Industrial Internet of Things main service platform data analysis cooperation system and method

Abstract

The invention relates to the technical field of intelligent manufacturing, and particularly discloses an industrial Internet of Things main service platform data analysis cooperation system and method, and the system comprises a user platform, a service platform, a management platform, a sensing network platform, and an object platform. The object platform comprises a plurality of production devices, each production device is provided with a sensing unit and a control unit, and the sensing units are used for collecting equipment operation physical quantities and product processing parameters, performing local preliminary screening, caching, marking emergency key data, removing invalid data and then transmitting the data to the sensing network platform to form preliminary sensing data; the sensing network platform comprises a plurality of sensing network sub-platforms and corresponding sensing network databases. The invention aims to provide the industrial Internet of Things main service platform data analysis cooperation system and method, so as to solve the technical problems of large data processing time delay, poor production real-time performance and low troubleshooting efficiency of the existing service platform.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent manufacturing, and specifically discloses a data analysis and collaboration system and method for an industrial Internet of Things main service platform. Background Art

[0002] With the rapid development of intelligent manufacturing technology, the application of industrial Internet of Things in the field of industrial production is becoming more and more extensive. A typical industrial Internet of Things system covers five major sections: user platform, service platform, management platform, sensor platform and object platform. They work together to realize intelligent detection and control of production lines.

[0003] For example, a Chinese invention patent discloses (publication number: CN114629940A) an industrial Internet of Things system and control method that is conducive to system scalability. The system shown includes a user platform, a service platform, a management platform, a sensor network platform and an object platform; the service platform and the sensor network platform are arranged independently; the service platform and the sensor network platform both include multiple sub-platforms, and each sub-platform is provided with a database, a processor and / or an information channel; the object platform includes a production line, and the production line is equipped with multiple sensors; in each sub-platform of the sensor network platform, the sensors connected to the same sensor network platform sub-platform use the same communication protocol; any sensor network platform sub-platform is connected to a unique service platform sub-platform through the management platform.

[0004] Under the above-mentioned industrial Internet of Things architecture, although each platform performs its own duties, many problems are gradually exposed when facing complex and changeable industrial production scenarios. For example, in large-scale interactive scenarios with large amounts of data, such as large-scale machinery manufacturing factories, a large number of production devices in the object platform (such as CNC machine tools, automated assembly robotic arms, etc.) continue to generate massive amounts of operating status data, product quality monitoring data, etc., and these data rush to the service platform. Since traditional service platforms mostly adopt a centralized architecture and lack effective data diversion and collaborative processing mechanisms, the service platform is overwhelmed when facing such a huge data torrent. This not only leads to a significant increase in data processing latency, seriously affecting production efficiency, but may also cause system downtime due to data backlogs, causing production lines to stagnate, and bring huge economic losses to enterprises.

[0005] For another example, in a real-time control scenario, taking the high-precision packaging production line of electronic chips as an example, the production process has extremely high real-time requirements. The production equipment of the object platform must be adjusted in real time based on the precise data such as temperature, humidity, and equipment vibration frequency collected in real time by the sensor network platform to ensure product quality. However, the data analysis between the platforms in the current system lacks coordination. The production process adjustment instructions issued by the user platform, the resource allocation information of the management platform, and the real-time perception data of the sensor network platform cannot be quickly and effectively summarized and converted into precise control signals on the main service platform. This makes it difficult for production equipment to respond in a timely manner, which easily causes an increase in the defective rate of products and cannot meet the stringent requirements of high-end manufacturing for product quality stability.

[0006] Another example is in the fault diagnosis and maintenance scenario, when a link in the industrial Internet of Things system fails, such as the reactor temperature control system failure on the chemical production line, the sensor network platform can capture abnormal temperature data, but there is a lack of close data analysis and coordination between the main service platform and other entities. The main service platform cannot timely cooperate with the management platform to retrieve the historical maintenance records and operating parameter change curves of the reactor, and it is difficult to quickly communicate with the user platform about the impact of production interruptions on order delivery. It is even more impossible to efficiently integrate the real-time fault-related data continuously transmitted by the sensor network platform (such as abnormal current fluctuations of heating elements, abnormal opening of temperature control valves, etc.), resulting in low efficiency in troubleshooting and extended maintenance time, further exacerbating production delays and cost increases.

[0007] In the above existing industrial Internet of Things systems, there are serious deficiencies in data analysis and coordination between the service platform and the user platform, management platform, sensor network platform and object platform. As a result, when faced with many key industrial production scenarios such as large data volume interaction, real-time control, fault diagnosis and maintenance, a series of technical problems such as large data processing delays, poor production real-time performance and low troubleshooting efficiency arise, which need to be solved urgently. Summary of the invention

[0008] The purpose of the present invention is to provide an industrial Internet of Things main service platform data analysis and collaboration system and method to solve the technical problems of large data processing delay, poor production real-time performance and low troubleshooting efficiency of the existing service platform.

[0009] In order to achieve the above-mentioned purpose, the basic scheme of the present invention is: data analysis and coordination system of the main service platform of the industrial Internet of Things, user platform, service platform, management platform, sensor network platform and object platform; The object platform includes a plurality of production devices, each of which is provided with a sensing unit and a control unit, wherein the sensing unit is used to collect physical quantities of equipment operation and product processing parameters, perform local preliminary screening and caching, mark urgent key data, remove invalid data and transmit to the sensor network platform to form preliminary sensing data; The sensor network platform includes multiple sensor network sub-platforms and corresponding sensor network databases, which are used to receive preliminary perception data, adaptively adjust the sampling frequency to prevent data transmission congestion, perform secondary screening and classification on the data, and store them in different databases according to rules, perform preliminary quantitative analysis to extract features, and then send the preliminary perception data after integration and analysis to the management platform; The management platform includes multiple management sub-platforms and an independent management database, which is used to receive the preliminary perception data, conduct in-depth interpretation based on the knowledge graph, judge the equipment operation status, predict the failure risk, and store the preliminary perception data together with the equipment status, failure prediction, and resource allocation plan in the management database, and send it to the service platform after forming management data; The service platform adopts an optimized combination and post-fractional layout, including a main service database and several sub-service databases. Each sub-service database is connected to a corresponding service sub-platform and is equipped with an intelligent shunting collaborative cache mechanism. The main service database receives management data and summarizes and preliminarily classifies it, and distributes it to the sub-service database according to type and priority. The service sub-platform processes and mines data value in parallel, generates refined data service products and stores them in the sub-service database; The user platform is used to receive data from the sub-service database, and to track and update the load status of the sub-service database and issue an overtime delay threshold warning, helping users obtain useful data to guide decision-making and adjust processes. The user platform is also used to modify the sub-service database data and feed back requirements and instructions to the service platform to form a closed-loop collaborative process.

[0010] The working principle and beneficial effects of this basic solution are: after the raw data is collected by the perception unit of the object platform, it is preliminarily screened and cached locally, which effectively reduces the transmission of invalid data and reduces the network burden. After receiving the preliminary perception data, the sensor network platform adaptively adjusts the sampling frequency to prevent data transmission congestion, and performs secondary screening and classification of the data and stores it in different databases, which helps to quickly respond to and process key data. This diversion mechanism reduces data congestion during transmission, improves the efficiency of data processing, and reduces latency.

[0011] The sensor network platform performs preliminary quantitative analysis of the data and extracts key features, such as statistical temperature trends, which enables subsequent platforms to conduct in-depth analysis and decision-making more quickly. Feature extraction simplifies the data, allowing important information to be quickly identified and processed, and improving the system's response speed and real-time performance.

[0012] The management platform deeply interprets the data sent by the sensor network platform based on the knowledge graph, determines the equipment operation status, predicts the risk of failure, and plans resource allocation, forming management data and sending it to the service platform. This in-depth interpretation and management decision-making improves the intelligence level of the system, enables the system to more accurately predict and respond to potential failures, and improves the efficiency of troubleshooting.

[0013] The service platform adopts a combined post-fractional layout of the main service database and the sub-service database. The sub-service database is connected to the service sub-platform with a high-speed processing chip, which realizes the intelligent diversion and parallel processing of data and improves the data processing speed. The intelligent diversion collaborative caching mechanism and parallel processing capability significantly reduce the data processing delay and improve the real-time performance of production.

[0014] The user platform tracks data updates and over-delay thresholds, provides early warnings, and helps users obtain useful data in a timely manner to guide decision-making and adjust processes. Real-time monitoring and early warning mechanisms improve users' control over the production process and enhance the real-time and flexibility of production.

[0015] In summary, the present invention achieves optimization in each link of data generation, transmission, processing and application. This solution improves the speed and accuracy of data processing, enhances the real-time response capability of the system, and improves the efficiency of fault diagnosis and maintenance. These technical means work together to form an efficient, intelligent and collaborative data analysis and collaboration system for the main service platform of the Industrial Internet of Things. The above technical means effectively solve the technical problems of large data processing delay, poor production real-time performance and low troubleshooting efficiency in the service platform of the Industrial Internet of Things.

[0016] Furthermore, the main service database or the management sub-platform is used to generate a timestamp while sending information to the user platform, and attach it to the information; the user platform is used to calculate the time delay between the time when the information from the sub-service database is actually received and the timestamp, and perform corresponding processing according to different ranges of the time delay, and the corresponding processing includes: When the latency is less than 30 milliseconds, the user platform displays to the user that the service sub-platform is working normally; When the latency is between 30 milliseconds and 800 milliseconds, the user platform displays service sub-platform network abnormality information to the user; When the latency exceeds 800 milliseconds, the user platform accesses the service sub-platform and enters fault handling mode.

[0017] When the main service database or management sub-platform sends information to the user platform, it generates a timestamp. This timestamp records the exact moment when the information was sent, usually in milliseconds. The timestamp can be generated by the high-precision clock built into the system to ensure its accuracy and consistency. The generated timestamp will be attached to the information data packet and sent to the user platform together. After the user platform receives the information, it will obtain the current system time, which represents the actual moment when the information arrives. Then, the user platform will extract the attached timestamp from the received information data packet, that is, the moment when the information was sent. By calculating the time difference between the current time and the timestamp, the user platform can obtain the time spent in the entire transmission process from sending to receiving the information, which is the latency. This calculation process can be implemented by a simple subtraction operation, or more accurately: latency = current time - timestamp.

[0018] The user platform compares the calculated latency value with different preset time ranges to determine the working status of the service sub-platform. Specifically: when the latency is less than 30 milliseconds, it means that the data transmission of the service sub-platform is very fast and the system runs smoothly. The user platform will display the "Service Sub-platform Normal Working Information" to the user to let the user know that the system is in good working condition and can perform various operations normally. When the latency is between 30 milliseconds and 800 milliseconds, it indicates that there may be a certain delay in the network of the service sub-platform, but it has not reached a serious failure level. At this time, the user platform will display the "Service Sub-platform Network Abnormal Information" to the user, reminding the user to pay attention to the network status, and may need to take some measures, such as checking the network connection, restarting the device, etc., to try to improve the network delay problem. When the latency exceeds 800 milliseconds, it means that the network delay of the service sub-platform is very serious and may have affected the normal operation of the system. In this case, the user platform will automatically access the service sub-platform and enter the "fault handling mode". In the fault handling mode, the user platform may execute a series of diagnostic procedures, such as detecting the connection status of the service sub-platform, querying error logs, analyzing data transmission paths, etc., to determine the specific cause of the fault and try to take corresponding repair measures, such as re-establishing the connection, adjusting the network configuration, etc., to restore the normal operation of the service sub-platform as soon as possible.

[0019] By attaching a timestamp when sending information and calculating the delay when receiving it, the user platform can monitor the data transmission status of the service sub-platform in real time. This real-time monitoring mechanism enables users to understand the operation of the system in a timely manner without waiting for regular system reports or manual checks, which improves the system's response speed and user experience. The judgment of the delay range provides a clear reference standard for the user platform, which can accurately locate the working status of the service sub-platform. When network delays or failures occur, the user platform can quickly identify the problem and give corresponding prompt information to help users quickly find the root cause of the problem, reducing the time and effort required to troubleshoot the problem. When the delay exceeds the preset threshold, the user platform will automatically enter the fault handling mode and take proactive measures to diagnose and repair the fault. This automated fault handling mechanism improves the reliability of the system, reduces the risk of system interruption caused by network delays or failures, and ensures the stable operation of the industrial Internet of Things system, thereby ensuring the continuity of industrial production and product quality.

[0020] The user platform displays different information to users according to different latency ranges, providing intuitive and easy-to-understand system status feedback. Users can adjust their operation strategies or take corresponding measures in a timely manner based on this information, which enhances their control and satisfaction with the system and improves the overall user experience.

[0021] Furthermore, the user platform is further configured to perform the following steps in the fault handling mode: Calculate the delay growth rate between the starting delay and the final delay of the delay per unit time; When the delay growth rate is less than 0, the user platform continues to make judgments in the next unit time until the delay growth rate is not less than 0 or the fault handling mode is exited; When the latency growth rate is between 0 and 30%, the user platform continues to make judgments in the next unit time until the latency growth rate is not less than 0 or the fault handling mode is exited; When the latency growth rate is between 30% and 80%, the user platform generates service sub-platform network anomaly information to the user; When the delay growth rate is greater than 80%, the user platform generates fault information of the corresponding management sub-platform to the user.

[0022] During the operation of the data analysis and collaboration system of the main service platform of the Industrial Internet of Things, when the user platform detects that the information delay from the sub-service database exceeds 800 milliseconds, it enters the fault handling mode. At this time, in order to accurately locate the root cause of the fault and track the development trend of the fault in real time, the user platform introduces the key indicator of the delay growth rate for in-depth analysis.

[0023] Specifically, the user platform will record the starting and final values ​​of the delay within a unit time (for example, set to 1 second), and obtain the delay change within the period through a specific calculation logic: delay growth rate = (final delay - starting delay) / starting delay × 100%. This delay growth rate presented in percentage form can intuitively reflect the performance fluctuation trend of the service sub-platform or related links in a short period of time.

[0024] When the calculated delay growth rate is less than 0, it means that the system has experienced a delay reduction in the unit time, which may be due to temporary self-repair of the network, partial congestion relief or other temporary factors. However, to ensure that the system stability continues to recover, the user platform will not immediately determine that the fault has been resolved, but will continue to repeat the above calculation and judgment process in the next unit time until the delay growth rate is stable and not less than 0, indicating that the system has returned to normal, or until other preset conditions for exiting the fault handling mode are met.

[0025] Similarly, when the latency growth rate is between 0 and 30%, the system is in a relatively mild fault fluctuation range. Although the latency has increased, the growth rate is slow. This may be caused by a slight increase in network load and less intense resource competition. The user platform continuously monitors and calculates the latency growth rate again in the next unit time to capture the fault development trend in a timely manner until the system recovers or meets the exit conditions.

[0026] When the latency growth rate is between 30% and 80%, it indicates that the network link on which the service sub-platform relies has experienced obvious performance degradation and data transmission efficiency has dropped significantly. This is most likely due to factors such as network congestion, excessive load on some nodes, or frequent minor failures at the software level. At this time, the user platform generates service sub-platform network abnormality information to users, prompting them to pay attention to network conditions and may need to take preliminary troubleshooting or optimization measures.

[0027] The most serious situation is that when the latency growth rate is greater than 80%, such a sharp latency growth indicates that the system has deep-seated, structural fault risks. It is very likely that the key link between the management sub-platform and the main service database is interrupted, or the management sub-platform itself has a serious fault (such as database crash, core processor overheating and freezing), resulting in data transmission obstruction and processing stagnation. Based on this, the user platform accurately generates corresponding management sub-platform fault information to the user, guiding the user to quickly focus on the core area of ​​the fault and carry out targeted repair work.

[0028] Through the multi-interval judgment mechanism of the delay growth rate, the fault type can be refined from broad delay exceeding the standard to specific network anomalies, management sub-platform failures, etc., helping users to quickly identify the root cause of the problem in complex industrial Internet of Things systems, reduce the blindness of troubleshooting, and greatly improve maintenance efficiency. For example, in the Internet of Things system of a large-scale automated factory production line, once production stagnation, data update lag and other problems occur, users can quickly determine whether the network wiring is loose (corresponding to network anomalies) or the key management server is frozen (corresponding to management sub-platform failure) based on the fault direction feedback from the user platform, avoiding indiscriminate troubleshooting of the entire system.

[0029] Continuous unit-time monitoring and judgment make the user platform like a diligent "fault detective", which can grasp the whole process of the system from the occurrence to the development of the fault in real time. Whether the fault is relieved by itself, gradually worsened, or maintained in an unstable state, users can obtain information at the first time, adjust production strategies or deploy maintenance resources in time. For example, in the electronic chip packaging workshop, production has extremely high requirements for the real-time performance of environmental temperature and humidity. If the IoT system has a delay problem, through the real-time tracking of the user platform, the operator can decide whether to suspend production and urgently deploy spare temperature control equipment according to the dynamics of the delay growth rate to ensure that product quality is not affected.

[0030] Based on accurate fault location and real-time tracking, the enterprise management can make more scientific production decisions based on user platform feedback. When faced with small network fluctuations that can be tolerated for a short period of time (delay growth rate less than 30%), production can be maintained to avoid unnecessary downtime losses; when faced with serious fault hazards (delay growth rate greater than 80%), production can be stopped for maintenance to prevent the generation of a large number of defective products, further damage to equipment and other adverse consequences, effectively balancing production efficiency and product quality, and improving the economic benefits of the enterprise.

[0031] Furthermore, the local preliminary screening, caching, marking of urgent key data, and transmission to the sensor network platform after removing invalid data include the following: Conduct local preliminary screening of the collected data to remove irrelevant or redundant data; Cache filtered data and mark urgent and critical data; Remove invalid data, including erroneous data, extreme values, and missing data; The processed data is transmitted to the sensor network platform to form preliminary perception data; Among them, invalid data is removed by using a box plot to find outliers in the data and by replacing missing values; The MQTT protocol is used during the transmission process to adapt to low-bandwidth scenarios and achieve effective data transmission; Before data transmission, edge computing is performed to filter and aggregate the data for pre-processing.

[0032] Furthermore, the sensor network platform adaptively adjusts the sampling frequency including the following contents; Set the acquisition duration, initial sampling frequency, maximum sampling frequency and detection window length; Generate a control signal according to the acquisition duration, the initial sampling frequency, the maximum sampling frequency and the detection window length, the amplitude and the period of the excitation signal; and sending the excitation signal to the sensor; The sensor works according to the excitation signal and outputs an analog signal; The sensor network platform generates a control signal according to the acquisition duration, initial sampling frequency, maximum sampling frequency, detection window length, amplitude and period of the excitation signal, and sends the control signal to the excitation module; at the same time, the analog signal output by the sensor is preprocessed, and the sampling frequency is adjusted in real time by using an adaptive sampling frequency adjustment algorithm for data acquisition; The sensor network platform is also used to calculate the sampling data change rate and cumulative change in the detection window by calculating the number of sampling points. By sliding the time window, the data change rate and cumulative change ratio in adjacent time windows are processed accordingly to adjust the real-time sampling frequency. When the data changes faster, the sampling frequency is increased; when the data changes slower, the sampling frequency is reduced.

[0033] Furthermore, the intelligent traffic distribution system cache mechanism includes the following contents: The service sub-platform identifies and classifies the received preliminary perception data; The main service database intelligently selects the sub-service database for data routing based on real-time load and performance indicators; The service platform monitors the load of the sub-service database and dynamically adjusts the data flow to achieve load balancing; The sub-service database determines the data caching strategy based on the preset caching rules and intelligent algorithms; The service sub-platforms share cache status information through a collaborative mechanism and perform collaborative cache management; When the cached data is updated, the relevant service sub-platforms synchronously update their caches to maintain data consistency and real-time performance.

[0034] Furthermore, the intelligent offloading collaborative caching mechanism of the sub-service database and the service sub-platform includes: The sub-service database uses a preset hybrid intelligent algorithm of least recently used, least frequently used, and first-in-first-out to determine the data cache strategy; The service sub-platforms share cache status information through message queues or distributed cache systems to achieve consistency and collaborative management of cache data; The hybrid intelligent algorithm dynamically adjusts the cache strategy based on data access frequency, heat and time sensitivity; The collaborative cache management includes when a service sub-platform updates or deletes cache data, other service sub-platforms receive change notifications and update their own cache states to maintain data consistency; The cache strategy includes increasing the priority of frequently accessed data in the cache, giving priority to caching critical business data to reduce the caching of non-critical data, and setting a shorter cache time for data with high timeliness requirements.

[0035] Further, the sub-service database A receives the preliminary sensing data from the sensor network platform and decides to cache the most recent preliminary sensing data according to the preset caching rules and hybrid intelligent algorithm; The main service database monitors the load of all sub-service databases and intelligently routes the initial perception data to the sub-service database A with lower load; The service sub-platforms share cache status information through the distributed cache system. After receiving the cache status information of the sub-service database A, the service sub-platform B decides not to cache the data repeatedly. When a new batch of preliminary perception data arrives, sub-service database A updates its cache and notifies service sub-platform B. After receiving the update notification, service sub-platform B synchronously updates its cache view to maintain data consistency; The service platform monitors the load of all sub-service databases. When it finds that the load of sub-service database B suddenly increases, it dynamically adjusts the data flow and routes part of the data to sub-service database C with a lower load to achieve load balancing.

[0036] Furthermore, the intelligent offloading collaborative caching mechanism includes: The sub-service database uses a hybrid intelligent algorithm of the preset least recently used LRU algorithm, the least frequently used LFU algorithm, and the first-in first-out FIFO algorithm, combined with a machine learning model to predict the data cache strategy. The specific formula is: The LRU algorithm sorts data based on access time and eliminates the least recently used data; The LFU algorithm sorts data based on access frequency and eliminates the data with the lowest access frequency; The FIFO algorithm eliminates the earliest data according to the time order in which the data enters the cache; The hybrid intelligent algorithm dynamically adjusts the cache strategy according to data access frequency F, heat H and time sensitivity T. The calculation formula is: C = αF + βH + γT, where α, β, and γ are weight coefficients, and α + β + γ = 1; The machine learning model predicts the future access pattern and importance of data based on historical data and real-time data, and adjusts the weight coefficients α, β, and γ to optimize the cache strategy; The main service database intelligently selects the sub-service database for data routing based on the real-time load L and performance index P, combined with the load trend and performance changes predicted by the prediction model. The selection formula is: S = f ( L , P , P pred), where f is the comprehensive evaluation function of load and performance indicators, P pred is the predicted performance change; The service platform monitors the load of the sub-service database and dynamically adjusts the data flow to achieve load balancing. The adjustment formula is: Δ D = k ( L max- L min)+Δ P pred, where ΔD is the data flow adjustment amount, k is the adjustment coefficient, Lmax and Lmin are the maximum and minimum loads of the sub-service database, Δ P pred is the load adjustment caused by the predicted performance change; The service sub-platforms share cache status information through message queues or distributed cache systems to achieve consistency and collaborative management of cache data. The sharing formula is: S sync= S 1∪ S 2∪…∪ Sn ,in S sync is a shared cache status information collection, S 1, S 2, ..., Sn Cache status information for each service platform; When the cached data is updated, the relevant service sub-platforms synchronously update their caches to maintain data consistency and real-time performance. The update formula is: C new= C old∪Δ C ,in C new is the updated cache data, C old is the cache data before update, Δ C For newly added or modified data; The cache strategy also includes increasing the priority of frequently accessed data in the cache, caching key business data first to reduce the cache of non-key data, and setting a shorter cache time for data with high timeliness requirements. The specific formula is: P = f ( F , K , T , P pred), where P is the priority of the data in the cache, F is the data access frequency, K is the key indicator of the data, and T is the timeliness requirement of the data. P pred is the predicted data priority change, and f is the priority calculation function.

[0037] After collecting a large amount of raw data, the sensing unit of the object platform production device immediately starts the local preliminary screening process. Due to the complex production environment, sensors will capture various types of information, but not all data are valuable for subsequent decision-making and control. For example, in a machining workshop, sensors may simultaneously collect data such as the ambient temperature of the equipment casing, the brightness of the workshop lighting, and the vibration frequency of key equipment components. At this time, according to the preset production process correlation rules, identify and remove data such as lighting brightness that is not directly related to the equipment operating status and product processing accuracy or is redundant, and only retain the core information that can reflect the health status of the equipment and product quality.

[0038] The filtered data enters the cache stage. The cache is like a temporary data "transfer station". It not only stores data, but also marks it according to the urgency of the data and the degree of its critical impact on the production process. For example, for a chemical reactor, if the temperature and pressure data are close to the threshold of chemical reaction out of control, these data will be marked as urgent critical data so that the subsequent system can give priority to processing and ensure production safety. The box plot, a statistical tool, uses the quartile distribution characteristics of the data to accurately locate abnormal values ​​that fall outside the reasonable range. For example, in the production of electronic equipment, if the temperature of a certain solder point instantly reaches a maximum value far exceeding the normal welding temperature range, it can be determined as invalid data and eliminated; at the same time, for missing values, reasonable replacement methods are used, such as replacement based on the mean, median or specific model prediction value of historical data, to ensure data integrity and availability.

[0039] The MQTT protocol is used in the transmission process. This protocol is designed for low-bandwidth and unstable network environments. It adopts a publish / subscribe model and can achieve effective transmission of data from the sensing unit to the sensor network platform with minimal overhead under limited network resources, ensuring timely delivery of data and avoiding the loss of key information due to network congestion.

[0040] Before the data is officially transmitted, edge computing comes into play. It filters the data locally in the perception unit close to the source of data generation to remove duplicate and low-value information. At the same time, it aggregates and pre-processes the relevant data, such as aggregating the temperature data of similar devices in multiple time series into the average temperature and temperature change trend over a period of time, etc., reducing the amount of transmitted data, improving transmission efficiency, and alleviating the processing burden of subsequent platforms.

[0041] By removing irrelevant, redundant and invalid data, the amount of data that needs to be transmitted is greatly reduced, preventing network resources from being occupied by useless information, allowing limited bandwidth to be focused on transmitting key data, ensuring smooth data transmission and preventing network congestion. This is especially effective in large-scale industrial Internet of Things deployment scenarios when many devices are transmitting data at the same time.

[0042] Accurate screening, marking and invalid data processing mechanisms ensure that the data entering the subsequent process is highly credible and highly relevant, providing a solid foundation for subsequent data-based analysis and decision-making, reducing the risk of misjudgment due to erroneous or low-quality data, and improving the accuracy of decision-making in the entire industrial Internet of Things system. For example, in the field of precision manufacturing, high-quality data ensures high-precision control of product quality.

[0043] The sampling frequency can be flexibly adjusted according to the actual dynamics of the production site to ensure that sufficiently detailed data can be collected at critical moments (such as precursors to equipment failures, critical changes in process parameters), providing a detailed basis for fault diagnosis and process optimization, avoiding missing out on key information due to fixed sampling frequencies, and improving the accuracy of data descriptions of the production process. For example, in aircraft engine monitoring, subtle performance fluctuations can be accurately captured to provide early warning of potential fault hazards.

[0044] As an important node for data inflow, the service sub-platform quickly identifies and classifies data based on preset data type labels, source identifiers, and production process association rules when receiving preliminary perception data from the sensor network platform. For example, in an automobile manufacturing production line, the body welding quality monitoring data, engine assembly parameter data, and component paint thickness data are classified according to their respective process links to facilitate subsequent targeted processing.

[0045] The main service database monitors the load of each sub-service database in real time. The load information covers multi-dimensional indicators such as data storage capacity, processing task queue length, current processing rate, etc. At the same time, combined with performance indicators such as data read and write delay, system throughput, etc., the built-in intelligent algorithm (comprehensive evaluation function of load and performance indicators) calculates the optimal data routing solution, and intelligently distributes the initial perception data to the sub-service database with lower load and satisfactory performance, ensuring balanced and efficient data processing. For example, in the e-commerce logistics warehousing IoT system, inventory management data is reasonably distributed to different sub-service databases according to the load conditions such as the frequency of goods entering and leaving the warehouse in different areas of the warehouse and the amount of inventory counting tasks.

[0046] The service platform continuously monitors the load of the sub-service database. Once a load imbalance is found, such as when the load of several sub-service databases exceeds the average level due to a sudden influx of large amounts of data or concentrated processing tasks, the service platform immediately modifies the data flow rules based on the preset dynamic adjustment strategy to divert part of the data to the sub-service database with a lighter load, thereby achieving global load balancing and ensuring stable system operation.

[0047] The sub-service database determines the data caching strategy based on the preset caching rules, taking into account factors such as data access frequency, update frequency, and data importance, and combining intelligent algorithms (such as hybrid intelligent algorithms). For example, frequently queried product standard specification data and recent key parameters of production processes are cached to improve data acquisition speed and reduce database repeated query overhead. Service sub-platforms use a collaborative mechanism to build information sharing channels using message queues or distributed cache systems to exchange cache status information in real time, including cache data content, cache validity period, data popularity, etc. When a service sub-platform caches data, it immediately notifies other related service sub-platforms through a shared channel to synchronize cache updates to ensure data consistency and avoid data processing errors or duplication of work due to cache inconsistencies. The intelligent diversion and caching mechanism reduces the transmission delay and processing waiting time of data within the system. Each sub-service database processes the corresponding type of data based on its own advantages to avoid resource competition and centralized processing bottlenecks. For example, in a large-scale steel production Internet of Things system, data from different process links is diverted and processed to speed up production and improve overall production capacity.

[0048] Collaborative cache management ensures that no matter which service platform the data is obtained from, its content and timeliness remain consistent, avoiding production decision-making errors, equipment control conflicts and other problems caused by inconsistent data, and improving the reliability of the operation of the Industrial Internet of Things system. For example, in multi-robot collaborative operation scenarios, it ensures that each robot acts according to the same and latest task instruction data.

[0049] Hybrid intelligent algorithms are combined with machine learning models to accurately predict data needs, optimize cache configuration, significantly improve cache hit rates, reduce database query time, and increase data acquisition speed. For example, in a financial transaction risk control IoT system with large amounts of data, key risk indicator data can be quickly retrieved to ensure transaction timeliness and security. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 This is a process framework diagram of the data analysis and collaboration system of the industrial Internet of Things main service platform according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The following is further described in detail through specific implementation methods: The embodiment is basically as shown in the attached Figure 1 Shown: Data analysis and collaboration system of the main service platform of the Industrial Internet of Things, user platform, service platform, management platform, sensor network platform and object platform; The object platform includes a plurality of production devices, each of which is provided with a sensing unit and a control unit, wherein the sensing unit is used to collect physical quantities (including temperature and pressure) of equipment operation and product processing parameters, perform local preliminary screening and caching, mark urgent key data, remove invalid data and transmit to the sensor network platform to form preliminary sensing data; The sensor network platform includes multiple sensor network sub-platforms and corresponding sensor network databases, which are used to receive preliminary perception data, adaptively adjust the sampling frequency to prevent data transmission congestion, perform secondary screening and classification on the data, and store them in different databases according to rules, perform preliminary quantitative analysis to extract features, and then send the preliminary perception data after integration and analysis to the management platform; The management platform includes multiple management sub-platforms and an independent management database, which is used to receive the preliminary perception data, conduct in-depth interpretation based on the knowledge graph, judge the equipment operation status, predict the failure risk, and store the preliminary perception data together with the equipment status, failure prediction, and resource allocation plan in the management database, and send it to the service platform after forming management data; The service platform adopts an optimized combination and post-fractional layout, including a main service database and several sub-service databases. Each sub-service database is connected to a corresponding service sub-platform and is equipped with an intelligent shunting collaborative cache mechanism. The main service database receives management data and summarizes and preliminarily classifies it, and distributes it to the sub-service database according to type and priority. The service sub-platform processes and mines data value in parallel, generates refined data service products and stores them in the sub-service database; The user platform is used to receive data from the sub-service database, and to track and update the load status of the sub-service database and issue an overtime delay threshold warning, helping users obtain useful data to guide decision-making and adjust processes. The user platform is also used to modify the sub-service database data and feed back requirements and instructions to the service platform to form a closed-loop collaborative process.

[0052] The main service database or the management sub-platform is used to generate a timestamp while sending information to the user platform, and attach it to the information; the user platform is used to calculate the time delay between the time when the information from the sub-service database is actually received and the timestamp, and perform corresponding processing according to different ranges of the time delay, and the corresponding processing includes: When the latency is less than 30 milliseconds, the user platform displays to the user that the service sub-platform is working normally; When the latency is between 30 milliseconds and 800 milliseconds, the user platform displays service sub-platform network abnormality information to the user; When the latency exceeds 800 milliseconds, the user platform accesses the service sub-platform and enters fault handling mode.

[0053] The user platform is also used to perform the following steps in the fault handling mode: Calculate the delay growth rate between the starting delay and the final delay of the delay per unit time; When the delay growth rate is less than 0, the user platform continues to make judgments in the next unit time until the delay growth rate is not less than 0 or the fault handling mode is exited; When the latency growth rate is between 0 and 30%, the user platform continues to make judgments in the next unit time until the latency growth rate is not less than 0 or the fault handling mode is exited; When the latency growth rate is between 30% and 80%, the user platform generates service sub-platform network anomaly information to the user; When the delay growth rate is greater than 80%, the user platform generates fault information of the corresponding management sub-platform to the user.

[0054] Specifically, in actual industrial production scenarios, taking automobile manufacturing plants as an example, the object platform is distributed in various workshop production lines, including a large number of production devices such as automobile engine assembly equipment, body welding robots, and painting devices. Each production device is equipped with an advanced sensing unit, which integrates a variety of high-precision sensors. For example, the temperature sensor uses a platinum resistance thermometer, which can accurately measure the heating of key parts of the equipment, and the pressure sensor uses a high-precision strain gauge sensor to accurately capture the pressure changes of the hydraulic and pneumatic systems. Acceleration sensors, displacement sensors, etc. are used to obtain operating physical quantities, and there are also sensors using laser measurement, machine vision and other technologies to collect product processing parameters, such as the dimensional accuracy of body welds and the tolerance of engine cylinder bores.

[0055] During operation, the perception unit collects thousands of sets of raw data per second. First, through the built-in primary screening algorithm, according to the preset key indicator range of the automobile production process, the data that is not related to the normal operation of the equipment and the qualification of the product is quickly eliminated, such as the brightness of the workshop lighting, the slight vibration data of non-critical parts, etc. The screened valid data is stored in the local cache chip. The cache capacity is designed to store high-frequency collected data within 10 minutes according to the data generation rate of the equipment. At the same time, using the priority marking algorithm, once urgent key data such as engine overheating and abnormal welding current are detected, they are immediately marked as high priority to ensure priority transmission. For invalid data, the box plot is combined with the statistical characteristics of historical data in the automotive industry to identify and remove the maximum or minimum values ​​caused by temporary sensor failures, as well as the missing data caused by communication interference, and the missing values ​​are replaced by interpolation based on adjacent data or historical mean values ​​of the same period. Finally, the processed data is transmitted to the sensor network platform through the wireless access point in the workshop according to the MQTT protocol to form preliminary perception data. The transmission frequency is dynamically adjusted according to the urgency of the data and the network bandwidth of the workshop, up to 100 times per second.

[0056] The sensor network platform sets up multiple sensor network sub-platforms in different areas of the factory workshop. Each sub-platform is connected to a group of sensing units of production equipment nearby to receive preliminary sensing data from the object platform. Taking the body workshop as an example, the sensor network sub-platform connects to the sensing units of dozens of welding robots and assembly equipment.

[0057] The sub-platform has a built-in adaptive sampling frequency adjustment module, with the initial setting of acquisition duration of 5 minutes, initial sampling frequency of once every 10 seconds, maximum sampling frequency of once every 1 second and detection window length of 1 minute. The control signal is generated according to factors such as the real-time production task volume of the workshop and the stability of equipment operation. For example, in the trial production stage of new models, the equipment is frequently debugged and the probability of abnormality is high. The control signal drives the excitation module to increase the amplitude and period of the excitation signal, so that the sensor increases the sampling frequency; in the stable production period of mature models, the sampling frequency is appropriately reduced. At the same time, the analog signal output by the sensor is pre-processed with a high-speed ADC chip. After being converted into a digital signal, the data is secondary screened and classified according to the automobile production process and quality control nodes, such as storing the body welding quality data in the welding process database and the assembly accuracy data in the assembly process database. The data mining algorithm is used for preliminary quantitative analysis and feature extraction, such as statistical welding spot temperature fluctuation range, assembly gap change trend, etc. The preliminary perception data after integration and analysis is sent to the management platform through the workshop backbone network every 30 seconds.

[0058] The management platform is located in the factory's central control room and includes multiple management sub-platforms. Each management sub-platform is responsible for the management of one or several production lines. It is connected to an independent management database and stores all historical data, process standards, equipment maintenance manuals and other knowledge graphs related to the production line.

[0059] After receiving the preliminary perception data from the sensor network platform, the management sub-platform runs the intelligent analysis software based on the knowledge graph. Taking the engine production line as an example, by comparing the real-time temperature and pressure data with the temperature-pressure curve model of the normal operation of the engine, combined with the equipment's cumulative operating time, recent maintenance records and other information, the equipment's operating status can be accurately judged and potential failure risks can be predicted, such as 30 minutes in advance warning of the power drop problem that may be caused by piston ring wear. At the same time, according to the production order task volume and the real-time status of the equipment, a resource allocation plan is formulated, such as adjusting the order of material distribution and arranging the online time of spare equipment. The preliminary perception data, equipment status assessment, fault prediction, and resource allocation plan are packaged as management data and sent to the service platform every 2 minutes through the factory's internal high-speed local area network.

[0060] The service platform is set up in the factory's data center, using an optimized combination and post-fractional layout. The main service database is equipped with a high-performance server cluster, with massive storage and high-speed data processing capabilities, and is connected to multiple sub-service databases. The sub-service databases are divided according to the automobile production business modules, such as powertrain, body, chassis, electronic and electrical sub-service databases. Each sub-service database is connected to a corresponding service sub-platform equipped with a GPU accelerator card to achieve data parallel processing.

[0061] After receiving the management data from the management platform, the main service database completes the aggregation and preliminary classification within 10 seconds, and intelligently selects the sub-service database for routing allocation based on factors such as data business type, urgency, and priority of impact on production. For example, engine fault warning data is preferentially allocated to the powertrain sub-service database. The service sub-platform corresponding to the sub-service database uses deep learning algorithms to mine the value of data, such as building a fault prediction model based on historical engine fault data and real-time operating parameters, and generating refined data service products such as engine preventive maintenance recommendation reports and body paint quality optimization solutions, which are stored in the sub-service database for user platforms to call at any time.

[0062] The user platform is deployed on the office terminals of factory managers and engineers and handheld smart devices at the production site to facilitate information acquisition at any time. The user platform establishes a high-speed connection with the sub-service database and receives data from the sub-service database in real time.

[0063] Take the workshop supervisor as an example. The terminal he uses refreshes the data pushed by the sub-service database every 5 seconds, and runs the load monitoring software at the same time to track the load status of the sub-service database, such as monitoring data query response time and the number of concurrent data processing tasks. Once it is found that the delay between the actual time of receiving the information from the sub-service database and the timestamp attached to the information exceeds the threshold, the corresponding processing mechanism is immediately activated. When the delay is less than 30 milliseconds, the normal working information of the service sub-platform is displayed in a green icon on the terminal interface; when the delay is between 30 milliseconds and 800 milliseconds, a yellow warning box pops up to display the network abnormality information of the service sub-platform; when the delay exceeds 800 milliseconds, it automatically connects to the corresponding service sub-platform and enters the fault handling mode.

[0064] In the fault handling mode, the user platform calculates the delay growth rate between the starting delay and the final delay within a unit time (set to 1 second). If the workshop supervisor finds that the delay growth rate is less than 0, it means that the situation may be improving. The user platform continues to judge in the next unit time until the delay growth rate is stable and not less than 0 or meets other conditions for exiting the fault handling mode, such as the network restores normal bandwidth and the service sub-platform load drops back to the normal range; when the delay growth rate is between 0 and 30%, it also continues to monitor; when the delay growth rate is between 30 and 80%, a red warning message is pushed to the workshop supervisor to inform the service sub-platform of network abnormality information; when the delay growth rate is greater than 80%, the root cause of the problem is accurately located, and a serious fault alarm is pushed to the supervisor, indicating that the corresponding management sub-platform has fault information, and assisting him to quickly organize a maintenance team to check the key parts of the management sub-platform involved in the server, network link, database, etc.

[0065] Through the specific implementation methods of close coordination among the above platforms, the data analysis and coordination system of the main service platform of the Industrial Internet of Things can operate efficiently, effectively solving the technical problems of large data processing delays, poor production real-time performance, and low troubleshooting efficiency of the existing service platform, and ensuring the intelligent and efficient advancement of industrial production.

[0066] In other embodiments, local preliminary screening and caching are performed, urgent key data is marked, and invalid data is removed before transmission to the sensor network platform, including the following: Conduct local preliminary screening of the collected data to remove irrelevant or redundant data; Cache filtered data and mark urgent and critical data; Remove invalid data, including erroneous data, extreme values, and missing data; The processed data is transmitted to the sensor network platform to form preliminary perception data; Among them, invalid data is removed by using a box plot to find outliers in the data and by replacing missing values; The MQTT protocol is used during the transmission process to adapt to low-bandwidth scenarios and achieve effective data transmission; Before data transmission, edge computing is performed to filter and aggregate the data for pre-processing.

[0067] The sensor network platform adaptively adjusts the sampling frequency including the following contents; Set the acquisition duration, initial sampling frequency, maximum sampling frequency and detection window length; Generate a control signal according to the acquisition duration, the initial sampling frequency, the maximum sampling frequency and the detection window length, the amplitude and the period of the excitation signal; and sending the excitation signal to the sensor; The sensor works according to the excitation signal and outputs an analog signal; The sensor network platform generates a control signal according to the acquisition duration, initial sampling frequency, maximum sampling frequency, detection window length, amplitude and period of the excitation signal, and sends the control signal to the excitation module; at the same time, the analog signal output by the sensor is preprocessed, and the sampling frequency is adjusted in real time by using an adaptive sampling frequency adjustment algorithm for data acquisition; The sensor network platform is also used to calculate the sampling data change rate and cumulative change in the detection window by calculating the number of sampling points. By sliding the time window, the data change rate and cumulative change ratio in adjacent time windows are processed accordingly to adjust the real-time sampling frequency. When the data changes faster, the sampling frequency is increased; when the data changes slower, the sampling frequency is reduced.

[0068] When used specifically: The sensing unit is equipped with an advanced microprocessor and a built-in intelligent screening algorithm. For CNC machine tools, its sensors can collect hundreds of data points per second, covering information such as the displacement of each axis of the equipment, cutting force, spindle speed, motor current, and dimensional accuracy of the workpiece. The algorithm quickly identifies and removes irrelevant data based on preset production process rules. For example, data such as the intensity of ambient light in the workshop and slight vibration of the equipment casing will be eliminated at the first time because they have no direct correlation with the machining accuracy of the machine tool and the stability of the equipment operation. At the same time, redundant data cannot escape screening. For example, in the stable cutting stage, if multiple almost identical spindle speed values ​​are continuously collected, only representative samples are retained, which greatly reduces the amount of data for subsequent transmission and processing.

[0069] The filtered valid data is immediately stored in the local cache chip, and the cache capacity is dynamically allocated according to the device data generation rate and the urgency of the production task. Taking the automated assembly robot as an example, when performing a critical assembly task, if it detects that the assembly pressure of a key component is close to the preset upper limit, the microprocessor will immediately mark the pressure data and the related component identification code, assembly time and other information as urgent critical data to ensure priority processing in subsequent transmissions to ensure assembly quality and equipment safety.

[0070] When using the box plot method to remove outliers, take the cutting force data of CNC machine tools as an example. The system constructs a box plot model based on historical cutting force data and sets reasonable upper and lower quartile intervals. When the cutting force data collected in real time exceeds this interval, such as when the cutting force suddenly soars due to tool breakage, the outlier will be judged as invalid data and removed. For missing values, if the motor current value collected at a certain moment is missing, the system will replace it by linear interpolation or taking the mean of adjacent time periods based on the current value change trend at the previous moment and the current operating status of the machine tool to ensure the continuity and integrity of the data.

[0071] Before data transmission, the edge computing module plays a key role. Taking the CNC machine tool as an example, it will filter the cached data, remove repeated data that is not very meaningful for the current processing task, and aggregate and pre-process the relevant data. Aggregate the displacement data of each axis over a period of time (such as the past 10 seconds) into a displacement change curve, and statistically calculate the average, maximum and minimum values ​​of the cutting force and other characteristic quantities to reduce the dimension and total amount of transmitted data. The transmission process adopts the MQTT protocol. In the complex electromagnetic environment of the factory workshop and the presence of many low-bandwidth areas, the perception unit of the CNC machine tool acts as a publisher and publishes the processed data to the corresponding MQTT proxy server according to the topic (such as device number, data type, etc.). The sensor network platform receives it on demand as a subscriber to ensure that the data can be transmitted efficiently and stably, and avoid data loss or delay due to network congestion.

[0072] Taking the environmental monitoring sensor network sub-platform as an example, it is necessary to accurately collect environmental parameters such as temperature, humidity, and harmful gas concentration in the workshop. The initial collection duration is set to 30 minutes, aiming to obtain a relatively complete trend of environmental changes; the initial sampling frequency is set to once every 5 minutes, which can not only not over-occupy resources in the initial stage, but also initially capture environmental changes; the maximum sampling frequency is increased to once every 1 minute to respond to sudden emergencies such as harmful gas leaks. The detection window length is set to 5 minutes to facilitate timely analysis of data changes. At the same time, according to the operating status of the ventilation system in the workshop (such as fan speed, ventilation volume), whether the production process has special temperature and humidity requirements and other factors, the amplitude and period of the excitation signal are determined, and then the control signal is generated through the built-in signal generation algorithm to drive the external excitation module to generate the corresponding excitation signal and send it to the sensor.

[0073] Temperature and humidity sensors, harmful gas sensors, etc. start working after receiving the excitation signal and output analog signals. On the one hand, the sensor network sub-platform synchronously feeds back the control signal to the excitation module to dynamically adjust the working state of the sensor to ensure the acquisition accuracy. On the other hand, the analog signal output by the sensor is converted and preprocessed using a high-speed ADC chip to remove signal noise and calibrate the signal baseline. The adaptive sampling frequency adjustment algorithm is used to monitor data changes in real time. By calculating the sampling data change rate and cumulative change amount in the sampling point detection window, the sliding time window technology is used to compare the data change rate and cumulative change ratio in adjacent time windows. If the concentration of harmful gases in a certain area of ​​the workshop rises rapidly due to chemical leakage, the data change rate increases sharply, and the ratio far exceeds the normal threshold, the system will immediately increase the sampling frequency from once every 5 minutes to once every 1 minute, or even higher, to capture the detailed process of harmful gas diffusion; on the contrary, during the environmental stability period, such as the workshop shutdown period, the data changes slowly, and the system automatically reduces the sampling frequency to once every 10 minutes, saving resources and avoiding unnecessary data collection and storage.

[0074] In another embodiment, the intelligent traffic distribution system cache mechanism includes the following contents: The service sub-platform identifies and classifies the received preliminary perception data; The main service database intelligently selects the sub-service database for data routing based on real-time load and performance indicators; The service platform monitors the load of the sub-service database and dynamically adjusts the data flow to achieve load balancing; The sub-service database determines the data caching strategy based on the preset caching rules and intelligent algorithms; The service sub-platforms share cache status information through a collaborative mechanism and perform collaborative cache management; When the cached data is updated, the relevant service sub-platforms synchronously update their caches to maintain data consistency and real-time performance.

[0075] The intelligent offloading and collaborative caching mechanism of the sub-service database and the service sub-platform includes: The sub-service database uses a preset hybrid intelligent algorithm of least recently used, least frequently used, and first-in-first-out to determine the data cache strategy; The service sub-platforms share cache status information through message queues or distributed cache systems to achieve consistency and collaborative management of cache data; The hybrid intelligent algorithm dynamically adjusts the cache strategy based on data access frequency, heat and time sensitivity; The collaborative cache management includes when a service sub-platform updates or deletes cache data, other service sub-platforms receive change notifications and update their own cache states to maintain data consistency; The cache strategy includes increasing the priority of frequently accessed data in the cache, giving priority to caching critical business data to reduce the caching of non-critical data, and setting a shorter cache time for data with high timeliness requirements.

[0076] Sub-service database A receives the preliminary perception data from the sensor network platform and decides to cache the most recent preliminary perception data based on the preset cache rules and hybrid intelligent algorithm; The main service database monitors the load of all sub-service databases and intelligently routes the initial perception data to the sub-service database A with lower load; The service sub-platforms share cache status information through the distributed cache system. After receiving the cache status information of the sub-service database A, the service sub-platform B decides not to cache the data repeatedly. When a new batch of preliminary perception data arrives, sub-service database A updates its cache and notifies service sub-platform B. After receiving the update notification, service sub-platform B synchronously updates its cache view to maintain data consistency; The service platform monitors the load of all sub-service databases. When it finds that the load of sub-service database B suddenly increases, it dynamically adjusts the data flow and routes part of the data to sub-service database C with a lower load to achieve load balancing.

[0077] The intelligent offloading collaborative caching mechanism includes: The sub-service database uses a hybrid intelligent algorithm of the preset least recently used LRU algorithm, the least frequently used LFU algorithm, and the first-in first-out FIFO algorithm, combined with a machine learning model to predict the data cache strategy. The specific formula is: The LRU algorithm sorts data based on access time and eliminates the least recently used data; The LFU algorithm sorts data based on access frequency and eliminates the data with the lowest access frequency; The FIFO algorithm eliminates the earliest data according to the time order in which the data enters the cache; The hybrid intelligent algorithm dynamically adjusts the cache strategy according to data access frequency F, heat H and time sensitivity T. The calculation formula is: C = αF + βH + γT, where α, β, and γ are weight coefficients, and α + β + γ = 1; The machine learning model predicts the future access pattern and importance of data based on historical data and real-time data, and adjusts the weight coefficients α, β, and γ to optimize the cache strategy; The main service database intelligently selects the sub-service database for data routing based on the real-time load L and performance index P, combined with the load trend and performance changes predicted by the prediction model. The selection formula is: S = f ( L , P , P pred), where f is the comprehensive evaluation function of load and performance indicators, P pred is the predicted performance change; The service platform monitors the load of the sub-service database and dynamically adjusts the data flow to achieve load balancing. The adjustment formula is: Δ D = k ( L max- L min)+Δ P pred, where ΔD is the data flow adjustment amount, k is the adjustment coefficient, Lmax and Lmin are the maximum and minimum loads of the sub-service database, Δ P pred is the load adjustment caused by the predicted performance change; The service sub-platforms share cache status information through message queues or distributed cache systems to achieve consistency and collaborative management of cache data. The sharing formula is: S sync= S 1∪ S 2∪…∪ Sn ,in S sync is a shared cache status information collection, S 1, S 2, ..., Sn Cache status information for each service platform; When the cached data is updated, the relevant service sub-platforms synchronously update their caches to maintain data consistency and real-time performance. The update formula is: C new= C old∪Δ C ,in Cnew is the updated cache data, C old is the cache data before update, Δ C For newly added or modified data; The cache strategy also includes increasing the priority of frequently accessed data in the cache, caching key business data first to reduce the cache of non-key data, and setting a shorter cache time for data with high timeliness requirements. The specific formula is: P = f ( F , K , T , P pred), where P is the priority of the data in the cache, F is the data access frequency, K is the key indicator of the data, and T is the timeliness requirement of the data. P pred is the predicted data priority change, and f is the priority calculation function.

[0078] When used specifically: The service sub-platform receives preliminary perception data from the sensor network platform, which contains key information such as the operating status of the equipment and environmental parameters. The service sub-platform identifies and classifies the data according to its type and characteristics, for example, classifying and storing temperature data, pressure data, vibration data, etc. for subsequent processing and analysis.

[0079] The main service database monitors the load of all sub-service databases in real time, including performance indicators such as data processing volume, storage space occupancy, and response time. When new preliminary perception data needs to be stored, the main service database will intelligently select a sub-service database with a lower load for data routing based on the current load and performance indicators. For example, if the load of sub-service database A is lower than that of other sub-service databases, the main service database will route the data to sub-service database A, thereby achieving load balancing and preventing a sub-service database from affecting data processing speed and system stability due to excessive load.

[0080] The sub-service database determines the data cache strategy based on the preset cache rules and hybrid intelligent algorithms. The hybrid intelligent algorithm combines the least recently used (LRU), least frequently used (LFU) and first-in-first-out (FIFO) algorithms, comprehensively considers factors such as data access frequency, popularity and time sensitivity, and dynamically adjusts the cache strategy. For example, data with high access frequency, such as real-time monitoring data, will be cached first; key business data, such as equipment failure warning data, will also be cached first to ensure fast access to key information; and for data with high timeliness requirements, such as production progress update data, a shorter cache time will be set to ensure the real-time nature of the data.

[0081] The machine learning model plays an important role in the formulation of cache strategies. It predicts the future access pattern and importance of data based on historical data and real-time data, thereby adjusting the weight coefficients α, β, and γ in the hybrid intelligent algorithm. For example, if the model predicts that the probability of failure of a certain device will increase, the priority of the data related to the device in the cache will be increased to ensure that the relevant data can be quickly obtained for processing when a failure occurs.

[0082] The service sub-platforms share cache status information through message queues or distributed cache systems to achieve consistency and collaborative management of cached data. When a service sub-platform updates or deletes cached data, it sends a change notification to other service sub-platforms. For example, after service sub-platform B receives the cache status information of sub-service database A, if it finds that sub-service database A has already cached some data, then service sub-platform B will decide not to cache the data repeatedly to avoid data redundancy and waste of storage space.

[0083] When cached data is updated, the relevant service sub-platforms will synchronously update their cached views to maintain data consistency and real-time performance. For example, when a new batch of preliminary perception data arrives, sub-service database A will update its cache and notify service sub-platform B. After receiving the update notification, service sub-platform B will synchronize the new data to its own cache to ensure that all service sub-platforms can access the latest data.

[0084] The service platform continuously monitors the load of all sub-service databases. When the load of a sub-service database is found to increase suddenly, for example, the load of sub-service database B exceeds the preset threshold, the service platform will dynamically adjust the data flow and route part of the data to sub-service database C with lower load. This adjustment is based on load balancing considerations and aims to optimize the overall performance of the system and prevent the operating efficiency of the entire system from being affected by the overload of a single sub-service database.

[0085] The above is only an embodiment of the present invention, and the common knowledge such as the known specific structure and characteristics in the scheme is not described in detail here. It should be pointed out that for those skilled in the art, several deformations and improvements can be made without departing from the structure of the present invention, which should also be regarded as the protection scope of the present invention, and these will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. Industrial Internet of Things main service platform data analysis and coordination system, including: The user platform, service platform, management platform, sensor network platform and object platform are characterized by: The object platform includes a plurality of production devices, each of which is provided with a sensing unit and a control unit, wherein the sensing unit is used to collect physical quantities of equipment operation and product processing parameters, perform local preliminary screening and caching, mark urgent key data, remove invalid data and transmit to the sensor network platform to form preliminary sensing data; The sensor network platform includes multiple sensor network sub-platforms and corresponding sensor network databases, which are used to receive preliminary perception data, adaptively adjust the sampling frequency to prevent data transmission congestion, perform secondary screening and classification on the data, and store them in different databases according to rules, perform preliminary quantitative analysis and extract features, and send the preliminary perception data after integration and analysis to the management platform; The management platform includes multiple management sub-platforms and an independent management database, which is used to receive the preliminary perception data, conduct in-depth interpretation based on the knowledge graph, judge the equipment operation status, predict the failure risk, and store the preliminary perception data together with the equipment status, failure prediction, and resource allocation plan in the management database, and send it to the service platform after forming management data; The service platform adopts an optimized combination and post-fractional layout, including a main service database and several sub-service databases. Each sub-service database is connected to a corresponding service sub-platform and is equipped with an intelligent shunting collaborative cache mechanism. The main service database receives management data and summarizes and preliminarily classifies it, and distributes it to the sub-service database according to type and priority. The service sub-platform processes and mines data value in parallel, generates refined data service products and stores them in the sub-service database; The user platform is used to receive data from the sub-service database, and to track and update the load status of the sub-service database and issue an overtime delay threshold warning, helping users obtain useful data to guide decision-making and adjust processes. The user platform is also used to modify the sub-service database data and feed back requirements and instructions to the service platform to form a closed-loop collaborative process.

2. The data analysis and collaboration system of the main service platform of the industrial Internet of Things according to claim 1 is characterized by: The main service database or the management sub-platform is used to generate a timestamp while sending information to the user platform, and attach it to the information; the user platform is used to calculate the time delay between the time when the information from the sub-service database is actually received and the timestamp, and perform corresponding processing according to different ranges of the time delay, and the corresponding processing includes: When the latency is less than 30 milliseconds, the user platform displays information to the user that the service sub-platform is working normally; When the latency is between 30 milliseconds and 800 milliseconds, the user platform displays service sub-platform network abnormality information to the user; When the delay exceeds 800 milliseconds, the user platform accesses the service sub-platform and enters the fault handling mode.

3. The data analysis and collaboration system of the main service platform of the industrial Internet of Things according to claim 2 is characterized by: The user platform is further configured to perform the following steps in the fault handling mode: Calculate the delay growth rate between the starting delay and the final delay of the delay per unit time; When the delay growth rate is less than 0 or between 0 and 30%, the user platform continues to make judgments in the next unit time until the delay growth rate is not less than 0 or the fault handling mode is exited; When the latency growth rate is between 30% and 80%, the user platform generates service sub-platform network anomaly information to the user; When the delay growth rate is greater than 80%, the user platform generates fault information of the corresponding management sub-platform to the user.

4. The data analysis and collaboration system of the main service platform of the industrial Internet of Things according to claim 3 is characterized by: The local preliminary screening, caching, marking of urgent key data, removal of invalid data and transmission to the sensor network platform include the following: Conduct local preliminary screening of the collected data to remove irrelevant or redundant data; Cache filtered data and mark urgent and critical data; Remove invalid data, including erroneous data, extreme values, and missing data; The processed data is transmitted to the sensor network platform to form preliminary perception data; Among them, invalid data is removed by using a box plot to find outliers in the data and by replacing missing values; The MQTT protocol is used during the transmission process to adapt to low-bandwidth scenarios and achieve effective data transmission; Before data transmission, edge computing is performed to filter and aggregate the data for pre-processing.

5. The data analysis and collaboration system of the main service platform of the industrial Internet of Things according to claim 4 is characterized by: The sensor network platform adaptively adjusts the sampling frequency including the following contents: Set the acquisition duration, initial sampling frequency, maximum sampling frequency and detection window length; Generate a control signal according to the acquisition duration, the initial sampling frequency, the maximum sampling frequency and the detection window length, the amplitude and the period of the excitation signal; and sending the excitation signal to the sensor; The sensor works according to the excitation signal and outputs an analog signal; The sensor network platform generates a control signal according to the acquisition duration, initial sampling frequency, maximum sampling frequency, detection window length, amplitude and period of the excitation signal, and sends the control signal to the excitation module; at the same time, the analog signal output by the sensor is preprocessed, and the sampling frequency is adjusted in real time by using an adaptive sampling frequency adjustment algorithm for data acquisition; The sensor network platform is also used to calculate the sampling data change rate and cumulative change in the detection window by calculating the number of sampling points. By sliding the time window, the data change rate and cumulative change ratio in adjacent time windows are processed accordingly to adjust the real-time sampling frequency. When the data changes faster, the sampling frequency is increased; when the data changes slower, the sampling frequency is reduced.

6. The data analysis and collaboration system of the main service platform of the industrial Internet of Things according to claim 5 is characterized by: The intelligent offloading collaborative caching mechanism includes the following contents: The service sub-platform identifies and classifies the received preliminary perception data; The main service database intelligently selects the sub-service database for data routing based on real-time load and performance indicators; The service platform monitors the load of the sub-service database and dynamically adjusts the data flow to achieve load balancing; The sub-service database determines the data caching strategy based on the preset caching rules and intelligent algorithms; The service sub-platforms share cache status information through a collaborative mechanism and perform collaborative cache management; When the cached data is updated, the relevant service sub-platforms synchronously update their caches to maintain data consistency and real-time performance.

7. The data analysis and collaboration system of the main service platform of the industrial Internet of Things according to claim 6 is characterized by: The intelligent offloading and collaborative caching mechanism of the sub-service database and the service sub-platform includes: The sub-service database uses a preset hybrid intelligent algorithm of least recently used, least frequently used, and first-in-first-out to determine the data cache strategy; The service sub-platforms share cache status information through message queues or distributed cache systems to achieve consistency and collaborative management of cache data; The hybrid intelligent algorithm dynamically adjusts the cache strategy based on data access frequency, heat and time sensitivity; The collaborative cache management includes when a service sub-platform updates or deletes cache data, other service sub-platforms receive change notifications and update their own cache states to maintain data consistency; The cache strategy includes increasing the priority of frequently accessed data in the cache, giving priority to caching critical business data to reduce the caching of non-critical data, and setting a shorter cache time for data with high timeliness requirements.

8. The data analysis and collaboration system of the main service platform of the industrial Internet of Things according to claim 7 is characterized by: Sub-service database A receives the preliminary perception data from the sensor network platform and decides to cache the most recent preliminary perception data based on the preset cache rules and hybrid intelligent algorithm; The main service database monitors the load of all sub-service databases and intelligently routes the initial perception data to the sub-service database A with lower load; The service sub-platforms share cache status information through the distributed cache system. After receiving the cache status information of the sub-service database A, the service sub-platform B decides not to cache the data repeatedly. When a new batch of preliminary perception data arrives, sub-service database A updates its cache and notifies service sub-platform B. After receiving the update notification, service sub-platform B synchronously updates its cache view to maintain data consistency; The service platform monitors the load of all sub-service databases. When it finds that the load of sub-service database B suddenly increases, it dynamically adjusts the data flow and routes part of the data to sub-service database C with a lower load to achieve load balancing.

9. The data analysis and collaboration system of the main service platform of the industrial Internet of Things according to claim 8 is characterized by: The intelligent offloading collaborative caching mechanism includes: The sub-service database uses a hybrid intelligent algorithm of the preset least recently used LRU algorithm, the least frequently used LFU algorithm, and the first-in first-out FIFO algorithm, combined with a machine learning model to predict the data cache strategy. The specific formula is: The LRU algorithm sorts data based on access time and eliminates the least recently used data; The LFU algorithm sorts data based on access frequency and eliminates the data with the lowest access frequency; The FIFO algorithm eliminates the earliest data according to the time order in which the data enters the cache; The hybrid intelligent algorithm dynamically adjusts the cache strategy according to data access frequency F, heat H and time sensitivity T. The calculation formula is: C = αF + βH + γT, where α, β, and γ are weight coefficients, and α + β + γ = 1; The machine learning model predicts the future access pattern and importance of data based on historical data and real-time data, and adjusts the weight coefficients α, β, and γ to optimize the cache strategy; The main service database intelligently selects the sub-service database for data routing based on the real-time load L and performance index P, combined with the load trend and performance changes predicted by the prediction model. The selection formula is: S = f ( L , P , P pred), where f is the comprehensive evaluation function of load and performance indicators, P pred is the predicted performance change; The service platform monitors the load of the sub-service database and dynamically adjusts the data flow to achieve load balancing. The adjustment formula is: Δ D = k ( L max- L min)+Δ P pred, where ΔD is the data flow adjustment amount, k is the adjustment coefficient, Lmax and Lmin are the maximum and minimum loads of the sub-service database, Δ P pred is the load adjustment caused by the predicted performance change; The service sub-platforms share cache status information through message queues or distributed cache systems to achieve consistency and collaborative management of cache data. The sharing formula is: S sync= S 1∪ S 2∪…∪ Sn ,in S sync is a shared cache status information collection, S 1, S 2, ..., Sn Cache status information for each service platform; When the cached data is updated, the relevant service sub-platforms synchronously update their caches to maintain data consistency and real-time performance. The update formula is: C new= C old∪Δ C ,in C new is the updated cache data, C old is the cache data before update, Δ C For newly added or modified data; The cache strategy also includes increasing the priority of frequently accessed data in the cache, giving priority to caching key business data to reduce the cache of non-key data, and setting a shorter cache time for data with high timeliness requirements. The specific formula is: P = f ( F , K , T , P pred), where P is the priority of the data in the cache, F is the data access frequency, K is the key indicator of the data, and T is the timeliness requirement of the data. P pred is the predicted performance change, and f is the priority calculation function.

10. A data analysis and collaboration method for an industrial Internet of Things main service platform, characterized in that: A system as claimed in any one of claims 1 to 9 is employed.

Images