation graph attention to learn the features. Therefore, our model can encapsulate the information from different distance neighbors, and enable the embeddings of entities and relations to better capture all-sided semantic information. The experimental results on benchmark datasets verify the superiority of our model over the state-of-the-art ones.
In the era of the fifth-generation fixed network (F5G), optical networks must be developed to support large bandwidth, low latency, high reliability, and intelligent management. Studies have shown that software-defined optical networks (SDON) and artificial intelligence can help improve the performance and management capabilities of optical networks. Inside a large-scale optical network, many types of alarms are reported that indicate network anomalies. Relationships between the alarms are complicated, making it difficult to accurately locate the source of the fault(s). In this work, we propose a knowledge-guided fault localization method, using network alarm knowledge to analyze network abnormalities. Our method introduces knowledge graphs (KGs) into the alarm analysis process. We also propose a reasoning model based on graph neural network (GNN), to perform relational reasoning on alarm KGs and locate the network faults. We develop an ONOS-based SDON platform for experimental verification, which includes a set of processes for the construction and application of alarm KGs. The experimental results show the proposed method has high accuracy and provide motivation for the industry-scale use of KGs for alarm analysis and fault localization.
In recent years, knowledge graph representation learning has prompted extensive research. Machine learning models are used to map entity and relational data in knowledge graphs to vector representations in low-dimensional spaces to predict and analyze potential relationships. Current works mainly focus on the knowledge representation of the triple structure and relationship path in knowledge graphs without fully utilizing external textual information to semantically supplement knowledge representation. However, the existing knowledge inventory, such as that for smart health and emotion care systems, is relatively meager, and structural knowledge is incomplete; therefore, knowledge graph completion is essential. In this paper, we propose a novel joint representation learning model that introduces text description information and extracts reliable feature information from text data by using a convolutional neural network (CNN) model. Furthermore, being based on the attention mechanism, the proposed model distinguishes the characteristic credibility of different relationships, enhances the representation of the entity relationship structure vector in the existing knowledge graph, and obtains rich semantic information. Finally, the two-dimensional (2D) convolution operation is used to process the joint representation vectors of entities and relationships to obtain nonlinear features, and the knowledge graph is completed by completing the calculation of the score function of the joint representation vector of the entity and the relationship. Experiments performing tasks, such as link prediction and triple classification, on the FreeBase (FB15k), WordNet (WN18) and Yet Another Great Ontology (YAGO3-10) datasets reveal that our model performs better than the benchmark model and has some degree of scalability.
Named entity recognition (NER) aims to recognize mentions of rigid designators from text belonging to predefined semantic types, such as person, location, and organization. In this article, we focus on a fundamental subtask of NER, named entity boundary detection, which aims at detecting the start and end boundaries of an entity mention in the text, without predicting its semantic type. The entity boundary detection is essentially a sequence labeling problem. Existing sequence labeling methods either suffer from sparse boundary tags (i.e., entities are rare and nonentities are common) or they cannot well handle the issue of variable size output vocabulary (i.e., need to retrain models with respect to different vocabularies). To address these two issues, we propose a novel entity boundary labeling model that leverages pointer networks to effectively infer boundaries depending on the input sequence. On the other hand, training models on source domains that generalize to new target domains at the test time are a challenging problem because of the performance degradation. To alleviate this issue, we propose METABDRY, a novel domain generalization approach for entity boundary detection without requiring any access to target domain information. Especially, adversarial learning is adopted to encourage domain-invariant representations. Meanwhile, metalearning is used to explicitly simulate a domain shift during training so that metaknowledge from multiple resource domains can be effectively aggregated. As such, METABDRY explicitly optimizes the capability of "learning to generalize," resulting in a more general and robust model to reduce the domain discrepancy. We first conduct experiments to demonstrate the effectiveness of our novel boundary labeling model. We then extensively evaluate METABDRY on eight data sets under domain generalization settings. The experimental results show that METABDRY achieves state-of-the-art results against the recent seven baselines.
Named Entity Recognition (NER) in social media posts is challenging since texts are usually short and contexts are lacking. Most recent works show that visual information can boost the NER performance since images can provide complementary contextual information for texts. However, the image-level features ignore the mapping relations between fine-grained visual objects and textual entities, which results in error detection in entities with different types. To better exploit visual and textual information in NER, we propose an adversarial gated bilinear attention neural network (AGBAN). The model jointly extracts entity-related features from both visual objects and texts, and leverages an adversarial training to map two different representations into a shared representation. As a result, domain information contained in an image can be transferred and applied for extracting named entities in the text associated with the image. Experimental results on Tweets dataset demonstrate that our model outperforms the state-of-the-art methods. Moreover, we systematically evaluate the effectiveness of the proposed gated bilinear attention network in capturing the interactions of mutimodal features visual objects and textual words. Our results indicate that the adversarial training can effectively exploit commonalities across heterogeneous data sources, which leads to improved performance in NER when compared to models purely exploiting text data or combining the image-level visual features.
Conflict-Based Search (CBS) is a leading algorithm for optimal Multi-Agent Path Finding (MAPF). CBS variants typically compute MAPF solutions using some form of A* search. However, they often do so under strict time limits so as to avoid exhausting the available memory. In this paper, we present IDCBS, an iterative-deepening variant of CBS which can be executed without exhausting the memory and without strict time limits. IDCBS can be substantially faster than CBS due to incremental methods that it uses when processing CBS nodes.
The global manufacturing industry is entering an era of considerable growth and innovation due to radical developments in advanced technologies. Knowledge remains crucial in improving product design, decision making, productivity, and increasing the speed of innovation. Communication is at the heart of knowledge exchange between stakeholders, especially those with different technical and social backgrounds. However, significant barriers exist when sharing product design knowledge due to the lack of available formalized tools and methods. In this article, to facilitate knowledge exchange and collaboration among stakeholders during product design, a method to systematically improve the understanding of target objects, including the composition of the object, nature of the object, and tools to support its related communication, is proposed. Further, a knowledge representation framework is described which enables parties to clearly understand the characteristics of the object and greatly improve communication efficiency. A case study is presented to validate how the proposed framework provides valuable insights into the improvement of design communication and contributes to global manufacturing development.
Distant supervision is widely used to extract relational facts with automatically labeled datasets to reduce high cost of human annotation. However, current distantly supervised methods suffer from the common problems of word-level and sentence-level noises, which come from a large proportion of irrelevant words in a sentence and inaccurate relation labels for numerous sentences. The problems lead to unacceptable precision in relation extraction and are critical for the success of using distant supervision. In this paper, we propose a novel and robust neural approach to deal with both problems by reducing influences of the multi-granularity noises. Three levels of noises from word, sentence until knowledge type are carefully considered in this work. We first initiate a question-answering based relation extractor (QARE) to remove noisy words in a sentence. Then we use multi-focus multi-instance learning (MMIL) to alleviate the effects of sentence-level noise by utilizing wrongly labeled sentences properly. Finally, to enhance our method against all the noises, we initialize parameters in our method with a priori knowledge learned from the relevant task of entity type classification by transfer learning. Extensive experiments on both existing benchmark and an improved larger dataset demonstrate that our proposed approach remarkably achieves new state-of-the-art performance.
Joint extraction of entities and relations is to detect entities and recognize semantic relations simultaneously. However, some existing joint models predict relations on words, instead of entities. These models cannot make full use of the entity information when predicting relations, which will affect relation extraction. We propose an end-to-end model with a double-pointer module that can jointly extract whole entities and relations. The double-pointer module is combined with multiple decoders to predict the start and end positions of the entity in the input sentence. In addition, in order to learn the relevance between long-distance entities effectively, the multi-layer convolution and self-attention mechanism are used as an encoder, instead of Bi-RNN. We conduct experiments on two public datasets and our models outperform the baseline methods significantly.
Most existing methods determine relation types only after all the entities have been recognized, thus the interaction between relation types and entity mentions is not fully modeled. This paper presents a novel paradigm to deal with relation extraction by regarding the related entities as the arguments of a relation. We apply a hierarchical reinforcement learning (HRL) framework in this paradigm to enhance the interaction between entity mentions and relation types. The whole extraction process is decomposed into a hierarchy of two-level RL policies for relation detection and entity extraction respectively, so that it is more feasible and natural to deal with overlapping relations. Our model was evaluated on public datasets collected via distant supervision, and results show that it gains better performance than existing methods and is more powerful for extracting overlapping relations1.
This article presents a dynamic hazard identification methodology founded on an ontology-based knowledge modeling framework coupled with probabilistic assessment. The objective is to develop an efficient and effective knowledge-based tool for process industries to screen hazards and conduct rapid risk estimation. The proposed generic model can translate an undesired process event (state of the process) into a graphical model, demonstrating potential pathways to the process event, linking causation to the transition of states. The Semantic web-based Web Ontology Language (OWL) is used to capture knowledge about unwanted process events. The resulting knowledge model is then transformed into Probabilistic-OWL (PR-OWL) based Multi-Entity Bayesian Network (MEBN). Upon queries, the MEBNs produce Situation Specific Bayesian Networks (SSBN) to identify hazards and their pathways along with probabilities. Two open-source software programs, Protégé and UnBBayes, are used. The developed model is validated against 45 industrial accidental events extracted from the U.S. Chemical Safety Board's (CSB) database. The model is further extended to conduct causality analysis.
As an important and challenging problem, knowledge representation and inference are typically carried out in a knowledge embedding framework over a multi-relational knowledge graph, and thus have a wide range of applications such as semantic retrieval and question answering. In this paper, we propose a bilinear learning framework which performs cross-entity knowledge relation analysis in the continuous vector space (derived from knowledge embedding). In the framework, we effectively model the intrinsic correlations among different types of knowledge relations within a max-margin multi-relational ranking scheme, which jointly optimizes the tasks of entity embedding and cross-entity relation prediction in terms of multi-relational structures of the knowledge graph. Specifically, we devise a bilinear scoring function that aims to evaluate the confidence degree of semantic relation prediction for entity pairs through a multi-relational learning-to-rank pipeline. In essence, the pipeline formulates the problem of relation prediction for entity pairs as that of learning relation-specific ranking functions by max-margin optimization. Experimental results demonstrate the effectiveness of the proposed framework on two common benchmark datasets.
Vehicular Ad-hoc Networks (VANETs) are a challenging IoT scenario. While research is proposing increasingly sophisticated hardware and software solutions for on-board context detection, probably high-level context information sharing has not been adequately addressed so far. This paper proposes a novel logic-based framework enabling a contextual data management and mining in VANETs. It grounds on a knowledge fusion algorithm based on non-standard, non-monotonic inference services in Description Logics, adopting standard Semantic Web languages. Ontology-referred context annotations produced by individual VANET nodes are merged with automatic reconciliation of inconsistencies. An efficient information dissemination protocol complements the proposal. The approach has been implemented in a vehicular network simulator and early experimental results proved its effectiveness and feasibility.
Knowledge base completion (KBC) aims to predict missing information in a knowledge base. In this paper, we address the out-of-knowledge-base (OOKB) entity problem in KBC: how to answer queries concerning test entities not observed at training time. Existing embedding-based KBC models assume that all test entities are available at training time, making it unclear how to obtain embeddings for new entities without costly retraining. To solve the OOKB entity problem without retraining, we use graph neural networks (Graph-NNs) to compute the embeddings of OOKB entities, exploiting the limited auxiliary knowledge provided at test time. The experimental results show the effectiveness of our proposed model in the OOKB setting. Additionally, in the standard KBC setting in which OOKB entities are not involved, our model achieves state-of-the-art performance on the WordNet dataset.
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.0 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data.
Knowledge graph completion aims to perform link prediction between entities. In this paper, we consider the approach of knowledge graph embeddings. Recently, models such as TransE and TransH build entity and relation embeddings by regarding a relation as translation from head entity to tail entity. We note that these models simply put both entities and relations within the same semantic space. In fact, an entity may have multiple aspects and various relations may focus on different aspects of entities, which makes a common space insufficient for modeling. In this paper, we propose TransR to build entity and relation embeddings in separate entity space and relation spaces. Afterwards, we learn embeddings by first projecting entities from entity space to corresponding relation space and then building translations between projected entities. In experiments, we evaluate our models on three tasks including link prediction, triple classification and relational fact extraction. Experimental results show significant and consistent improvements compared to state-of-the-art baselines including TransE and TransH.
Graph-structured data appears frequently in domains including chemistry,
natural language semantics, social networks, and knowledge bases. In this work,
we study feature learning techniques for graph-structured inputs. Our starting
point is previous work on Graph Neural Networks (Scarselli et al., 2009), which
we modify to use gated recurrent units and modern optimization techniques and
then extend to output sequences. The result is a flexible and broadly useful
class of neural network models that has favorable inductive biases relative to
purely sequence-based models (e.g., LSTMs) when the problem is
graph-structured. We demonstrate the capabilities on some simple AI (bAbI) and
graph algorithm learning tasks. We then show it achieves state-of-the-art
performance on a problem from program verification, in which subgraphs need to
be matched to abstract data structures.
We consider the problem of embedding entities and relationships of multi relational data in low-dimensional vector spaces. Our objective is to propose a canonical model which is easy to train, contains a reduced number of parameters and can scale up to very large databases. Hence, we propose TransE, a method which models relationships by interpreting them as translations operating on the low-dimensional embeddings of the entities. Despite its simplicity, this assumption proves to be powerful since extensive experiments show that TransE significantly outperforms state-of-the-art methods in link prediction on two knowledge bases. Besides, it can be successfully trained on a large scale data set with 1M entities, 25k relationships and more than 17M training samples.