A Method and System for Predicting Mechanical Faults Based on Knowledge Graphs and Convolutional Neural Networks

Abstract

The application discloses a mechanical fault prediction method and system based on a knowledge graph and a convolutional neural network, constructs a basic mechanical knowledge graph, constructs a training set, adopts the training set to train a mechanical fault prediction model, and obtains the trained mechanical fault prediction model; the training set is mechanical part vibration signal data of a known mechanical fault type, and the data of the training set is obtained by querying the constructed basic mechanical knowledge graph; a to-be-identified mechanical part vibration signal is acquired, the to-be-identified mechanical part vibration signal is input into the trained mechanical fault prediction model, and a mechanical fault type corresponding to the to-be-identified mechanical part vibration signal is output; a relationship between the to-be-identified mechanical part vibration signal and the mechanical fault type is updated into the basic mechanical knowledge graph, and an optimized mechanical knowledge graph is obtained; a mechanical fault problem is acquired, a query result is found from the optimized mechanical knowledge graph, and the query result is output as an answer corresponding to the mechanical fault problem.

Description

Technical Field

[0001] This invention relates to the field of mechanical fault diagnosis and prediction technology, and in particular to a mechanical fault prediction method and system based on knowledge graphs and convolutional neural networks. Background Technology

[0002] The statements in this section merely refer to the background art related to this invention and do not necessarily constitute prior art.

[0003] Knowledge graphs (KG) are a popular form of knowledge representation, released by Google in 2012. They focus on entities and their relationships to represent static knowledge. Although the concept of knowledge graphs was only released by Google in 2012, research on knowledge graphs began as early as 1965. Knowledge graphs are categorized into early knowledge bases, open knowledge graphs, Chinese common sense knowledge graphs, and domain knowledge graphs. However, a vast amount of event information in the world conveys dynamic and procedural knowledge. Therefore, event-centric knowledge representations such as event knowledge graphs (EKG) are also essential. Similar to the construction methods of entity knowledge graphs, event knowledge graphs typically use ontology-based construction methods for knowledge modeling, first building the top-level representation schema of the event knowledge graph, and then refining and supplementing it with instances.

[0004] In 2006, Hinton et al. proposed the following viewpoints: (1) multi-hidden-layer artificial neural networks have excellent feature learning capabilities; (2) the training difficulties of deep neural networks can be effectively overcome through "layer-wise pre-training". Since then, experts and scholars have conducted research on deep learning and artificial neural networks. Compared with shallow models, deep learning has obvious advantages in feature extraction and modeling. Deep learning is better at mining abstract feature representations from raw data, and these feature representations usually have good generalization ability. As the main form of deep learning, deep convolutional neural networks (DCNNs) are inspired by the structure of the visual cortex of animals in terms of the connection between neurons. This convolutional neural network is widely used in computer vision, natural language processing and other fields such as image classification, face recognition, font recognition, and audio recognition.

[0005] Fault diagnosis technology is an applied interdisciplinary field, its theoretical foundation involving multiple disciplines such as modern control theory, computer engineering, mathematical statistics, fuzzy set theory, signal processing, and pattern recognition. The task of fault diagnosis is to identify the characteristic description of a system failure based on the different characteristics exhibited by various quantities (measurable or unmeasurable) or some of their quantities compared to the normal state, and to detect and isolate the fault. Early methods relied on manual diagnosis; however, in practice, traditional fault diagnosis techniques are no longer sufficient to meet the actual requirements of complex mechanical equipment operation. Predictive models are a core component of mechanical fault prediction technology. Currently, commonly used models in fault prediction include curve fitting models, filtering models, time series models, grey modeling, artificial neural network models, and fuzzy models. How to establish a mechanical fault prediction model is a key aspect of this research. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a mechanical fault prediction method and system based on knowledge graphs and convolutional neural networks. This enables the prediction of mechanical faults, promotes predictive maintenance of machinery, and reduces the occurrence of accidents. Based on the signal data of the machinery collected by the user, and combined with an event knowledge graph, relationships are established between mechanical parts and signal data, mechanical parts and fault types, and mechanical parts and solutions. Furthermore, a convolutional neural network is used to provide the function of predicting mechanical faults.

[0007] In a first aspect, the present invention provides a mechanical fault prediction method based on knowledge graphs and convolutional neural networks;

[0008] Mechanical fault prediction methods based on knowledge graphs and convolutional neural networks include:

[0009] Construct a basic mechanical knowledge graph, which includes: the relationship between mechanical parts and mechanical part vibration signal data, the relationship between mechanical parts and mechanical fault types, the relationship between mechanical part vibration signal data and mechanical fault types, and the relationship between mechanical parts and solutions;

[0010] A training set is constructed, and the mechanical fault prediction model is trained using the training set to obtain the trained mechanical fault prediction model; the training set consists of vibration signal data of mechanical parts with known mechanical fault types, and the data of the training set is obtained by querying the constructed basic mechanical knowledge graph.

[0011] The vibration signal of the mechanical part to be identified is obtained, and the vibration signal of the mechanical part to be identified is input into the trained mechanical fault prediction model. The mechanical fault type corresponding to the vibration signal of the mechanical part to be identified is output. The relationship between the vibration signal of the mechanical part to be identified and the mechanical fault type is updated in the basic mechanical knowledge graph to obtain the optimized mechanical knowledge graph.

[0012] The system retrieves mechanical fault questions, finds query results from the optimized mechanical knowledge graph, and outputs the query results as the answers corresponding to the mechanical fault questions.

[0013] Secondly, the present invention provides a mechanical fault prediction system based on knowledge graphs and convolutional neural networks;

[0014] A mechanical fault prediction system based on knowledge graphs and convolutional neural networks includes:

[0015] The construction module is configured to: construct a basic mechanical knowledge graph, which includes: the relationship between mechanical parts and mechanical part vibration signal data, the relationship between mechanical parts and mechanical fault types, the relationship between mechanical part vibration signal data and mechanical fault types, and the relationship between mechanical parts and solutions;

[0016] The training module is configured to: construct a training set, use the training set to train the mechanical fault prediction model, and obtain the trained mechanical fault prediction model; the training set consists of vibration signal data of mechanical parts with known mechanical fault types, and the data in the training set is obtained by querying the constructed basic mechanical knowledge graph;

[0017] The prediction module is configured to: acquire the vibration signal of the mechanical part to be identified; input the vibration signal of the mechanical part to be identified into the trained mechanical fault prediction model; output the mechanical fault type corresponding to the vibration signal of the mechanical part to be identified; update the relationship between the vibration signal of the mechanical part to be identified and the mechanical fault type to the basic mechanical knowledge graph to obtain the optimized mechanical knowledge graph.

[0018] The question-and-answer module is configured to: retrieve mechanical fault questions, find query results from the optimized mechanical knowledge graph, and output the query results as the answers corresponding to the mechanical fault questions.

[0019] Thirdly, the present invention also provides an electronic device, comprising:

[0020] Memory, used for non-transitory storage of computer-readable instructions; and

[0021] Processor, for executing the computer-readable instructions,

[0022] When the computer-readable instructions are executed by the processor, they perform the method described in the first aspect above.

[0023] Fourthly, the present invention also provides a storage medium for non-transitory storage of computer-readable instructions, wherein, when the non-transitory computer-readable instructions are executed by a computer, the instructions for executing the method described in the first aspect are executed.

[0024] Fifthly, the present invention also provides a computer program product, including a computer program that, when run on one or more processors, is used to implement the method described in the first aspect above.

[0025] Compared with the prior art, the beneficial effects of the present invention are:

[0026] A mechanical fault event knowledge graph is constructed, which provides query functions for various parameters of mechanical faults. A convolutional neural network is established, and the overall structure, parameters of each layer, and optimization algorithm of the convolutional neural network are determined through multiple experiments. A suitable dataset is selected to train and establish the determined convolutional neural network. Based on the mechanical fault event knowledge graph and the established convolutional neural network, fault prediction is realized for newly collected data, and corresponding entities and their related relationships are established in the event knowledge graph.

[0027] Based on the characteristics of the mechanical failure field, a combined bottom-up and top-down approach was used to construct a knowledge graph of rolling bearing failure events, and four types of visual query functions were implemented based on the constructed event knowledge graph. Secondly, a convolutional neural network (WDCNN) fault diagnosis method was studied, and a deep convolutional neural network with a wide first-layer convolutional kernel was established to diagnose the operating status of rolling bearings. The experimentally established WDCNN model achieved an accuracy of over 99.8% on the training dataset and over 97% on the test dataset, demonstrating high accuracy. Finally, by combining the event knowledge graph and the deep convolutional neural network, bearing failure prediction was achieved. This study explored intelligent fault diagnosis and prediction technologies for rolling bearings, enabling predictive maintenance of complex mechanical equipment and avoiding losses caused by downtime due to failures. Attached Figure Description

[0028] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0029] Figure 1 This is the bottom-up construction process of the knowledge graph in Embodiment 1 of this application;

[0030] Figure 2 This is the seven-step ontology construction process of Embodiment 1 of this application;

[0031] Figure 3 This is the rolling bearing data acquisition system of Case Western Reserve University as described in Embodiment 1 of this application;

[0032] Figure 4 This is an ontology class hierarchy diagram of the reference knowledge graph for Embodiment 1 of this application;

[0033] Figure 5 This is a hierarchy diagram of classes and classes in Embodiment 1 of this application. Detailed Implementation

[0034] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0035] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0036] All data acquisition in this embodiment is carried out in accordance with laws and regulations and with user consent, and the data is used legally.

[0037] Example 1

[0038] This embodiment provides a mechanical fault prediction method based on knowledge graphs and convolutional neural networks;

[0039] Mechanical fault prediction methods based on knowledge graphs and convolutional neural networks include:

[0040] S101: Construct a basic mechanical knowledge graph, which includes: the relationship between mechanical parts and mechanical part vibration signal data, the relationship between mechanical parts and mechanical fault types, the relationship between mechanical part vibration signal data and mechanical fault types, and the relationship between mechanical parts and solutions;

[0041] S102: Construct a training set and use the training set to train the mechanical fault prediction model to obtain the trained mechanical fault prediction model; the training set consists of vibration signal data of mechanical parts with known mechanical fault types, and the data of the training set is obtained by querying the constructed basic mechanical knowledge graph.

[0042] S103: Obtain the vibration signal of the mechanical part to be identified, input the vibration signal of the mechanical part to be identified into the trained mechanical fault prediction model, and output the mechanical fault type corresponding to the vibration signal of the mechanical part to be identified; update the relationship between the vibration signal of the mechanical part to be identified and the mechanical fault type to the basic mechanical knowledge graph to obtain the optimized mechanical knowledge graph.

[0043] S104: Obtain the mechanical fault problem, find the query results from the optimized mechanical knowledge graph, and output the query results as the answer corresponding to the mechanical fault problem.

[0044] Further, S101: Constructing a basic mechanical knowledge graph, the basic mechanical knowledge graph including: the relationship between mechanical parts and mechanical part vibration signal data, the relationship between mechanical parts and mechanical fault types, the relationship between mechanical part vibration signal data and mechanical fault types, and the relationship between mechanical parts and solutions, specifically including:

[0045] The running parts of the rolling bearing include: the drive end and the fan end;

[0046] The physical components of a rolling bearing include: an inner ring, an outer ring, and rolling elements;

[0047] The failure diameters of rolling bearings include: 7 mils, 14 mils, and 21 mils;

[0048] The operating status of rolling bearings includes: normal and faulty;

[0049] The operating conditions of rolling bearings include: motor load horsepower (HP) and motor speed (RPM). These two items are grouped together, with a total of 4 combinations: (0HP, 1797rpm), (1HP, 1772rpm), (2HP, 1750rpm) and (3HP, 1730rpm).

[0050] Protective measures for rolling bearings include: daily maintenance, regular inspection and maintenance, and periodic replacement;

[0051] Using fault name, fault type, fault solution, bearing physical composition, and maintenance plan entity as nodes, and with the relationships between mechanical parts and mechanical part vibration signal data, mechanical parts and mechanical fault types, mechanical part vibration signal data and mechanical fault types, fault type and fault solution, fault type and bearing physical composition, and bearing physical composition and maintenance plan as edges, a bearing fault event knowledge graph is constructed.

[0052] It should be understood that the rolling bearing dataset used is the rolling bearing dataset published by the Rolling Bearing Data Center of Case Western Reserve University (CWRU), and the data in the 12k Drive End Bearing Fault and Normal Baseline files of the dataset were selected for ten-class experiments after screening.

[0053] Furthermore, a basic mechanical knowledge graph is constructed, including:

[0054] S101-1: Modify the dataset;

[0055] S101-2: Create entities and visualize them within the neo4j graph database;

[0056] S101-3: Import data attributes;

[0057] S101-4: Create relationships between entities.

[0058] Further, S101-1: Modifying the dataset includes:

[0059] S101-11: Read the data content from the driver-side data file: file creation time, version information, global information, driver-side acceleration data, fan-side acceleration data, and motor speed; global information includes the software version used to create the file and the platform configuration information used to create the file;

[0060] S101-12: Set the fault name, which includes: normal state and nine fault states. The nine fault states include: nine fault states formed by combining two of the following three fault occurrence locations: {inner race, ball, outer race} and three fault diameters: {0.007 inches, 0.014 inches, 0.021 inches}.

[0061] S101-13: Add data columns to the dataset corresponding to the bearing location, motor load horsepower, motor speed, bearing operating status, fault diameter (fault status), fault location (fault status), and fault name, and output the modified dataset.

[0062] Further, S101-2: Creating entities and visualizing them within the neo4j graph database includes:

[0063] The bearing operating part class contains two entities: the drive end and the fan end;

[0064] The physical component class of a bearing contains three entities: the inner ring, rolling elements, and outer ring of a rolling bearing.

[0065] The subclasses of the bearing operating_condition class have four types of entities for motor load horsepower and motor speed: {0HP, 1HP, 2HP, 3HP} and {1797rpm, 1772rpm, 1750rpm, 1730rpm}.

[0066] There are 10 types of fault names for rolling bearings, and there are also 10 subclasses in the corresponding fault type class. Among them, fault_0 represents the fault name of the normal state. The fault diameter class in the subclasses fault_1-fault_9 contains three entities: 0.007 inches, 0.014 inches and 0.021 inches.

[0067] The solution class for bearing failure contains three entities: replace inner ring, replace rolling elements, and replace outer ring.

[0068] Further, S101-3: Importing data attributes includes:

[0069] Import the dataset where the bearing is in normal operating condition, and import the vibration signal data (vibration_signal_data) in the dataset into the vibration_signal_data entities corresponding to each subclass under the Normal state class. The signal data is used as entity attributes, and the operating condition, motor load, motor speed, and fault name are used as entity attributes.

[0070] Import the dataset where the bearing is in a faulty operating state. Import the fault vibration signal data (fault_vibration_signal_data) into the fault_vibration_signal_data entity corresponding to each subclass under the fault state (Abnormal) class. The signal data is used as the entity attribute, and the operating state, motor load, fault name, and fault diameter are used as the entity attributes.

[0071] Further, S101-4: Creating relationships between entities includes:

[0072] The relationship between fault name and fault type, the relationship between fault type and solution, the relationship between fault type and physical components of rolling bearing, and the relationship between physical components of rolling bearing and maintenance plan.

[0073] Further, in step S102: a training set is constructed, and the mechanical fault prediction model is trained using the training set to obtain the trained mechanical fault prediction model; the training set consists of vibration signal data of mechanical parts with known mechanical fault types, and the data in the training set is obtained by querying the constructed basic mechanical knowledge graph, specifically including:

[0074] Construct a dataset consisting of vibration signal data of mechanical parts with known mechanical fault types;

[0075] The dataset is divided into training and testing sets according to a set ratio;

[0076] Convolutional neural networks are constructed, the training set is input into the convolutional neural network, the network is trained, and a trained mechanical fault prediction model is obtained.

[0077] Furthermore, the convolutional neural network (WDCNN, Deep Convolution Neural Networks with Wide First-layer Kernels) includes: a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer, a third convolutional layer, a third pooling layer, a fourth convolutional layer, a fourth pooling layer, a fifth convolutional layer, a fifth pooling layer, a fully connected layer, and a softmax layer, all connected in sequence.

[0078] Furthermore, the parameters of the convolutional neural network are set as follows: the kernel size of the first convolutional layer is 64×1, and the stride is 16×1; the kernel size of the remaining convolutional layers is 3×1, and the stride is 2×1; and the five pooling layers are all set with a kernel size of 2×1 and a stride of 2×1. The activation function is set to ReLU. In addition, the Softmax layer is set to produce 10 outputs, corresponding to 10 bearing states.

[0079] Further, in step S103: acquiring the vibration signal of the mechanical part to be identified, inputting the vibration signal of the mechanical part to be identified into the trained mechanical fault prediction model, and outputting the mechanical fault type corresponding to the vibration signal of the mechanical part to be identified; updating the relationship between the vibration signal of the mechanical part to be identified and the mechanical fault type to the basic mechanical knowledge graph to obtain the optimized mechanical knowledge graph, specifically including:

[0080] Preprocess the vibration signals of the mechanical parts to be identified;

[0081] The trained mechanical fault prediction model is used to predict the vibration signal of the mechanical part to be identified, and the predicted mechanical fault type of the vibration signal of the mechanical part to be identified is obtained.

[0082] Based on the predicted mechanical failure type, create a new node under the corresponding class in the neo4j graph database;

[0083] Establish the relationship between the new node and the corresponding type entity under the Fault type class.

[0084] Further, S104: Obtaining the mechanical fault problem, finding the query result from the optimized mechanical knowledge graph, and outputting the query result as the answer corresponding to the mechanical fault problem, specifically includes:

[0085] S104-1: Query the operating status and physical composition parameters of rolling bearings;

[0086] S104-2: Inquire about the type and location of rolling bearing failures;

[0087] S104-3: Inquire about rolling bearing fault types and solutions;

[0088] S104-4: Inquire about the physical components and maintenance procedures of rolling bearings.

[0089] Furthermore, S104-1: Query the operating status and physical composition parameters of the rolling bearing, including:

[0090] S104-11: Find entities in the normal or abnormal state under the subclass of the operating_status class and their vibration signal data / fault vibration signal data vibration_signal_data / fault_vibration_signal_data.

[0091] S104-12: Enter the corresponding query statement, MATCH(n:operating_status)RETURN nLIMIT 25;

[0092] S104-13: Obtain the normal or abnormal state class node and the entity nodes under the class node, and then display the query results of the rolling bearing's operating status and physical composition.

[0093] Further, S104-2: querying the rolling bearing fault type and fault location includes:

[0094] S104-21: There are two methods for querying:

[0095] (1) Directly query the fault type and directly return the fault type node;

[0096] (2) Convert to a query on the "happened_at" relation, and the query returns the type of failure that occurred on the physical component;

[0097] S104-22: Enter the corresponding query statement: MATCH p=()-[r:happened_at]→()RETURNpLIMIT 25 or MATCH(n:fault_details)RETURN n LIMIT 25;

[0098] S104-23: Obtain the fault type node, detailed parameters of the fault, and the rolling bearing entity where the fault occurred.

[0099] Furthermore, S104-3: querying rolling bearing fault types and solutions includes:

[0100] S104-31: Finding solutions based on fault type is transformed into a query of the "solution_is" relationship for a specified starting point;

[0101] S104-32: Enter the corresponding query statement to obtain the query results.

[0102] Furthermore, S104-4: querying the physical components and maintenance plan of rolling bearings, including:

[0103] S104-41: Find the corresponding maintenance method based on the physical components of the rolling bearing, specifying the starting point as a certain physical component entity and finding the endpoint with the relationship "maintenance method is" ("maintenance_method_is");

[0104] S104-42: Enter the corresponding query statement to get the query results.

[0105] This invention constructs a bearing failure event knowledge graph based on the bearing dataset from Case Western Reserve University and the neo4j graph database. On the other hand, it trains and establishes a convolutional neural network with prediction capabilities using Tensorflow and the bearing dataset. Finally, the two are combined to realize the mechanical failure prediction method based on the event knowledge graph and the convolutional neural network.

[0106] This invention proposes a mechanical fault prediction method based on event knowledge graphs and convolutional neural networks. Taking a key component of rotating machinery as an example, the invention first establishes an event knowledge graph by manually modifying the dataset and then filling the modified data into the event knowledge graph. Second, using the rolling bearing dataset published by the Rolling Bearing Data Center at Case Western Reserve University (CWRU), a convolutional neural network is trained and built. Then, this network is combined with the event knowledge graph to establish ontology-ontology, ontology-event, and event-event relationships. Finally, the established convolutional neural network is used to predict faults.

[0107] This method is based on Python 3.8, Anaconda, neo4j graph database, and the Case Western Reserve University (CWRU) bearing database, and was developed in an environment of Tensorflow 2.10.0, scipy 1.10.0, scikit-learn 1.2.1, and py2neo 2021.2.3.

[0108] This method takes bearings as the research object:

[0109] First, such as Figure 1 , Figure 2As shown, based on the bottom-up construction process of knowledge graphs and the seven-step ontology construction process, using, as Figure 3 The CWRU rolling bearing data acquisition system uses two datasets (Table 1) from its bearing database, Normal BaselineData and 12k Drive End Bearing Fault Data, to establish a bearing fault event knowledge graph (e.g., ...). Figure 4 ), its class and class hierarchy diagram (such as Figure 5 ).

[0110] Secondly, a deep convolutional neural network model with a wide kernel layer was built and trained using Python in the Tensorflow environment. This convolutional neural network was then used to classify the dataset into 10 categories, as shown in Table 2. Finally, the newly collected data was tested using the convolutional neural network to obtain the average loss and accuracy values ​​for the newly collected data. At this point, a threshold was set to determine whether the bearing's operating state belonged to the data was Normal or Abnormal.

[0111] Table 1. Storage data for Normal Baseline Data and 12k Drive End Bearing Fault Data

[0112]

[0113] Table 2 10 Fault Names

[0114]

[0115] Example 2

[0116] This embodiment provides a mechanical fault prediction system based on knowledge graphs and convolutional neural networks;

[0117] A mechanical fault prediction system based on knowledge graphs and convolutional neural networks includes:

[0118] The construction module is configured to: construct a basic mechanical knowledge graph, which includes: the relationship between mechanical parts and mechanical part vibration signal data, the relationship between mechanical parts and mechanical fault types, the relationship between mechanical part vibration signal data and mechanical fault types, and the relationship between mechanical parts and solutions;

[0119] The training module is configured to: construct a training set, use the training set to train the mechanical fault prediction model, and obtain the trained mechanical fault prediction model; the training set consists of vibration signal data of mechanical parts with known mechanical fault types, and the data in the training set is obtained by querying the constructed basic mechanical knowledge graph;

[0120] The prediction module is configured to: acquire the vibration signal of the mechanical part to be identified; input the vibration signal of the mechanical part to be identified into the trained mechanical fault prediction model; output the mechanical fault type corresponding to the vibration signal of the mechanical part to be identified; update the relationship between the vibration signal of the mechanical part to be identified and the mechanical fault type to the basic mechanical knowledge graph to obtain the optimized mechanical knowledge graph.

[0121] The question-and-answer module is configured to: retrieve mechanical fault questions, find query results from the optimized mechanical knowledge graph, and output the query results as the answers corresponding to the mechanical fault questions.

[0122] It should be noted that the aforementioned construction module, training module, prediction module, and question-answering module correspond to steps S101 to S104 in Embodiment 1. The examples and application scenarios implemented by these modules and their corresponding steps are the same, but they are not limited to the content disclosed in Embodiment 1. It should also be noted that these modules, as part of the system, can be executed in a computer system, such as a set of computer-executable instructions.

[0123] The descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0124] The proposed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and the division of modules described above is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed.

[0125] Example 3

[0126] This embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein, the processor is connected to the memory, and the one or more computer programs are stored in the memory. When the electronic device is running, the processor executes the one or more computer programs stored in the memory to cause the electronic device to perform the method described in Embodiment 1.

[0127] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0128] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0129] In the implementation process, each step of the above method can be completed by the integrated logic circuits in the processor hardware or by software instructions.

[0130] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0131] Those skilled in the art will recognize that the units and algorithm steps described in connection with the various examples of this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0132] Example 4

[0133] This embodiment also provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the method described in Embodiment 1.

[0134] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A mechanical failure prediction method based on a knowledge graph and a convolutional neural network, characterized in that, The application comprises the following steps: A basic mechanical knowledge graph is constructed, which comprises the relationships between mechanical parts and mechanical part vibration signal data, the relationships between mechanical parts and mechanical fault types, the relationships between mechanical part vibration signal data and mechanical fault types, and the relationships between mechanical parts and solutions; The basic mechanical knowledge graph is constructed in detail as follows: taking fault name, fault type, fault solution, bearing physical composition, and maintenance scheme entity as nodes, and taking the relationships between mechanical parts and mechanical part vibration signal data, the relationships between mechanical parts and mechanical fault types, the relationships between mechanical part vibration signal data and mechanical fault types, the relationships between fault type and fault solution, the relationships between fault type and bearing physical composition, and the relationships between bearing physical composition and maintenance scheme as edges, a bearing fault event knowledge graph is constructed; A training set is constructed, and a mechanical fault prediction model is trained using the training set to obtain a trained mechanical fault prediction model; the training set is mechanical part vibration signal data of known mechanical fault types, and the data of the training set is obtained by querying the constructed basic mechanical knowledge graph; A to-be-identified mechanical part vibration signal is obtained, the to-be-identified mechanical part vibration signal is input into the trained mechanical fault prediction model, and a mechanical fault type corresponding to the to-be-identified mechanical part vibration signal is output; the relationship between the to-be-identified mechanical part vibration signal and the mechanical fault type is updated into the basic mechanical knowledge graph to obtain an optimized mechanical knowledge graph; A mechanical fault problem is obtained, a query result is found from the optimized mechanical knowledge graph, and the query result is output as an answer corresponding to the mechanical fault problem. 2.The knowledge graph and convolutional neural network-based mechanical failure prediction method of claim 1, wherein, The basic mechanical knowledge graph is constructed, comprising: The data set is modified, entities are created in the neo4j graph database and visualized, and data attributes are imported; and the relationships between entities are created. 3.The knowledge graph and convolutional neural network-based mechanical failure prediction method of claim 2, wherein, The data set is modified, comprising: The data content in the driving end data file is read: file creation time, version information, global information, driving end acceleration data, fan end acceleration data, and motor speed; the global information includes the software version used for creating the file and the platform configuration information for creating the file; The fault name is set, and the fault name comprises a normal state and nine fault states, wherein the nine fault states comprise nine fault states obtained by combining three fault occurrence positions {inner ring, rolling body, outer ring} and three fault diameters {0.007 inch, 0.014 inch, and 0.021 inch} in pairs; The corresponding data columns of the bearing position, motor load horsepower, motor speed, bearing running state, fault diameter, fault occurrence position, and fault name are added to the data set, and the modified data set is output. 4.The knowledge graph and convolutional neural network-based mechanical failure prediction method of claim 2, wherein, Entities are created in the neo4j graph database and visualized, comprising: The bearing running part class of the bearing comprises two entities: the driving end and the fan end; The physical composition class of the bearing comprises three entities: the inner ring, the rolling body, and the outer ring of the rolling bearing; The entities of the subclasses of the operating condition class of the bearing are four: {0 HP, 1 HP, 2 HP, 3 HP} and {1797 rpm, 1772 rpm, 1750 rpm, 1730 rpm}; The fault name of the rolling bearing has 10 types, and the corresponding fault type class also has 10 subclasses, among which fault_0 represents the fault name of the normal state, and the fault diameter class in the subclasses fault_1-fault_9 contains three entities: 0.007 inches, 0.014 inches and 0.021 inches; The corresponding solution class of the bearing fault contains three entities: replace the inner ring, replace the rolling body and replace the outer ring.

5. The mechanical fault prediction method based on knowledge graphs and convolutional neural networks as described in claim 2, characterized in that, it imports... Data attributes include: Import the data set of the bearing operating state as normal, import the vibration signal data in the data set into the corresponding entity of each subclass under the normal state class, and the signal data as the entity attribute, and the operating state, motor load, motor speed, fault name as the entity attribute; Import the data set of the bearing operating state as normal, import the vibration signal data in the data set into the corresponding entity of each subclass under the normal state class, and the signal data as the entity attribute, and the operating state, motor load, motor speed, fault name as the entity attribute. 6.The knowledge graph and convolutional neural network-based mechanical failure prediction method of claim 1, wherein, Obtain the mechanical fault problem, find the query result from the optimized mechanical knowledge graph, and output the query result as the answer corresponding to the mechanical fault problem, which specifically includes: Query the running state and physical composition parameters of the rolling bearing; query the fault type and fault position of the rolling bearing; query the fault type and solution of the rolling bearing; query the physical composition parts and maintenance scheme of the rolling bearing; Among them, querying the running state and physical composition parameters of the rolling bearing includes: finding the entity and its vibration signal data / fault vibration signal data attribute under the normal state or fault state of the running state class; input the corresponding query statement; get the normal state or fault state class node and entity node under the class node, and then display the query result of the running state and physical composition of the rolling bearing; Among them, querying the fault type and fault position of the rolling bearing includes: Directly query the fault type, and directly return the fault type node; input the corresponding query statement; get the fault type node and detailed parameters of the fault and the rolling bearing entity that occurs fault; Among them, querying the fault type and solution of the rolling bearing includes: finding the solution according to the fault type, which is converted into a query of the "solution is" relationship of the specified starting point; input the corresponding query statement to get the query result; Querying the physical composition parts and maintenance scheme of the rolling bearing includes: finding the corresponding maintenance scheme according to the physical composition parts of the rolling bearing, specifying the starting point as a certain physical composition entity and finding the end point with the relationship "maintenance scheme is"; input the corresponding query statement to get the query result.

7. A mechanical failure prediction system based on a knowledge graph and a convolutional neural network, characterized in that, It includes: The construction module is configured to construct a basic mechanical knowledge graph, and the basic mechanical knowledge graph includes relationships among mechanical parts, mechanical part vibration signal data, mechanical failure types, relationships between mechanical part vibration signal data and mechanical failure types, and relationships between mechanical parts and solutions. The construction of the basic mechanical knowledge graph specifically includes: taking a fault name, a fault type, a fault solution, a bearing physical composition, and a maintenance scheme as nodes, taking relationships among mechanical parts, mechanical part vibration signal data, mechanical failure types, relationships between mechanical part vibration signal data and mechanical failure types, a fault type and a fault solution, a fault type and a bearing physical composition, and a bearing physical composition and a maintenance scheme as edges, and completing construction of a bearing fault event knowledge graph. The training module is configured to construct a training set, train a mechanical fault prediction model by using the training set, and obtain a trained mechanical fault prediction model. The training set is mechanical part vibration signal data of a known mechanical fault type, and data of the training set is obtained by querying the constructed basic mechanical knowledge graph. The prediction module is configured to obtain to-be-identified mechanical part vibration signal, input the to-be-identified mechanical part vibration signal into the trained mechanical fault prediction model, and output a mechanical fault type corresponding to the to-be-identified mechanical part vibration signal. Relationships between the to-be-identified mechanical part vibration signal and the mechanical fault type are updated into the basic mechanical knowledge graph to obtain an optimized mechanical knowledge graph. The question and answer module is configured to obtain a mechanical fault question, find a query result from the optimized mechanical knowledge graph, and output the query result as an answer corresponding to the mechanical fault question.

8. An electronic device, comprising: a memory for non-transiently storing computer readable instructions; and a processor for running the computer readable instructions, wherein the computer readable instructions, when run by the processor, perform the method of any one of claims 1-6.

9. A storage medium characterized by, non-transiently storing computer readable instructions, wherein when the non-transient computer readable instructions are executed by a computer, instructions for performing the method of any one of claims 1-6.

Images