Smart factory production line management and control method, device, equipment and medium

Abstract

The invention relates to the technical field of information management, and provides a smart factory production line management and control method and device, equipment and a medium, and the method comprises the steps: obtaining work data streams and category information corresponding to a plurality of pieces of production equipment on a production line in a smart factory; obtaining a target state judgment rule according to the work data streams and the category information corresponding to the plurality of production devices; according to a target state judgment rule, performing state judgment on the working data flow corresponding to each production device to obtain a state judgment result corresponding to each production device; carrying out fault trend prediction on the production line according to the state judgment result corresponding to each piece of production equipment to obtain a target fault trend; and according to the target fault trend, generating a production line maintenance plan so as to maintain a plurality of production devices on the production line according to the production line maintenance plan. According to the scheme, the production stability and efficiency of the smart factory production line are improved.

Description

Technical Field

[0001] The present application relates to the field of information management technology, and more specifically, to a method, device, equipment and medium for controlling production lines in a smart factory. Background Art

[0002] Smart factory, also known as intelligent factory or Industry 4.0 factory, refers to a modern factory that uses advanced information technology and operational technology to achieve highly automated, intelligent and information-based production processes. Smart factory is a core component of the digital transformation of the manufacturing industry, aiming to enhance the competitiveness of enterprises by optimizing resource allocation, improving production efficiency, enhancing product quality and reducing operating costs.

[0003] However, in the existing technology, the monitoring of production lines in smart factories usually relies on fixed thresholds and rules to detect abnormal situations. This method lacks flexibility, cannot cope well with the dynamic changes of the production line, and has poor stability. In addition, traditional data reporting methods are often carried out after the problem occurs, and cannot provide preventive maintenance suggestions. Delayed production line maintenance will reduce production efficiency.

[0004] Therefore, how to improve the stability and efficiency of smart factory production lines is an urgent problem that needs to be solved. Summary of the invention

[0005] In view of the above problems existing in the prior art, the purpose of this application is to propose a smart factory production line management method, device, equipment and medium to at least solve the technical problem of how to improve the stability and efficiency of smart factory production lines.

[0006] To achieve the above objectives and other related objectives, the present application provides a smart factory production line control method, the method comprising: Obtain the work data flow and category information corresponding to multiple production equipment on the production line in the smart factory; Obtaining a target state judgment rule according to the work data flows and category information corresponding to the plurality of production devices; According to the target state judgment rule, the state of the work data stream corresponding to each of the production equipment is judged to obtain the state judgment result corresponding to each of the production equipment; According to the status judgment results corresponding to each of the production equipment, the fault trend of the production line is predicted to obtain the target fault trend; A production line maintenance plan is generated according to the target failure trend, so as to maintain the plurality of production equipment on the production line according to the production line maintenance plan.

[0007] In some embodiments, the category information includes device type and device model, the work data stream includes temperature time series data, vibration time series data, pressure time series data, and current time series data. After obtaining the work data stream and category information corresponding to multiple production devices on the production line in the smart factory, the following is also included: By setting up edge computing nodes around the production equipment, according to the equipment type and equipment model, a classification mark is added to the work data stream corresponding to each of the production equipment, and the classification mark is used to obtain the target state judgment rule; The temperature time series data, vibration time series data, pressure time series data, and current time series data are cleaned through the edge computing node, and the temperature time series data, vibration time series data, pressure time series data, and current time series data after data cleaning are standardized to obtain a preprocessed working data stream, and the preprocessed working data stream is used to obtain the target state judgment rule.

[0008] In some implementations, obtaining a target state judgment rule according to the work data flows and category information corresponding to the plurality of production devices includes: Extracting common features from the preprocessed working data stream by means of a convolutional neural network and a time series network, wherein the common features include common time domain features and frequency domain features; According to the classification mark, extracting multiple features corresponding to the classification mark from the common features, combining the multiple features to obtain a feature set corresponding to each of the classification marks; By using a preset weighted multi-task loss function, a plurality of task weights are obtained according to the number of the feature sets and the error change of the feature sets; According to each of the task weights, a target state judgment rule corresponding to each of the classification marks is generated.

[0009] In some implementations, the state judgment of the work data stream corresponding to each of the production devices is performed according to the target state judgment rule to obtain the state judgment result corresponding to each of the production devices, including: Extracting time domain features and frequency domain features corresponding to the working data stream; Combining the time domain features and frequency domain features corresponding to the working data stream to obtain a multi-dimensional feature vector; The multidimensional feature vector and the target state judgment rule are transmitted to a pre-trained decision tree for state judgment to obtain a state judgment result corresponding to each of the production equipment, wherein the decision tree is trained by finding the optimal segmentation point based on the features corresponding to the acquired historical work data stream.

[0010] In some implementations, the predicting of the fault trend of the production line according to the status judgment results corresponding to each of the production equipment to obtain a target fault trend includes: Obtaining historical failure events and environmental data corresponding to each of the production equipment; Integrate the state judgment results, historical fault events and environmental data corresponding to each of the production equipment to obtain an initial feature matrix, wherein the fault event includes the fault time and the fault condition, the row of the initial feature matrix includes the fault time, and the column of the initial feature matrix includes the state judgment results, the fault condition and the environmental data; Analyzing the state change rate, the fault frequency, and the environmental correlation according to the initial feature matrix, and constructing a target feature matrix according to the state change rate, the fault frequency, and the environmental correlation; The target feature matrix is ​​transmitted to a pre-built fault trend prediction model to obtain the target fault trend.

[0011] In some implementations, generating a production line maintenance plan according to the target failure trend includes: Dividing the target fault trend according to a plurality of preset time windows to obtain a fault probability distribution corresponding to each of the preset time windows; Determining a risk score of the production equipment within each of the preset time windows according to the failure probability distribution corresponding to each of the preset time windows; Sorting the maintenance priorities of the plurality of production equipment according to the risk scores of the production equipment in each of the preset time windows to obtain an initial maintenance priority; According to the category information, adjusting the initial maintenance priority to obtain a target maintenance priority; According to the preset production plan and the target maintenance priority, the maintenance time window of the production equipment is determined to obtain the production line maintenance plan.

[0012] In some implementations, after generating a production line maintenance plan according to the target failure trend, the method further includes: Determining a failure probability according to the target failure trend; Comparing the failure probability with a preset failure probability threshold to obtain a comparison result; When the comparison result shows that the fault probability is greater than or equal to the preset fault probability threshold, increasing the frequency of obtaining the work data stream corresponding to the production equipment; When the comparison result is that the fault probability is less than the preset fault probability threshold, the frequency of obtaining the work data stream corresponding to the production equipment is reduced.

[0013] In one embodiment of the present application, a smart factory production line control device is also provided, the device comprising: The data acquisition module is used to obtain the work data flow and category information corresponding to multiple production equipment on the production line in the smart factory; A rule generation module, used to obtain a target state judgment rule according to the work data flow and category information corresponding to the plurality of production devices; A state judgment module, used to judge the state of the work data stream corresponding to each of the production equipment according to the target state judgment rule, and obtain the state judgment result corresponding to each of the production equipment; A fault trend prediction module is used to predict the fault trend of the production line according to the status judgment results corresponding to each of the production equipment to obtain a target fault trend; The production line maintenance plan generating module is used to generate a production line maintenance plan according to the target failure trend, so as to maintain the multiple production equipment on the production line according to the production line maintenance plan.

[0014] In one embodiment of the present application, a computer-readable storage medium is further provided, wherein the computer-readable storage medium includes a stored computer program, wherein the computer program executes the above-mentioned smart factory production line management and control method when running.

[0015] In one embodiment of the present application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor is configured to execute the above-mentioned smart factory production line management and control method through the computer program.

[0016] Beneficial effects of the present invention: First, obtain the working data streams and category information corresponding to multiple production equipment on the production line in the smart factory; then, obtain the target state judgment rule based on the working data streams and category information corresponding to the multiple production equipment; then, according to the target state judgment rule, perform state judgment on the working data stream corresponding to each of the production equipment to obtain the state judgment result corresponding to each of the production equipment; then, according to the state judgment result corresponding to each of the production equipment, predict the fault trend of the production line to obtain the target fault trend; finally, generate a production line maintenance plan based on the target fault trend, so as to maintain the multiple production equipment on the production line according to the production line maintenance plan. In this application, by acquiring the working data stream of production equipment in real time, the operating status of the equipment can be captured in time to avoid misjudgment or delays caused by data lag; by extracting valuable features from complex working data streams and generating accurate status judgment rules, the ability can be discovered earlier to prevent small problems from turning into major failures, thereby improving production stability; by analyzing the status judgment results of production equipment, future failure trends can be predicted. This predictive maintenance method is more efficient than traditional time-based maintenance. Preventive measures can be taken before a failure occurs, reducing the possibility of unexpected downtime, thereby improving production efficiency; based on the target failure trend, a personalized maintenance plan can be generated to ensure that maintenance work is performed at the most appropriate time, which not only reduces unnecessary maintenance costs, but also avoids waste of resources caused by excessive maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings: Figure 1 is a schematic diagram of an application environment of a smart factory production line control method shown in an exemplary embodiment of the present application; Figure 2 It is a flowchart of a smart factory production line control method shown in an exemplary embodiment of the present application; Figure 3 is a schematic diagram of a smart factory production line control device shown in an exemplary embodiment of the present application; Figure 4 It is a schematic diagram of the structure of an electronic device shown in an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0019] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0020] In one embodiment of the present application, a smart factory production line control method, device, equipment and medium are provided. Optionally, as an optional implementation, the smart factory production line control method can be but is not limited to being applied to Figure 1 in the environment shown. Figure 1 is a schematic diagram of an application environment of a smart factory production line control method shown in an exemplary embodiment of the present application, see Figure 1 , the implementation environment includes production equipment 101, data acquisition terminal 102, production line management and control terminal 103, production equipment 101, data acquisition terminal 102, production line management and control terminal 103 can, but are not limited to, communicate through a network, and production line management and control terminal 103 can, but are not limited to, perform operations on a database, such as write data operations or read data operations. The above-mentioned production line management and control terminal 103 can, but are not limited to, include a human-computer interaction screen, a processor, and a memory. The above-mentioned human-computer interaction screen can, but are not limited to, be used to display a production line maintenance plan. The above-mentioned processor can, but is not limited to, be used to respond to the above-mentioned human-computer interaction operation, perform corresponding operations, or generate corresponding instructions, and send the generated instructions to the production equipment 101. The above-mentioned memory is used to store relevant storage data, such as the work data flow and category information corresponding to the production equipment.

[0021] As an optional method, data can be collected through the data collection terminal 102, for example, the work data flow and category information corresponding to multiple production equipment on the production line in the smart factory can be collected.

[0022] As an optional method, the following steps in the smart factory production line control method may be performed on the production line control terminal 103: Obtain the work data flow and category information corresponding to multiple production equipment on the production line in the smart factory; Obtaining a target state judgment rule according to the work data flows and category information corresponding to the plurality of production devices; According to the target state judgment rule, the state of the work data stream corresponding to each of the production equipment is judged to obtain the state judgment result corresponding to each of the production equipment; According to the status judgment results corresponding to each of the production equipment, the fault trend of the production line is predicted to obtain the target fault trend; A production line maintenance plan is generated according to the target failure trend, so as to maintain the plurality of production equipment on the production line according to the production line maintenance plan.

[0023] The above method, by acquiring the working data stream of production equipment in real time, can capture the operating status of the equipment in time and avoid misjudgment or delays caused by data lag; by extracting valuable features from complex working data streams and generating accurate status judgment rules, the ability can be discovered earlier to prevent small problems from turning into major failures, thereby improving production stability; by analyzing the status judgment results of production equipment, it can predict future failure trends. This predictive maintenance method is more efficient than traditional time-based maintenance. Preventive measures can be taken before a failure occurs, reducing the possibility of unexpected downtime, thereby improving production efficiency; according to the target failure trend, a personalized maintenance plan can be generated to ensure that maintenance work is carried out at the most appropriate time, which not only reduces unnecessary maintenance costs, but also avoids waste of resources caused by excessive maintenance.

[0024] Optionally, in this embodiment, the above network may include but is not limited to: a wireless network, wherein the wireless network includes: Bluetooth, WIFI and other networks that implement wireless communication.

[0025] As an optional example, this embodiment does not limit the execution subject of the above-mentioned smart factory production line control method, and some or all steps of the above-mentioned smart factory production line control method can be executed on the production line control terminal 103.

[0026] In one embodiment of the present application, a smart factory production line management and control method is provided. Figure 2 This is a flow chart of a smart factory production line control method shown in an exemplary embodiment of the present application, see Figure 2 The smart factory production line control method includes the following steps S210 to S250: In step S210, the work data flow and category information corresponding to multiple production equipment on the production line in the smart factory are obtained.

[0027] Among them, multiple production equipment on the production line refers to various types of equipment involved in the production process in the smart factory, such as CNC machine tools, welding robots, conveyor belts, injection molding machines, etc. These devices together constitute the production line and are responsible for completing specific production tasks. Different types of equipment have different functions and operating characteristics, and may involve multiple equipment such as mechanical, electrical, and hydraulic. Through the Internet of Things technology, these devices can be connected to each other to form an intelligent production network.

[0028] Among them, the working data stream refers to various operating parameters collected from production equipment in real time, including but not limited to time series data such as temperature, vibration, pressure, current, voltage, speed, flow, etc. These data reflect the operating status and working performance of the equipment. Each device may generate multiple different types of data, forming a multi-dimensional time series data stream. The data is collected continuously and can capture the operation of the equipment at different time points. The data collection frequency is high, usually collected at intervals of seconds or shorter, to ensure that abnormalities can be discovered in a timely manner.

[0029] Among them, category information refers to metadata related to production equipment, which can include information such as equipment type, model, manufacturer, installation location, maintenance history, etc. Category information is used to distinguish different types of equipment and provide background support for subsequent status judgment and fault prediction. Category information is usually static and does not change frequently, but can be updated as the equipment is updated or modified. Category information is usually stored in a structured manner for easy query and management. Category information is closely related to the work data flow and can better understand the operating status of the equipment.

[0030] In step S220, a target state judgment rule is obtained according to the work data flows and category information corresponding to the plurality of production equipment.

[0031] Among them, the status judgment rules refer to a series of rules generated based on the working data flow and category information of the production equipment to judge the current status of the equipment. These rules can be simple rules based on thresholds or complex rules generated by machine learning models. Different types of equipment may require different status judgment rules, so the rules are personalized and can adapt to the operating characteristics of different equipment. The rules can be dynamically adjusted based on the historical data and real-time data of the equipment to ensure that it is always in the best state.

[0032] In step S230, according to the target state judgment rule, the state of the work data stream corresponding to each of the production equipment is judged to obtain the state judgment result corresponding to each of the production equipment.

[0033] Among them, state judgment refers to analyzing the working data flow of production equipment according to the generated state judgment rules to obtain the current state of the equipment (such as normal, slight abnormality, and serious abnormality). The results of state judgment are the basis for subsequent fault trend prediction and maintenance plan. State judgment is carried out in real time and can immediately issue an alarm when the equipment is abnormal. The state judgment process is automated, reducing the need for manual intervention and improving efficiency.

[0034] In step S240, according to the status judgment results corresponding to each of the production equipment, the fault trend of the production line is predicted to obtain a target fault trend.

[0035] Among them, fault trend prediction refers to predicting the probability of failure in the future based on the status judgment results of each production equipment. Fault trend prediction can help take preventive measures in advance to avoid unexpected downtime. Fault trend prediction is not just an assessment of the current status, but a prediction of possible future failures, which is forward-looking. Fault trend prediction can take into account multiple factors, including the equipment's historical fault records, current status, environmental conditions, etc., to form a multi-dimensional prediction model.

[0036] In step S250, a production line maintenance plan is generated according to the target failure trend, so as to maintain the plurality of production equipment on the production line according to the production line maintenance plan.

[0037] Among them, the production line maintenance plan refers to the specific maintenance task arrangement generated based on the fault trend prediction results, including maintenance time, maintenance content, required tools and personnel, etc. The purpose of the maintenance plan is to ensure that the production equipment operates in the best condition and reduce the risk of failure. The maintenance plan is personalized and customized according to the specific situation of each equipment to ensure the effectiveness and pertinence of the maintenance work. The maintenance plan can be flexibly adjusted according to actual production needs to ensure that the maintenance work will not affect the normal production progress.

[0038] For example, suppose there is a smart factory with the following key equipment on the production line: CNC machine tools, used to process metal parts, with key parameters including temperature, vibration, pressure, current, etc.; welding robots, used to weld metal parts, with key parameters including current, voltage, welding speed, etc.; conveyor belts, used to transport materials, with key parameters including speed, load, vibration, etc. Through the sensor network, the working data stream of each device is collected in real time, for example, the temperature, vibration, pressure, current and other data of CNC machine tools, the current, voltage, welding speed and other data of welding robots, the speed, load, vibration and other data of conveyor belts, and at the same time, the category information of each device is obtained, such as the equipment type, model, manufacturer, installation location, etc. According to the working data stream and category information of each device, convolutional neural networks and time series networks are used to extract common features and generate target state judgment rules corresponding to each classification mark. For example, for CNC machine tools, the generated state judgment rules may be: normal state: vibration ≤ 0.05 and temperature ≤ 30°C; slightly abnormal state: vibration > 0.05 or temperature > 30°C; seriously abnormal state: vibration > 0.1 or temperature > 40°C. According to the generated state judgment rules, the working data stream of each device is analyzed in real time to obtain the current state of the device. For example, the state judgment result of the CNC machine tool shows that it is currently in a slightly abnormal state because the vibration exceeds 0.05. According to the state judgment result of each device, the prediction model is used to predict the failure probability in the future. For example, the prediction result shows that the failure probability of the CNC machine tool will gradually increase in the next few days and is expected to reach 0.9 on the 5th day, which means there is a high risk of failure. A specific maintenance plan is generated based on the failure trend prediction results. For example, for CNC machine tools with a high failure risk, maintenance is scheduled on the 4th day. Specific tasks include checking vibration sensors, replacing worn bearings, etc. The maintenance plan will also consider the maintenance needs of other equipment to ensure that maintenance work on the entire production line is carried out in an orderly manner.

[0039] By adopting the above-mentioned embodiments provided by the present application, by acquiring the working data stream of the production equipment in real time, the operating status of the equipment can be captured in time to avoid misjudgment or delays caused by data lag; by extracting valuable features from the complex working data stream and generating accurate status judgment rules, the ability can be discovered earlier to prevent small problems from turning into major failures, thereby improving production stability; by analyzing the status judgment results of the production equipment, future failure trends can be predicted. This predictive maintenance method is more efficient than traditional time-based maintenance. Preventive measures can be taken before a failure occurs, reducing the possibility of unexpected downtime, thereby improving production efficiency; according to the target failure trend, a personalized maintenance plan can be generated to ensure that maintenance work is carried out at the most appropriate time, which not only reduces unnecessary maintenance costs, but also avoids waste of resources caused by excessive maintenance.

[0040] In one embodiment of the present application, the category information includes device type and device model, the work data stream includes temperature time series data, vibration time series data, pressure time series data, and current time series data. After obtaining the work data stream and category information corresponding to multiple production devices on the production line in the smart factory, it also includes: By setting up edge computing nodes around the production equipment, according to the equipment type and equipment model, a classification mark is added to the work data stream corresponding to each of the production equipment, and the classification mark is used to obtain the target state judgment rule; The temperature time series data, vibration time series data, pressure time series data, and current time series data are cleaned through the edge computing node, and the temperature time series data, vibration time series data, pressure time series data, and current time series data after data cleaning are standardized to obtain a preprocessed working data stream, and the preprocessed working data stream is used to obtain the target state judgment rule.

[0041] Among them, edge computing nodes refer to computing devices deployed near production equipment, which can perform data processing, analysis and decision-making locally without transmitting all data to the cloud. Edge computing nodes can be small servers, embedded devices, etc. Since data processing is performed locally, the time for data transmission is reduced, and lower response delays can be achieved, which is suitable for application scenarios with high real-time requirements. By pre-processing data locally and transmitting only necessary data to the cloud, the network bandwidth usage is greatly reduced. Even when the network connection is unstable, the edge computing node can still continue to work to ensure the continuity of production.

[0042] Among them, classification tags are added to the work data streams corresponding to each of the production equipment to distinguish different types of equipment and their corresponding characteristics. Classification tags can better understand the operating characteristics of the equipment and provide a basis for the subsequent generation of status judgment rules. Different types of equipment may have different operating characteristics and failure modes, so the classification tags are personalized and can adapt to the needs of different equipment. The classification tags are closely related to the category information of the equipment and can better understand the background information of the equipment. With the addition of new equipment or the transformation of existing equipment, the classification tags can be dynamically updated to ensure flexibility.

[0043] Among them, data cleaning refers to the preprocessing of the collected working data stream to remove invalid information such as outliers, noise, and duplicate data to ensure the quality and reliability of the data. Data cleaning is an important step in data preprocessing and can improve the accuracy of subsequent analysis. For example: through statistical analysis or machine learning algorithms, identify and remove data points that are obviously not within the normal range to avoid their impact on subsequent analysis; for missing data points, interpolation, mean filling and other methods can be used to fill in to ensure data integrity; through filters or other signal processing technologies, remove noise in the data and improve the signal-to-noise ratio of the data; remove duplicate data records to avoid redundant information interfering with subsequent analysis.

[0044] Among them, standardization processing refers to converting data of different dimensions (such as temperature, vibration, pressure, and current) into the same numerical range, so that different types of sensor data can be uniformly processed in the same model. Through standardization processing, data of different dimensions are converted into the same numerical range, eliminating the impact of dimensional differences on the model. Standardized data is easier to be processed by machine learning models, which can improve the training speed and prediction accuracy of the model.

[0045] Exemplarily, the working data stream of each device is collected in real time through the sensor network, for example, the temperature, vibration, pressure, current and other data of the CNC machine tool; the current, voltage, welding speed and other data of the welding robot; the speed, load, vibration and other data of the conveyor belt. At the same time, the category information of each device, such as the device type and model, is obtained. Edge computing nodes are set around the production equipment, and classification tags are added to the working data stream of each device according to the device type and model. The edge computing node performs data cleaning on the collected working data stream. For example, for the temperature time series data of the CNC machine tool, obvious outliers (such as temperature readings exceeding 100°C) are removed, missing temperature data points are filled (such as using linear interpolation), and noise is removed by a low-pass filter. The edge computing node performs standard processing on the cleaned time series data such as temperature, vibration, pressure, current, etc., for example, the temperature data is converted to a standard normal distribution with a mean of 0 and a standard deviation of 1; the vibration data is converted to the range of [0, 1].

[0046] In this embodiment, by removing outliers, filling missing values, and denoising, the quality and reliability of the data are ensured, and the impact of invalid data on subsequent analysis is avoided. By converting data of different dimensions into the same numerical range, the impact of dimensional differences on the model is eliminated, and the training speed and prediction accuracy of the model are improved. By preprocessing data locally, the time of data transmission is reduced, and a lower response delay is achieved, which is suitable for application scenarios with high real-time requirements. The edge computing node can perform preliminary status judgment and fault warning locally, discover potential problems in time, and avoid small problems from turning into major faults. By performing data cleaning and standardization on the edge computing node, only the necessary data is transmitted to the cloud, which greatly reduces the network bandwidth occupation and reduces the communication cost. Even in the case of unstable network connection, the edge computing node can still continue to work to ensure the continuity of production. Sensitive data can be processed locally, reducing the risk of data leakage and improving security. By preprocessing data on the edge computing node, it is avoided to upload the original data directly to the cloud, and privacy protection is further enhanced. By adding classification tags to the working data stream of each device, the operating characteristics of the device can be better understood, and more accurate status judgment rules can be generated, thereby improving maintenance efficiency. Based on classification tags and status judgment rules, a personalized maintenance plan can be generated for each device to ensure the effectiveness and pertinence of maintenance work.

[0047] In one embodiment of the present application, obtaining a target state judgment rule according to the work data streams and category information corresponding to the plurality of production devices includes: Extracting common features from the preprocessed working data stream by means of a convolutional neural network and a time series network, wherein the common features include common time domain features and frequency domain features; According to the classification mark, extracting multiple features corresponding to the classification mark from the common features, combining the multiple features to obtain a feature set corresponding to each of the classification marks; By using a preset weighted multi-task loss function, a plurality of task weights are obtained according to the number of the feature sets and the error change of the feature sets; According to each of the task weights, a target state judgment rule corresponding to each of the classification marks is generated.

[0048] Among them, convolutional neural network is a deep learning model that is particularly suitable for processing data with grid structure, such as images, audio, etc. In industrial scenarios, convolutional neural network can be used to extract time domain features and capture local patterns in time series data. Through the convolution layer, convolutional neural network can capture local features in time series data, such as temperature changes and vibration fluctuations in a short period of time. The convolution kernel slides over the entire time series, reducing the number of model parameters and improving computational efficiency.

[0049] Among them, the time series network is a type of neural network specifically used to process time series data. Common examples include time series long short-term memory networks and gated recurrent units. These networks can capture long-term dependencies in time series and are suitable for processing industrial data with time dependencies. Time series networks can remember past information and combine it with current inputs to make more accurate predictions.

[0050] For example, common features are extracted from the preprocessed working data stream through convolutional neural networks and time series networks. Convolutional neural networks are mainly used to extract time domain features, while time series networks are used to extract frequency domain features. The combination of these two features can fully capture the operating status of the device.

[0051] Among them, time domain features refer to features directly extracted from time series data, which reflect the instantaneous state of the device at different time points, for example, the change of sensor data such as temperature, vibration, pressure, and current over time. Frequency domain features refer to features obtained by Fourier transform or other frequency domain analysis methods on time series data, which reflect the energy distribution of the device at different frequencies, for example, the spectrum diagram and power spectrum density of the vibration signal. Time domain features and frequency domain features complement each other and can describe the operating status of the device from different angles. Time domain features are suitable for capturing short-term changes, while frequency domain features are suitable for capturing long-term trends. By combining time domain and frequency domain features, the operating status of the device can be more comprehensively understood and the accuracy of status judgment can be improved.

[0052] Among them, the weighted multi-task loss function is a loss function for multi-task learning, which can assign different weights between multiple tasks so that the model can optimize multiple objectives at the same time. In this embodiment, each classification label corresponds to a task, and the weighted multi-task loss function assigns different weights according to the importance of each task. Through the weighted multi-task loss function, the state judgment rules corresponding to multiple classification labels can be optimized at the same time to ensure that the rules of each device are fully trained. The weight of the loss function can be dynamically adjusted according to the difficulty and importance of the task to ensure that the model can better adapt to different equipment types and working conditions. By reasonably allocating weights, it is possible to balance the allocation of resources between different tasks and avoid certain tasks being neglected.

[0053] Among them, the error change of the feature set refers to how the error between the feature set and the true label changes over time during the training process. By monitoring the error change, the training effect of the model can be evaluated and the parameters or structure of the model can be adjusted as needed. By observing the error change curve, it can be determined whether the model has converged or whether there is an overfitting or underfitting problem. According to the trend of the error change, the hyperparameters of the model, such as the learning rate, regularization coefficient, etc., can be dynamically adjusted to improve the performance of the model. If the error of a task is large, the weight of the task can be appropriately increased to ensure that it receives more attention.

[0054] Exemplarily, through the weighted multi-task loss function, the weights of each task are dynamically adjusted according to the number of each feature set and the error change. The adjustment of task weights is intended to balance the resource allocation between different tasks and ensure that each task can be fully trained. Task weights are not fixed, but are dynamically adjusted according to the number of feature sets and the error change. If the error of a task is large, the weight of the task will be increased to ensure that it receives more attention. By reasonably allocating weights, it is possible to balance the resource allocation between different tasks and avoid certain tasks being neglected. Task weights can be automatically adjusted according to actual conditions to ensure that the model can better adapt to different equipment types and working conditions.

[0055] For example, according to the weight of each task, the corresponding status judgment rules are generated. These rules are used to judge the current status of the equipment (such as normal, slightly abnormal, and seriously abnormal), and provide a basis for subsequent fault prediction and maintenance planning. Each classification mark corresponds to a specific equipment type and model, so the generated status judgment rules are personalized and can adapt to the needs of different equipment. With the addition of new data and the continuous optimization of the model, the status judgment rules can be dynamically updated to ensure that they are always in the best state.

[0056] In this embodiment, by combining the convolutional neural network and the timing network, time domain and frequency domain features can be extracted simultaneously, and the complex patterns in the operation of the equipment can be captured, thereby improving the accuracy of state judgment. According to the classification mark, a personalized state judgment rule is generated for each device to ensure that the rule can adapt to the needs of different devices and further improve the accuracy of judgment. Through the weighted multi-task loss function, the task weight can be dynamically adjusted to ensure that each task can be fully trained and avoid certain tasks from being ignored. By extracting common time domain and frequency domain features, the commonalities between different devices can be captured, and the generalization ability of the model can be improved, so that it can better adapt to new equipment types and working conditions. Through multi-task learning, multiple tasks can be optimized at the same time to ensure that each task can be fully trained and avoid the overfitting problem of a single task. By reasonably allocating task weights, the resource allocation between different tasks can be balanced to ensure that each task can obtain sufficient computing resources and avoid certain tasks from being ignored. According to the error changes of the task, the task weight can be dynamically adjusted to ensure that the resource allocation is more reasonable and improve the overall training efficiency. With the addition of new data and the continuous optimization of the model, the state judgment rule can be dynamically updated to ensure that it is always in the best state and adapt to the ever-changing production environment.

[0057] In one embodiment of the present application, the state judgment of the working data stream corresponding to each of the production equipment is performed according to the target state judgment rule to obtain the state judgment result corresponding to each of the production equipment, including: Extracting time domain features and frequency domain features corresponding to the working data stream; Combining the time domain features and frequency domain features corresponding to the working data stream to obtain a multi-dimensional feature vector; The multidimensional feature vector and the target state judgment rule are transmitted to a pre-trained decision tree for state judgment to obtain a state judgment result corresponding to each of the production equipment, wherein the decision tree is trained by finding the optimal segmentation point based on the features corresponding to the acquired historical work data stream.

[0058] Among them, the time domain features and frequency domain features extracted from the working data stream are merged to form a multidimensional feature vector. This multidimensional feature vector contains the operation information of the equipment at different time scales and can describe the status of the equipment more comprehensively. Time domain features (such as temperature, vibration, pressure, current, etc.) reflect the instantaneous state of the equipment in a short period of time, while frequency domain features (such as the spectrum diagram and power spectrum density of the vibration signal) reflect the changes in the equipment in the long-term trend. By merging these two features, the complex patterns in the operation of the equipment can be captured. The time domain features and frequency domain features complement each other and can describe the operating status of the equipment from different angles, thereby improving the accuracy of status judgment.

[0059] Among them, the decision tree is a supervised learning algorithm for classification and regression. It gradually divides the input data into different categories or output values ​​through a series of conditional judgments (i.e. "nodes"). Each internal node represents a test of a feature, each branch represents a test result, and each leaf node represents a final classification or prediction result. The structure of the decision tree is intuitive and easy to understand, and the splitting conditions of each node are clearly visible, which makes it easy for operators to understand and explain the decision-making process of the model. The decision tree does not need to assume the distribution form of the data and is applicable to various types of data. The decision tree can process multi-dimensional feature vectors and is suitable for complex industrial scenarios.

[0060] Among them, the pre-trained decision tree refers to a decision tree model that has been trained using historical data. The model can quickly make state judgments based on the input multi-dimensional feature vector and output the current state of the device (such as normal, slightly abnormal, and seriously abnormal). The decision tree can be trained offline to ensure the accuracy and stability of the model; while in the online state, the model can quickly make state judgments to meet real-time requirements. The reasoning process of the decision tree is simple and efficient, and can complete state judgments within milliseconds, which is suitable for application scenarios with high real-time requirements.

[0061] Among them, the optimal split point refers to the best feature segmentation threshold selected during the training process of the decision tree. By finding the optimal split point, the decision tree can divide the data set into subsets that are as pure as possible, so that the probability that the samples in each subset belong to the same category is maximized. Decision trees usually use indicators such as information gain and Gini coefficient to evaluate the quality of split points, and select the split point that maximizes information gain or minimizes Gini coefficient as the optimal split point. The decision tree recursively finds the optimal split point at each node, and gradually divides the data set into multiple subsets until the preset termination condition is reached (such as the number of samples at the leaf node is less than a certain threshold). Through pruning technology, decision trees can remove unnecessary branches, avoid model overfitting, and improve generalization ability.

[0062] The training process of the decision tree is based on the features corresponding to the historical work data stream. By analyzing historical data, the decision tree can learn the relationship between different features and equipment status, and find the optimal split point, thereby building a model that can accurately judge the equipment status. The training of the decision tree relies entirely on historical data and can automatically learn the complex relationship between different features and equipment status. As new data is added, the decision tree can be retrained regularly to ensure that the model can adapt to the ever-changing production environment. By finding the optimal split point, the decision tree can divide the data set into subsets that are as pure as possible, improving the accuracy of status judgment.

[0063] In this embodiment, by combining time domain features and frequency domain features, the operating status of the device can be described more comprehensively, and the complex patterns in the operation of the device can be captured, thereby improving the accuracy of state judgment. By finding the optimal segmentation point, the decision tree can divide the data set into subsets that are as pure as possible, ensuring that the probability that the samples in each subset belong to the same category is the highest, further improving the accuracy of state judgment. The decision tree trained based on historical data can learn the relationship between the characteristics and states of different devices, generate personalized state judgment rules, and ensure that the rules can adapt to the needs of different devices. The reasoning process of the decision tree is simple and efficient, and can complete state judgment within milliseconds, meet real-time requirements, and ensure that potential problems can be discovered in time. The decision tree can be trained offline to ensure the accuracy and stability of the model; while in the online state, the model can quickly make state judgments to meet real-time requirements.

[0064] In one embodiment of the present application, the fault trend prediction of the production line is performed according to the state judgment result corresponding to each of the production equipment to obtain the target fault trend, including: Obtaining historical failure events and environmental data corresponding to each of the production equipment; Integrate the state judgment results, historical fault events and environmental data corresponding to each of the production equipment to obtain an initial feature matrix, wherein the fault event includes the fault time and the fault condition, the row of the initial feature matrix includes the fault time, and the column of the initial feature matrix includes the state judgment results, the fault condition and the environmental data; Analyzing the state change rate, the fault frequency, and the environmental correlation according to the initial feature matrix, and constructing a target feature matrix according to the state change rate, the fault frequency, and the environmental correlation; The target feature matrix is ​​transmitted to a pre-built fault trend prediction model to obtain the target fault trend.

[0065] Among them, historical fault events refer to the fault records of production equipment in the past period of time. Each fault event includes the fault time (i.e. the specific time point when the fault occurred) and the fault situation (such as fault type, fault description, maintenance measures, etc.). Each fault event has a clear timestamp, which records the specific time when the fault occurred. The fault situation usually contains a detailed description, such as the fault type (mechanical fault, electrical fault, etc.), fault location (specific components), fault cause (overload, wear, etc.) and maintenance measures taken.

[0066] Among them, environmental data refers to the external environmental conditions in which production equipment is operating, such as temperature, humidity, vibration, noise, air quality, etc. These data reflect the external factors affecting the operation of the equipment, which may have a significant impact on the performance and failure risk of the equipment. Environmental data covers information in multiple dimensions and can fully reflect the external conditions when the equipment is operating. Environmental data is dynamic and changes with time, season, production tasks and other factors. Environmental data is closely related to the operating status of the equipment, and certain environmental factors (such as high temperature and high humidity) may increase the risk of equipment failure.

[0067] Among them, the state judgment results of production equipment, historical fault events and environmental data are integrated to form an initial feature matrix. Each row of the matrix represents a time point (usually the fault time), and each column represents a different feature (such as state judgment results, fault conditions, environmental data, etc.). In this way, multi-source data can be uniformly represented, which is convenient for subsequent analysis and modeling.

[0068] The initial feature matrix integrates data from different sources, including status judgment results, historical fault events, and environmental data, and can more comprehensively describe the operating status of the equipment. The initial feature matrix is ​​a structured data representation that facilitates subsequent feature extraction and model training. The initial feature matrix has time series characteristics and can capture the status changes and environmental impacts of the equipment at different time points.

[0069] Among them, the state change rate refers to the speed at which the device state changes over time. By performing time series analysis on the state judgment results, the change rate of the device state can be calculated to reflect the changing trend of the device state. For example, if the device gradually changes from a normal state to a slightly abnormal state and then to a seriously abnormal state, the state change rate can reflect the speed of this change. The state change rate can help identify the changing trend of the device state and discover potential failure risks in advance. The state change rate is a dynamic indicator that can reflect the changes in the device state in real time and is suitable for real-time monitoring and early warning. The state change rate is very sensitive to changes in the device state and can capture small changes, improving the accuracy of fault prediction.

[0070] Among them, the failure frequency refers to the number of times a device fails per unit time. By statistically analyzing historical failure events, the failure frequency of the device can be calculated to reflect the frequency of device failures. For example, if a device has 10 failures in the past year, its annual failure frequency is 10 times / year. The failure frequency can reflect the long-term failure trend of the device and help evaluate the reliability and stability of the device. By performing periodic analysis of the failure frequency, the periodic laws of equipment failures can be discovered and preventive measures can be taken in advance.

[0071] Environmental correlation refers to the degree of association between environmental data and equipment failure. By performing correlation analysis on environmental data and historical failure events, it is possible to identify which environmental factors have a significant impact on equipment failure. For example, a high temperature environment may increase the failure risk of some equipment, while a low humidity environment may cause failure of other equipment. Environmental correlation analysis can reveal the causal relationship between environmental factors and equipment failure, helping to better understand the cause of the failure. By identifying key environmental factors, the risk of equipment failure can be predicted in advance and corresponding preventive measures can be taken.

[0072] Among them, based on the state change rate, fault frequency and environmental correlation in the initial feature matrix, the target feature matrix is ​​further extracted and constructed. The target feature matrix contains more refined and targeted features, which can better reflect the failure trend of the equipment. Through feature selection technology, the target feature matrix retains the most representative and predictive features, removes redundant information, and improves the training efficiency and prediction accuracy of the model. The target feature matrix further processes and transforms the original features through feature engineering technology, enhancing the expressiveness of the features and the generalization ability of the model.

[0073] Among them, the fault trend prediction model is a machine learning model used to predict the future fault trend of equipment. The model is trained based on historical data and can predict the probability of equipment failure in the future based on the current state characteristics and environmental conditions. The fault trend prediction model can include random forests, gradient boosting trees, long short-term memory networks, etc.

[0074] Exemplarily, the historical fault event records of each device are obtained from the factory's historical database, including the fault time and fault condition. At the same time, the environmental data of the equipment during operation, such as temperature, humidity, vibration, etc., are obtained. For example, for a CNC machine tool, historical fault events may include the following: Fault time: 2024-01-15 10:00; Fault condition: spindle overheating, resulting in shutdown; environmental data may include: temperature: 35°C; Humidity: 70%; Vibration: 0.08 g. The state judgment results of the CNC machine tool, historical fault events, and environmental data are integrated to form an initial feature matrix. For example, the initial feature matrix can be shown in Table 1: Table 1 Initial feature matrix time Status judgment result Fault conditions Temperature (°C) humidity(%) Vibration(g) 2024-01-15 10:00 Slight abnormality Spindle overheating 35 70 0.08 2024-02-20 14:30 normal none 30 65 0.05 2024-03-10 09:45 Severe abnormalities Motor burnt out 40 80 0.12 .

[0075] By analyzing the initial feature matrix, the state change rate, fault frequency and environmental correlation are calculated. For example, in the past three months, the state of the CNC machine tool has gradually changed from normal to slightly abnormal, and then to seriously abnormal, with a state change rate of 0.05 / day; the CNC machine tool has had 3 failures in the past year, with an annual failure frequency of 3 times / year; through correlation analysis, it is found that there is a significant correlation between temperature and vibration and failure, especially when the temperature exceeds 35°C and the vibration exceeds 0.08 g, the probability of failure increases significantly.

[0076] Based on the state change rate, fault frequency and environmental relevance, the target feature matrix is ​​further extracted and constructed. The target feature matrix is ​​transmitted to a pre-built fault trend prediction model (such as a random forest model) for fault trend prediction.

[0077] In this embodiment, by analyzing the state change rate, failure frequency and environmental correlation, the potential failure risk of the equipment can be discovered in advance, small problems can be prevented from turning into major failures, and the possibility of unexpected downtime can be reduced. The state change rate is a dynamic indicator that can reflect the changes in the equipment state in real time. It is suitable for real-time monitoring and early warning to ensure timely response. By constructing a target feature matrix, the features of multiple dimensions such as state change rate, failure frequency and environmental correlation are integrated, which can more comprehensively describe the operating status of the equipment and improve the accuracy of fault prediction. The fault trend prediction model relies entirely on historical data for training, and can automatically learn the complex relationship between different features and equipment failures, further improving the accuracy of the prediction.

[0078] In one embodiment of the present application, generating a production line maintenance plan according to the target failure trend includes: Dividing the target fault trend according to a plurality of preset time windows to obtain a fault probability distribution corresponding to each of the preset time windows; Determining a risk score of the production equipment within each of the preset time windows according to the failure probability distribution corresponding to each of the preset time windows; Sorting the maintenance priorities of the plurality of production equipment according to the risk scores of the production equipment in each of the preset time windows to obtain an initial maintenance priority; According to the category information, adjusting the initial maintenance priority to obtain a target maintenance priority; According to the preset production plan and the target maintenance priority, the maintenance time window of the production equipment is determined to obtain the production line maintenance plan.

[0079] The preset time window refers to dividing the future time into multiple fixed time periods, which is used to evaluate the probability of equipment failure in each time period. Common preset time windows can be hours, days, weeks or months, depending on production requirements and equipment types. The preset time window can be flexibly set according to different production cycles and equipment characteristics to ensure that the maintenance plan can adapt to different application scenarios. By dividing multiple preset time windows, future failure trends can be continuously evaluated to ensure the comprehensiveness and accuracy of the maintenance plan.

[0080] Among them, the failure probability distribution refers to the probability of a production device failing in each preset time window. By analyzing the target failure trend, the failure probability distribution in each time window can be calculated to reflect the failure risk of the device in different time periods in the future. The failure probability distribution changes dynamically in different time windows, reflecting the changing trend of the device status. For example, some equipment may have a higher probability of failure during peak workloads and a lower probability of failure during off-peak hours. By dividing time into multiple windows, the failure risk of the equipment can be refined and potential high-risk time periods can be discovered in advance.

[0081] Among them, risk scoring refers to the quantitative assessment of the failure risk of production equipment in each preset time window based on the failure probability distribution. The risk score is usually a numerical value that reflects the likelihood and severity of equipment failure during that time period. The risk score not only considers the probability of failure, but may also be combined with other factors (such as the severity of the failure, the impact on production, etc.) for a comprehensive assessment. As the status of the equipment changes, the risk score will be dynamically adjusted to ensure that the maintenance plan can respond to the latest failure trends in a timely manner.

[0082] Among them, the initial maintenance priority refers to the result of preliminary sorting of production equipment according to risk scores. However, different types of equipment have different importance and maintenance requirements, so it is necessary to adjust the initial maintenance priority according to the category information of the equipment (such as equipment type, model, criticality, etc.) to obtain the final target maintenance priority. According to the category information of different equipment, the initial maintenance priority can be adjusted individually to ensure that key equipment is given priority maintenance. By adjusting the maintenance priority, maintenance resources can be reasonably allocated to ensure that maintenance work can be carried out when it is most needed and avoid waste of resources.

[0083] Among them, the production line maintenance plan refers to determining the specific maintenance time window for each production equipment based on the preset production plan and target maintenance priority, and generating a detailed maintenance task schedule. The selection of the maintenance time window should try to avoid affecting normal production while ensuring that maintenance work can be carried out at the best time. The maintenance plan is closely coordinated with the preset production plan to ensure that maintenance work does not interfere with the normal production progress. The selection of the maintenance time window has a certain degree of flexibility and can be dynamically adjusted according to the actual production situation and equipment status.

[0084] In this embodiment, by dividing the target fault trend into multiple preset time windows and calculating the fault probability distribution within each time window, potential fault risks can be discovered in advance, maintenance work can be arranged in time, and the possibility of unexpected downtime can be reduced. As the equipment status changes, the maintenance plan can be dynamically adjusted to ensure that maintenance work can be carried out when it is most needed, avoiding insufficient or excessive maintenance. Through comprehensive consideration of risk scores and category information, production equipment can be reasonably prioritized for maintenance to ensure that key equipment is given priority maintenance and avoid waste of resources. By adjusting the maintenance priority, maintenance resources can be reasonably allocated to ensure that maintenance work can be carried out at the most appropriate time and improve maintenance efficiency. The maintenance plan is closely coordinated with the preset production plan to ensure that maintenance work does not interfere with the normal production schedule and avoid production interruptions caused by maintenance work. The selection of the maintenance time window has a certain degree of flexibility and can be dynamically adjusted according to the actual production situation and equipment status to ensure that maintenance work can be carried out at the best time. By planning maintenance work in advance, the possibility of unexpected downtime can be reduced, the continuity of production can be ensured, and the stability of production can be improved. By optimizing the maintenance plan, product quality problems caused by equipment failure can be reduced, the qualified rate of products can be improved, and the overall production quality can be improved.

[0085] In an embodiment of the present application, after generating a production line maintenance plan according to the target failure trend, the method further includes: Determining a failure probability according to the target failure trend; Comparing the failure probability with a preset failure probability threshold to obtain a comparison result; When the comparison result shows that the fault probability is greater than or equal to the preset fault probability threshold, increasing the frequency of obtaining the work data stream corresponding to the production equipment; When the comparison result is that the fault probability is less than the preset fault probability threshold, the frequency of obtaining the work data stream corresponding to the production equipment is reduced.

[0086] For example, based on the previously generated target failure trend, the failure probability of each device in the future is calculated. The failure probability of each time window is compared with the preset failure probability threshold. According to the comparison results, the working data flow acquisition frequency of the CNC machine tool is dynamically adjusted: when the failure probability is less than the threshold (25%), the data acquisition frequency is reduced from once per minute to once every 5 minutes to save resources; when the failure probability is greater than the threshold (25%), the data acquisition frequency is increased from once per minute to once every 10 seconds to monitor the device status more frequently and detect potential problems in time.

[0087] In this embodiment, by dynamically adjusting the data collection frequency, data collection can be reduced during low-risk periods, reducing the computing load of edge computing nodes and cloud servers, and saving computing resources. Reducing the data collection frequency during low-risk periods can significantly reduce the amount of data transmitted to the cloud, reduce network bandwidth usage, and save communication costs. Reducing unnecessary data collection can extend the service life of sensors and reduce the maintenance and replacement costs of sensors.

[0088] It can be seen from the above embodiments that by acquiring the working data stream of production equipment in real time, the operating status of the equipment can be captured in time to avoid misjudgment or delays caused by data lag; by extracting valuable features from complex working data streams and generating accurate status judgment rules, potential problems can be discovered earlier to avoid small problems from turning into major failures, thereby improving production stability; by analyzing the status judgment results of production equipment, future failure trends can be predicted. This predictive maintenance method is more efficient than traditional time-based maintenance. Preventive measures can be taken before a failure occurs, reducing the possibility of unexpected downtime, thereby improving production efficiency; according to the target failure trend, a personalized maintenance plan can be generated to ensure that maintenance work is carried out at the most appropriate time, which not only reduces unnecessary maintenance costs, but also avoids waste of resources caused by excessive maintenance.

[0089] In one embodiment of the present application, a smart factory production line management and control device is also provided. Figure 3 is a schematic diagram of a smart factory production line control device shown in an exemplary embodiment of the present application, see Figure 3 , the device comprises: The data acquisition module 301 is used to obtain the work data flow and category information corresponding to multiple production equipment on the production line in the smart factory; A rule generation module 302, used to obtain a target state judgment rule according to the work data flows and category information corresponding to the plurality of production devices; The state judgment module 303 is used to judge the state of the work data stream corresponding to each of the production equipment according to the target state judgment rule, and obtain the state judgment result corresponding to each of the production equipment; A fault trend prediction module 304 is used to predict the fault trend of the production line according to the state judgment results corresponding to each of the production equipment to obtain a target fault trend; The production line maintenance plan generating module 305 is used to generate a production line maintenance plan according to the target failure trend, so as to maintain the multiple production equipment on the production line according to the production line maintenance plan.

[0090] The smart factory production line control device in the embodiment of the present application can capture the operating status of the equipment in a timely manner by acquiring the working data stream of the production equipment in real time, avoiding misjudgment or delays caused by data lag; it can extract valuable features from complex working data streams and generate accurate status judgment rules, so as to discover potential problems earlier and avoid small problems from turning into major failures, thereby improving production stability; by analyzing the status judgment results of the production equipment, it can predict future failure trends. This predictive maintenance method is more efficient than traditional time-based maintenance. Preventive measures can be taken before a failure occurs, reducing the possibility of unexpected downtime, thereby improving production efficiency; according to the target failure trend, a personalized maintenance plan can be generated to ensure that maintenance work is carried out at the most appropriate time, which not only reduces unnecessary maintenance costs, but also avoids waste of resources caused by excessive maintenance.

[0091] The specific embodiments of the smart factory production line control device in this application can refer to the examples shown in the above-mentioned smart factory production line control method, which will not be repeated here in this example.

[0092] In one embodiment of the present application, an electronic device for implementing the above-mentioned smart factory production line control method is also provided. The electronic device includes a memory and a processor, the memory stores a computer program, and the processor is configured to execute the above-mentioned smart factory production line control method through the computer program.

[0093] See also Figure 4 , Figure 44 is a schematic diagram of the structure of an electronic device shown in an exemplary embodiment of the present application. The computer system 400 includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 402 or the program loaded from the storage part 408 to the random access memory (RAM) 403, such as executing the method described in the above embodiment. In RAM 403, various programs and data required for system operation are also stored. CPU 401, ROM 402 and RAM 403 are connected to each other through bus 404. Input / output (I / O) interface 405 is also connected to bus 404.

[0094] The following components are connected to the I / O interface 405: an input section 406 including a keyboard, a mouse, etc.; an output section 407 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the I / O interface 405 as needed. A removable medium 411, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 410 as needed so that a computer program read therefrom is installed into the storage section 408 as needed.

[0095] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 409, and / or installed from a removable medium 411. When the computer program is executed by the central processing unit (CPU) 401, various functions defined in the system of the present application are executed.

[0096] It should be noted that the computer-readable medium shown in the embodiment of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier, which carries a computer-readable computer program. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. A computer program contained on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0097] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. Wherein, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0098] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.

[0099] Another aspect of the present application further provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program, wherein the computer program executes the above-mentioned smart factory production line control method when it is run. The computer-readable storage medium may be included in the electronic device described in the above embodiment, or may exist independently without being assembled into the electronic device.

[0100] Another aspect of the present application further provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the smart factory production line management and control method provided in each of the above embodiments.

[0101] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the technology may modify or change the above embodiments without violating the spirit and scope of the present invention. Therefore, all equivalent modifications or changes made by a person of ordinary skill in the art without departing from the spirit and technical ideas disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A smart factory production line management and control method, characterized in that: The method comprises: Obtain the work data flow and category information corresponding to multiple production equipment on the production line in the smart factory; Obtaining a target state judgment rule according to the work data flows and category information corresponding to the plurality of production devices; According to the target state judgment rule, the state of the work data stream corresponding to each of the production equipment is judged to obtain the state judgment result corresponding to each of the production equipment; According to the status judgment results corresponding to each of the production equipment, the fault trend of the production line is predicted to obtain the target fault trend; A production line maintenance plan is generated according to the target failure trend, so as to maintain the plurality of production equipment on the production line according to the production line maintenance plan.

2. The smart factory production line control method according to claim 1, characterized in that: The category information includes the device type and device model, the working data stream includes temperature time series data, vibration time series data, pressure time series data, and current time series data. After obtaining the working data streams and category information corresponding to the multiple production devices on the production line in the smart factory, it also includes: By setting up edge computing nodes around the production equipment, according to the equipment type and equipment model, a classification mark is added to the work data stream corresponding to each of the production equipment, and the classification mark is used to obtain the target state judgment rule; The temperature time series data, vibration time series data, pressure time series data, and current time series data are cleaned through the edge computing node, and the temperature time series data, vibration time series data, pressure time series data, and current time series data after data cleaning are standardized to obtain a preprocessed working data stream, and the preprocessed working data stream is used to obtain the target state judgment rule.

3. The smart factory production line control method according to claim 2, characterized in that: The target state judgment rule is obtained according to the work data streams and category information corresponding to the plurality of production devices, including: Extracting common features from the preprocessed working data stream by means of a convolutional neural network and a time series network, wherein the common features include common time domain features and frequency domain features; According to the classification mark, extracting multiple features corresponding to the classification mark from the common features, combining the multiple features to obtain a feature set corresponding to each of the classification marks; By using a preset weighted multi-task loss function, a plurality of task weights are obtained according to the number of the feature sets and the error change of the feature sets; According to each of the task weights, a target state judgment rule corresponding to each of the classification marks is generated.

4. The smart factory production line control method according to claim 1, characterized in that: The step of performing status judgment on the work data stream corresponding to each of the production equipment according to the target status judgment rule to obtain the status judgment result corresponding to each of the production equipment includes: Extracting time domain features and frequency domain features corresponding to the working data stream; Combining the time domain features and frequency domain features corresponding to the working data stream to obtain a multi-dimensional feature vector; The multidimensional feature vector and the target state judgment rule are transmitted to a pre-trained decision tree for state judgment to obtain a state judgment result corresponding to each of the production equipment, wherein the decision tree is trained by finding the optimal segmentation point based on the features corresponding to the acquired historical work data stream.

5. The smart factory production line control method according to claim 1, characterized in that: The method of predicting the fault trend of the production line according to the state judgment result corresponding to each of the production equipment to obtain the target fault trend includes: Obtaining historical failure events and environmental data corresponding to each of the production equipment; Integrate the state judgment results, historical fault events and environmental data corresponding to each of the production equipment to obtain an initial feature matrix, wherein the fault event includes the fault time and the fault condition, the row of the initial feature matrix includes the fault time, and the column of the initial feature matrix includes the state judgment results, the fault condition and the environmental data; Analyzing the state change rate, the fault frequency, and the environmental correlation according to the initial feature matrix, and constructing a target feature matrix according to the state change rate, the fault frequency, and the environmental correlation; The target feature matrix is ​​transmitted to a pre-built fault trend prediction model to obtain the target fault trend.

6. The smart factory production line control method according to claim 1, characterized in that: Generating a production line maintenance plan according to the target failure trend includes: Dividing the target fault trend according to a plurality of preset time windows to obtain a fault probability distribution corresponding to each of the preset time windows; Determining a risk score of the production equipment within each of the preset time windows according to the failure probability distribution corresponding to each of the preset time windows; Sorting the maintenance priorities of the plurality of production equipment according to the risk scores of the production equipment in each of the preset time windows to obtain an initial maintenance priority; According to the category information, adjusting the initial maintenance priority to obtain a target maintenance priority; According to the preset production plan and the target maintenance priority, the maintenance time window of the production equipment is determined to obtain the production line maintenance plan.

7. The smart factory production line control method according to claim 1, characterized in that: After generating a production line maintenance plan according to the target failure trend, the method further includes: Determining a failure probability according to the target failure trend; Comparing the failure probability with a preset failure probability threshold to obtain a comparison result; When the comparison result shows that the fault probability is greater than or equal to the preset fault probability threshold, increasing the frequency of obtaining the work data stream corresponding to the production equipment; When the comparison result is that the fault probability is less than the preset fault probability threshold, the frequency of obtaining the work data stream corresponding to the production equipment is reduced.

8. A smart factory production line control device, characterized in that: The device comprises: The data acquisition module is used to obtain the work data flow and category information corresponding to multiple production equipment on the production line in the smart factory; A rule generation module, used to obtain a target state judgment rule according to the work data flow and category information corresponding to the plurality of production devices; A state judgment module, used to judge the state of the work data stream corresponding to each of the production equipment according to the target state judgment rule, and obtain the state judgment result corresponding to each of the production equipment; A fault trend prediction module is used to predict the fault trend of the production line according to the status judgment results corresponding to each of the production equipment to obtain a target fault trend; The production line maintenance plan generating module is used to generate a production line maintenance plan according to the target failure trend, so as to maintain the multiple production equipment on the production line according to the production line maintenance plan.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein the computer program executes the smart factory production line management and control method described in any one of claims 1 to 7 when running.

10. An electronic device comprising a memory and a processor, characterized in that: The memory stores a computer program, and the processor is configured to execute the smart factory production line management and control method according to any one of claims 1 to 7 through the computer program.

Images