Monitoring and early warning system and method based on digital eagle eyes

Abstract

The invention, which belongs to the technical field of equipment monitoring, discloses a digital eagle-eye-based monitoring and early warning system comprising a data acquisition module used for acquiring equipment data of equipment in a factory; the data preprocessing module is used for preprocessing the equipment data to obtain a second feature sequence; the topological graph structure construction module is used for constructing a topological graph structure according to the second feature sequence; the equipment state prediction module is used for constructing the equipment state prediction module according to the topological graph structure and predicting the equipment state score of the equipment; and the maintenance measure generation module is used for generating a maintenance measure according to the equipment state score of the equipment. According to the invention, accurate monitoring and prediction of the factory equipment state are realized by constructing the equipment state prediction model, so that the safety and efficiency of factory operation are remarkably improved.

Description

Technical Field

[0001] The invention belongs to the technical field of equipment monitoring, and in particular relates to a digital eagle eye-based monitoring and early warning system and method. Background Art

[0002] With the continuous advancement of Industry 4.0, the level of intelligence and automation of factories has been greatly improved, and the health status monitoring of equipment during factory operations has become particularly critical. The normal operation of equipment is not only directly related to production efficiency and product quality, but also closely related to production safety. Traditional equipment monitoring usually uses independent sensor systems to monitor specific equipment or parameters (such as temperature, vibration, and pressure). This isolated monitoring method makes it difficult to capture the mutual influence between equipment and cannot form a global equipment status assessment. In addition, although many factories have deployed sensor networks to collect equipment operation data, data processing and analysis are usually performed offline. This delayed data analysis method may result in equipment failures that cannot be discovered in a timely manner, thereby affecting production continuity and even causing safety accidents.

[0003] Therefore, there is an urgent need for a digital eagle eye monitoring and early warning system and method to solve the above problems. Summary of the invention

[0004] The present invention provides a digital eagle eye-based monitoring and early warning system to solve the technical problem that the equipment monitoring method in the related technology cannot capture the mutual influence between equipment and cannot process in real time.

[0005] The present invention provides a digital eagle eye monitoring and early warning system, comprising:

[0006] A data collection module, which is used to collect device data of M devices in the factory within a first preset time T and at a preset time interval t, and represent the device data at each time point as a first feature sequence;

[0007] The mth sequence unit of the first feature sequence at the nth time point represents the device data of the mth device at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T / t;

[0008] Equipment data includes: equipment temperature, equipment vibration, equipment pressure, equipment current and equipment voltage;

[0009] Equipment vibration is represented by a time domain waveform, where the horizontal axis represents the time point and the vertical axis represents the intensity of the vibration, expressed as acceleration;

[0010] A data preprocessing module, which is used to preprocess the first feature sequence to obtain a second feature sequence;

[0011] A topological graph structure building module, which is used to build a topological graph structure according to the second feature sequence;

[0012] The topological graph structure includes: nodes, initial features of nodes and edges;

[0013] Each device establishes a mapping relationship with a node;

[0014] The device data of each device is used as the initial feature of the node to which the mapping relationship is established;

[0015] A device status prediction module is used to construct a device status prediction module according to the topology structure and predict the device status score of the device;

[0016] The input of the device status prediction module is the topological graph structure, and the output is the device status scores of M devices within the second preset time P;

[0017] The value range of the device status score is an integer from 0 to 10, indicating the status of the device from normal to severe failure;

[0018] The maintenance measure generating module is used to generate maintenance measures according to the device status scores of M devices within a second preset time P.

[0019] Furthermore, T, t, M, m, N, n and P are all custom parameters.

[0020] Furthermore, the pre-processing step includes:

[0021] Step S201, for missing values ​​in the device data, filling processing is performed by taking the average value of the value of the previous sequence unit and the value of the next sequence unit;

[0022] Step S202, convert the time domain waveform of the device vibration into a frequency domain waveform through fast Fourier transform. In the frequency domain waveform, the horizontal axis represents the frequency and the vertical axis represents the amplitude. The frequency component with the highest amplitude is extracted from the frequency domain waveform as the main frequency. The main frequency energy is obtained according to the main frequency, and the main frequency energy feature is created. It is spliced ​​after the device data. The formula of the main frequency energy is: E mf =|A(f mf )| 2 , where A(f mf ) represents the main frequency f mf The corresponding amplitude;

[0023] Step S203, adaptively denoising the device data using a Bayesian denoising method;

[0024] Step S204, normalizing the device data by using the maximum-minimum normalization method.

[0025] Furthermore, the ways to construct edges include:

[0026] Build edges based on physical connections between devices;

[0027] When the distance between two devices is less than a first threshold, an edge is built between the two devices, and the first threshold is a custom parameter.

[0028] Further, the device state prediction model includes: a first hidden layer, a combination layer, a second hidden layer, a segmentation layer and a first classifier;

[0029] The first hidden layer includes N first units, the nth first unit inputs the regional graph structure and outputs the first updated features of M nodes;

[0030] The combination layer includes N combination units, and the nth combination unit is used to combine the first updated features output by the N first units into a combination vector;

[0031] The second hidden layer includes N second units, the nth second unit inputs the combination vector output by the nth combination unit, and outputs the second update feature;

[0032] The segmentation layer is used to segment the second update feature output by the Nth second unit into M third update features, where each third update feature corresponds to a node update feature of a node;

[0033] The node update feature of each node is input into the first classifier, and the classification space of the first classifier represents the device status score of each device.

[0034] Furthermore, the first hidden layer is used to process the topological graph structure at each time point to obtain updated node features. The calculation formula of each time point of the first hidden layer includes:

[0035] Perform a linear transformation on the initial features of the mth node of the kth attention head:

[0036]

[0037] in represents the feature of the mth node of the kth attention head after linear transformation, represents the first weight parameter of the kth attention head, h m represents the initial features of the mth node;

[0038] Calculate the attention score between the mth node and its neighbor node j:

[0039]

[0040] in represents the attention score between the mth node of the kth attention head and its neighbor node j, and represents the second weight parameter, the third weight parameter and the fourth weight parameter of the kth attention head, concat represents the concatenation operation, T represents the transposition operation, and LeakyReLU represents the rectified linear unit activation function;

[0041] Normalize the attention score to get the attention weight:

[0042]

[0043] in represents the attention weights of nodes m and j of the kth attention head, N m represents the neighbor set of node m, Represents the attention score between the node m of the kth attention head and its adjacent node h;

[0044] Perform independent operations on each attention head to obtain their respective node representations

[0045] in represents the feature representation of node m of the k-th attention head, that is, the first updated feature of node m, and sigmoid represents the sigmoid activation function.

[0046] Furthermore, the second hidden layer is constructed based on the long short-term memory network unit, and is used to process a sequence consisting of combined vectors spliced ​​from nodes of the topological graph structure at N time points to obtain a second updated feature, which is composed of node update features of M nodes.

[0047] Furthermore, the device data during device operation and the historical data of device operation are collected in real time as sample data of the device state prediction model, and are manually labeled as sample labels of the device state prediction model.

[0048] Furthermore, the equipment status score is divided into three levels. When the equipment status score is between 0 and 3, it means that the equipment is in normal or slightly worn state, and the maintenance measure is to take no measures; when the equipment status score is between 4 and 7, it means that the equipment has obvious wear or abnormality, and the maintenance measure is to replace worn parts, conduct detailed functional tests on the equipment, find out potential problems or hidden dangers, and repair them; when the equipment status is between 8 and 10, it means that the equipment has a serious risk of failure, and the maintenance measure is to immediately stop the equipment operation and conduct a comprehensive inspection. When the equipment has failed, an accident report and analysis will be conducted to find out the cause of the failure and formulate improvement measures to prevent similar problems from happening again.

[0049] The present invention provides a digital eagle eye monitoring and early warning method, comprising the following steps:

[0050] Step S401, collecting device data of M devices in the factory at a preset time interval t within a first preset time T;

[0051] Step S402, preprocessing the device data of the M devices to obtain a second feature sequence;

[0052] Step S403, constructing a topological graph structure according to the second characteristic sequence;

[0053] Step S404, constructing a device status prediction module according to the topology structure and predicting the device status score of the device;

[0054] Step S405: generating maintenance measures according to the device status scores of the M devices within the second preset time P.

[0055] The beneficial effects of the present invention are as follows: the present invention effectively captures the complex relationships between devices by constructing a topological graph structure in both time and space dimensions, and realizes accurate monitoring and prediction of the status of factory equipment by constructing an equipment status prediction model, thereby significantly improving the safety and efficiency of factory operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a module schematic diagram of a digital eagle eye monitoring and early warning system of the present invention;

[0057] Figure 2 is a flow chart of preprocessing device data of the present invention;

[0058] Figure 3 The present invention is a flow chart of a digital eagle eye monitoring and early warning method.

[0059] In the figure: a data acquisition module 101, a data preprocessing module 102, a topology structure building module 103, an equipment status prediction module 104 and a maintenance measure generation module 105. DETAILED DESCRIPTION

[0060] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that the discussion of these embodiments is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the contents of this specification. Each example may omit, replace or add various processes or components as needed. In addition, the features described relative to some examples may also be combined in other examples.

[0061] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0062] like Figure 1-Figure 3 As shown, a digital eagle eye monitoring and early warning system includes:

[0063] The data collection module 101 is used to collect device data of M devices in the factory within a first preset time T according to a preset time interval t, and represent the device data at each time point as a first feature sequence;

[0064] The mth sequence unit of the first feature sequence at the nth time point represents the device data of the mth device at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T / t;

[0065] Equipment data includes: equipment temperature, equipment vibration, equipment pressure, equipment current and equipment voltage;

[0066] Equipment vibration is represented by a time domain waveform, where the horizontal axis represents the time point and the vertical axis represents the intensity of the vibration, expressed as acceleration;

[0067] A data preprocessing module 102, which is used to preprocess the first feature sequence to obtain a second feature sequence;

[0068] A topology structure building module 103, which is used to build a topology structure according to the second feature sequence;

[0069] The topological graph structure includes: nodes, initial features of nodes and edges;

[0070] Each device establishes a mapping relationship with a node;

[0071] The device data of each device is used as the initial feature of the node to which the mapping relationship is established;

[0072] The device state prediction module 104 is used to construct a device state prediction module according to the topological graph structure and predict the device state score of the device;

[0073] The input of the device status prediction module is the topological graph structure, and the output is the device status scores of M devices within the second preset time P;

[0074] The value range of the device status score is an integer from 0 to 10, indicating the status of the device from normal to severe failure;

[0075] The maintenance measure generating module 105 is used to generate maintenance measures according to the device status scores of the M devices within the second preset time P.

[0076] In one embodiment of the present invention, T, t, M, m, N, n and P are all custom parameters. Preferably, T is set to 2 hours, t is set to 5 minutes, and P is set to 2 hours.

[0077] In one embodiment of the present invention, a temperature sensor is installed on the motor, bearing or heat-sensitive area of ​​the equipment to monitor the operating temperature of the equipment; a vibration sensor is installed on the bearing of the equipment to monitor the vibration of the equipment; a pressure sensor is installed on the pipeline part of the equipment to monitor the pressure data change of the equipment; a current sensor is installed on the motor part of the equipment to monitor the current consumption of the equipment; a voltage sensor is installed on the circuit of the equipment to monitor the stability of the voltage; the collected data is transmitted to the central data processing system by wireless transmission through the sensors, and the collected data is classified, marked and managed using the data management platform, so as to facilitate subsequent preprocessing operations.

[0078] In one embodiment of the present invention, Figure 2 As shown, the preprocessing steps include:

[0079] Step S201, for missing values ​​in the device data, filling processing is performed by taking the average value of the value of the previous sequence unit and the value of the next sequence unit;

[0080] Step S202, convert the time domain waveform of the device vibration into a frequency domain waveform through fast Fourier transform. In the frequency domain waveform, the horizontal axis represents the frequency and the vertical axis represents the amplitude. The frequency component with the highest amplitude is extracted from the frequency domain waveform as the main frequency. The main frequency energy is obtained according to the main frequency, and the main frequency energy feature is created. It is spliced ​​after the device data. The formula of the main frequency energy is: E mf =|A(f mf )| 2 , where A(f mf ) represents the main frequency f mf The corresponding amplitude;

[0081] Step S203, adaptively denoising the device data by using a Bayesian denoising method. The Bayesian denoising method calculates the most likely signal estimate, i.e., the maximum a posteriori probability estimate, given a noise model and a signal prior. This method can adaptively adjust the denoising process, retain key features of the signal, and effectively remove noise.

[0082] Step S204, normalizing the device data by using the maximum-minimum normalization method.

[0083] In one embodiment of the present invention, the method of constructing an edge includes:

[0084] Building edges based on the physical connection between devices, for example, when two devices are connected by a wire, building an edge between the two devices;

[0085] When the distance between two devices is less than a first threshold, an edge is built between the two devices, and the first threshold is a custom parameter.

[0086] In one embodiment of the present invention, the device state prediction model includes: a first hidden layer, a combination layer, a second hidden layer, a segmentation layer and a first classifier;

[0087] The first hidden layer includes N first units, the nth first unit inputs the regional graph structure and outputs the first updated features of M nodes;

[0088] The combination layer includes N combination units, and the nth combination unit is used to combine the first updated features output by the N first units into a combination vector;

[0089] The second hidden layer includes N second units, the nth second unit inputs the combination vector output by the nth combination unit, and outputs a second updated feature;

[0090] The segmentation layer is used to segment the second update feature output by the Nth second unit into M third update features, where each third update feature corresponds to a node update feature of a node;

[0091] The node update feature of each node is input into the first classifier, and the classification space of the first classifier represents the device status score of each device.

[0092] In one embodiment of the present invention, the first hidden layer is used to process the topological graph structure at each time point to obtain updated node features. The calculation formula of each time point of the first hidden layer includes:

[0093] Perform a linear transformation on the initial features of the mth node of the kth attention head:

[0094]

[0095] in represents the feature of the mth node of the kth attention head after linear transformation, represents the first weight parameter of the kth attention head, h m represents the initial features of the mth node;

[0096] Calculate the attention score between the mth node and its neighbor node j:

[0097]

[0098] in represents the attention score between the mth node of the kth attention head and its neighbor node j, and represents the second weight parameter, the third weight parameter and the fourth weight parameter of the kth attention head, concat represents the concatenation operation, T represents the transposition operation, and LeakyReLU represents the rectified linear unit activation function;

[0099] Normalize the attention score to get the attention weight:

[0100]

[0101] in represents the attention weights of nodes m and j of the kth attention head, N m represents the neighbor set of node m, Represents the attention score between the node m of the kth attention head and its adjacent node h;

[0102] Perform independent operations on each attention head to obtain their respective node representations

[0103]

[0104] in represents the feature representation of node m of the k-th attention head, that is, the first updated feature of node m, and sigmoid represents the sigmoid activation function.

[0105] In one embodiment of the present invention, the weights of the edges of the topological graph structure are adjusted by a greedy algorithm to improve the performance of the device status prediction model. The specific steps are:

[0106] Step S301, setting an initial weight for each edge of the topological graph structure;

[0107] Step S302: Run the model according to the current initial weights to calculate the prediction accuracy of the model. The objective function is: in represents the state of the mth device predicted by the model, y m represents the true state label of the mth device, Q represents the total number of predicted samples, Indicates when When , the value is 1, otherwise it is 0;

[0108] Step S303, make a slight adjustment to the weight of each edge and recalculate the prediction accuracy of the model. If the prediction accuracy is improved after the adjustment, the adjustment is retained; otherwise, the original weight is restored.

[0109] Step S304, if the current number of iterations is greater than or equal to the maximum number of iterations, the weight configuration of the edge of the current number of iterations is used as the final weight configuration of the edge, otherwise return to step S303, where the maximum number of iterations is a custom parameter, preferably, the maximum number of iterations is set to 100.

[0110] In one embodiment of the present invention, the second hidden layer is constructed based on a long short-term memory network unit, and is used to process a sequence consisting of combined vectors spliced ​​from nodes of a topological graph structure at N time points to obtain a second updated feature, which is composed of node update features of M nodes.

[0111] In one embodiment of the present invention, the loss function of the device state prediction model uses a cross entropy loss function.

[0112] In one embodiment of the present invention, the device data during device operation and the historical data of device operation are collected in real time as sample data of the device state prediction model, and are manually annotated as sample labels of the device state prediction model.

[0113] In one embodiment of the present invention, the equipment status score of the equipment is divided into three levels. When the equipment status score is between 0 and 3, it indicates that the equipment is in a normal or slightly worn state, and the maintenance measure is to take no measures; when the equipment status score is between 4 and 7, it indicates that the equipment has obvious wear or abnormality, and the maintenance measure is to replace the worn parts, and conduct detailed functional tests on the equipment to find out potential problems or hidden dangers and repair them; when the equipment status is between 8 and 10, it indicates that the equipment has a serious risk of failure, and the maintenance measure is to immediately stop the equipment operation and conduct a comprehensive inspection. When the equipment has failed, an accident report and analysis is conducted to find out the cause of the failure and formulate improvement measures to prevent similar problems from happening again.

[0114] In one embodiment of the present invention, Figure 3 As shown, the present invention provides a digital eagle eye monitoring and early warning method, comprising the following steps:

[0115] Step S401, collecting device data of M devices in the factory at a preset time interval t within a first preset time T;

[0116] Step S402, preprocessing the device data of the M devices to obtain a second feature sequence;

[0117] Step S403, constructing a topological graph structure according to the second characteristic sequence;

[0118] Step S404, constructing a device status prediction module according to the topology structure and predicting the device status score of the device;

[0119] Step S405: generating maintenance measures according to the device status scores of the M devices within the second preset time P.

[0120] The above describes an embodiment of the present embodiment, but the present embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present embodiment, ordinary technicians in this field can also make many forms, all of which are within the protection of the present embodiment.

Claims

1. A digital eagle eye monitoring and early warning system, characterized in that: include: A data collection module, which is used to collect device data of M devices in the factory within a first preset time T and at a preset time interval t, and represent the device data at each time point as a first feature sequence; The mth sequence unit of the first feature sequence at the nth time point represents the device data of the mth device at the nth time point, where 1≤m≤M, 1≤n≤N, and N=T / t; Equipment data includes: equipment temperature, equipment vibration, equipment pressure, equipment current and equipment voltage; Equipment vibration is represented by a time domain waveform, where the horizontal axis represents the time point and the vertical axis represents the intensity of the vibration, expressed as acceleration; A data preprocessing module, which is used to preprocess the first feature sequence to obtain a second feature sequence; A topological graph structure building module, which is used to build a topological graph structure according to the second feature sequence; The topological graph structure includes: nodes, initial features of nodes and edges; Each device establishes a mapping relationship with a node; The device data of each device is used as the initial feature of the node to which the mapping relationship is established; A device status prediction module is used to construct a device status prediction module according to the topology structure and predict the device status score of the device; The input of the device status prediction module is the topological graph structure, and the output is the device status scores of M devices within the second preset time P; The value range of the device status score is an integer from 0 to 10, indicating the status of the device from normal to severe failure; The maintenance measure generating module is used to generate maintenance measures according to the device status scores of M devices within a second preset time P.

2. According to claim 1, a digital eagle eye monitoring and early warning system is characterized in that: T, t, M, m, N, n, and P are custom parameters.

3. According to claim 1, a digital eagle eye monitoring and early warning system is characterized in that: The preprocessing steps include: Step S201, for missing values ​​in the device data, filling processing is performed by taking the average value of the value of the previous sequence unit and the value of the next sequence unit; Step S202, convert the time domain waveform of the device vibration into a frequency domain waveform through fast Fourier transform. In the frequency domain waveform, the horizontal axis represents the frequency and the vertical axis represents the amplitude. The frequency component with the highest amplitude is extracted from the frequency domain waveform as the main frequency. The main frequency energy is obtained according to the main frequency, and the main frequency energy feature is created. It is spliced ​​after the device data. The formula of the main frequency energy is: E mf =|A(f mf )| 2 , where A(f mf ) represents the main frequency f mf The corresponding amplitude; Step S203, adaptively denoising the device data using a Bayesian denoising method; Step S204, normalizing the device data by using the maximum-minimum normalization method.

4. According to claim 1, a digital eagle eye monitoring and early warning system is characterized in that: Ways to construct edges include: Build edges based on physical connections between devices; When the distance between two devices is less than a first threshold, an edge is built between the two devices, and the first threshold is a custom parameter.

5. The digital eagle eye monitoring and early warning system according to claim 1 is characterized in that: The equipment state prediction model includes: a first hidden layer, a combination layer, a second hidden layer, a segmentation layer and a first classifier; The first hidden layer includes N first units, the nth first unit inputs the regional graph structure and outputs the first updated features of M nodes; The combination layer includes N combination units, and the nth combination unit is used to combine the first updated features output by the N first units into a combination vector; The second hidden layer includes N second units, the nth second unit inputs the combination vector output by the nth combination unit, and outputs a second updated feature; The segmentation layer is used to segment the second update feature output by the Nth second unit into M third update features, where each third update feature corresponds to a node update feature of a node; The node update feature of each node is input into the first classifier, and the classification space of the first classifier represents the device status score of each device.

6. The digital eagle eye monitoring and early warning system according to claim 5 is characterized in that: The first hidden layer is used to process the topological graph structure at each time point to obtain updated node features. The calculation formula for each time point of the first hidden layer includes: Perform a linear transformation on the initial features of the mth node of the kth attention head: in represents the feature of the mth node of the kth attention head after linear transformation, represents the first weight parameter of the kth attention head, h m represents the initial features of the mth node; Calculate the attention score between the mth node and its neighbor node j: in represents the attention score between the mth node of the kth attention head and its neighbor node j, and represents the second weight parameter, the third weight parameter and the fourth weight parameter of the kth attention head, concat represents the concatenation operation, T represents the transposition operation, and LeakyReLU represents the rectified linear unit activation function; Normalize the attention score to get the attention weight: in represents the attention weights of nodes m and j of the kth attention head, N m represents the neighbor set of node m, Represents the attention score between the node m of the kth attention head and its adjacent node h; Perform independent operations on each attention head to obtain their respective node representations in represents the feature representation of node m of the k-th attention head, that is, the first updated feature of node m, and sigmoid represents the sigmoid activation function.

7. The digital eagle eye monitoring and early warning system according to claim 5 is characterized in that: The second hidden layer is constructed based on a long short-term memory network unit, and is used to process a sequence consisting of a combination vector concatenated from nodes of a topological graph structure at N time points to obtain a second update feature consisting of node update features of M nodes.

8. The digital eagle eye monitoring and early warning system according to claim 1 is characterized in that: The device data during device operation and the historical data of device operation are collected in real time as sample data of the device status prediction model, and are manually labeled as sample labels of the device status prediction model.

9. The digital eagle eye monitoring and early warning system according to claim 1 is characterized in that: The equipment status score is divided into three levels. When the equipment status score is between 0 and 3, it means that the equipment is in normal or slightly worn state, and the maintenance measure is to take no measures; when the equipment status score is between 4 and 7, it means that the equipment has obvious wear or abnormality, and the maintenance measure is to replace the worn parts, and conduct detailed functional tests on the equipment to find out potential problems or hidden dangers and repair them; when the equipment status is between 8 and 10, it means that the equipment has a serious risk of failure, and the maintenance measure is to immediately stop the equipment operation and conduct a comprehensive inspection. When the equipment has failed, the accident report and analysis shall be carried out to find out the cause of the failure and formulate improvement measures to prevent similar problems from happening again.

10. A digital eagle eye monitoring and early warning method, characterized in that: The digital eagle eye monitoring and early warning system according to any one of claims 1 to 9 is used, comprising the following steps: Step S401, collecting device data of M devices in the factory at a preset time interval t within a first preset time T; Step S402, preprocessing the device data of the M devices to obtain a second feature sequence; Step S403, constructing a topological graph structure according to the second characteristic sequence; Step S404, constructing a device status prediction module according to the topology structure and predicting the device status score of the device; Step S405: generating maintenance measures according to the device status scores of the M devices within the second preset time P.

Images