Multi-target variable measurement system based on knowledge enhancement

Abstract

The invention relates to the technical field of measurement, in particular to a multi-target variable measurement system based on knowledge enhancement. The system comprises a multi-target tracking measurement module, a measurement knowledge graph construction module, an entity semantic association mapping module and a knowledge enhancement measurement correction module. A production tracking measurement target can be obtained, and an image acquisition device, an industrial sensor and a data processing center are connected to generate a multi-target variable tracking measurement network. Performing multi-target measurement on the production target area by using the multi-target variable tracking measurement network to obtain production multi-target variable data; obtaining production and measurement target domain knowledge, and performing target knowledge entity extraction and measurement target knowledge graph construction to generate a production and measurement target knowledge graph; and performing entity semantic association mapping and knowledge enhancement measurement correction on corresponding target variables in the production multi-target variable data to obtain a multi-target variable measurement correction result. According to the invention, efficient and accurate measurement of a plurality of target variables can be realized.

Description

Technical Field

[0001] The invention relates to the field of measurement technology, and in particular to a multi-objective variable measurement system based on knowledge enhancement. Background Art

[0002] In many practical application scenarios, it is often necessary to measure multiple target variables at the same time to obtain comprehensive and accurate information. For example, in the industrial production process, it is necessary to monitor multiple variables such as temperature, pressure, flow, etc. at the same time to ensure production stability and product quality; in environmental monitoring, it is necessary to measure multiple variables such as air quality indicators and water quality parameters at the same time to evaluate environmental conditions.

[0003] However, traditional multi-objective variable measurement methods mainly rely on various sensors and simple data processing algorithms. On the one hand, the sensors themselves have measurement errors and are easily disturbed in complex environments, resulting in reduced accuracy and reliability of the measured data. On the other hand, traditional data processing algorithms often only focus on the data itself, lack effective use of domain knowledge, and are difficult to deeply analyze and correct the measured data, thus failing to fully explore the knowledge associations between multi-objective variables. Summary of the invention

[0004] Based on this, it is necessary for the present invention to provide a multi-objective variable measurement system based on knowledge enhancement to solve at least one of the above technical problems.

[0005] To achieve the above purpose, a multi-objective variable measurement system based on knowledge enhancement includes the following modules: The multi-target tracking and measurement module is used to obtain the production tracking and measurement targets, and based on the production tracking and measurement targets, dynamically configure the data acquisition frequency and build the corresponding data acquisition transmission link to connect the image acquisition equipment, industrial sensors and data processing center to generate a multi-target variable tracking and measurement network; use the multi-target variable tracking and measurement network to perform multi-target measurement on the production target area to obtain production multi-target variable data; The measurement knowledge graph construction module is used to obtain the corresponding production measurement target domain knowledge through the production tracking measurement target, and extract the target knowledge entity of the production measurement target domain knowledge to obtain the production measurement target knowledge entity, which includes production tracking variables, production tracking equipment and production scene phenomena; the measurement target knowledge graph is constructed according to the production measurement target knowledge entity to generate the production measurement target knowledge graph; An entity semantic association mapping module is used to perform entity semantic association mapping on the corresponding target variables in the production multi-target variable data based on the corresponding knowledge entities in the production measurement target knowledge graph, so as to generate multi-target variable measurement data corresponding to the knowledge entity semantic annotations; The knowledge-enhanced measurement correction module is used to obtain the corresponding production measurement knowledge constraints and production measurement reasoning mechanisms through the production measurement target knowledge graph, and based on the production measurement knowledge constraints and production measurement reasoning mechanisms, the corresponding knowledge entities in the production measurement target knowledge graph are used as additional supervision information in combination with the multi-objective variable tracking measurement network to perform knowledge-enhanced measurement correction on the multi-objective variable measurement data corresponding to the knowledge entity semantic annotations, so as to obtain the multi-objective variable measurement correction results.

[0006] Furthermore, the multi-target tracking and measurement module includes the following functions: Obtain production tracking measurement targets, including the physical attributes of each target, the role of the production process, and the frequency and amplitude of its production status changes; Perform target cluster analysis on production tracking measurement targets, so as to classify the key targets with frequent state changes in the production process into one category according to the corresponding production characteristics, and classify the relatively stable targets with stable state changes into another category, so as to obtain a production tracking target classification system; Based on the production tracking target classification system, the data acquisition frequency corresponding to the image acquisition equipment and industrial sensors is dynamically configured to generate a data acquisition frequency dynamic configuration strategy; Based on the data acquisition frequency dynamic configuration strategy, the image acquisition equipment and industrial sensors are adapted and optimized for equipment parameters. For the image acquisition equipment, the corresponding resolution, frame rate and exposure time are adjusted according to the acquisition frequency requirements of different targets. For the industrial sensors, their sensitivity and response time are optimized according to the physical properties of the targets, and the adapted and optimized image acquisition equipment and industrial sensors are obtained. By combining wired transmission and wireless transmission, the acquisition and transmission links between the adapted and optimized image acquisition devices and industrial sensors and the data processing center are constructed, so that the corresponding devices that are close to the data processing center and have large data volumes and high real-time requirements are transmitted using high-speed wired Ethernet, and the devices that are far from the data processing center and distributed in complex environments or mobile are transmitted using low-power, self-organizing wireless transmission technology to construct the corresponding data acquisition and transmission links; the data acquisition and transmission links are used to connect the adapted and optimized image acquisition devices and industrial sensors to the data processing center to generate a multi-target variable tracking and measurement network; The multi-objective variable tracking measurement network is used to perform multi-objective measurement on the production target area to obtain production multi-objective variable data.

[0007] Furthermore, the industrial sensor includes a temperature sensor, a pressure sensor and a vibration sensor.

[0008] Furthermore, the data collection frequency dynamic configuration strategy is specifically for target categories that are critical and frequently changing in the production process, to dynamically adjust the data collection frequency corresponding to 1 to 2 time intervals according to the information entropy corresponding to their production status changes; and for target categories that are relatively stable and have smooth status changes, the data collection frequency corresponding to 3 to 5 time intervals is dynamically adjusted based on the time period or production event trigger, where the time interval is specifically 2s.

[0009] Furthermore, the measurement knowledge graph construction module includes the following functions: Obtain the corresponding production measurement target domain knowledge through production tracking measurement targets, including production process parameters, equipment operation specifications and production environment experience knowledge; Extract target knowledge entities from the production measurement target domain knowledge to obtain production measurement target knowledge entities, including production tracking variables, production tracking equipment, and production scenario phenomena; Perform entity attribute mining and analysis between various production measurement target knowledge entities to generate entity attribute relationships between various production measurement knowledge entities, including causal relationships, association relationships, and constraint relationships; Based on the entity attribute relationship between each production measurement knowledge entity, a measurement target knowledge graph is constructed for each production measurement target knowledge entity to generate a production measurement target knowledge graph.

[0010] Furthermore, the entity semantic association mapping module includes the following functions: Perform entity segmentation processing on the corresponding knowledge entities in the production measurement target knowledge graph to obtain production measurement knowledge entity segmentation; Based on the production measurement target knowledge graph, the segmentation index of each production measurement knowledge entity segmentation is measured to obtain the knowledge entity degree and knowledge entity centrality corresponding to each knowledge entity segmentation in the knowledge graph; Based on the knowledge entity degree and knowledge entity centrality of each knowledge entity segmentation in the knowledge graph, the corresponding knowledge entity in the production measurement target knowledge graph is subjected to description association mining analysis to generate description association paths between each production measurement knowledge entity; Based on the description association path between each production measurement knowledge entity, the corresponding knowledge entity in the production measurement target knowledge graph is described and semantically embedded to obtain the production measurement knowledge entity description embedding vector, which includes the attributes, text features and semantic relationships between each production measurement knowledge entity and other knowledge entities. Based on the production measurement knowledge entity description embedding vector, entity semantic association mapping is performed on the corresponding target variables in the production multi-target variable data to generate multi-target variable measurement data corresponding to the knowledge entity semantic annotations.

[0011] Furthermore, the entity semantic association mapping of the corresponding target variables in the production multi-target variable data based on the production measurement knowledge entity description embedding vector includes: Use the corresponding text encoder in the CLIP model to perform text encoding embedding on the corresponding target variables in the production multi-target variable data to generate the target variable text prompt embedding vector; Based on the text features corresponding to the embedding vector of the production measurement knowledge entity description, the entity text semantic association calculation formula is used to measure the entity semantic association of the target variable text prompt corresponding to the target variable text prompt embedding vector to obtain the text semantic association between the target variable and the knowledge graph entity; Based on the textual semantic association between the target variable and the knowledge graph entity and combined with the corresponding knowledge entities in the production measurement target knowledge graph, entity semantic association mapping is performed on the corresponding target variables in the production multi-target variable data to generate multi-target variable measurement data with corresponding knowledge entity semantic annotations.

[0012] Furthermore, the entity text semantic association calculation formula is specifically: ; In the formula, The target variable With knowledge graph entities The semantic correlation between the texts The total number of text feature correspondences in the embedding vector of the knowledge entity description is measured for production, is the total number of target variable text prompts in the target variable text prompt embedding vector, To produce measurement knowledge entity descriptions, the first text features, The first target variable text prompt, For the The text features and The vector inner product between the target variable text prompts, For the The text features and The correlation weight coefficient between the target variable text prompts, is an exponential function, For the The text features and The squared Euclidean distance between the target variable text prompts, For the The text features and The distance decay factor between target variable text prompts, is the correction coefficient of text semantic relevance.

[0013] Furthermore, the knowledge-enhanced measurement correction module includes the following functions: Obtain the corresponding production measurement knowledge constraints and production measurement reasoning mechanism through the production measurement target knowledge graph; A measurement model based on knowledge enhancement is constructed using convolutional neural networks. The corresponding convolutional layer structure and connection hyperparameters in the measurement model based on knowledge enhancement are optimized by introducing production measurement knowledge constraints and production measurement reasoning mechanisms, so that the measurement model based on knowledge enhancement can capture the intrinsic measurement relationship between multiple objective variables and generate a measurement optimization model based on knowledge enhancement. The corresponding knowledge entities in the production measurement target knowledge graph are used as additional supervision information, and the multi-objective variable measurement data corresponding to the knowledge entity semantic annotations are input into the measurement optimization model based on knowledge enhancement and combined with the multi-objective variable tracking measurement network to perform knowledge enhanced measurement correction to obtain the multi-objective variable measurement correction results.

[0014] Furthermore, the process of knowledge-enhanced measurement correction is specifically to use the measurement optimization model based on knowledge enhancement to perform comparative measurements between the corresponding knowledge entities in the production measurement target knowledge graph and the multi-objective variables corresponding to the knowledge entity semantic annotations. When the multi-objective variables corresponding to the knowledge entity semantic annotations conflict with the corresponding knowledge entities in the production measurement target knowledge graph, the multi-objective variable tracking measurement network is used to re-measure the target variables corresponding to the conflict to correct them, so as to obtain the multi-objective variable measurement correction results.

[0015] Beneficial effects of the present invention: The multi-target variable measurement system based on knowledge enhancement proposed in the present invention is composed of a multi-target tracking measurement module, a measurement knowledge graph construction module, an entity semantic association mapping module and a knowledge enhancement measurement correction module. Compared with the prior art, the beneficial effect of the present application is that by acquiring the production tracking measurement target, it is possible to construct a highly flexible and intelligent data acquisition network based on the specific requirements of the production target by dynamically configuring the data acquisition frequency. This process mainly forms a multi-target variable tracking measurement network by integrating image acquisition equipment, industrial sensors and data processing centers. The tracking of each target can not only collect data from multiple sensors in real time, but also combine the visual information provided by the image acquisition equipment to achieve comprehensive monitoring of various targets in the production environment. By accurately adjusting the acquisition frequency, the system can adjust the frequency and method of data acquisition in real time according to changes in the production process, thereby improving the accuracy and reliability of measurement data acquisition, so that various types of information on the production site can be fully captured, and through the collaborative work between different sensors and devices, the limitations of a single sensor or device are reduced, the comprehensiveness and diversity of the data are ensured, and the dynamic changes in the production process can be quickly and accurately monitored, thereby providing sufficient basis for subsequent data analysis and decision-making. Secondly, by extracting the corresponding production measurement target domain knowledge and generating specific knowledge entities through the extraction of target knowledge entities to analyze the various entities involved in the production environment, a detailed production measurement target knowledge graph can be constructed. This graph includes information such as production tracking variables, production tracking equipment and its operating status, and various phenomena in the production scene. The extraction of production measurement target domain knowledge can help the system better understand the relationship, mutual influence and mutual restriction between different targets. The systematic and structured organization of this domain knowledge not only provides a rich semantic background for subsequent data analysis and knowledge reasoning, but also provides a theoretical basis for the complex situations that arise in the production site, which can achieve more efficient data association and utilization, thereby improving the overall intelligent measurement level of the production process.Then, through the existing knowledge entities in the production measurement target knowledge graph, the multi-target variable data in the production process can be semantically associated and mapped. This process generates multi-target variable measurement data with entity semantic annotations by associating the data with the entities in the knowledge graph. Through this semantic annotation method, the originally purely numerical measurement data is endowed with rich semantic information, so that the data is no longer an isolated number, but can correspond to actual production scenarios, equipment, operations and other factors. The role of this entity semantic association mapping is that it can help the system understand the correlation between different measurement variables and how they interact with different factors, equipment, operating strategies, etc. in the production process. Through this mapping, changes in the production process can be more easily compared with entities in the knowledge graph to enhance the perception and response capabilities of the intelligent system, so as to more fully explore the knowledge association relationship between multi-target variables. Finally, by combining the knowledge entities in the production measurement target knowledge graph, the multi-objective variable measurement data is corrected for knowledge enhancement. The core of this process is to use the production measurement constraints and reasoning mechanisms as additional supervisory information to guide the correction of the measurement data. By introducing these knowledge entities and constraints, the measurement data that originally had certain errors will be corrected through the rules and reasoning mechanisms in the knowledge graph, thereby improving the accuracy of the measurement results. This can continuously optimize the quality of measurement data in large-scale production processes and ensure that the data in the production process is more credible. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments thereof made with reference to the following drawings: Figure 1 It is a module schematic diagram of the multi-objective variable measurement system based on knowledge enhancement of the present invention; Figure 2 for Figure 1 Functional flow diagram of the multi-target tracking and measurement module; Figure 3 for Figure 1 Schematic diagram of the functional flow of the measurement knowledge graph construction module. DETAILED DESCRIPTION

[0017] The technical system of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0018] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor systems and / or microcontroller systems.

[0019] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0020] To achieve this, please refer to Figures 1 to 3 The present invention provides a multi-objective variable measurement system based on knowledge enhancement, the system comprising the following modules: The multi-target tracking and measurement module is used to obtain the production tracking and measurement targets, and based on the production tracking and measurement targets, dynamically configure the data acquisition frequency and build the corresponding data acquisition transmission link to connect the image acquisition equipment, industrial sensors and data processing center to generate a multi-target variable tracking and measurement network; use the multi-target variable tracking and measurement network to perform multi-target measurement on the production target area to obtain production multi-target variable data; The measurement knowledge graph construction module is used to obtain the corresponding production measurement target domain knowledge through the production tracking measurement target, and extract the target knowledge entity of the production measurement target domain knowledge to obtain the production measurement target knowledge entity, which includes production tracking variables, production tracking equipment and production scene phenomena; the measurement target knowledge graph is constructed according to the production measurement target knowledge entity to generate the production measurement target knowledge graph; An entity semantic association mapping module is used to perform entity semantic association mapping on the corresponding target variables in the production multi-target variable data based on the corresponding knowledge entities in the production measurement target knowledge graph, so as to generate multi-target variable measurement data corresponding to the knowledge entity semantic annotations; The knowledge-enhanced measurement correction module is used to obtain the corresponding production measurement knowledge constraints and production measurement reasoning mechanisms through the production measurement target knowledge graph, and based on the production measurement knowledge constraints and production measurement reasoning mechanisms, the corresponding knowledge entities in the production measurement target knowledge graph are used as additional supervision information in combination with the multi-objective variable tracking measurement network to perform knowledge-enhanced measurement correction on the multi-objective variable measurement data corresponding to the knowledge entity semantic annotations, so as to obtain the multi-objective variable measurement correction results.

[0021] In the embodiment of the present invention, please refer to Figure 1 FIG. 1 is a schematic diagram of modules of a multi-objective variable measurement system based on knowledge enhancement of the present invention. In this example, the multi-objective variable measurement system based on knowledge enhancement includes the following modules: S1: Multi-target tracking and measurement module, used to obtain production tracking and measurement targets, and based on the production tracking and measurement targets, dynamically configure the data acquisition frequency and build the corresponding data acquisition transmission link to connect the image acquisition equipment, industrial sensors and data processing center to generate a multi-target variable tracking and measurement network; use the multi-target variable tracking and measurement network to perform multi-target measurement on the production target area to obtain production multi-target variable data; In an embodiment of the present invention, by setting the production tracking measurement target, the production target area and target variables to be tracked are clarified. Then, according to the characteristics of the production task, the data acquisition frequency is dynamically configured to ensure that the acquisition equipment can cover the changing needs of the production tracking measurement target. The acquisition equipment includes image acquisition equipment, industrial sensors, etc. These devices are connected through a dedicated industrial communication protocol and network to build a stable acquisition transmission link with a suitable data transmission rate. This link effectively connects the acquisition equipment with the data processing center to ensure efficient data transmission. On this basis, by establishing a multi-target variable tracking and measurement network, multiple variables in the production target area are monitored simultaneously. These target variables include target images, temperature, humidity, pressure, speed, position, etc. The data collected by sensors and cameras will be transmitted to the data processing center in real time for data analysis and processing. After data processing, the data of various target variables in the production process will be aggregated in real time to finally form complete production multi-target variable data.

[0022] S2: measurement knowledge graph construction module, used to obtain the corresponding production measurement target domain knowledge through production tracking measurement targets, and extract target knowledge entities from the production measurement target domain knowledge to obtain production measurement target knowledge entities, including production tracking variables, production tracking equipment, and production scenario phenomena; construct measurement target knowledge graphs based on production measurement target knowledge entities to generate production measurement target knowledge graphs; In an embodiment of the present invention, by analyzing the data of the production tracking measurement target, the domain knowledge of the production measurement target in the relevant field is extracted. The domain knowledge includes various important factors in the production process, such as the type and function of the equipment and the changing factors in the production environment. In order to further organize and extract the relevant knowledge in the field of the production measurement target, the natural language processing (NLP) technology is used to process these data to extract the target knowledge entity. This process involves identifying and extracting key entities, such as production tracking variables (such as temperature, pressure, speed, etc.), production tracking equipment (such as temperature sensors, cameras, etc.) and production scene phenomena (such as production line status, equipment failure, etc.). The extraction of these knowledge entities is achieved by constructing a named entity recognition model (NER), which can automatically identify the core content of the relevant field. Subsequently, the extracted knowledge entities are used to organize into a production measurement target knowledge graph by constructing a knowledge graph. The knowledge graph is presented in a graph structure, each node represents an entity, and each edge represents the relationship between entities. In this way, the domain knowledge related to the production measurement target can be effectively integrated, and finally a production measurement target knowledge graph is generated.

[0023] S3: Entity semantic association mapping module, used to perform entity semantic association mapping on the corresponding target variables in the production multi-target variable data based on the corresponding knowledge entities in the production measurement target knowledge graph, so as to generate multi-target variable measurement data corresponding to the knowledge entity semantic annotations; In an embodiment of the present invention, based on the generated production measurement target knowledge graph, entity semantic association mapping technology is used to semantically associate knowledge entities in the graph with target variables in production multi-target variable data. Specifically, the production multi-target variable data is matched with entities in the corresponding production measurement target knowledge graph, and the value of the corresponding variable is annotated with the semantic label of the knowledge entity related thereto. For example, if a variable collected by the data is a temperature value, the temperature value will be connected to related entities such as temperature sensors and heat conduction processes of production lines, and target variable measurement data with semantic annotations are generated through association mapping. This process mainly relies on a graph-based knowledge reasoning model for association. Through multiple rounds of reasoning and relationship graph analysis, it is ensured that each content of the production data can correspond to the relevant entity in the knowledge graph, and then give it accurate semantic annotations. This operation effectively combines production data with domain knowledge, improves the interpretability and usability of the data, facilitates subsequent intelligent decision support and further reasoning analysis, and finally generates multi-target variable measurement data corresponding to the knowledge entity semantic annotations.

[0024] S4: Knowledge-enhanced measurement correction module, which is used to obtain the corresponding production measurement knowledge constraints and production measurement reasoning mechanisms through the production measurement target knowledge graph, and based on the production measurement knowledge constraints and production measurement reasoning mechanisms, use the corresponding knowledge entities in the production measurement target knowledge graph as additional supervision information and combine the multi-objective variable tracking measurement network to perform knowledge-enhanced measurement correction on the multi-objective variable measurement data corresponding to the knowledge entity semantic annotations to obtain the multi-objective variable measurement correction results.

[0025] In an embodiment of the present invention, after acquiring multi-objective variable measurement data and completing entity semantic annotation, production measurement knowledge constraints and production measurement reasoning mechanisms are further introduced to perform knowledge enhancement and correction. First, through further analysis of the production measurement target knowledge graph, constraints and reasoning rules related to the production measurement targets are obtained. These constraints may involve information such as the working range of the equipment, the normal fluctuation range of the variables, and the validity of the measurement data. Next, combined with these constraints, the multi-objective variable measurement data with semantic annotations is corrected through knowledge enhancement through reasoning mechanisms (such as rule-based reasoning or neural network reasoning). This process uses the knowledge entities in the production measurement target knowledge graph as additional supervisory information and uses the reasoning mechanism to correct and optimize the data. For example, if the measured temperature value exceeds the safe operating temperature range of the equipment, the temperature value can be adjusted to a reasonable range by combining the knowledge in the graph through the reasoning mechanism and establishing the corresponding model. In this way, the accuracy and credibility of the measurement data can be greatly improved, providing more accurate data support for quality control and equipment maintenance in the production process. When the model detects a conflict between the measurement data and the knowledge entity (for example, the measurement result exceeds the allowable range or does not conform to the historical law), the multi-objective variable tracking measurement network is enabled for further correction. The tracking network will re-measure the conflicting target variables and dynamically adjust the measurement results according to the knowledge graph reasoning mechanism and real-time data to output the corrected multi-objective variable measurement results. The whole process emphasizes the deep integration and intelligent correction of the model to the production environment and measurement data to ensure the accuracy and consistency of the measurement results, and finally obtain the multi-objective variable measurement correction results.

[0026] Furthermore, the multi-target tracking and measurement module includes the following functions: Obtain production tracking measurement targets, including the physical attributes of each target, the role of the production process, and the frequency and amplitude of its production status changes; Perform target cluster analysis on production tracking measurement targets, so as to classify the key targets with frequent state changes in the production process into one category according to the corresponding production characteristics, and classify the relatively stable targets with stable state changes into another category, so as to obtain a production tracking target classification system; Based on the production tracking target classification system, the data acquisition frequency corresponding to the image acquisition equipment and industrial sensors is dynamically configured to generate a data acquisition frequency dynamic configuration strategy; Based on the data acquisition frequency dynamic configuration strategy, the image acquisition equipment and industrial sensors are adapted and optimized for equipment parameters. For the image acquisition equipment, the corresponding resolution, frame rate and exposure time are adjusted according to the acquisition frequency requirements of different targets. For the industrial sensors, their sensitivity and response time are optimized according to the physical properties of the targets, and the adapted and optimized image acquisition equipment and industrial sensors are obtained. By combining wired transmission and wireless transmission, the acquisition and transmission links between the adapted and optimized image acquisition devices and industrial sensors and the data processing center are constructed, so that the corresponding devices that are close to the data processing center and have large data volumes and high real-time requirements are transmitted using high-speed wired Ethernet, and the devices that are far from the data processing center and distributed in complex environments or mobile are transmitted using low-power, self-organizing wireless transmission technology to construct the corresponding data acquisition and transmission links; the data acquisition and transmission links are used to connect the adapted and optimized image acquisition devices and industrial sensors to the data processing center to generate a multi-target variable tracking and measurement network; The multi-objective variable tracking measurement network is used to perform multi-objective measurement on the production target area to obtain production multi-objective variable data.

[0027] As an embodiment of the present invention, refer to Figure 2 As shown, Figure 1 The functional flow diagram of the multi-target tracking and measurement module in this embodiment includes the following functions: S11: Obtaining production tracking measurement targets, including the physical attributes corresponding to each target, the role of the production process, and the frequency and amplitude corresponding to the change of its production status; In an embodiment of the present invention, each production target is identified and defined through data collected in the production process. The physical properties involved in the target, such as temperature, pressure, vibration, etc., need to be clarified. Each target is classified according to its role and influence in the production process. For example, some targets are critical and directly affect the quality, efficiency or safety of the production process, while other targets are auxiliary. For each target, the corresponding production state change frequency and amplitude are also measured and recorded. This can be achieved by analyzing historical data and combining real-time information collected by sensors and monitoring systems to establish a state change model for each target. For targets that change frequently and with large amplitudes, the cycle and characteristics of their state changes are recorded; and for targets that change less frequently, their stable change patterns are recorded. The data collection method uses temperature sensors, pressure sensors, vibration sensors and other equipment to monitor the targets in real time for a long time. Through sensor feedback information, each production target is accurately measured, and finally a production tracking measurement target is obtained, which includes the physical properties corresponding to each target, the role of the production process, and the frequency and amplitude corresponding to its production state changes.

[0028] S12: Performing target cluster analysis on the production tracking measurement targets, so as to classify the key targets with frequent state changes in the production process into one category according to the corresponding production characteristics, and classify the relatively stable targets with stable state changes into another category, so as to obtain a production tracking target classification system; In an embodiment of the present invention, production tracking targets are classified through cluster analysis. First, the collected information of each production target is processed through a machine learning algorithm (such as K-means clustering or hierarchical clustering). The input of this analysis process is the physical properties, state change frequency and state change amplitude of each target. By analyzing the state change frequency and amplitude of the target, the production targets can be divided into two categories: one is critical targets with frequent state changes, and the other is relatively stable targets with smooth state changes. For targets with frequent state changes, they will be adjusted or optimized many times during the production process, so they require higher frequency monitoring; for stable targets, the monitoring frequency is relatively low. The purpose of cluster analysis is to give a clear target classification based on the production characteristics of each target, so as to facilitate subsequent data collection and processing strategies.

[0029] S13: dynamically configuring the data acquisition frequencies corresponding to the image acquisition devices and industrial sensors based on the production tracking target classification system to generate a data acquisition frequency dynamic configuration strategy; In an embodiment of the present invention, the data collection frequency is adjusted according to the state change characteristics of the production target. For those target categories with frequent state changes, their data collection frequency is dynamically adjusted. The data collection frequency corresponding to the image acquisition device and the industrial sensor (including temperature sensor, pressure sensor and vibration sensor) is evaluated by information entropy. The information entropy reflects the complexity of the target state change. The target with higher information entropy will adopt a higher collection frequency (for example, 1 to 2 time intervals (for example, 2s) for data collection) to ensure that the target state change is captured quickly; while for the target category with smooth and stable state changes, the collection frequency can be relatively low, usually 3 to 5 time intervals for collection. For example, the collection frequency of these targets is once every 2 seconds, or dynamically adjusted when a production event is triggered. This dynamic configuration strategy of the collection frequency can effectively balance the real-time performance and storage pressure of data collection, and improve the efficiency of data analysis, and finally generate a dynamic configuration strategy for data collection frequency.

[0030] S14: Adapt and optimize the device parameters of the image acquisition device and the industrial sensor based on the data acquisition frequency dynamic configuration strategy. For the image acquisition device, the corresponding resolution, frame rate and exposure time are adjusted according to the acquisition frequency requirements of different targets. For the industrial sensor, its sensitivity and response time are optimized according to the physical properties of the target, so as to obtain the image acquisition device and industrial sensor after adaptation and optimization. In an embodiment of the present invention, the device parameters of the image acquisition device and the industrial sensor are adapted and optimized according to the previously configured data acquisition frequency. For image acquisition devices, such as cameras or video surveillance devices, it is necessary to adjust the resolution, frame rate and exposure time according to the target change frequency. For example, for target categories that change frequently, the frame rate of the image acquisition device can be increased and the resolution can be appropriately improved to ensure the clarity of the image and the capture of details; the exposure time needs to be optimized according to the lighting conditions of the production environment. For industrial sensors (such as temperature, pressure, and vibration sensors), their sensitivity and response time also need to be adjusted according to the physical properties of the target. If the target category status changes frequently, it is necessary to improve the response speed and sensitivity of the sensor to ensure that subtle changes are captured; and for stable targets, the sensor sensitivity is appropriately reduced and the response time can be relatively extended to reduce unnecessary data fluctuations. The adapted device can effectively improve the acquisition accuracy and data analysis efficiency, and finally obtain an adapted and optimized image acquisition device and industrial sensor.

[0031] S15: By combining wired transmission and wireless transmission, a data acquisition and transmission link is constructed between the adapted and optimized image acquisition device and the industrial sensor and the data processing center, so that the corresponding devices close to the data processing center, with large data volume and high real-time requirements are transmitted using high-speed wired Ethernet, and the devices far from the data processing center and distributed in complex environments or mobile are transmitted using low-power, self-organizing wireless transmission technology to construct the corresponding data acquisition and transmission link; the adapted and optimized image acquisition device and the industrial sensor are connected to the data processing center using the data acquisition and transmission link to generate a multi-target variable tracking and measurement network; In an embodiment of the present invention, an efficient data acquisition and transmission link is constructed to connect the optimized device to the data processing center. Different transmission modes are adopted in consideration of the distance of the device from the data processing center and the amount of data and the size of the data. For devices that are close to the data processing center and have large amounts of data and high real-time requirements, high-speed wired Ethernet is adopted for data transmission. High-speed Ethernet can provide a stable and low-latency transmission channel, which is suitable for processing large-scale, real-time data. For example, when the image acquisition device has a large amount of data and requires high-frequency transmission, the use of a wired network can avoid the bandwidth limitation of wireless transmission. For devices that are far away from the data processing center and distributed in complex environments or mobile locations, low-power, self-organizing wireless transmission technology is adopted. Wireless networks can better adapt to harsh environments, especially in industrial field applications. For example, wireless protocols such as LoRa and ZigBee are used to ensure stable connections between devices. Finally, through the combination of wired and wireless technologies, the construction of a multi-target variable tracking and measurement network is completed to ensure smooth data acquisition and transmission, and finally a multi-target variable tracking and measurement network is connected and generated.

[0032] S16: Perform multi-objective measurement on the production target area using the multi-objective variable tracking measurement network to obtain production multi-objective variable data.

[0033] In an embodiment of the present invention, multi-target measurement of the production target area is performed based on an established multi-target variable tracking and measurement network. The network analyzes the data collected in real time to extract multi-variable information of the production target area, such as changes in physical quantities such as target image, temperature, pressure, and vibration. Advanced data processing techniques, such as time series analysis and trend prediction, are used to process the collected multi-target variable data to further optimize the production process. For example, by analyzing the changing trends of temperature and pressure, it is possible to detect in real time whether the production equipment has abnormalities or potential failures. The key to this step lies in real-time and accuracy. Through the multi-target measurement network, each target can be continuously monitored during the production process, and changes in the production environment can be responded to in a timely manner. Decision support is provided to ensure the stability and efficiency of production multi-variable measurement, and finally the production multi-target variable data is measured.

[0034] Furthermore, the industrial sensor includes a temperature sensor, a pressure sensor and a vibration sensor.

[0035] Furthermore, the data collection frequency dynamic configuration strategy is specifically for target categories that are critical and frequently changing in the production process, to dynamically adjust the data collection frequency corresponding to 1 to 2 time intervals according to the information entropy corresponding to their production status changes; and for target categories that are relatively stable and have smooth status changes, the data collection frequency corresponding to 3 to 5 time intervals is dynamically adjusted based on the time period or production event trigger, where the time interval is specifically 2s.

[0036] Furthermore, the measurement knowledge graph construction module includes the following functions: Obtain the corresponding production measurement target domain knowledge through production tracking measurement targets, including production process parameters, equipment operation specifications and production environment experience knowledge; Extract target knowledge entities from the production measurement target domain knowledge to obtain production measurement target knowledge entities, including production tracking variables, production tracking equipment, and production scenario phenomena; Perform entity attribute mining and analysis between various production measurement target knowledge entities to generate entity attribute relationships between various production measurement knowledge entities, including causal relationships, association relationships, and constraint relationships; Based on the entity attribute relationship between each production measurement knowledge entity, a measurement target knowledge graph is constructed for each production measurement target knowledge entity to generate a production measurement target knowledge graph.

[0037] As an embodiment of the present invention, refer to Figure 3 As shown, Figure 1 Schematic diagram of the functional flow of the measurement knowledge graph construction module in the embodiment. In this embodiment, the measurement knowledge graph construction module includes the following functions: S21: Obtain the corresponding production measurement target domain knowledge through production tracking measurement targets, including production process parameters, equipment operation specifications and production environment experience knowledge; In an embodiment of the present invention, domain knowledge related to the production measurement target is obtained from industry standards, expert experience, and historical data, including production process parameters, equipment operation specifications, and production environment experience knowledge. This process is achieved through a production tracking system. The system collects production data related to the target in real time, such as process parameters such as temperature, pressure, vibration, and rotation speed, and records equipment status information (such as equipment startup, shutdown, and maintenance records). In addition, it also combines production environment conditions, such as workshop temperature and humidity, and airflow distribution in the production area. These data sources provide an important basis for subsequent analysis. All relevant data are managed uniformly through the production tracking system, and the accuracy and timeliness of the data are ensured to form a complete production data set, and finally the corresponding production measurement target domain knowledge is obtained.

[0038] S22: extracting target knowledge entities from the production measurement target domain knowledge to obtain production measurement target knowledge entities, including production tracking variables, production tracking equipment, and production scenario phenomena; In an embodiment of the present invention, by analyzing the previously acquired production measurement target domain knowledge, target knowledge entity extraction is implemented. This process uses natural language processing (NLP) and machine learning technology to automatically label and extract key variables involved in production data. Specifically, the system will identify key entities related to production measurement targets through defined rules or models, such as production tracking variables (such as temperature, humidity, pressure, etc.), production tracking equipment (such as equipment ID, equipment type, equipment operating status, etc.) and phenomena in production scenarios (such as failure occurrence, product quality fluctuation, etc.). Through the extraction process, a list of knowledge entities of production measurement targets is generated, and the position and attributes of each entity in the data are marked, and finally the production measurement target knowledge entity is obtained.

[0039] S23: Perform entity attribute mining analysis on each production measurement target knowledge entity to generate entity attribute relationships between each production measurement knowledge entity, including causal relationships, association relationships, and constraint relationships; In an embodiment of the present invention, by mining and analyzing the entity attribute relationship between production measurement target knowledge entities, first, association analysis and statistical analysis methods are applied to explore the internal connection between different knowledge entities. For example, by statistically analyzing the relationship between equipment operating status and production variables, the causal relationship between equipment failure and production process parameter fluctuations is mined. This process also involves using association rule mining algorithms to identify the correlation between different entities, such as the change pattern of certain variables under different equipment states. Next, constraint condition analysis is used to define the constraint relationship between different entities. For example, when a production variable changes within a certain range, the equipment working state must be maintained within a specific range. Through multi-level analysis, the causal relationship, association relationship and constraint relationship between different entities are obtained. These relationships provide a theoretical basis for constructing a production measurement target knowledge graph, and finally generate entity attribute relationships between various production measurement knowledge entities.

[0040] S24: constructing a measurement target knowledge graph for each production measurement target knowledge entity based on the entity attribute relationship between each production measurement knowledge entity to generate a production measurement target knowledge graph.

[0041] In an embodiment of the present invention, a production measurement target knowledge graph is constructed based on the entity attribute relationship previously mined. This graph is a structured graphical representation, in which each node represents a production measurement target knowledge entity, and each edge represents the attribute relationship between entities. Graph database technology (such as Neo4j) is used to implement the storage and management of the knowledge graph to ensure that the relationship between entities can be efficiently queried and updated. The specific operations include: first, the system connects the identified causal relationships, association relationships, and constraint relationships to the relevant entity nodes in the form of edges; then, the graph is optimized and corrected according to existing rules to ensure the accuracy and completeness of the graph, so as to display the knowledge graph through a graph visualization tool to facilitate further analysis and decision-making by experts or systems. The knowledge graph can not only reflect the changes in measurement targets in the production process in real time, but also provide decision support, help optimize the production process measurement process, and finally generate a production measurement target knowledge graph.

[0042] Furthermore, the entity semantic association mapping module includes the following functions: Perform entity segmentation processing on the corresponding knowledge entities in the production measurement target knowledge graph to obtain production measurement knowledge entity segmentation; In an embodiment of the present invention, entity segmentation is performed on each knowledge entity in the knowledge graph. The operation firstly segments the text contained in the knowledge graph through natural language processing technology. This process performs fine-grained word segmentation on each knowledge entity in the production measurement target knowledge graph by combining a segmentation tool (such as Jieba segmentation or spaCy). During the segmentation, the attributes, parameters, categories and other details of the production measurement target knowledge entity involved are identified and cut, thereby obtaining the segmentation sequence of each knowledge entity. The entities after segmentation can effectively support subsequent analysis and mining work. For example, in the subsequent steps, further calculations and association analysis will be required based on these segmentation results, and finally the production measurement knowledge entity segmentation will be obtained.

[0043] Preferably, the segmentation index of each production measurement knowledge entity segmentation is measured based on the production measurement target knowledge graph to obtain the knowledge entity degree and knowledge entity centrality corresponding to each knowledge entity segmentation in the knowledge graph; In an embodiment of the present invention, after obtaining the word segmentation results of the production measurement target knowledge entity, the word segmentation index of each production measurement knowledge entity is measured based on these word segmentation results to obtain the entity degree and centrality. The knowledge entity degree refers to the degree of connection between the entity and other entities in the knowledge graph, and the centrality reflects the important position of the entity in the knowledge graph. For this purpose, the centrality measurement method in graph theory is adopted, such as degree centrality, closeness centrality or betweenness centrality. Through algorithms (such as PageRank or HITS algorithm), the importance of each knowledge entity in the graph can be effectively evaluated. Based on the calculated entity degree and centrality data, a weight value can be assigned to each knowledge entity. The weight value will become a key indicator for subsequent description association mining and entity embedding analysis, and finally the knowledge entity degree and knowledge entity centrality corresponding to each knowledge entity word segmentation in the knowledge graph are obtained.

[0044] Preferably, based on the knowledge entity degree and knowledge entity centrality corresponding to each knowledge entity segmentation in the knowledge graph, a description association mining analysis is performed on the knowledge entities corresponding to the production measurement target knowledge graph to generate a description association path between each production measurement knowledge entity; In an embodiment of the present invention, after the degree and centrality of each production measurement knowledge entity are quantified, a description association mining analysis is performed next. The purpose of this step is to reveal the semantic relationship and description path between different knowledge entities. Through graph analysis tools, such as Gephi or NetworkX, graph algorithms (such as shortest path algorithm or graph clustering) can be used based on the calculated entity degree and centrality to discover the association path between entities. In this process, the connectivity and structural characteristics of the graph are used to mine the direct or indirect semantic relationship between each production measurement knowledge entity, and further generate an association path. The association path represents a possible relationship chain from one entity to another, and finally generates a description association path between each production measurement knowledge entity.

[0045] Preferably, the corresponding knowledge entities in the production measurement target knowledge graph are described and semantically embedded based on the description association path between each production measurement knowledge entity to obtain a production measurement knowledge entity description embedding vector, which includes the attributes and text features corresponding to each production measurement knowledge entity and the semantic relationship between it and other knowledge entities; In an embodiment of the present invention, after generating the description association path between each production measurement knowledge entity, a description semantic embedding analysis is performed. This step uses a semantic embedding model (such as Word2Vec, GloVe or BERT model) to semantically represent the production measurement knowledge entity. Through these models, the description of each knowledge entity and its association path can be converted into a low-dimensional vector representation. These vectors not only contain the attribute information of each entity itself (such as the characteristics and categories of the entity), but also include the semantic relationship with other entities. This semantic embedding process is trained in massive data through a deep learning model to capture the potential semantic connection between entities, thereby generating a high-quality embedding vector. The vector can provide a basis for further association and analysis of target variable data, and finally obtain the production measurement knowledge entity description embedding vector.

[0046] Preferably, entity semantic association mapping is performed on the corresponding target variables in the production multi-target variable data based on the production measurement knowledge entity description embedding vector to generate multi-target variable measurement data corresponding to the knowledge entity semantic annotation.

[0047] In an embodiment of the present invention, after completing the semantic embedding of the production measurement knowledge entity, the entity semantic association mapping of the target variable is then performed. This process realizes the combination of the knowledge graph and the multi-target variable data by connecting the description embedding vector of the production measurement knowledge entity with the target variable in the actual production measurement data. Specifically, by utilizing the correlation between the semantic embedding vector of the knowledge graph and the target variable, machine learning or statistical analysis methods (such as regression analysis, clustering algorithm, matching algorithm, etc.) are used for association mapping, and finally multi-target variable measurement data annotated with semantic labels are generated. This mapping process can not only enhance the semantic understanding of the measurement data, but also provide deeper intelligent analysis and decision support for various target variables appearing in the production measurement process.

[0048] Furthermore, the entity semantic association mapping of the corresponding target variables in the production multi-target variable data based on the production measurement knowledge entity description embedding vector includes: Use the corresponding text encoder in the CLIP model to perform text encoding embedding on the corresponding target variables in the production multi-target variable data to generate the target variable text prompt embedding vector; In an embodiment of the present invention, the text encoder of the CLIP (Contrastive Language-Image Pretraining) model is used to perform text encoding on the target variables in the production multi-target variable data. This process involves inputting the text description of the target variable into the text encoder part of the CLIP model. The text encoder converts the text of the target variable into a vector representation, which is a distributed representation of the target variable in a multi-dimensional semantic space. In a specific implementation, it is first ensured that the data of each target variable contains sufficient descriptive information, such as detailed descriptions through texts such as product names, function descriptions, and performance parameters. Then, the pre-trained text encoder in the CLIP model is used to perform text embedding on each target variable, and the text of the target variable is converted into a high-dimensional vector. This vector captures the semantic information contained in the target variable text, and finally obtains the target variable text prompt embedding vector.

[0049] Preferably, based on the text features corresponding to the embedding vector of the production measurement knowledge entity description, the entity semantic association calculation formula is used to measure the entity semantic association of the target variable text prompt corresponding to the embedding vector of the target variable text prompt, so as to obtain the text semantic association between the target variable and the knowledge graph entity; In an embodiment of the present invention, a suitable entity text semantic association calculation formula is formed by combining the text features in the production measurement knowledge entity description embedding vector, the target variable text prompt in the target variable text prompt embedding vector, the vector inner product, the correlation weight coefficient, the Euclidean distance square value, the distance decay factor and related parameters to perform association measurement calculation to calculate the semantic similarity between the text prompt embedding vector of the target variable and the knowledge graph entity description embedding vector, and finally obtain the text semantic association between the target variable and the knowledge graph entity. In addition, the entity text semantic association calculation formula can also use any text similarity measurement algorithm in this field to replace the entity semantic association measurement process, and is not limited to the entity text semantic association calculation formula.

[0050] Preferably, entity semantic association mapping is performed on the corresponding target variables in the production multi-target variable data based on the textual semantic association between the target variable and the knowledge graph entity and in combination with the corresponding knowledge entity in the production measurement target knowledge graph to generate multi-target variable measurement data corresponding to the knowledge entity semantic annotations.

[0051] In an embodiment of the present invention, entity semantic association mapping is performed on the target variables in the production multi-target variable data by combining the text semantic association between the target variable and the knowledge graph entity obtained previously, and the entity information in the production measurement target knowledge graph. Specifically, the calculated text semantic association is used to match and map it with the corresponding knowledge entity in the knowledge graph. When the semantic association between the text prompt embedding vector of the target variable and the description embedding vector of a knowledge graph entity reaches a certain threshold, it indicates that the target variable has a high correlation with the knowledge entity. Therefore, the semantic knowledge corresponding to the entity can be mapped to the target variable. Through this mapping process, the generated multi-target variable measurement data will be attached with the semantic annotation information of the corresponding knowledge entity. The annotation information may include the category, attribute and relationship between the knowledge entity and the target variable, etc., thereby enhancing the semantic expression ability and interpretability of the measurement data. The obtained multi-target variable measurement data not only contains the original variable data, but also integrates the semantic information of the knowledge graph entity, forming a data set with rich semantic annotations, and finally generating multi-target variable measurement data corresponding to the knowledge entity semantic annotations.

[0052] Furthermore, the entity text semantic association calculation formula is specifically: ; In the formula, The target variable With knowledge graph entities The semantic correlation between the texts The total number of text feature correspondences in the embedding vector of the knowledge entity description is measured for production, is the total number of target variable text prompts in the target variable text prompt embedding vector, To produce measurement knowledge entity descriptions, the first text features, The first target variable text prompt, For the The text features and The vector inner product between the target variable text prompts, For the The text features and The correlation weight coefficient between the target variable text prompts, is an exponential function, For the The text features and The squared Euclidean distance between the target variable text prompts, For the The text features and The distance decay factor between target variable text prompts, is the correction coefficient of text semantic relevance.

[0053] The present invention obtains an entity text semantic association calculation formula by using a specific mathematical model and after verification, which is used to measure the entity semantic association of the target variable text prompt corresponding to the target variable text prompt embedding vector. The entity text semantic association calculation formula can accurately measure the text semantic association between the target variable and the knowledge graph entity by combining the text prompt of the target variable and the description embedding vector of the production measurement knowledge entity, which helps to better understand the semantic relationship between the target variable and the entity, making the prediction or analysis of multi-target variable data more reliable and accurate. The correlation weight coefficient and distance decay factor in the formula make the association between the text feature and the target variable more refined. Through the weighted inner product and distance decay, the correlation between different target variables and knowledge graph entities can be effectively distinguished, avoiding simple matching or overfitting, and improving the accuracy of semantic matching. The inner product operation captures the intrinsic relationship between the entity description and the target variable text prompt, reflecting the potential connection between different text features. This mechanism can not only identify the similarity of text content, but also adapt to different types of text expressions, thereby improving the versatility and robustness of the model. In addition, the introduction of the correction coefficient can provide an adjustment mechanism for the formula, which can flexibly fine-tune the calculated text semantic relevance. It helps to eliminate possible errors or deviations and ensure that the final semantic relevance can truly reflect the actual relationship between the target variable and the knowledge graph entity. By combining the inner product of the text embedding vector and the Euclidean distance, this formula can handle high-dimensional text data and complex semantic relationships. For production data with multiple target variables, it can deeply explore the complex relationships between each target and between them and entities in the knowledge graph, thereby providing strong support for subsequent analysis and modeling. In summary, this formula fully takes into account the target variable With knowledge graph entities The semantic correlation between the text , the total number of text features corresponding to the embedding vector of the knowledge entity description produced , the total number of target variable text prompts in the target variable text prompt embedding vector , the production measurement knowledge entity describes the first Text features , the target variable text hint embedding vector Target variable text prompt , No. The text features and The vector inner product between the target variable text prompts , No. The text features and The correlation weight coefficient between the target variable text prompts , exponential function , No. The text features and The squared Euclidean distance between the target variable text prompts , No. The text features and The distance decay factor between the target variable text prompts , correction coefficient of text semantic relevance , according to the target variable With knowledge graph entities The semantic correlation between the text The correlation between the above parameters constitutes a functional relationship This formula can realize the entity semantic association measurement process of the target variable text prompt corresponding to the target variable text prompt embedding vector, and at the same time, through the correction coefficient of text semantic association The introduction of can be adjusted according to the errors that occur in the calculation process, thereby improving the accuracy and applicability of the entity text semantic association calculation formula.

[0054] Furthermore, the knowledge-enhanced measurement correction module includes the following functions: Obtain the corresponding production measurement knowledge constraints and production measurement reasoning mechanism through the production measurement target knowledge graph; In an embodiment of the present invention, knowledge modeling of the production measurement process is performed through a production measurement target knowledge graph (KM). The knowledge graph contains a variety of entities and relationships, representing important concepts and laws in the field of production measurement. Through semantic analysis, relevant production measurement knowledge constraints (for example, measurement accuracy requirements, equipment limitations, measurement environment conditions, etc.) and production measurement reasoning mechanisms (such as the relationship between multiple target variables and reasoning rules) are obtained. On this basis, a multi-level and multi-dimensional knowledge graph that can reflect the intrinsic relationship between different variables is constructed. Specifically, a graph database or a graph convolutional network (GCN) can be used to extract rules and relationships from the production environment and historical data. The extracted knowledge will form strict constraints and reasoning mechanisms for the measurement process, and finally the corresponding production measurement knowledge constraints and production measurement reasoning mechanisms are obtained.

[0055] Preferably, a measurement model based on knowledge enhancement is constructed using a convolutional neural network, and the corresponding convolutional layer structure and connection hyperparameters in the measurement model based on knowledge enhancement are optimized by introducing production measurement knowledge constraints and production measurement reasoning mechanisms, so that the measurement model based on knowledge enhancement captures the intrinsic measurement relationship between multiple objective variables, and generates a measurement optimization model based on knowledge enhancement; In an embodiment of the present invention, a measurement model based on knowledge enhancement is constructed by using a convolutional neural network (CNN). First, a convolutional neural network architecture is designed, with the focus on carefully setting and optimizing the convolutional layers in the network structure. These convolutional layers will enhance the perception of the intrinsic measurement relationship between multiple objective variables by learning local feature extraction of input data. For knowledge enhancement, a reasonable loss function is designed, integrating production measurement knowledge constraints and reasoning mechanisms. This loss function will prompt the model to learn how to adjust the network structure according to the known knowledge graph, especially the weights and connection hyperparameters of the convolutional layer. In this process, the hyperparameters involved (such as convolution kernel size, step size, activation function, etc.) need to be tuned according to the characteristics of the production measurement task. Through the training process, the parameters of the model are continuously optimized so that it can capture and model the precise measurement relationship between multiple objective variables, and finally optimize and generate a measurement optimization model based on knowledge enhancement.

[0056] Preferably, the corresponding knowledge entities in the production measurement target knowledge graph are used as additional supervision information, and the multi-objective variable measurement data corresponding to the knowledge entity semantic annotations are input into the measurement optimization model based on knowledge enhancement and combined with the multi-objective variable tracking measurement network to perform knowledge enhanced measurement correction to obtain the multi-objective variable measurement correction results.

[0057] In an embodiment of the present invention, knowledge entities in the production measurement target knowledge graph are input as additional supervisory information into the measurement optimization model based on knowledge enhancement. These knowledge entities may include specific production processes, equipment configurations, environmental factors and other information. The semantic annotations of the knowledge entities are combined with the multi-objective variable measurement data to form an enhanced learning mechanism. Next, the measurement data of the multi-objective variables (such as target images, temperature, pressure, vibration, etc.) are input into the optimized measurement model together with their corresponding knowledge entity semantic annotations to perform measurement correction. To achieve this correction process, the measurement optimization model based on knowledge enhancement will perform measurement correction on the input multi-objective variables and The knowledge entities are measured comparatively to determine whether the current measurement results conform to the constraints and reasoning rules in the knowledge graph. When the model detects a conflict between the measurement data and the knowledge entity (for example, the measurement result exceeds the allowable range or does not conform to historical rules), the multi-objective variable tracking measurement network is enabled for further correction. The tracking network will re-measure the conflicting target variables and dynamically adjust the measurement results according to the knowledge graph reasoning mechanism and real-time data to output the corrected multi-objective variable measurement results. The whole process emphasizes the deep integration and intelligent correction of the model to the production environment and measurement data to ensure the accuracy and consistency of the measurement results, and finally obtain the multi-objective variable measurement correction results.

[0058] Furthermore, the process of knowledge-enhanced measurement correction is specifically to use the measurement optimization model based on knowledge enhancement to perform comparative measurements between the corresponding knowledge entities in the production measurement target knowledge graph and the multi-objective variables corresponding to the knowledge entity semantic annotations. When the multi-objective variables corresponding to the knowledge entity semantic annotations conflict with the corresponding knowledge entities in the production measurement target knowledge graph, the multi-objective variable tracking measurement network is used to re-measure the target variables corresponding to the conflict to correct them, so as to obtain the multi-objective variable measurement correction results.

[0059] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0060] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A multi-objective variable measurement system based on knowledge enhancement, characterized in that: Includes the following modules: The multi-target tracking and measurement module is used to obtain the production tracking and measurement targets, and based on the production tracking and measurement targets, dynamically configure the data acquisition frequency and build the corresponding data acquisition transmission link to connect the image acquisition equipment, industrial sensors and data processing center to generate a multi-target variable tracking and measurement network; Use the multi-objective variable tracking measurement network to perform multi-objective measurement on the production target area to obtain production multi-objective variable data; The measurement knowledge graph construction module is used to obtain the corresponding production measurement target domain knowledge through the production tracking measurement target, and extract the target knowledge entity of the production measurement target domain knowledge to obtain the production measurement target knowledge entity, which includes production tracking variables, production tracking equipment and production scenario phenomena; Constructing a measurement target knowledge graph based on the production measurement target knowledge entity to generate a production measurement target knowledge graph; An entity semantic association mapping module is used to perform entity semantic association mapping on the corresponding target variables in the production multi-target variable data based on the corresponding knowledge entities in the production measurement target knowledge graph, so as to generate multi-target variable measurement data corresponding to the knowledge entity semantic annotations; The knowledge-enhanced measurement correction module is used to obtain the corresponding production measurement knowledge constraints and production measurement reasoning mechanisms through the production measurement target knowledge graph, and based on the production measurement knowledge constraints and production measurement reasoning mechanisms, the corresponding knowledge entities in the production measurement target knowledge graph are used as additional supervision information in combination with the multi-objective variable tracking measurement network to perform knowledge-enhanced measurement correction on the multi-objective variable measurement data corresponding to the knowledge entity semantic annotations, so as to obtain the multi-objective variable measurement correction results.

2. The multi-objective variable measurement system based on knowledge enhancement according to claim 1 is characterized in that: The multi-target tracking and measurement module includes the following functions: Obtain production tracking measurement targets, including the physical attributes of each target, the role of the production process, and the frequency and amplitude of its production status changes; Perform target cluster analysis on production tracking measurement targets, so as to classify the key targets with frequent state changes in the production process into one category according to the corresponding production characteristics, and classify the relatively stable targets with stable state changes into another category, so as to obtain a production tracking target classification system; Based on the production tracking target classification system, the data acquisition frequency corresponding to the image acquisition equipment and industrial sensors is dynamically configured to generate a data acquisition frequency dynamic configuration strategy; Based on the data acquisition frequency dynamic configuration strategy, the image acquisition equipment and industrial sensors are adapted and optimized for equipment parameters. For the image acquisition equipment, the corresponding resolution, frame rate and exposure time are adjusted according to the acquisition frequency requirements of different targets. For the industrial sensors, their sensitivity and response time are optimized according to the physical properties of the targets, and the adapted and optimized image acquisition equipment and industrial sensors are obtained. By combining wired transmission and wireless transmission, the acquisition and transmission links between the adapted and optimized image acquisition devices and industrial sensors and the data processing center are constructed, so that the corresponding devices that are close to the data processing center and have large data volumes and high real-time requirements are transmitted using high-speed wired Ethernet, and the devices that are far from the data processing center and distributed in complex environments or mobile are transmitted using low-power, self-organizing wireless transmission technology to construct the corresponding data acquisition and transmission links; the data acquisition and transmission links are used to connect the adapted and optimized image acquisition devices and industrial sensors to the data processing center to generate a multi-target variable tracking and measurement network; The multi-objective variable tracking measurement network is used to perform multi-objective measurement on the production target area to obtain production multi-objective variable data.

3. The multi-objective variable measurement system based on knowledge enhancement according to claim 2 is characterized in that: The industrial sensors include temperature sensors, pressure sensors and vibration sensors.

4. The multi-objective variable measurement system based on knowledge enhancement according to claim 2 is characterized in that: The data collection frequency dynamic configuration strategy is specifically for target categories that are critical and frequently changing in the production process, to dynamically adjust the data collection frequency corresponding to 1 to 2 time intervals according to the information entropy corresponding to their production status changes; and for target categories that are relatively stable and have smooth status changes, the data collection frequency corresponding to 3 to 5 time intervals is dynamically adjusted based on the time period or production event trigger, where the time interval is specifically 2s.

5. The multi-objective variable measurement system based on knowledge enhancement according to claim 1 is characterized in that: The measurement knowledge graph building module includes the following functions: Obtain the corresponding production measurement target domain knowledge through production tracking measurement targets, including production process parameters, equipment operation specifications and production environment experience knowledge; Extract target knowledge entities from the production measurement target domain knowledge to obtain production measurement target knowledge entities, including production tracking variables, production tracking equipment, and production scenario phenomena; Perform entity attribute mining and analysis between various production measurement target knowledge entities to generate entity attribute relationships between various production measurement knowledge entities, including causal relationships, association relationships, and constraint relationships; Based on the entity attribute relationship between each production measurement knowledge entity, a measurement target knowledge graph is constructed for each production measurement target knowledge entity to generate a production measurement target knowledge graph.

6. The multi-objective variable measurement system based on knowledge enhancement according to claim 1 is characterized in that: The entity semantic association mapping module includes the following functions: Perform entity segmentation processing on the corresponding knowledge entities in the production measurement target knowledge graph to obtain production measurement knowledge entity segmentation; Based on the production measurement target knowledge graph, the segmentation index of each production measurement knowledge entity segmentation is measured to obtain the knowledge entity degree and knowledge entity centrality corresponding to each knowledge entity segmentation in the knowledge graph; Based on the knowledge entity degree and knowledge entity centrality of each knowledge entity segmentation in the knowledge graph, the corresponding knowledge entity in the production measurement target knowledge graph is subjected to description association mining analysis to generate description association paths between each production measurement knowledge entity; Based on the description association path between each production measurement knowledge entity, the corresponding knowledge entity in the production measurement target knowledge graph is described and semantically embedded to obtain the production measurement knowledge entity description embedding vector, which includes the attributes, text features and semantic relationships between each production measurement knowledge entity and other knowledge entities. Based on the production measurement knowledge entity description embedding vector, entity semantic association mapping is performed on the corresponding target variables in the production multi-target variable data to generate multi-target variable measurement data corresponding to the knowledge entity semantic annotations.

7. The multi-objective variable measurement system based on knowledge enhancement according to claim 6 is characterized in that: The method performs entity semantic association mapping on the corresponding target variables in the production multi-target variable data based on the production measurement knowledge entity description embedding vector include: Use the corresponding text encoder in the CLIP model to perform text encoding embedding on the corresponding target variables in the production multi-target variable data to generate the target variable text prompt embedding vector; Based on the text features corresponding to the embedding vector of the production measurement knowledge entity description, the entity text semantic association calculation formula is used to measure the entity semantic association of the target variable text prompt corresponding to the target variable text prompt embedding vector to obtain the text semantic association between the target variable and the knowledge graph entity; Based on the textual semantic association between the target variable and the knowledge graph entity and combined with the corresponding knowledge entities in the production measurement target knowledge graph, entity semantic association mapping is performed on the corresponding target variables in the production multi-target variable data to generate multi-target variable measurement data with corresponding knowledge entity semantic annotations.

8. The multi-objective variable measurement system based on knowledge enhancement according to claim 7 is characterized in that: The entity text semantic association calculation formula is specifically: ； In the formula, The target variable With knowledge graph entities The semantic correlation between the texts, The total number of text feature correspondences in the embedding vector of the knowledge entity description is measured for production, is the total number of target variable text prompts in the target variable text prompt embedding vector, To produce measurement knowledge entity descriptions, the first text features, The first target variable text prompt, For the The text features and The vector inner product between the target variable text prompts, For the The text features and The correlation weight coefficient between the target variable text prompts, is an exponential function, For the The text features and The squared Euclidean distance between the target variable text prompts, For the The text features and The distance decay factor between target variable text prompts, is the correction coefficient of text semantic relevance.

9. The multi-objective variable measurement system based on knowledge enhancement according to claim 1 is characterized in that: The knowledge-enhanced measurement correction module includes the following functions: Obtain the corresponding production measurement knowledge constraints and production measurement reasoning mechanism through the production measurement target knowledge graph; A measurement model based on knowledge enhancement is constructed using convolutional neural networks. The corresponding convolutional layer structure and connection hyperparameters in the measurement model based on knowledge enhancement are optimized by introducing production measurement knowledge constraints and production measurement reasoning mechanisms, so that the measurement model based on knowledge enhancement can capture the intrinsic measurement relationship between multiple objective variables and generate a measurement optimization model based on knowledge enhancement. The corresponding knowledge entities in the production measurement target knowledge graph are used as additional supervision information, and the multi-objective variable measurement data corresponding to the knowledge entity semantic annotations are input into the measurement optimization model based on knowledge enhancement and combined with the multi-objective variable tracking measurement network to perform knowledge enhanced measurement correction to obtain the multi-objective variable measurement correction results.

10. The multi-objective variable measurement system based on knowledge enhancement according to claim 9, characterized in that: The process of knowledge-enhanced measurement correction is specifically to use the measurement optimization model based on knowledge enhancement to compare and measure the corresponding knowledge entities in the production measurement target knowledge graph with the multi-objective variables corresponding to the knowledge entity semantic annotations. When the multi-objective variables corresponding to the knowledge entity semantic annotations conflict with the corresponding knowledge entities in the production measurement target knowledge graph, the multi-objective variable tracking measurement network is used to re-measure the target variables corresponding to the conflict to correct them, so as to obtain the multi-objective variable measurement correction results.

Images