Digital intelligent production platform management method and system combined with Internet of Things

Abstract

The invention discloses a digital intelligent production platform management method and system combined with the Internet of Things, and mainly relates to the technical field of platform management. Comprising the steps of collecting operation data of a plurality of production point locations in a preset window in a target production line corresponding to the digital intelligent production platform through an Internet of Things sensor, and obtaining an operation data sequence of the plurality of production point locations; performing anomaly observation through a random network distillation mechanism, and performing supplementary data verification based on anomaly observation results to obtain anomaly observation results of a plurality of production point locations; carrying out production abnormity trend feature identification; and performing abnormal monitoring channel resource management on the digital intelligent production platform to obtain a target platform management scheme. The method has the beneficial effects that the technical problem that the production efficiency is affected due to unreasonable platform management resource allocation caused by lack of collaboration and independence analysis of production point locations in the prior art is solved, and the technical effect of improving the production platform management reliability is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of platform management, and particularly relates to a digital wisdom production platform management method and system combined with the Internet of Things. BACKGROUND

[0002] In modern industrial production, with the improvement of production automation, digital wisdom production platforms have gradually become important tools for realizing intelligent production. Internet of Things technology is widely used in digital wisdom production platforms, and through sensor and device interconnection, various types of data on the production line, such as temperature, humidity, pressure, and flow, are collected in real time. These data are used for real-time monitoring, analysis, and decision-making. Currently, many digital production platforms use Internet of Things data for device monitoring, fault diagnosis, and production scheduling.

[0003] For example, the invention patent with the announcement number CN109949007A discloses a mine production management platform and a management method, which includes a collection and control system, a distribution and storage system, and a visual management and control system. The collection and control system is used to realize real-time collection and control of automatic data and information data of the management platform. The distribution and storage system is used to publish and store the automatic data and information data in real time using a message queue method. The visual management and control system is used to receive the published automatic data and information data, and to perform visual management, remote control, and big data analysis on the mine.

[0004] For example, the invention patent with the announcement number CN119356269A discloses a cloud management Internet of Things system and a management method for a smart factory. The user platform processes first attention information to generate a first control instruction. The service platform forwards the first attention information and sends it to the user platform. The management platform processes production-related information to generate a first attention information to generate a second control instruction. The sensing network platform is used to forward the production-related information to the management platform. The object platform controls the production line for intelligent production according to the first control instruction and / or the second control instruction obtained and forwarded in turn by the service platform, the management platform, and the sensing network platform. The object platform is also used to receive production-related information and send it to the management platform and the service platform, thereby generating the first control instruction and the second control instruction.

[0005] However, in the process of implementing the technical scheme of the present application, the above-mentioned technology at least has the following technical problems: In the process of processing production point operation data, there is a lack of targeted abnormal observation mechanism, and only simple data threshold judgment or single algorithm analysis can hardly accurately capture potential abnormalities in complex production scenarios. Meanwhile, without collaborative and independent classification and identification of abnormal points, it is difficult to distinguish multi-point linkage abnormalities from single-point independent abnormalities, thereby affecting the accuracy of abnormal trend feature extraction. In the abnormal monitoring process, the dynamic management of channel resources is not combined with trend features, which may cause unreasonable resource allocation and fail to adapt to the processing needs of different types of abnormalities, ultimately leading to insufficient pertinence and effectiveness of the platform management scheme. SUMMARY

[0006] The present application provides a combination of Internet of Things digital smart production platform management method and system, which is used to solve the technical problem of lack of collaborative and independent analysis of production points in the prior art, leading to unreasonable allocation of platform management resources and affecting production efficiency.

[0007] In view of the above problems, the present application provides a combination of Internet of Things digital smart production platform management method and system.

[0008] The first aspect of the present application provides a combination of Internet of Things digital smart production platform management method, which comprises: Collecting operation data of a plurality of production points in a target production line corresponding to a digital smart production platform within a preset window through Internet of Things sensors to obtain a plurality of production point operation data sequences; Performing abnormal observation on the plurality of production point operation data sequences through a random network distillation mechanism, and performing supplementary data verification based on the abnormal observation results to obtain a plurality of production point abnormal observation results; Performing collaborative point identification and independent point identification on the plurality of production point abnormal observation results to obtain a plurality of collaborative abnormal production point sets and independent abnormal production point sets; Performing production abnormal trend feature identification on the plurality of collaborative abnormal production point sets and independent abnormal production point sets respectively to obtain a plurality of collaborative abnormal trend features and independent abnormal trend feature sets; Performing abnormal monitoring channel resource management on the digital smart production platform according to the plurality of collaborative abnormal trend features and independent abnormal trend feature sets to obtain a target platform management scheme.

[0009] In one possible implementation, performing abnormal observation on the plurality of production point operation data sequences through a random network distillation mechanism, and performing supplementary data verification based on the abnormal observation results to obtain a plurality of production point abnormal observation results, comprises: Calling a target network and a prediction network in the random distillation mechanism; respectively, the target network and the prediction network are used to perform state prediction on the plurality of production point operation data sequences, to obtain a plurality of target operation state prediction results and a plurality of real-time prediction operation state prediction results; The mapping Euclidean distance calculation is performed on the plurality of target operation state prediction results and the plurality of real-time prediction operation state prediction results, to determine a plurality of initial abnormal observation results; When the plurality of initial abnormal observation results are greater than or equal to a preset threshold, supplementary data verification is performed, the plurality of initial abnormal observation results are corrected according to the supplementary data verification result, and the plurality of initial abnormal observation results are added to the plurality of production point abnormal observation results; When the plurality of initial abnormal observation results are less than the preset threshold, the plurality of initial abnormal observation results are added to the plurality of production point abnormal observation results.

[0010] In a possible implementation, the target network is a pre-trained network, which is used to represent the behavior mode of the production point under normal operation conditions, and the output is a prediction result of the standard behavior mode; The prediction network is a network that is continuously updated according to real-time production conditions, which is used to represent the behavior prediction of the production point at the current moment, and the output is a prediction result of the real-time behavior mode.

[0011] In a possible implementation, when the plurality of initial abnormal observation results are greater than or equal to a preset threshold, supplementary data verification is performed, the plurality of initial abnormal observation results are corrected according to the supplementary data verification result, and the plurality of initial abnormal observation results are added to the plurality of production point abnormal observation results, including: When the plurality of initial abnormal observation results are greater than or equal to a preset threshold, the initial abnormal observation results of the plurality of production points in the preset neighborhood bandwidth are obtained, to obtain a plurality of neighborhood initial abnormal observation result sets; The supplementary data verification is performed on the plurality of neighborhood initial abnormal observation result sets and the plurality of initial abnormal observation results, to obtain a supplementary data verification result; When the supplementary data verification result is consistent, the plurality of initial abnormal observation results are added to the plurality of production point abnormal observation results; When the supplementary data verification result is inconsistent, a plurality of secondary window supplementary data are obtained, the plurality of initial abnormal observation results are corrected based on the plurality of secondary window supplementary data, and the plurality of corrected abnormal observation results are added to the plurality of production point abnormal observation results.

[0012] In a possible implementation, the plurality of production point abnormal observation results are subjected to cooperative point identification and independent point identification, to obtain a plurality of cooperative abnormal production point sets and a plurality of independent abnormal production point sets, including: randomly extracting, without replacement, a first production point anomaly observation result from the plurality of production point anomaly observation results; extracting, without replacement, from the plurality of production point anomaly observation results, a production point corresponding to a production point anomaly observation result whose Euclidean distance to the first production point anomaly observation result is within a preset distance threshold, and adding the first production point corresponding to the first production point anomaly observation result into a first collaborative anomaly production point set; randomly extracting, without replacement, a second production point anomaly observation result from the plurality of production point anomaly observation results again, and constructing a second collaborative anomaly production point set, and so on until a preset extraction number is met, to obtain a plurality of collaborative anomaly production point sets; adding, into an independent anomaly production point set, production points in the plurality of production points other than the plurality of collaborative anomaly production point sets.

[0013] In a possible implementation, production anomaly trend feature identification is respectively performed on the plurality of collaborative anomaly production point sets and the independent anomaly production point set to obtain a plurality of collaborative anomaly trend features and an independent anomaly trend feature set, including: obtaining a plurality of historical production point anomaly observation result sequence sets and a historical independent anomaly production point anomaly observation result sequence set of the plurality of collaborative anomaly production point sets and the independent anomaly production point set within a preset historical window; performing production anomaly trend feature identification on the plurality of historical production point anomaly observation result sequence sets to obtain a plurality of historical production anomaly trend feature sets; calculating a mean value of the plurality of historical production anomaly trend feature sets to obtain a plurality of collaborative anomaly trend features; respectively performing production anomaly trend feature identification on the historical independent anomaly production point anomaly observation result sequence set to obtain an independent anomaly trend feature set.

[0014] In a possible implementation, performing production anomaly trend feature identification on the plurality of historical production point anomaly observation result sequence sets to obtain a plurality of historical production anomaly trend feature sets includes: calling an anomaly trend feature identifier to identify the plurality of historical production point anomaly observation result sequence sets to obtain a plurality of historical production anomaly trend feature sets; The anomaly trend feature identifier is obtained based on a feedforward neural network after supervised training.

[0015] In a possible implementation, the digital intelligent production platform is monitored in an anomaly monitoring channel resource management manner according to the plurality of collaborative anomaly trend features and the independent anomaly trend feature set to obtain a target platform management scheme, including: The collaborative monitoring weight and the independent monitoring weight are obtained by combining the collaborative abnormal production point set and the independent abnormal production point set for collaborative monitoring and independent monitoring weight analysis, wherein the collaborative monitoring weight and the independent monitoring weight add up to 1. The abnormality degree analysis is performed on the plurality of collaborative abnormality degrees and the independent abnormality degrees to obtain a plurality of collaborative abnormality degrees and independent abnormality degrees. The collaborative monitoring sub-weights and the independent monitoring sub-weights are obtained by splitting the collaborative monitoring weight and the independent monitoring weight based on the plurality of collaborative abnormality degrees and the independent abnormality degrees. The target platform management scheme is obtained by performing abnormal monitoring channel resource management on the digital intelligent production platform according to the plurality of collaborative monitoring sub-weights and the independent monitoring sub-weights.

[0016] In one possible implementation, the collaborative monitoring sub-weights and the independent monitoring sub-weights are obtained by splitting the collaborative monitoring weight and the independent monitoring weight based on the plurality of collaborative abnormality degrees and the independent abnormality degrees, comprising: The sum of the plurality of collaborative abnormality degrees and the independent abnormality degrees is calculated to obtain a plurality of collaborative abnormality degree sums and independent abnormality degree sums. The ratio of each collaborative abnormality degree to the plurality of collaborative abnormality degree sums is taken as a plurality of collaborative monitoring sub-weights. The ratio of each independent abnormality degree to the independent abnormality degree sum is taken as an independent monitoring sub-weight to obtain an independent monitoring sub-weight set.

[0017] In a second aspect of the present application, a system for managing an Internet of Things digital intelligent production platform is provided, comprising: A running data collection module is configured to collect running data of a plurality of production points in a target production line corresponding to a digital intelligent production platform within a preset window through an Internet of Things sensor to obtain a plurality of production point running data sequences. A supplementary data verification module is configured to perform abnormal observation on the plurality of production point running data sequences through a random network distillation mechanism, and perform supplementary data verification based on the abnormal observation results to obtain a plurality of production point abnormal observation results. A point identification module is configured to perform collaborative point identification and independent point identification on the plurality of production point abnormal observation results to obtain a plurality of collaborative abnormal production point sets and independent abnormal production point sets. A feature identification module is configured to perform production abnormal trend feature identification on the plurality of collaborative abnormal production point sets and independent abnormal production point sets respectively to obtain a plurality of collaborative abnormal trend features and independent abnormal trend features. The resource management module is used to manage the resources of the digital intelligent production platform's anomaly monitoring channels based on the multiple sets of collaborative and independent anomaly trend characteristics, and to obtain a target platform management solution.

[0018] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application collects operational data from multiple production points within a preset window on a target production line corresponding to a digital intelligent production platform using IoT sensors. This yields operational data sequences for multiple production points. Then, through a random network distillation mechanism, anomalies are observed in these data sequences. Supplementary data is used to verify the anomaly observations, resulting in multiple anomaly observation results for each production point. Furthermore, collaborative and independent anomaly identification is performed on these results, resulting in sets of collaborative and independent anomaly production points. Anomaly trend features are then identified for both sets, yielding sets of collaborative and independent anomaly trend features. Finally, based on these features, anomaly monitoring channel resource management is implemented for the digital intelligent production platform, resulting in a target platform management solution. This achieves the technical effect of targeted analysis of actual production conditions and improved reliability of production platform management. Attached Figure Description

[0019] Appendix Figure 1 This is a schematic diagram of the management method for a digital intelligent production platform combined with the Internet of Things provided in an embodiment of the present invention.

[0020] Appendix Figure 2 This is a schematic diagram of the structure of the IoT-integrated digital intelligent production platform management system provided in an embodiment of the present invention.

[0021] The labels shown in the attached diagram: The system includes a data collection module 11, a supplementary data verification module 12, a location identification module 13, a feature identification module 14, and a resource management module 15. Detailed Implementation

[0022] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims. It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices.

[0023] Example 1, as shown in the appendix Figure 1 As shown, this application provides a management method combining an IoT-based digital smart production platform, wherein the method includes: S1: Collect the operating data of multiple production points in the target production line corresponding to the digital intelligent production platform within a preset window through IoT sensors to obtain the operating data sequence of multiple production points; It should be noted that IoT sensors are terminal devices with data acquisition and wireless transmission capabilities, capable of capturing physical quantities or operating parameters such as temperature, pressure, rotational speed, and energy consumption in real time during the production process. The preset window is a time interval pre-defined by those skilled in the art, such as 5 minutes or 10 minutes, used to limit the time range of data acquisition. The production point operation data sequence is a dataset formed by arranging continuously collected data from a single production point within the preset window in chronological order.

[0024] First, based on the production process and control requirements of the target production line, corresponding IoT sensors are deployed at key production stages, such as raw material transportation, processing and assembly, and finished product inspection, with each sensor corresponding to a production point. Then, a preset window for data acquisition is set, and the IoT sensors collect operational data from each point at a preset frequency, such as once per second, including robotic arm speed, motor current, and processing accuracy at processing and assembly points. Finally, the data collected from each production point within each preset window is organized chronologically to form multiple production point operational data sequences.

[0025] Exemplarily, a certain automobile parts production line includes four production points of raw material conveying, stamping processing, welding assembly and quality detection, each of which is deployed with 3-5 Internet of Things sensors, specifically including a conveying belt speed sensor and a material weight sensor at the raw material conveying point, a stamping pressure sensor and a mold temperature sensor at the stamping processing point, and so on. The preset window is set to 10 minutes, the sliding step is 5 minutes, and the Internet of Things sensor collects data every second. In the first 10-minute window, the data collected by the conveying belt speed sensor at the raw material conveying point is arranged in time sequence as [2.5 m / s, 2.5 m / s,..., 2.6 m / s], the data sequence collected by the stamping pressure sensor at the stamping processing point is [15 MPa, 15.2 MPa,..., 14.8 MPa], and so on. Finally, a plurality of running data sequences corresponding to the four production points are obtained.

[0026] S2: performing anomaly observation on the plurality of production point running data sequences through a random network distillation mechanism, and performing supplementary data verification based on the anomaly observation result to obtain a plurality of production point anomaly observation results; Further, anomaly observation is performed on the plurality of production point running data sequences through a random network distillation mechanism, and supplementary data verification is performed based on the anomaly observation result to obtain a plurality of production point anomaly observation results. The step S2 of the embodiment of the present application further includes: calling a target network and a prediction network in the random distillation mechanism; performing state prediction on the plurality of production point running data sequences by using the target network and the prediction network respectively to obtain a plurality of target running state prediction results and a plurality of real-time prediction running state prediction results; mapping Euclidean distance calculation is performed on the plurality of target running state prediction results and the plurality of real-time prediction running state prediction results to determine a plurality of initial anomaly observation results; When the plurality of initial anomaly observation results are greater than or equal to a preset threshold, supplementary data verification is performed, the plurality of initial anomaly observation results are corrected according to the supplementary data verification result, and the plurality of initial anomaly observation results are added to the plurality of production point anomaly observation results; When the plurality of initial anomaly observation results are less than the preset threshold, the plurality of initial anomaly observation results are added to the plurality of production point anomaly observation results.

[0027] Further, the target network is a pre-trained network for representing the behavior mode of the production point under normal running conditions, and the output thereof is the prediction result of the standard behavior mode; The prediction network is a network that is continuously updated according to real-time production conditions, and is used for representing the behavior prediction of the production point at the current moment, and the output thereof is the prediction result of the real-time behavior mode.

[0028] Further, when the plurality of initial anomaly observations is greater than or equal to a preset threshold, supplementary data verification is performed, the plurality of initial anomaly observations is corrected according to the supplementary data verification result, and the plurality of initial anomaly observations is added to the plurality of production point anomaly observations. The embodiment of the application step S2 further comprises: When the plurality of initial anomaly observations is greater than or equal to a preset threshold, the initial anomaly observations of a plurality of production points in a preset neighborhood bandwidth are obtained, and a plurality of neighborhood initial anomaly observation sets are obtained. The plurality of neighborhood initial anomaly observation sets and the plurality of initial anomaly observations are subjected to supplementary data verification, and a supplementary data verification result is obtained. When the supplementary data verification result is consistent, the plurality of initial anomaly observations is added to the plurality of production point anomaly observations. When the supplementary data verification result is inconsistent, a plurality of secondary window supplementary data is obtained, the plurality of initial anomaly observations is corrected based on the plurality of secondary window supplementary data, and a plurality of corrected anomaly observations is added to the plurality of production point anomaly observations.

[0029] It should be noted that the target network is a model pre-trained by a large amount of normal production data, which solidifies the standard behavior pattern of the production point under normal operating conditions, and outputs a target operating state prediction result consistent with the normal law. The prediction network is a model that dynamically updates parameters according to real-time production data of the production line, that is, a model trained by some data of abnormal operating conditions, which is used to capture the actual behavior characteristics of the production point at the current time, and outputs a real-time prediction operating state prediction result. The preset threshold is an abnormality judgment threshold determined by a person skilled in the art based on historical data statistics, which is used to distinguish between normal fluctuations and abnormal deviations.

[0030] The target network and the prediction network in the random network distillation mechanism are called, and a plurality of production point operating data sequences are input into the two networks. The target network outputs a target operating state prediction result based on the normal behavior pattern, and the prediction network outputs a real-time prediction operating state prediction result based on the real-time state. The mapping Euclidean distance of the two groups of prediction results is calculated to obtain a plurality of initial anomaly observations representing the degree of deviation.

[0031] The initial anomaly observations are compared with the preset threshold respectively. If greater than or equal to the preset threshold, there is potential anomaly, which needs to be verified by supplementary data. Specifically, the neighborhood point data, historical same period data, etc. are called to correct the initial result, and the corrected result is added to the production point anomaly observation result. If less than the preset threshold, the deviation belongs to normal fluctuation, and the initial anomaly observation is directly added to the production point anomaly observation result.

[0032] Specifically, the preset neighborhood bandwidth is a range set around a production point, and the data points in the neighborhood bandwidth are production points adjacent to the point, which are used to filter out the near-neighbor production points related to the current point. By using the Euclidean distance calculation formula, the similarity distance of the multiple initial abnormal observation result sets and the multiple initial abnormal observation results is calculated. When the calculation result is less than the preset distance, the verification is passed, and the supplementary data verification result is consistent.

[0033] When the calculation result is greater than or equal to the preset distance, the verification is not passed, and when the supplementary data verification result is inconsistent, the supplementary data verification needs to be analyzed again. The multiple secondary window supplementary data are respectively the monitoring data of the production points in the next preset window, and then the multiple secondary window supplementary data and the multiple production point running data sequences corresponding to the multiple initial abnormal observation results are summarized, and then the target network and the prediction network are used for analysis again, the output results are compared, the multiple corrected abnormal observation results are obtained, and then the multiple corrected abnormal observation results are added to the multiple production point abnormal observation results.

[0034] S3: Cooperating point identification and independent point identification are performed on the multiple production point abnormal observation results to obtain a multiple cooperating abnormal production point set and an independent abnormal production point set; Further, cooperating point identification and independent point identification are performed on the multiple production point abnormal observation results to obtain a multiple cooperating abnormal production point set and an independent abnormal production point set. The step S3 of the embodiment of the application further comprises: The first production point abnormal observation result is randomly extracted from the multiple production point abnormal observation results without replacement; The production point abnormal observation result corresponding to the production point whose Euclidean distance to the first production point abnormal observation result is within the preset distance threshold is extracted from the multiple production point abnormal observation results without replacement, and the first production point corresponding to the first production point abnormal observation result is added to the first cooperating abnormal production point set; The second production point abnormal observation result is randomly extracted from the multiple production point abnormal observation results without replacement, and a second cooperating abnormal production point set is constructed. In this way, until the preset extraction times are met, a multiple cooperating abnormal production point set is obtained; The production points other than the multiple cooperating abnormal production point set in the multiple production points are added to the independent abnormal production point set.

[0035] It should be noted that the cooperating point identification is a process of identifying multiple production points with similar abnormal characteristics caused by the same or related abnormal factors and grouping them into a set. The independent point identification is a process of identifying production points with independent abnormal characteristics and no obvious correlation with other points.

[0036] From all production point anomaly observation results, a result is randomly extracted as the first production point anomaly observation result in a non-replacement manner, then other anomaly observation results are continuously extracted in a non-replacement manner, the Euclidean distances of these results and the first production point anomaly observation result are calculated, and the production points corresponding to the results within the preset distance threshold are collectively attributed to the first collaborative anomaly production point set together with the first production point.

[0037] According to the same logic, the second production point anomaly observation result is extracted again in a non-replacement manner, and the second collaborative anomaly production point set is constructed, and the operation is repeated until the preset extraction number is reached, and a plurality of collaborative anomaly production point sets are obtained. All points in all production points that are not included in any collaborative set are all attributed to the independent anomaly production point set. Through quantitative anomaly feature similarity, the accurate classification of anomaly points is realized, and the multi-point collaborative anomaly and single-point independent anomaly are distinguished, so as to provide a classification basis for subsequent targeted extraction of trend features and allocation of monitoring resources, and to avoid the technical effect that the management scheme is unreliable due to confusion of anomaly types.

[0038] For example, there are 8 production points with abnormal observation results in the automobile parts production line, numbered 1-8, the preset distance threshold is 0.5, and the preset extraction number is 2. The abnormal observation result of point 3 is randomly extracted in a non-replacement manner for the first time, the Euclidean distances of the remaining points and point 3 are calculated, it is found that the distance of point 2 is 0.3 and the distance of point 4 is 0.4. Within the threshold, the first collaborative anomaly production point set {2, 3, 4} is constructed. The abnormal observation result of point 7 is extracted in a non-replacement manner for the second time, the distances of points 1, 5, 6, and 8 to point 7 are calculated, and the distance of point 8 is 0.2 within the threshold, the second collaborative anomaly production point set {7, 8} is constructed. Finally, points 1, 5, and 6 that are not in the set are attributed to the independent anomaly production point set {1, 5, 6}.

[0039] S4: respectively performing production anomaly trend feature identification on the plurality of collaborative anomaly production point sets and the independent anomaly production point set to obtain a plurality of collaborative anomaly trend features and an independent anomaly trend feature set; Further, respectively performing production anomaly trend feature identification on the plurality of collaborative anomaly production point sets and the independent anomaly production point set to obtain a plurality of collaborative anomaly trend features and an independent anomaly trend feature set, the step S4 of the embodiment of the application further includes: obtaining a plurality of historical production point anomaly observation result sequence sets and a historical independent anomaly production point anomaly observation result sequence set of the plurality of collaborative anomaly production point sets and the independent anomaly production point set within a preset historical window; performing production anomaly trend feature identification on the plurality of historical production point anomaly observation result sequence sets to obtain a plurality of historical production anomaly trend feature sets; calculate the mean of the plurality of historical production anomaly trend feature sets to obtain a plurality of collaborative anomaly trend features; respectively, the historical independent anomaly production point anomaly observation result sequence set is subjected to production anomaly trend feature identification to obtain an independent anomaly trend feature set.

[0040] Further, the plurality of historical production point anomaly observation result sequence sets are subjected to production anomaly trend feature identification to obtain a plurality of historical production anomaly trend feature sets, and the embodiment S4 of the present application further comprises: calling an anomaly trend feature identifier to identify the plurality of historical production point anomaly observation result sequence sets to obtain a plurality of historical production anomaly trend feature sets; The anomaly trend feature identifier is obtained based on a feedforward neural network after supervised training.

[0041] Preferably, a feedforward neural network model is constructed to determine an input layer for receiving the historical production point anomaly observation result sequence set, a hidden layer for feature conversion and mining, and an output layer for outputting the identified anomaly trend features. Then a large number of historical production point anomaly observation result sequences with labeled anomaly trend features are selected as training samples, the samples are input into the feedforward neural network, the weights and bias parameters of the network are continuously optimized through the back propagation algorithm until the recognition accuracy of the model reaches the preset standard, the supervised training is completed and the anomaly trend feature identifier is formed. The plurality of historical production point anomaly observation result sequence sets are input into the trained anomaly trend feature identifier, the identifier automatically mines the trend features such as anomaly duration, change rate and diffusion speed from each sequence set through the learned feature extraction rules, and finally outputs a plurality of historical production anomaly trend feature sets. Further, the mean of the plurality of historical production anomaly trend feature sets is calculated to obtain a plurality of collaborative anomaly trend features.

[0042] Similarly, based on the same principle as obtaining the plurality of historical production anomaly trend feature sets, the historical independent anomaly production point anomaly observation result sequence set is subjected to production anomaly trend feature identification to obtain an independent anomaly trend feature set.

[0043] S5: performing anomaly monitoring channel resource management on the digital intelligent production platform according to the plurality of collaborative anomaly trend features and independent anomaly trend feature sets to obtain a target platform management scheme.

[0044] Further, according to the plurality of collaborative anomaly trend features and independent anomaly trend feature sets, the anomaly monitoring channel resource management is performed on the digital intelligent production platform to obtain a target platform management scheme, and the embodiment S5 of the present application further comprises: The collaborative monitoring weight and the independent monitoring weight are obtained by combining the collaborative abnormal production point set and the independent abnormal production point set for collaborative monitoring and independent monitoring weight analysis, wherein the collaborative monitoring weight and the independent monitoring weight add up to 1; The abnormality degree analysis is performed on the multiple collaborative abnormal trend characteristics and the independent abnormal trend characteristics set to obtain a multiple collaborative abnormality degree and independent abnormality degree set; The collaborative monitoring weight and the independent monitoring weight are split based on the multiple collaborative abnormality degree and independent abnormality degree set to obtain a multiple collaborative monitoring sub-weight and independent monitoring sub-weight set; The digital intelligent production platform is monitored according to the multiple collaborative monitoring sub-weight and independent monitoring sub-weight set to obtain a target platform management scheme.

[0045] Further, the collaborative monitoring weight and the independent monitoring weight are split based on the multiple collaborative abnormality degree and independent abnormality degree set to obtain a multiple collaborative monitoring sub-weight and independent monitoring sub-weight set, and the embodiment S5 further includes: The sum of the multiple collaborative abnormality degree and independent abnormality degree set is calculated respectively to obtain a multiple collaborative abnormality degree sum and independent abnormality degree sum; The ratio of the multiple collaborative abnormality degree to the multiple collaborative abnormality degree sum is taken as the multiple collaborative monitoring sub-weight; The ratio of each independent abnormality degree in the independent abnormality degree set to the independent abnormality degree sum is taken as the independent monitoring sub-weight to obtain an independent monitoring sub-weight set.

[0046] When necessary, the hardware resources for abnormal monitoring in the digital intelligent production platform, such as sensor data acquisition channels, data transmission channels, and software resources, such as data processing threads, algorithm calculation resources, and storage resources, need to be dynamically allocated according to the actual situation of each production point in the production line.

[0047] The collaborative monitoring weight and the independent monitoring weight are analyzed by combining the number of the collaborative abnormal production point set and the independent abnormal production point set, the importance of the covered production link, and the historical abnormal influence degree, and the weight distribution of the two is determined, such as the collaborative monitoring weight 0.6 and the independent monitoring weight 0.4. The multiple collaborative abnormal trend characteristics and the independent abnormal trend characteristics set are traversed, and the abnormality degree analysis is carried out by quantifying the abnormal duration, the change rate, and the influence range, and the multiple collaborative abnormality degree is obtained, such as the abnormality degree of each collaborative set is 0.3, 0.2, 0.1, and the independent abnormality degree set, such as the abnormality degree of each independent point is 0.15, 0.1, 0.05.

[0048] Calculate the sum of collaborative anomalies and the sum of independent anomalies. Use the ratio of each collaborative anomaly to the sum of collaborative anomalies as the collaborative monitoring sub-weight, and use the ratio of each independent anomaly to the sum of independent anomalies as the independent monitoring sub-weight, thus forming a set of independent monitoring sub-weights.

[0049] Based on the set of collaborative monitoring sub-weights and independent monitoring sub-weights, the platform's anomaly monitoring channel resources are allocated in a refined manner. More data processing threads and transmission channels are allocated to the collaborative set with a sub-weight of 0.5, ultimately forming a target platform management scheme that includes details of resource allocation, monitoring frequency, and anomaly handling priority.

[0050] By assigning weights and splitting sub-weights, the monitoring resources are accurately matched with the severity and scope of the anomalies, avoiding resource waste or under-allocation and improving the platform's response efficiency and management effectiveness in response to anomalies.

[0051] Example 2, based on the same inventive concept as the IoT-based digital intelligent production platform management method in the foregoing examples, as shown in the appendix. Figure 2 As shown, this application provides a management system integrating an IoT-based digital intelligent production platform. The system and method embodiments in this application are based on the same inventive concept. The system includes: The data collection module 11 is used to collect the operation data of multiple production points in the target production line corresponding to the digital intelligent production platform within a preset window through IoT sensors, and obtain the operation data sequence of multiple production points. The supplementary data verification module 12 is used to perform anomaly observation on the operation data sequence of the multiple production points through a random network distillation mechanism, and perform supplementary data verification based on the anomaly observation results to obtain anomaly observation results of multiple production points. The point identification module 13 is used to perform collaborative point identification and independent point identification on the abnormal observation results of the multiple production points, and obtain multiple sets of collaborative abnormal production points and sets of independent abnormal production points; Feature recognition module 14 is used to identify production anomaly trend features of the multiple sets of coordinated abnormal production points and the set of independent abnormal production points respectively, and obtain multiple sets of coordinated abnormal trend features and independent abnormal trend features. Resource management module 15 is used to manage the abnormal monitoring channel resources of the digital intelligent production platform based on the multiple sets of collaborative abnormal trend characteristics and independent abnormal trend characteristics, and to obtain a target platform management scheme.

[0052] Furthermore, the supplementary data verification module 12 is used to perform the following steps: Invoke the target network and prediction network in the random distillation mechanism; The target network and the prediction network are respectively used for state prediction on the plurality of production point operation data sequences, to obtain a plurality of target operation state prediction results and a plurality of real-time prediction operation state prediction results; The plurality of target operation state prediction results and the plurality of real-time prediction operation state prediction results are subjected to mapping Euclidean distance calculation, to determine a plurality of initial abnormal observation results; When the plurality of initial abnormal observation results are greater than or equal to a preset threshold, supplementary data verification is performed, and the plurality of initial abnormal observation results are corrected according to a supplementary data verification result, and added to the plurality of production point abnormal observation results; When the plurality of initial abnormal observation results are less than the preset threshold, the plurality of initial abnormal observation results are added to the plurality of production point abnormal observation results.

[0053] Further, the target network is a pre-trained network, used to represent the behavior mode of the production point under normal operation condition, and the output is a prediction result of the standard behavior mode; The prediction network is a network that is continuously updated according to real-time production conditions, used to represent the behavior prediction of the production point at the current time, and the output is a prediction result of the real-time behavior mode.

[0054] Further, the supplementary data verification module 12 is used to perform the following steps: When the plurality of initial abnormal observation results are greater than or equal to a preset threshold, a plurality of initial abnormal observation results of the production points in a preset neighborhood bandwidth are obtained, to obtain a plurality of neighborhood initial abnormal observation result sets; The plurality of neighborhood initial abnormal observation result sets and the plurality of initial abnormal observation results are subjected to supplementary data verification, to obtain a supplementary data verification result; When the supplementary data verification result is consistent, the plurality of initial abnormal observation results are added to the plurality of production point abnormal observation results; When the supplementary data verification result is inconsistent, a plurality of secondary window supplementary data are obtained, the plurality of initial abnormal observation results are corrected based on the plurality of secondary window supplementary data, and the plurality of corrected abnormal observation results are added to the plurality of production point abnormal observation results.

[0055] Further, the point identification module 13 is used to perform the following steps: A first production point abnormal observation result is randomly extracted from the plurality of production point abnormal observation results without replacement; The production point corresponding to the production point anomaly observation result in the plurality of production point anomaly observation results is not put back into the first production point anomaly observation result, and the first production point anomaly observation result is added to the first cooperative anomaly production point set. The second production point anomaly observation result is randomly extracted from the plurality of production point anomaly observation results without replacement, and a second cooperative anomaly production point set is constructed. In this way, the plurality of cooperative anomaly production point sets are obtained by repeating the above steps until the preset extraction times are met. The production points in the plurality of production points except the plurality of cooperative anomaly production point sets are added to the independent anomaly production point set.

[0056] Further, the feature recognition module 14 is configured to perform the following steps: Obtain the plurality of historical production point anomaly observation result sequence sets and the historical independent anomaly production point anomaly observation result sequence set of the plurality of cooperative anomaly production point sets and the independent anomaly production point set in the preset historical window; Perform production anomaly trend feature recognition on the plurality of historical production point anomaly observation result sequence sets to obtain a plurality of historical production anomaly trend feature sets; Calculate the mean of the plurality of historical production anomaly trend feature sets to obtain a plurality of cooperative anomaly trend features; Perform production anomaly trend feature recognition on the historical independent anomaly production point anomaly observation result sequence set to obtain an independent anomaly trend feature set.

[0057] Further, the feature recognition module 14 is configured to perform the following steps: Call the anomaly trend feature recognizer to recognize the plurality of historical production point anomaly observation result sequence sets to obtain a plurality of historical production anomaly trend feature sets; The anomaly trend feature recognizer is obtained based on a feedforward neural network after supervised training.

[0058] Further, the resource management module 15 is configured to perform the following steps: Perform cooperative monitoring and independent monitoring weight analysis on the plurality of cooperative anomaly production point sets and the independent anomaly production point set to obtain a cooperative monitoring weight and an independent monitoring weight, wherein the sum of the cooperative monitoring weight and the independent monitoring weight is 1; Perform anomaly degree analysis on the plurality of cooperative anomaly trend features and the independent anomaly trend feature set to obtain a plurality of cooperative anomaly degrees and independent anomaly degrees; split the cooperative monitoring weight and the independent monitoring weight based on the plurality of cooperative anomaly degrees and the independent anomaly degree set, to obtain a plurality of cooperative monitoring sub-weights and an independent monitoring sub-weight set; According to the plurality of cooperative monitoring sub-weights and the independent monitoring sub-weight set, the digital intelligent production platform is monitored for abnormal monitoring channel resources, to obtain a target platform management scheme.

[0059] Further, the resource management module 15 is configured to perform the following steps: The sum of the plurality of cooperative anomaly degrees and the independent anomaly degree set is calculated respectively, to obtain a plurality of cooperative anomaly degree sums and independent anomaly degree sums; The ratio of each cooperative anomaly degree to the plurality of cooperative anomaly degree sums is taken as a plurality of cooperative monitoring sub-weights; The ratio of each independent anomaly degree in the independent anomaly degree set to the independent anomaly degree sum is taken as an independent monitoring sub-weight, to obtain an independent monitoring sub-weight set.

[0060] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. Moreover, the above-mentioned embodiments of the present application are described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0061] The above-mentioned only for the preferred embodiments of the present application, and not to limit the present application, any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application, should be included in the protection scope of the present application.

[0062] The present application and the drawings are only exemplary descriptions of the present application, and are considered to cover any and all modifications, changes, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalent technology, the present application intends to include these modifications and changes.

Claims

1. A management method combining an IoT-based digital intelligent production platform, characterized in that: The method includes: By collecting operational data from multiple production points within a preset window in the target production line corresponding to the digital intelligent production platform through IoT sensors, a sequence of operational data from multiple production points is obtained. Anomalies were observed in the operational data sequences of the multiple production points through a random network distillation mechanism, and supplementary data was used to verify the anomalies based on the observation results, thereby obtaining anomaly observation results for multiple production points. The abnormal observation results of the multiple production points are used to identify both coordinated and independent production points, thereby obtaining a set of coordinated abnormal production points and a set of independent abnormal production points. The production anomaly trend features are identified for the multiple sets of coordinated abnormal production points and the sets of independent abnormal production points, respectively, to obtain multiple sets of coordinated abnormal trend features and sets of independent abnormal trend features. Based on the multiple sets of collaborative and independent abnormal trend features, the digital intelligent production platform is managed for abnormal monitoring channel resources to obtain a target platform management solution.

2. The management method for a digital intelligent production platform combined with the Internet of Things as described in claim 1, characterized in that, Anomaly observations were performed on the operational data sequences of the multiple production sites using a random network distillation mechanism. Supplementary data was then used to verify the anomaly observations, resulting in anomaly observation results for multiple production sites, including: Invoke the target network and prediction network in the random distillation mechanism; The target network and the prediction network are used to predict the state of the multiple production point operation data sequences, respectively, to obtain multiple target operation state prediction results and multiple real-time prediction operation state prediction results. The Euclidean distance is calculated by mapping the multiple target operation state prediction results and the multiple real-time predicted operation state prediction results to determine multiple initial abnormal observation results; When the multiple initial abnormal observation results are greater than or equal to a preset threshold, supplementary data verification is performed, and the multiple initial abnormal observation results are corrected based on the supplementary data verification results and added to the multiple production point abnormal observation results. When the multiple initial abnormal observation results are less than a preset threshold, the multiple initial abnormal observation results are added to the multiple production point abnormal observation results.

3. The management method for a digital intelligent production platform combined with the Internet of Things as described in claim 2, characterized in that, The target network is a pre-trained network used to represent the behavior patterns of production points under normal operating conditions, and its output is the prediction result of the standard behavior pattern. The prediction network is a network that is continuously updated based on real-time production conditions. It is used to represent the behavior prediction of production points at the current moment, and its output is the prediction result of real-time behavior patterns.

4. The management method for a digital intelligent production platform combined with the Internet of Things as described in claim 2, characterized in that, When the multiple initial abnormal observation results are greater than or equal to a preset threshold, supplementary data verification is performed. Based on the supplementary data verification results, the multiple initial abnormal observation results are corrected and added to the multiple production point abnormal observation results, including: When the multiple initial anomaly observation results are greater than or equal to a preset threshold, the initial anomaly observation results of multiple production points within a preset neighborhood bandwidth are obtained, and a set of multiple neighborhood initial anomaly observation results is obtained. Supplementary data verification is performed on the multiple sets of initial anomaly observation results in the neighborhood and the multiple initial anomaly observation results to obtain supplementary data verification results; When the supplementary data verification results are consistent, multiple initial abnormal observation results will be added to the multiple production point abnormal observation results. When the supplementary data verification results are inconsistent, multiple secondary window supplementary data are obtained, and the multiple initial abnormal observation results are corrected based on the multiple secondary window supplementary data. The multiple corrected abnormal observation results are then added to the multiple production point abnormal observation results.

5. The management method for a digital intelligent production platform combined with the Internet of Things as described in claim 1, characterized in that, The abnormal observation results of the multiple production points are used to perform coordinated point identification and independent point identification to obtain multiple sets of coordinated abnormal production points and sets of independent abnormal production points, including: The anomaly observation result of the first production point is randomly extracted without replacement from the anomaly observation results of the multiple production points; Without replacement, extract the production points corresponding to the abnormal production point observation results of the multiple production point abnormal observation results whose Euclidean distance to the first production point abnormal observation result is within a preset distance threshold, and add them to the first collaborative abnormal production point set in combination with the first production point abnormal observation result. Without replacement, a second abnormal observation result of a production point is randomly extracted from the multiple abnormal observation results of production points, and a second set of collaboratively abnormal production points is constructed. This process is repeated until a preset number of extractions is met, resulting in multiple sets of collaboratively abnormal production points. Add production points other than the aforementioned set of multiple collaborative abnormal production points to the set of independent abnormal production points.

6. The management method for a digital intelligent production platform combined with the Internet of Things as described in claim 1, characterized in that, Production anomaly trend features are identified for both the multiple sets of collaboratively anomalous production points and the sets of independently anomalous production points, resulting in multiple sets of collaboratively anomalous trend features and independently anomalous trend features, including: Obtain a set of historical anomaly observation results for multiple production points and a set of historical independent anomaly observation results for the multiple sets of coordinated and independent anomaly production points within a preset historical window; The anomaly trend feature is identified by performing production anomaly trend feature identification on the set of anomaly observation results of multiple historical production points to obtain multiple sets of historical production anomaly trend features. Calculate the mean of the multiple sets of historical production anomaly trend features to obtain multiple coordinated anomaly trend features; Production anomaly trend features were identified by analyzing the set of historical independent anomaly observation results for each production point, thus obtaining a set of independent anomaly trend features.

7. The management method for a digital intelligent production platform combined with the Internet of Things as described in claim 6, characterized in that, The anomaly trend feature identification is performed on the set of anomaly observation results from multiple historical production sites to obtain multiple sets of historical production anomaly trend features, including: The abnormal trend feature recognizer is invoked to identify the abnormal observation result sequence set of the multiple historical production points, thereby obtaining multiple sets of historical production abnormal trend features; Among them, the abnormal trend feature recognizer is obtained based on a feedforward neural network after supervised training.

8. The management method for a digital intelligent production platform combined with the Internet of Things as described in claim 1, characterized in that, Based on the multiple sets of collaborative and independent anomaly trend characteristics, the digital intelligent production platform is used for anomaly monitoring channel resource management to obtain a target platform management scheme, including: By combining multiple sets of collaborative abnormal production points and sets of independent abnormal production points, a collaborative monitoring and independent monitoring weight analysis is performed to obtain the collaborative monitoring weight and the independent monitoring weight. The sum of the collaborative monitoring weight and the independent monitoring weight is 1. Anomaly degree analysis is performed by traversing the multiple sets of coordinated anomaly trend features and independent anomaly trend features to obtain multiple sets of coordinated anomaly degree and independent anomaly degree. Based on the multiple sets of collaborative anomalies and independent anomalies, the collaborative monitoring weights and independent monitoring weights are split to obtain multiple sets of collaborative monitoring sub-weights and independent monitoring sub-weights. Based on the set of multiple collaborative monitoring sub-weights and independent monitoring sub-weights, the digital intelligent production platform is managed for anomaly monitoring channel resources to obtain a target platform management scheme.

9. The management method for a digital intelligent production platform combined with the Internet of Things as described in claim 8, characterized in that, Based on the multiple sets of collaborative anomalies and independent anomalies, the collaborative monitoring weights and independent monitoring weights are split to obtain multiple sets of collaborative monitoring sub-weights and independent monitoring sub-weights, including: Calculate the sum of the multiple sets of cooperative anomalies and independent anomalies respectively to obtain the sum of multiple cooperative anomalies and the sum of independent anomalies; The ratio of each of the multiple collaborative anomalies to the sum of the multiple collaborative anomalies is used as the weight of the multiple collaborative monitoring sub-weights; The ratio of each independent anomaly in the set of independent anomalies to the sum of independent anomalies is used as the independent monitoring sub-weight to obtain the set of independent monitoring sub-weights.

10. A digital intelligent production platform management system integrated with the Internet of Things (IoT) is characterized by: The system is used to implement the management method of the IoT-integrated digital smart production platform as described in any one of claims 1-9, the system comprising: The data collection module is used to collect operational data from multiple production points within a preset window in the target production line corresponding to the digital smart production platform through IoT sensors, and obtain operational data sequences from multiple production points. The supplementary data verification module is used to perform anomaly observation on the operation data sequence of the multiple production points through a random network distillation mechanism, and to perform supplementary data verification based on the anomaly observation results to obtain anomaly observation results for multiple production points. The point identification module is used to perform collaborative point identification and independent point identification on the abnormal observation results of the multiple production points, and obtain multiple sets of collaborative abnormal production points and sets of independent abnormal production points; The feature recognition module is used to identify production anomaly trend features of the multiple sets of coordinated abnormal production points and the sets of independent abnormal production points, respectively, to obtain multiple sets of coordinated abnormal trend features and sets of independent abnormal trend features; The resource management module is used to manage the resources of the digital intelligent production platform's anomaly monitoring channels based on the multiple sets of collaborative and independent anomaly trend characteristics, and to obtain a target platform management solution.

Images