Digital factory monitoring data mining method and system

By using adaptive mode to capture multimodal data and perform edge computing and machine learning analysis in industrial monitoring systems, the problem of insufficient data fusion and prediction capabilities is solved, more efficient equipment status monitoring and production scheduling is achieved, and production is improved Efficiency and product quality.

CN118915591AActive Publication Date: 2024-11-08ZHIZHU TECH (JILIN PROVINCE) CO LTD

Patent Information

Application Number
CN202411127161.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-11-08
Estimated Expiration
2044-08-16

AI Technical Summary

Technical Problem

When facing complex and changing industrial environments and massive heterogeneous data, existing industrial monitoring systems lack effective data fusion mechanisms, resulting in low data analysis accuracy and efficiency, unable to respond to equipment failures or performance degradation in a timely manner, and lack of future trends. Predictive ability.

Method used

Adaptive mode is used to capture multimodal data of the operating state of the device in real time, and initially process and fusion through edge computing, map it to a unified data space, and use machine learning models to perform pattern recognition and abnormal detection, output analysis results and build a device health status model. , predict future trends, generate maintenance suggestions and production scheduling strategies.

Benefits of technology

Improve the accuracy of equipment status monitoring and fault prediction, enhance the efficiency of production scheduling decision-making, fully sense the equipment status, improve data quality and analysis efficiency, reduce unplanned downtime, and improve production efficiency and product quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118915591A_ABST
    Figure CN118915591A_ABST
Patent Text Reader

Abstract

The invention discloses a digital factory monitoring data mining method and system, and relates to the technical field of artificial intelligence, and the method comprises the steps: capturing multi-modal data in real time through an adaptive mode, carrying out the edge calculation preprocessing, carrying out the fusion mapping to a unified data space, employing a deep learning model to recognize the mode, and detecting the abnormality, the monitoring precision and the response speed are improved. And a preliminary analysis result is uploaded to a cloud, an equipment health state model is constructed, the operation trend is predicted, predictive maintenance is realized, and non-planned shutdown is reduced. And based on the prediction result, a maintenance suggestion and a production scheduling strategy are generated, the production process is optimized, efficiency is improved, and cost is reduced. According to the method, the data processing capability is enhanced, intelligent decision making is promoted, the method has an important value for improving the automation and intelligence level of a factory, production continuity and high efficiency are ensured, and the overall operation efficiency and market competitiveness are enhanced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a digital factory monitoring data mining method and system. Background Art

[0002] In recent years, with the rapid progress of the Internet of Things, big data and artificial intelligence technologies, the concept of Industry 4.0 has been widely promoted and applied around the world. Digital factories have achieved comprehensive monitoring and intelligent management of the production process. Traditional rule-based methods are gradually replaced by data-driven methods, which effectively improves the reliability and production efficiency of equipment. However, although existing technologies have improved the operation and maintenance management of equipment to a certain extent, there are still many challenges when facing complex and changing industrial environments and massive heterogeneous data.

[0003] At present, although most industrial monitoring systems can collect equipment operation data, they still have significant defects in data processing and analysis. The lack of an effective data fusion mechanism reduces the accuracy and efficiency of data analysis. Most existing anomaly detection technologies rely on single feature extraction or preset threshold judgment, which affects the accuracy of prediction. Traditional monitoring systems lack the ability to predict future trends when generating maintenance recommendations and production scheduling strategies, and cannot respond to possible equipment failures or performance degradation in a timely manner. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a digital factory monitoring data mining method and system to solve the problems of heterogeneity and real-time nature of multimodal data in digital factories, optimize the accuracy of equipment status monitoring and fault prediction, and improve the efficiency of production scheduling decisions.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, an embodiment of the present invention provides a digital factory monitoring data mining method, which includes, using an adaptive mode to capture multimodal data of physical parameters and environmental variables of the equipment operation status in real time, and uploading it to an edge computing device for preliminary processing; Fuse the preliminarily processed multimodal data and map them into a unified data space; Perform preliminary pattern recognition and anomaly detection on the fused multimodal data through machine learning models, output preliminary analysis results and upload them to the cloud platform; Based on the preliminary analysis results, build a health status model for the equipment, identify abnormal operation patterns of the equipment, and predict future operation trends; Based on the prediction results, maintenance recommendations and production scheduling strategies are generated.

[0007] As a preferred solution of the digital factory monitoring data mining method of the present invention, the method of using an adaptive mode to capture multimodal data of physical parameters and environmental variables of the equipment operating status in real time includes the following steps: Monitor temperature changes during equipment operation through temperature sensors to detect overheating and other temperature anomalies; Vibration sensors capture the vibration frequency and intensity of equipment during operation and analyze the mechanical stability of the equipment; Pressure sensors monitor the pressure of fluids and gases to maintain stable operation of equipment; The sound sensor records the sound of the device when it is running, and identifies abnormal noise through voiceprint analysis; Image sensors capture images of the equipment’s exterior and key parts for visual inspection and defect detection; Design an adaptive data acquisition frequency algorithm to automatically adjust the sensor sampling frequency based on the current load and operating status of the equipment; Upload the collected multimodal data to the edge computing device.

[0008] As a preferred solution of the digital factory monitoring data mining method described in the present invention, the preliminary processing includes data cleaning, denoising, filtering and data standardization processing on the collected data on the edge computing device.

[0009] As a preferred solution of the digital factory monitoring data mining method of the present invention, the fusion of the preliminarily processed multimodal data and mapping to a unified data space includes the following steps: Extract key features from the data of each modality based on the nature of the data and the analysis objectives; Use data fusion algorithm to stitch together the features of different modalities to form a high-dimensional feature vector; Create a unified data framework, and map all fused data into a unified data space using a mapping function or transformation matrix.

[0010] As a preferred solution of the digital factory monitoring data mining method of the present invention, the following steps are included: performing preliminary pattern recognition and anomaly detection on the fused multimodal data through a machine learning model, outputting preliminary analysis results and uploading them to the cloud platform: Adopt the convolutional neural network and long short-term memory network combination model in deep learning; The spatial features of different modal data are extracted through convolutional neural networks, and the long short-term memory network captures the dependencies in the time series; The multimodal data mapped into the unified data space after fusion is input into the model for preliminary pattern recognition and anomaly detection. The expression is: ; Where t is the current time, It's at the time The anomaly detection value of , N is the total number of modes, It is The importance weight of the modal data, The fusion function The calculated The mode at the time point The fusion value of It is based on the historical fusion data. The mode at the time point The predicted fusion value of and They are The mean and standard deviation of the modal data, is a balancing factor between statistical anomaly detection and prediction bias, is a non-negative exponent used to adjust the sensitivity of anomaly detection to outliers, is a non-negative exponent used to adjust the sensitivity of anomaly detection to prediction bias; The preliminary analysis results and processed data are compressed and packaged, and the data packets are encrypted using the SSL protocol and uploaded to the cloud platform for storage.

[0011] As a preferred solution of the digital factory monitoring data mining method of the present invention, the following steps are included: based on the preliminary analysis results, constructing a health status model of the equipment, identifying abnormal modes of equipment operation, and predicting future operation trends: The cloud platform decrypts the preliminary analysis results and the processed encrypted data packets; The hybrid expert network in deep learning is combined with the long short-term memory network to build the health status model of the equipment. The real-time health status and future trend prediction of the equipment are integrated to design the health status prediction function, which is expressed as: ; in, It's at the time The predicted value of the equipment health status, Z is the normalization constant, and They are The center position and width parameters of the abnormal detection value of the modal data, By long short-term memory network at time point The predicted The health status value of the modal data, It is The sensitivity of trend prediction of modal data, It is The threshold value for trend prediction of modal data, is the long short-term memory network at time point Predicted future trend value of the device.

[0012] As a preferred solution of the digital factory monitoring data mining method of the present invention, generating maintenance suggestions and production scheduling strategies based on the prediction results includes the following steps: In-depth analysis of the prediction results of the health status model to identify the changing trend of the equipment health status; Develop a schedule for preventive maintenance based on the forecast results; Assess the impact of failures of various equipment and components on the overall production process and determine maintenance priorities; Organize maintenance recommendations into reports based on maintenance priorities; Submit maintenance recommendation reports to the maintenance team for execution, and set up a tracking mechanism to monitor maintenance progress and results; Combine real-time monitoring data with predictive models to regularly assess the health status of all equipment; Predict possible future maintenance needs based on health status assessment; Re-plan production scheduling based on maintenance demand forecast; Introduce flexible production strategies to cope with equipment failures and maintenance needs.

[0013] In a second aspect, the present invention provides a digital factory monitoring data mining system, comprising: Multimodal data acquisition and edge computing preprocessing module: Adopts adaptive mode to capture multimodal data of physical parameters and environmental variables of equipment operation status in real time, and uploads them to edge computing devices for preliminary processing; Multimodal data fusion module: fuses the multimodal data after preliminary processing and maps it to a unified data space; Pattern recognition and anomaly detection module: performs preliminary pattern recognition and anomaly detection on the fused multimodal data through machine learning models, outputs preliminary analysis results and uploads them to the cloud platform; Cloud data analysis and prediction module: Based on the preliminary analysis results, it builds a health status model for the equipment, identifies abnormal operation patterns of the equipment, and predicts future operation trends; Decision-making and optimization module: Generates maintenance recommendations and production scheduling strategies based on prediction results.

[0014] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, any step of the digital factory monitoring data mining method as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the digital factory monitoring data mining method as described in the first aspect of the present invention is implemented.

[0016] The beneficial effects of the present invention are: the multimodal data of the equipment is captured in real time through an adaptive mode, and after edge computing preprocessing, it is integrated and mapped to a unified data space. The deep learning model is used to identify patterns and detect anomalies, and the analysis results are output to the cloud. Based on this, a device health status model is constructed to predict operating trends. This method comprehensively perceives the status of the equipment, improves data quality and analysis efficiency, and quantifies anomaly detection through the model to enhance prediction accuracy. The health status model combines deep learning with expert networks to achieve forward-looking maintenance and production scheduling optimization, reduce unplanned downtime, and improve production efficiency and product quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0018] Figure 1 This is a flow chart of the digital factory monitoring data mining method in Example 1. DETAILED DESCRIPTION

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0020] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0022] Example 1, reference Figure 1 , which is the first embodiment of the present invention, and provides a digital factory monitoring data mining method, comprising the following steps: S1. Using adaptive mode to capture multimodal data of physical parameters and environmental variables of equipment operation status in real time includes the following steps: Monitor temperature changes during equipment operation through temperature sensors to detect overheating or other temperature anomalies; Vibration sensors capture the vibration frequency and intensity of equipment during operation and are used to analyze the mechanical stability of the equipment; Pressure sensors monitor the pressure of fluids and gases to ensure stable system operation; The sound sensor records the sound of the device when it is running, and identifies abnormal noise through voiceprint analysis; Image sensors capture images of the equipment’s exterior and key parts for visual inspection and defect detection; Design an adaptive data acquisition frequency algorithm to automatically adjust the sensor sampling frequency based on the current load and operating status of the equipment; Upload the collected multimodal data to the edge computing device.

[0023] Furthermore, the expression of the adaptive data acquisition frequency algorithm is: ; in, For the current moment, For at the moment The frequency of data collection, For the device at time The load level, is the running status of the device (0 means stop, 1 means start, 2 means high load), and are the adjustment factor and the base load level, Controls the steepness of the effect of load changes on the acquisition frequency. Determined the load threshold at which the acquisition frequency starts to increase significantly; When the device is not started, =0, then =0, data collection is suspended; when the device is started but has not reached high load, =1, acquisition frequency Depending on the load level Automatically adjusts as near , the acquisition frequency gradually increases; when the device is under high load, =2, at this time What is the specific value of and the frequency of collection? will automatically increase to maximum to ensure all critical data is captured.

[0024] S2. Preliminary processing includes data cleaning, denoising, filtering and data standardization of the collected data on the edge computing device.

[0025] Furthermore, statistical methods (such as Z-score, IQR) or machine learning algorithms (such as IsolationForest, Local Outlier Factor) are used to identify and remove outliers caused by sensor failures; missing values ​​are filled by neighboring point interpolation (such as linear interpolation), model-based prediction (such as KNN Imputation), mean / median filling, etc.; low-pass filters are used to remove high-frequency noise and retain the basic trend of the signal, while high-pass filters are used to eliminate slowly changing background signals and highlight the rapidly changing parts of the signal; data are smoothed by calculating the average value within a sliding window of the data sequence; data are converted into a standard normal distribution with a mean of 0 and a standard deviation of 1, which is suitable for situations where the data has no obvious boundaries or follows a normal distribution; data are scaled to between 0 and 1, which is suitable for situations where the data has clear boundaries.

[0026] S3, fusing the preliminarily processed multimodal data and mapping them to a unified data space includes the following steps: Extract key features from the data of each modality based on the nature of the data and the analysis objectives; Use data fusion algorithms (such as weighted fusion, decision-level fusion, feature-level fusion, etc.) to stitch together features of different modalities to form a high-dimensional feature vector; Create a unified data framework, and map all fused data into a unified data space using a mapping function or transformation matrix.

[0027] Furthermore, a multivariate data fusion function F is constructed to map multimodal data into a unified space, which is expressed as: ; Where T represents the temperature data feature vector, V represents the vibration data feature vector, P represents the pressure data feature vector, S represents the sound data feature vector, I represents the image data feature vector, i is the modal data index, and N is the total number of modalities. is the eigenvector of the i-th mode, is the weight coefficient corresponding to the i-th modal data, indicating the importance of the modal data in the overall fusion. is the data fusion function of the i-th modal data.

[0028] Furthermore, for each data type, a specific fusion function is designed, and the expression is: ; in, is the eigenvector of the i-th modal data at the j-th time point, and are the mean and standard deviation of the temperature data, respectively, used to normalize the data.

[0029] S4. Performing preliminary pattern recognition and anomaly detection on the fused multimodal data through a machine learning model, outputting preliminary analysis results and uploading them to the cloud platform includes the following steps: Adopt the convolutional neural network and long short-term memory network combination model in deep learning; The spatial features of different modal data are extracted through convolutional neural networks, and the long short-term memory network captures the dependencies in the time series; The multimodal data mapped into the unified data space after fusion is input into the model for preliminary pattern recognition and anomaly detection. The expression is: ; Where t is the current time, It's at the time The anomaly detection value of , N is the total number of modes, It is The importance weight of the modal data, The fusion function The calculated The mode at the time point The fusion value of It is based on the historical fusion data. The mode at the time point The predicted fusion value of and They are The mean and standard deviation of the modal data, is a balancing factor between statistical anomaly detection and prediction bias, is a non-negative exponent used to adjust the sensitivity of anomaly detection to outliers, is a non-negative exponent used to adjust the sensitivity of anomaly detection to prediction bias; The preliminary analysis results and processed data are compressed and packaged, and the data packets are encrypted using the SSL protocol and uploaded to the cloud platform for storage.

[0030] Furthermore, the anomaly detection function The value range of is [0, +∞], and the threshold of anomaly detection is defined as ,when < When , it indicates that the equipment is operating normally without any abnormality; when ≥ , it indicates that the equipment is operating abnormally and needs further observation and maintenance measures if necessary.

[0031] S5. Based on the preliminary analysis results, construct a health status model of the equipment, identify abnormal modes of equipment operation, and predict future operation trends, including the following steps: The cloud platform decrypts the preliminary analysis results and the processed encrypted data packets; The hybrid expert network in deep learning is combined with the long short-term memory network to build the health status model of the equipment. The real-time health status and future trend prediction of the equipment are integrated to design the health status prediction function, which is expressed as: ; in, It's at the time The predicted value of the equipment health status, Z is the normalization constant, and They are The center position and width parameters of the abnormal detection value of the modal data, By long short-term memory network at time point The predicted The health status value of the modal data, It is The sensitivity of trend prediction of modal data, It is The threshold value for trend prediction of modal data, is the long short-term memory network at time point Predicted future trend value of the device.

[0032] Furthermore, the health status prediction function The value range of is generally designed to be 0 to 1. When it is close to 0, it means that the health status of the device is poor and there may be an impending failure. When it is close to 1, it means that the equipment is in good health, running stably, and the risk of failure is low.

[0033] S6. Based on the prediction results, generating maintenance suggestions and production scheduling strategies includes the following steps: In-depth analysis of the prediction results of the health status model to identify the changing trends of the equipment health status; based on the prediction results, formulate a schedule for preventive maintenance, including but not limited to regular inspections, component lubrication, cleaning and replacement of parts that are about to fail; evaluate the impact of the failure of each equipment and component on the overall production process and determine the maintenance priority order; based on the maintenance priority order, organize the maintenance recommendations into a report, including recommended maintenance activities, estimated maintenance time, required resources and spare parts list, and possible maintenance team assignments; submit the maintenance recommendation report to the maintenance team for execution, and set up a tracking mechanism to monitor the progress and effectiveness of maintenance; combine real-time monitoring data and predictive models to regularly evaluate the health status of all equipment; based on the health status assessment, predict possible future maintenance needs, including expected maintenance time windows and possible downtime; based on maintenance demand forecasts, re-plan production scheduling, which may include rescheduling tasks on the production line, adjusting work shifts, or purchasing necessary raw materials in advance; introduce flexible production strategies, such as activating spare production lines, arranging overtime, or outsourcing some production tasks to deal with equipment failures and maintenance needs.

[0034] This embodiment also provides a digital factory monitoring data mining system, including: Multimodal data acquisition and edge computing preprocessing module: Adopts adaptive mode to capture multimodal data of physical parameters and environmental variables of equipment operation status in real time, and uploads them to edge computing devices for preliminary processing; Multimodal data fusion module: fuses the multimodal data after preliminary processing and maps it to a unified data space; Pattern recognition and anomaly detection module: performs preliminary pattern recognition and anomaly detection on the fused multimodal data through machine learning models, outputs preliminary analysis results and uploads them to the cloud platform; Cloud data analysis and prediction module: Based on the preliminary analysis results, it builds a health status model for the equipment, identifies abnormal operation patterns of the equipment, and predicts future operation trends; Decision-making and optimization module: Generates maintenance recommendations and production scheduling strategies based on prediction results.

[0035] This embodiment also provides a computer device suitable for the case of a digital factory monitoring data mining method, including: a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the digital factory monitoring data mining method proposed in the above embodiment.

[0036] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covered on the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0037] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, the digital factory monitoring data mining method proposed in the above embodiment is implemented; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, referred to as EPROM), programmable read-only memory (Programmable Red-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0038] In summary, the present invention captures multimodal data of equipment in real time through adaptive mode, and after edge computing preprocessing, it is integrated and mapped to a unified data space. A deep learning model is used to identify patterns and detect anomalies, and the analysis results are output to the cloud. Based on this, a device health status model is constructed to predict operating trends. This method comprehensively perceives the status of the equipment, improves data quality and analysis efficiency, and quantifies anomaly detection through models to enhance prediction accuracy. The health status model combines deep learning with expert networks to achieve forward-looking maintenance and production scheduling optimization, reduce unplanned downtime, and improve production efficiency and product quality.

[0039] Example 2, referring to Table 1, is the second example of the present invention. To further verify the advancement of the present invention, experimental simulation data of a digital factory monitoring data mining method is provided.

[0040] First, the experimental object was a high-speed stamping machine, a key equipment on an automated production line in a manufacturing factory. The high-speed stamping machine was equipped with temperature sensors, vibration sensors, pressure sensors, sound sensors and image sensors.

[0041] The device is connected to an edge computing device with a built-in adaptive data acquisition frequency algorithm. The algorithm parameters are set as follows: α = 0.5, β = 80% of the maximum load. Data preprocessing and analysis software is installed on the edge computing device to perform data cleaning, denoising, standardization, and subsequent data fusion and pattern recognition. The data analysis results are uploaded to a secure cloud server for long-term storage and further deep learning analysis.

[0042] Before the equipment is started, all sensors are in standby mode and the data collection frequency is 0. When the equipment is started and enters normal operation, the adaptive data collection frequency algorithm automatically adjusts the sampling frequency of the sensor according to the equipment load level. Every 10 minutes, the edge computing device performs data preprocessing, including outlier identification, missing value filling, filtering and smoothing. The processed data is fused and mapped to a unified data space, and then a deep learning model is used for preliminary pattern recognition and anomaly detection. The anomaly detection results and preliminary analysis are uploaded to the cloud for further health status prediction and trend analysis. Finally, maintenance recommendations and production scheduling strategies are generated based on the prediction results, as shown in Table 1: Table 1 Equipment health status monitoring system performance evaluation table

[0043] By analyzing the above table data, it is obvious that the effectiveness of the adaptive data acquisition frequency algorithm and its significant improvement in the performance of the equipment health status monitoring system. As the equipment load level increases, the acquisition frequency is automatically adjusted upward, which ensures that more key data can be captured when the equipment is under high load, thereby improving the accuracy of anomaly detection. For example, at a load level of 90%, the acquisition frequency of punching machine C reached 8Hz, and the anomaly detection score also increased to 0.20, indicating that there is a potential health risk for the equipment at this stage.

[0044] The health status prediction value reflects the overall health status of the equipment. The closer the value is to 1, the better the health status of the equipment. From the data, we can see that the health status prediction values ​​of punching machines A and B are 0.95 and 0.90 respectively, which indicates that at lower load levels, the equipment operates stably and the risk of failure is low. However, as the load increases, such as punching machines D and E, the health status prediction values ​​drop to 0.60 and 0.55 respectively, indicating that the equipment may need preventive maintenance.

[0045] By comparing data collected at a fixed frequency, adaptive data collection not only improves data quality and analysis efficiency, but also significantly reduces unnecessary data storage and transmission costs. In addition, combined with the anomaly detection and health status prediction of the deep learning model, potential problems in equipment operation can be discovered in a timely manner, maintenance can be carried out in advance, unplanned downtime can be effectively avoided, and production efficiency and product quality can be improved.

[0046] In summary, the equipment health status monitoring system that adopts adaptive data acquisition frequency algorithm and multimodal data fusion has shown significant advantages in improving monitoring intelligence and optimizing maintenance decisions, and plays an important role in promoting industrial production automation and intelligence.

[0047] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A digital factory monitoring data mining method, Features: include, Adopting adaptive mode, multi-modal data of physical parameters and environmental variables of equipment operation status is captured in real time and uploaded to edge computing devices for preliminary processing; Fuse the preliminarily processed multimodal data and map them into a unified data space; Perform preliminary pattern recognition and anomaly detection on the fused multimodal data through machine learning models, output preliminary analysis results and upload them to the cloud platform; Based on the preliminary analysis results, build a health status model for the equipment, identify abnormal operation patterns of the equipment, and predict future operation trends; Based on the prediction results, maintenance recommendations and production scheduling strategies are generated.

2. The digital factory monitoring data mining method according to claim 1, Features: The method of using the adaptive mode to capture multimodal data of physical parameters and environmental variables of the equipment operation status in real time includes the following steps: Monitor temperature changes during equipment operation through temperature sensors to detect overheating and other temperature anomalies; Vibration sensors capture the vibration frequency and intensity of equipment during operation and analyze the mechanical stability of the equipment; Pressure sensors monitor the pressure of fluids and gases to maintain stable operation of equipment; The sound sensor records the sound of the device when it is running, and identifies abnormal noise through voiceprint analysis; Image sensors capture images of the equipment’s exterior and key parts for visual inspection and defect detection; Design an adaptive data acquisition frequency algorithm to automatically adjust the sensor sampling frequency based on the current load and operating status of the equipment; Upload the collected multimodal data to the edge computing device.

3. The digital factory monitoring data mining method according to claim 1, Features: The preliminary processing includes data cleaning, denoising, filtering and data standardization on the collected data on the edge computing device.

4. The digital factory monitoring data mining method according to claim 3, Features: The fusion of the preliminarily processed multimodal data and mapping it to a unified data space includes the following steps: Extract key features from the data of each modality based on the nature of the data and the analysis objectives; Use data fusion algorithm to stitch together the features of different modalities to form a high-dimensional feature vector; Create a unified data framework, and use the mapping function to map all the fused data into a unified data space.

5. The digital factory monitoring data mining method according to claim 4, Features: The following steps are used to perform preliminary pattern recognition and anomaly detection on the fused multimodal data through machine learning models, output preliminary analysis results and upload them to the cloud platform: Adopt the convolutional neural network and long short-term memory network combination model in deep learning; The spatial features of different modal data are extracted through convolutional neural networks, and the long short-term memory network captures the dependencies in the time series; The multimodal data mapped into the unified data space after fusion is input into the model for preliminary pattern recognition and anomaly detection. The expression is: ; Where t is the current time, It's at the time The anomaly detection value of , N is the total number of modes, It is The importance weight of the modal data, The fusion function The calculated The mode at the time point The fusion value of It is based on the historical fusion data. The mode at the time point The predicted fusion value of and They are The mean and standard deviation of the modal data, is a balancing factor between statistical anomaly detection and prediction bias, is a non-negative exponent used to adjust the sensitivity of anomaly detection to outliers, is a non-negative exponent used to adjust the sensitivity of anomaly detection to prediction bias; The preliminary analysis results and processed data are compressed and packaged, and the data packets are encrypted using the SSL protocol and uploaded to the cloud platform for storage.

6. The digital factory monitoring data mining method according to claim 5, Features: Based on the preliminary analysis results, building a health status model for the equipment, identifying abnormal modes of equipment operation, and predicting future operating trends includes the following steps: The cloud platform decrypts the preliminary analysis results and the processed encrypted data packets; The hybrid expert network in deep learning is used in combination with the long short-term memory network to build the health status model of the equipment. The real-time health status and future trend prediction of the equipment are integrated to design the health status prediction function, which is expressed as: ; in, It's at the time The predicted value of the equipment health status, Z is the normalization constant, and They are The center position and width parameters of the abnormal detection value of the modal data, By long short-term memory network at time point The predicted The health status value of the modal data, It is The sensitivity of trend prediction of modal data, It is The threshold value for trend prediction of modal data, is the long short-term memory network at time point Predicted future trend value of the device.

7. The digital factory monitoring data mining method according to claim 6, Features: Based on the prediction results, generating maintenance recommendations and production scheduling strategies includes the following steps: In-depth analysis of the prediction results of the health status model to identify the changing trend of the equipment health status; Develop a schedule for preventive maintenance based on the forecast results; Assess the impact of failures of various equipment and components on the overall production process and determine maintenance priorities; Organize maintenance recommendations into reports based on maintenance priorities; Submit maintenance recommendation reports to the maintenance team for execution, and set up a tracking mechanism to monitor maintenance progress and results; Combine real-time monitoring data with predictive models to regularly assess the health status of all equipment; Predict possible future maintenance needs based on health status assessment; Re-plan production scheduling based on maintenance demand forecast; Introduce flexible production strategies to cope with equipment failures and maintenance needs.

8. A digital factory monitoring data mining system, based on the digital factory monitoring data mining method according to any one of claims 1 to 7, Features: include, Multimodal data acquisition and edge computing preprocessing module: Adopts adaptive mode to capture multimodal data of physical parameters and environmental variables of equipment operation status in real time, and uploads them to edge computing devices for preliminary processing; Multimodal data fusion module: fuses the preliminarily processed multimodal data and maps them into a unified data space; Pattern recognition and anomaly detection module: Use machine learning models to perform preliminary pattern recognition and anomaly detection on the fused multimodal data, output preliminary analysis results and upload them to the cloud platform; Cloud data analysis and prediction module: Based on the preliminary analysis results, it builds a health status model for the equipment, identifies abnormal operation patterns of the equipment, and predicts future operation trends; Decision-making and optimization module: Generates maintenance recommendations and production scheduling strategies based on prediction results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program. Features: When the processor executes the computer program, the steps of the digital factory monitoring data mining method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, Features: When the computer program is executed by a processor, the steps of the digital factory monitoring data mining method according to any one of claims 1 to 7 are implemented.

Citation Information

Patent Citations

  • Highway bridge comprehensive monitoring system based on 5G and artificial intelligence

    CN117236930A

  • Method and system for monitoring health state of transformer of power station

    CN117630758A

  • Blast furnace condition intelligent evaluation method and system based on big data

    CN117787743A

  • Intelligent equipment monitoring and early warning system

    CN117826715A

  • Safety monitoring method for online operation state of electricity utilization information acquisition terminal

    CN117833464A

Cited By

  • Equipment management method and system for 3D digital factory

    CN119439937A

  • Architecture system for intelligent production of digital factory and implementation method thereof

    CN119556651A

  • Method and system for recording and storing multi-protocol communication messages of secondary equipment of transformer substation

    CN119728153A

  • Image and data internet-of-things combined equipment monitoring method

    CN119828534A

  • Gas power plant redundancy control method and system based on multi-modal perception and AI decision

    CN120178653A