Multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios

Abstract

This invention relates to the field of Industrial Internet of Things (IIoT) and discloses a multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios. The system includes a configuration management module for parsing encrypted configuration files and dynamically loading communication protocol drivers using a reflection mechanism; a data acquisition module for establishing connections with industrial field devices and acquiring raw data using the loaded communication protocol drivers; a data processing module for calculating the semantic entropy value of the raw data within a time window and marking data value levels based on the semantic entropy value; a data storage module for performing differentiated storage operations on the data according to the data value levels; a memory management module; and a data forwarding module. By constructing a data acquisition module with dual-mode hardware and software interfaces and combining it with an adaptive driver correction mechanism based on physical layer signal feature analysis, the system achieves the effect of maintaining communication connections even when standard protocol drivers fail or device aging parameters drift.

Description

Technical Field

[0001] This invention relates to the field of industrial Internet of Things (IoT) technology, specifically to a multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios. Background Technology

[0002] With the advancement of industrial digital transformation, the application of Industrial Internet of Things (IIoT) technology in the manufacturing process is becoming increasingly sophisticated. Edge acquisition devices, acting as the hub connecting the field control layer and the upper-level management system, undertake the task of connecting heterogeneous devices and aggregating data. In actual industrial settings, there are numerous types of production equipment, with inconsistent communication interfaces and protocol standards, and the equipment spans a wide range of years since its introduction. This places high demands on the compatibility and stability of the acquisition system.

[0003] Existing data acquisition systems typically rely on pre-built standard protocol libraries for device interoperability when dealing with complex industrial environments. However, industrial sites contain numerous older or non-standard customized devices with long service lives. Due to physical factors such as component aging, these devices are highly susceptible to low-level signal characteristic drift, such as baud rate offset and unstable signal levels. Traditional data acquisition drivers mostly perform application-layer parsing based on fixed logic rules, lacking the ability to perceive and adaptively adjust physical layer signals. If even a slight deviation occurs between the low-level signal characteristics and standard protocol specifications, the driver often directly determines communication failure, leading to data acquisition link interruption and making it difficult to guarantee connection stability under multi-protocol compatibility.

[0004] Furthermore, high-frequency data acquisition in industrial settings generates massive amounts of time-series data, while edge devices typically have limited storage space and network bandwidth. Existing data processing solutions often employ simple first-in-first-out (FIFO) or random discarding mechanisms to address resource bottlenecks, failing to differentiate the inherent information value of the data. During stable equipment operation, a large amount of redundant data consumes valuable cache resources. When equipment malfunctions and causes sudden data fluctuations, if network congestion or storage overflow occurs, this critical abnormal data is easily discarded or its upload delayed. This lack of intelligent hierarchical data processing reduces the response efficiency of the operation and maintenance system to abnormal conditions and fails to meet the timeliness requirements of intelligent operation and maintenance for critical information.

[0005] Meanwhile, industrial edge gateway devices are typically deployed in enclosed control cabinets, lacking convenient human-machine interface peripherals. Existing system configurations and protocol extensions often rely on complex online debugging or full firmware flashing. When new device models or new communication protocols need to be connected, maintenance personnel usually need to recompile and update the entire system firmware, which not only increases the time cost of on-site deployment but also introduces the risk of service interruption due to firmware version changes. Therefore, there is an urgent need for a data acquisition and maintenance system that can adapt to changes in underlying signals, possesses intelligent resource scheduling capabilities, and is easy to expand and maintain. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios. It solves the problems of poor communication compatibility and data disconnection caused by non-standard or aging equipment parameters drift in industrial scenarios, as well as the loss or transmission delay of key fault information due to lack of data value assessment when edge computing resources are limited.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios, comprising a configuration management module, a data acquisition module, a data processing module, a data storage module, a memory management module, and a data forwarding module.

[0008] The configuration management module parses encrypted configuration files and dynamically loads communication protocol drivers using reflection. Through this module, the system monitors external storage media access, reads and decrypts encrypted configuration files to obtain protocol type identifiers, and then retrieves locally stored dynamic link library files based on these identifiers. The system uses reflection to read the assembly metadata of the dynamic link library files, locates and instantiates driver classes that implement the standard driver interface, and assigns the instantiated driver class object to the driver interface variable, thus completing the dynamic loading of the communication protocol driver. Protocol extensions can be implemented without recompiling the system firmware.

[0009] The data acquisition module connects to the configuration management module and is used to establish a connection with industrial field devices and acquire raw data using the loaded communication protocol driver. To address communication anomalies caused by aging or non-standard customization of industrial field devices, the data acquisition module executes adaptive driver correction logic. When data acquisition using the communication protocol driver fails, the module acquires the physical signal waveform sequence of the communication port at a sampling frequency higher than a preset multiple of the communication baud rate, and extracts features from the physical signal waveform sequence to construct a signal feature vector. Subsequently, the system matches the signal feature vector with a preset fingerprint database template, determines the communication deviation mode based on the matching result, generates driver patch parameters, and uses the driver patch parameters to correct the protocol parsing logic to reacquire data.

[0010] During the fingerprint database template matching process, the system determines the matching result by calculating quantitative indicators. Specifically, it calculates the square of the difference between the currently collected physical feature component and the standard feature value in the fingerprint database template, performs a weighted summation of the squares of the differences in each dimension and takes the square root, takes the reciprocal of the square root result plus one as the similarity confidence score, and selects the template with the highest similarity confidence score as the matching result.

[0011] The data processing module is connected to the data acquisition module and is used to calculate the semantic entropy value of the raw data within a time window, and to label the data value level based on the semantic entropy value. This module introduces information theory and statistical methods to quantify data value. The process of calculating the semantic entropy value includes: calculating the probability distribution of the data sequence falling into discrete value intervals within the statistical time window; calculating the negative sum of the product of the probability of occurrence of each discrete value interval and the natural logarithm of the probability of occurrence, as the basic information entropy; calculating the absolute value of the difference between the data at the current time and the data at the previous time, and using the ratio of this difference to the standard deviation of the data sequence as the normalized mutation term; adding the basic information entropy to the weighted normalized mutation term to obtain the semantic entropy value. Based on this, the system compares the calculated semantic entropy value with a preset high entropy threshold. If the semantic entropy value is greater than or equal to the high entropy threshold, the data is marked as high-value data; if the semantic entropy value is less than the high entropy threshold, the data is marked as low-value data.

[0012] The data storage module is connected to the data processing module and is used to perform differentiated storage operations on data based on its value level. For data marked as high-value data, a lossless compression algorithm is used to write it into the core table of the database to preserve complete fault characteristics. For data marked as low-value data, statistical dimensionality reduction processing is performed before writing it into the circular buffer table of the database. When the circular buffer table is full, the earliest written data record is overwritten, thereby optimizing the limited storage resources at the edge.

[0013] The memory management module connects to the data processing module to monitor memory resources and calculate the eviction priority of cached data based on semantic entropy values, then performs memory cleanup. To prevent critical fault data from being accidentally deleted when memory overflows, the system runs an entropy-weighted resource scheduling algorithm. The specific steps are as follows: obtain the difference between the current system time and the last access time of the data cache block as the idle time; obtain the semantic entropy value corresponding to the data cache block; calculate the ratio of the idle time to the sum of the weighted semantic entropy values, and use this ratio as the eviction score; determine the cleanup order of cached data according to the eviction score from high to low, ensuring that low-value and long-term idle data is cleaned up first.

[0014] The data forwarding module connects to the data processing and data storage modules, and is used to send data to the upper-layer management system based on the transmission priority determined by the semantic entropy value. This module has a resume capability; after the network connection is restored, it iterates through the data packets in the queue, calculates a transmission priority score by weighting the semantic entropy value and freshness of each data packet, where the freshness of the data packet is inversely proportional to the elapsed time since its generation. The system reorders the queue of data packets in descending order according to the transmission priority score, and prioritizes sending the data packets at the head of the queue to ensure that high-value data is reported first. Furthermore, before data transmission, the data forwarding module uses a symmetric encryption algorithm to encrypt the data payload and calculates the message digest of the encrypted data packet, attaching the message digest to the message header for integrity verification.

[0015] This invention provides a multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios. It offers the following advantages:

[0016] 1. This invention achieves the effect of maintaining communication connectivity even when standard protocol drivers fail or device aging parameters drift by constructing a data acquisition module with dual-mode hardware and software interfaces and combining it with an adaptive driver correction mechanism based on physical layer signal feature analysis. This solution utilizes a high-speed analog-to-digital converter to capture the underlying waveform and matches it with a fingerprint library to dynamically generate driver patches, solving the problems of poor multi-protocol compatibility and data disconnection caused by non-standard customization or signal distortion in IoT acquisition processes for industrial scenarios, thus ensuring broad adaptability of the underlying acquisition.

[0017] 2. This invention introduces a data value assessment algorithm based on semantic entropy, and accordingly executes differentiated storage, entropy-weighted LRU memory cleaning, and high-entropy-value-first breakpoint resumption strategies, achieving intelligent data stream classification and critical information delivery at the edge. This solution addresses the pain point of traditional acquisition systems failing to identify and prioritize fault symptom data when network congestion or hardware resource constraints occur, improving the system's response speed to abnormal operating conditions and demonstrating the system's intelligent operation and maintenance capabilities at the resource scheduling level.

[0018] 3. This invention employs encrypted configuration parsing based on external physical media and a driver-driven dynamic loading architecture based on reflection technology, combined with RBAC permission management and full-link log auditing, to achieve rapid network deployment and new protocol expansion without the need for external screen devices. This solution addresses the difficulties in interaction caused by harsh industrial environments and the maintenance bottleneck of recompiling firmware when adding new supported protocols in traditional systems, thus meeting the urgent needs of multi-protocol compatible systems for rapid deployment and secure operation and maintenance in practical applications. Attached Figure Description

[0019] Figure 1This is a system module architecture diagram of the present invention. Detailed Implementation

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please refer to the appendix. Figure 1 This invention provides a multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios. The system is built on a distributed edge computing architecture and is mainly implemented through edge gateway devices.

[0022] The system includes a configuration management module 10, a data acquisition module 20, a data processing module 30, a data storage module 40, a memory management module 50, a data forwarding module 60, a security management module 70, and a dashboard display module 80.

[0023] The configuration management module 10 is used to manage the initialization parameters and driver loading of the edge gateway device. The configuration management module 10 is equipped with a universal serial bus interface listener program to detect the access status of the external storage medium. The configuration management module 10 is configured to read the encrypted configuration file in the external storage medium, parse the network address, location table, and protocol parameters, and dynamically load the corresponding dynamic link library file as the driver program according to the configuration file, thereby achieving device initialization without a display screen.

[0024] The data acquisition module 20 is used to establish communication connections with industrial field equipment and acquire raw data. The data acquisition module 20 supports both PLC software acquisition and hardware signal acquisition modes, and integrates Modbus, OPC UA, and MQTT protocol stacks. The data acquisition module 20 has physical layer signal analysis capabilities. When standard protocol-driven acquisition fails, the data acquisition module 20 captures the physical electrical signal characteristics of the communication link, uses feature fingerprint matching technology to identify the device's communication mode, and dynamically corrects the drive logic to restore data access.

[0025] The data processing module 30 is used to clean and value-assess the collected raw data. The data processing module 30 runs lightweight artificial intelligence algorithms to perform data deduplication and anomaly detection. The data processing module 30 is further configured to calculate the semantic entropy value of the data sequence within a specific time window, and to label the data as high-value or low-value data based on the semantic entropy value, providing a basis for subsequent storage and transmission decisions.

[0026] The data storage module 40 manages read and write operations on the local SQLite database. The data storage module 40 implements a differentiated storage strategy, performing lossless compression storage on data marked as high-value by the data processing module 30, and lossy compression storage on low-value data. The data storage module 40 is configured with an automatic backup mechanism for cyclic overwriting and persistent data storage.

[0027] The memory management module 50 is used to monitor the system's memory resource usage. The memory management module 50 runs an entropy-weighted least recently used algorithm, calculating an eviction score based on the semantic entropy value of the data and its access time, dynamically cleaning up low-value cached data in memory to prevent the system from stopping due to memory overflow.

[0028] The data forwarding module 60 is used to send the processed data to the upper-layer management system. The data forwarding module 60 calls the AES-256 hardware encryption chip to encrypt the data payload and generates a message digest for integrity verification. The data forwarding module 60 has a breakpoint resumption function; after network interruption is recovered, it reorders the cache queue according to the semantic entropy value of the data, prioritizing the transmission of high-value data.

[0029] The security management module 70 is used for system access control and auditing. The security management module 70 enforces role-based access control policies and records operation logs for configuration changes, driver loading, and data transmission.

[0030] The Kanban display module 80 provides a visual operation and maintenance interface. It renders the device's operating status, real-time data streams, and key temperature indicators through chart components and provides a report export interface.

[0031] After the system starts, the configuration management module 10 executes the initialization process. After detecting the connection of a Universal Serial Bus device containing a configuration file, the configuration management module 10 automatically parses the file content and sets the system parameters, while loading the protocol driver plugin adapted to the target device.

[0032] After initialization, the data acquisition module 20 executes the data acquisition task. The data acquisition module 20 polls the industrial field devices using the loaded driver. If a communication anomaly occurs, the data acquisition module 20 initiates physical layer feature analysis, extracts signal waveform features and matches them with a pre-set fingerprint database, generates a driver patch to correct the parsing logic, and completes data access.

[0033] The raw data is transmitted to the data processing module 30. The data processing module 30 performs deduplication on the data and calculates the semantic entropy of the data stream. Based on the calculation results, the data processing module 30 marks data containing fault symptoms or sudden state changes as high-value data and data in a stable state as low-value data.

[0034] Subsequently, the data storage module 40 performs hierarchical storage based on the value of the data. High-value data is written to the lossless storage area, while low-value data is written to the circular storage area. Simultaneously, the memory management module 50 monitors the memory level in real time and dynamically releases cached low-value data that has not been accessed for a long time based on the data entropy value.

[0035] The data forwarding module 60 encapsulates the stored or processed data. It encrypts the data using an AES-256 hardware encryption chip and sends it to the upper-layer management system via network protocols. When a network connection interruption is detected, the data is temporarily stored in a local database. After the network is restored, the data forwarding module 60 scans the local cache, prioritizes extracting data packets with high semantic entropy values ​​for transmission, and then retransmits low-entropy data.

[0036] Throughout the system's operation, the dashboard display module 80 updates the device status view in real time. The security management module 70 records all operations and system events in the background, ensuring traceability throughout the entire process.

[0037] The configuration management module 10 is mainly used to configure the initialization parameters of the edge gateway device and dynamically load the communication protocol driver. Addressing the common problem of configuration difficulties in industrial settings due to a lack of display peripherals, this embodiment employs a configuration file parsing mechanism based on external physical media and a plug-in driver architecture based on reflection technology.

[0038] The configuration management module 10 automatically runs the USB interface listening service after system startup. This service monitors the hardware connection status of the USB port in real time by calling the operating system's underlying system message interfaces (such as the WM_DEVICECHANGE message in Windows or Netlink sockets in Linux). When an external storage device is detected, the configuration management module 10 automatically mounts the device and traverses its root directory to retrieve the encrypted configuration file with a preset filename. To ensure the security of the configuration data, the configuration management module 10 uses a symmetric decryption algorithm to decrypt the read file stream and verifies whether the digital signature in the file header matches. If the verification passes, the subsequent parsing process is executed.

[0039] The encrypted configuration file organizes data using Extensible Markup Language (XML) or JavaScript Object Notation (JSON) format. This file contains a network configuration node and a channel configuration node. The network configuration node defines the static IP address, subnet mask, gateway address, and domain name resolution server address of the edge gateway device. The channel configuration node defines the connection parameters of the device to be collected, including the protocol type identifier, device IP address, port number, slave address, and point mapping table. The point mapping table further details the register address, data type (e.g., floating-point, integer, boolean), and collection frequency of each physical point.

[0040] The configuration management module 10 maps the parsed parameters to a configuration object model in system memory. For network parameters, the configuration management module 10 calls the operating system's network interface controller API for real-time setting, updating the network status without restarting the system. After configuration, the configuration management module 10 controls the status indicator lights on the edge gateway device panel to execute preset flashing logic, such as controlling the LEDs to switch from a slow flashing red light to a solid green light, providing feedback to on-site personnel that the configuration is complete.

[0041] To address the need for multi-protocol compatibility, the configuration management module 10 executes a driver loading process based on plug-in technology. This process specifically includes the following steps:

[0042] Step S201: Establish a standard driver interface contract.

[0043] The system predefines a unified driver interface (e.g., named IDriverInterface), which declares abstract methods such as OpenConnection (establish connection), ReadTag (read point), WriteTag (write point), and CloseConnection (disconnect). All drivers for specific protocols must implement this interface.

[0044] Step S202: Retrieve and locate the driver.

[0045] Based on the protocol type identifier parsed from the configuration file, the configuration management module 10 retrieves the corresponding dynamic link library file (DLL) from the local driver storage path. The system maintains a mapping table between protocol identifiers and DLL filenames, which is used to accurately locate the target driver file.

[0046] Step S203: Dynamic reflection loading.

[0047] Using reflection in C#, the configuration management module 10 reads the assembly metadata of the target DLL file, iterates through all exported types in the assembly, and searches for the concrete class that implements the aforementioned IDriverInterface interface. After locating the target class, it uses the Activator.CreateInstance method to create an instance of the driver class in memory and assigns its reference to the main program's driver manager interface variable.

[0048] Step S204: Driver initialization.

[0049] The initialization method of the driver instance is called, passing in parameters such as the device IP, port, and timeout from the configuration file, to complete the initialization and construction of the underlying communication socket.

[0050] Through the above mechanism, the configuration management module 10 decouples the driver from the main application. When it is necessary to support new industrial protocols or update existing driver logic, it is only necessary to store the DLL file containing the new driver logic into an external storage device and re-execute the configuration process to complete the system protocol extension, without recompiling or replacing the entire system firmware. The specific implementation code logic of the reflection mechanism and the method of generating dynamic link libraries are conventional techniques in the field of software engineering, and those skilled in the art can implement them based on the above logical description, so they will not be elaborated here.

[0051] To adapt to complex industrial environments, the data acquisition module 20 supports both software protocol acquisition and hardware signal acquisition modes. In software acquisition mode, the data acquisition module 20 calls the driver instance loaded by the configuration management module 10, sends request frames conforming to Modbus, OPC UA, or MQTT protocol specifications via Ethernet or serial bus interface, and parses the response frames returned by the device. In hardware acquisition mode, for older devices without standard communication interfaces, the data acquisition module 20 directly reads the 4-20mA analog quantity or 0-10V voltage signal output by the sensor through the analog-to-digital converter (ADC) interface and converts it into a digital quantity.

[0052] For some non-standard devices or devices whose communication parameters drift due to aging, the data acquisition module 20 runs an adaptive driver correction mechanism. This mechanism relies on a signal conditioning circuit and a high-speed analog-to-digital converter connected in parallel at the communication interface of the edge gateway device. It identifies the device's communication mode by analyzing the physical layer signal characteristics and dynamically corrects the driver logic. The specific execution flow is as follows:

[0053] Step S301: Communication anomaly monitoring.

[0054] The data acquisition module 20 monitors the execution status of the standard driver in real time. When the number of consecutive data read timeouts, cyclic redundancy check (CRC) errors, or frame format errors exceeds a preset threshold, the system does not directly disconnect the connection, but instead activates the high-speed analog-to-digital conversion unit to switch the working mode to physical layer monitoring state.

[0055] Step S302: Physical feature extraction.

[0056] The data acquisition module 20 samples the electrical signal at the communication port at a sampling frequency at least five times higher than the communication baud rate, acquiring a raw signal waveform sequence. The system then extracts features from this sequence to construct a signal feature vector. This feature vector contains physical parameters in multiple dimensions, specifically represented as follows:

[0057] ;

[0058] In the formula, For the first The physical characteristic components include the average slope of the waveform rising edge, the average voltage drift of the high and low levels, the timing jitter variance of the signal period, and the statistical distribution value of the bit width. This represents the total number of dimensions of the features.

[0059] Step S303: Fingerprint matching analysis.

[0060] The data acquisition module 20 calls the locally stored fault fingerprint database, which contains signal feature templates for various known non-standard or aging devices. The system calculates the currently acquired feature vector. With the fingerprint database Template The degree of matching between them. To quantify the similarity between the two, the confidence score is calculated using the inverse form of the weighted Euclidean distance. The calculation formula is as follows:

[0061] ;

[0062] In the formula, For the first The weight coefficients of each feature dimension are used to adjust their influence in matching based on the feature's discriminative power. This is the k-th feature value of the currently acquired signal; For the first The standard feature values ​​corresponding to each template; , where is the confidence level.

[0063] Step S304: Drive logic correction.

[0064] Data acquisition module 20 selects confidence level The template with the highest value exceeding the adaptive threshold is selected as the matching result. Based on the type of the matched template, the system determines the current communication deviation pattern of the device, such as a specific byte order reversal, a non-standard baud rate offset, or a special frame header check rule. The system then generates corresponding driver patch parameters in memory, dynamically modifies the protocol parser's logic rules (e.g., adjusting the baud rate setting or rearranging the byte order of the receive buffer), and attempts to re-establish the communication connection using the corrected driver.

[0065] Through the above mechanism, the data acquisition module 20 can automatically adapt to communication anomalies caused by equipment aging or non-standard customization without manual intervention, ensuring the continuity and integrity of data acquisition. The specific message formats and CRC check algorithms of standard protocols such as Modbus are well-known technologies in the field of industrial communication and will not be elaborated upon here.

[0066] The data processing module 30 cleans and evaluates the value of the raw data stream from the acquisition module, aiming to extract key information from massive amounts of redundant data and provide a basis for decision-making for subsequent hierarchical storage and priority transmission.

[0067] The data processing module 30 first executes a basic cleaning process. This process includes deduplication and anomaly detection. Deduplication involves comparing continuously sampled data of the same variable. If the current value is exactly the same as the previous value or the change is within a preset dead zone, the current data is discarded, and only the timestamp update is retained. Anomaly detection uses the Laida criterion (3σ criterion) in statistics to calculate the mean and standard deviation of the data stream within a sliding window. If the current sampled value deviates from the mean by more than three times the standard deviation, it is marked as an outlier.

[0068] Following basic cleaning, data processing module 30 introduces a semantic entropy model to quantify and evaluate the value density of the data. This model, based on information theory principles and incorporating the temporal characteristics of industrial data, utilizes statistical analysis algorithms to calculate the semantic entropy value of the data within a specific time window. The specific implementation process is as follows:

[0069] Step S401: Data windowing sequence construction.

[0070] Data processing module 30 is set to a length of The time window, the data sequence contained within this window is represented as... ,set up Representing the The sampled values ​​at each time point.

[0071] Step S402: Calculate the probability distribution.

[0072] The system divides the range of data values ​​into: Given discrete intervals, count the number of samples falling into each interval in the sequence D, and calculate the number of samples in each interval. probability of occurrence .

[0073] Step S403: Calculate the comprehensive semantic entropy.

[0074] To simultaneously reflect both the discreteness and the severity of transient changes in the data, the data processing module 30 uses the following formula to calculate the comprehensive semantic entropy. :

[0075] ;

[0076] In the formula, For the first The probability of each range of values ​​occurring; The data sample value at the current moment; This is the data sample value from the previous moment; This is the standard deviation of the data sequence within the current time window, used to normalize the magnitude of change. It is a very small positive number, used to prevent the denominator from being zero; This is the mutation gain coefficient, which is a preset constant used to adjust the contribution weight of transient changes to the overall entropy value.

[0077] Step S404: Value grading labeling.

[0078] The data processing module 30 will calculate the results. With the preset high entropy threshold Compare. If This indicates that the data is currently in a high-information-content state, typically corresponding to sudden changes in equipment operating conditions, fault symptoms, or non-steady-state operation; the system marks this as "high-value data." This indicates that the data is in a low-information-content state, which usually corresponds to the device being in stable standby or running at a constant speed. The system marks it as "low-value data".

[0079] Through the above steps, the data processing module 30 transforms the undifferentiated raw data stream into an information stream with value labels. This entropy-based quantification method enables the system to identify key data points whose values, while not exceeding limits, have undergone abrupt changes in form, solving the problem that traditional threshold methods struggle to detect early, subtle fault characteristics. The specific calculation steps of the Raida criterion and the basic methods of probability statistics are common knowledge in the field of data analysis and will not be detailed here.

[0080] The data storage module 40 and the memory management module 50 together constitute the system's resource scheduling center, responsible for ensuring the persistence of critical data and the stability of system operation under limited edge computing resources.

[0081] The data storage module 40 is built on a local storage architecture based on an embedded SQLite database engine. To address the storage pressure caused by massive amounts of industrial data, this module implements a differentiated storage strategy based on data value. For information marked as "high-value data" by the data processing module 30, the data storage module 40 uses a lossless compression algorithm (such as the Deflate algorithm) to completely write it into the core table of the database. This core table has an anti-overwrite flag, and the system only allows the table to be cleaned up after storage space is critical and manual confirmation is required. For information marked as "low-value data," the data storage module 40 performs lossy processing, for example, only retaining the average, maximum, and minimum statistics for each minute, and writes them into a circular buffer table. This circular table has a fixed record limit, and when new data is written and the table is full, it automatically overwrites the oldest record. In addition, the data storage module 40 has a built-in automatic backup service, which exports incremental data from the core table as an encrypted file every morning and transfers it to external storage media or a cloud server.

[0082] The memory management module 50 is responsible for monitoring the random access memory (RAM) usage of the edge gateway device to prevent memory leaks or overflows due to prolonged operation. The memory management module 50 runs an improved Least Recently Used (LRU) algorithm, which is an entropy-weighted LRU memory purging strategy.

[0083] When the system detects that the memory usage exceeds a preset warning threshold (e.g., 85%), the memory management module 50 triggers a memory cleanup process. This process determines the cleanup order by calculating the eviction score of cached data blocks. The specific calculation logic is as follows:

[0084] Step S501: Traverse the cached objects.

[0085] The memory management module 50 scans all data cache blocks currently residing in memory and obtains the metadata of each data block, including the timestamp of the last access and the semantic entropy value corresponding to the data block.

[0086] Step S502: Calculate the elimination priority.

[0087] The system calculates the elimination score for each data block using the following formula. :

[0088] ;

[0089] In the formula, This is the current system's real-time timestamp; The timestamp of the most recent query, processing, or upload of this data block; This is the semantic entropy value corresponding to the data block, which is calculated by the data processing module 30. This is an adjustment constant used to prevent the denominator from being zero and to adjust the weight of the entropy value.

[0090] Step S503: Perform cleanup.

[0091] According to the calculation The memory management module 50 sorts all cache blocks in descending order. A higher value indicates that the data block has been idle for a longer period and contains less information, thus having a higher priority for eviction. The system releases the memory space occupied by the data block with the highest eviction score in sequence until the memory usage rate falls below a safe threshold (e.g., 60%).

[0092] Through the above mechanism, when resources are scarce, the system can prioritize retaining critical cache data that, although temporarily unaccessed, contains high information (such as fault waveform data), while prioritizing the discarding of low-value redundant data generated during stable operation. This strategy effectively avoids the risk of accidentally deleting important historical fault data, which may occur with the traditional LRU algorithm, ensuring the system's ability to trace back critical events under limited hardware resources. The specific SQL operation commands and the underlying implementation of the lossless compression algorithm for the SQLite database are well-known technologies in the field of computer software and will not be elaborated upon here.

[0093] The data forwarding module 60 serves as a communication hub between the edge device and the cloud or upper-level management systems (such as MES and ERP). To break down protocol barriers between the underlying devices and upper-level software, the data forwarding module 60 incorporates a multi-protocol adaptation engine. This engine can serialize the internal data objects standardized by the data processing module 30 into a payload in JavaScript Object Notation (JSON) format, and encapsulate it into a Message Queuing Telemetry Transport (MQTT) publish message or a Hypertext Transfer Protocol (HTTP) POST request message according to configuration requirements, thus achieving cross-system interconnection.

[0094] Before data is sent to the public network, the data forwarding module 60 performs a strict security encapsulation process. The system calls the onboard AES-256 hardware encryption chip to send the plaintext data payload to the encryption coprocessor, which generates a ciphertext data stream using a preset 256-bit key. Simultaneously, the data forwarding module 60 calculates an MD5 message digest for each encrypted data packet and appends this digest to the message header. Upon receiving the data, the receiving system recalculates the digest and compares it with the message header. If they do not match, it is determined that the data has been tampered with or lost during transmission, triggering a retransmission request, thereby ensuring the confidentiality and integrity of the data.

[0095] To address the pain point of unstable industrial field networks, the data forwarding module 60 is designed with an intelligent breakpoint resumption mechanism based on semantic entropy priority. This mechanism abandons the traditional simple first-in-first-out (FIFO) queue logic and instead adopts a value-driven scheduling strategy. The specific implementation steps are as follows:

[0096] Step S601: Link status monitoring and caching.

[0097] The data forwarding module 60 monitors the connection status with the upper-layer server in real time through a heartbeat mechanism. When a network connection interruption is detected (such as a Ping timeout or TCP connection reset), the system automatically suspends the sending task and stores the subsequently generated data packets into the pending queue of the local SQLite database, while recording the generation time and semantic entropy value of each data packet.

[0098] Step S602: Calculate transmission priority.

[0099] When the network connection is restored, faced with the accumulated historical data locally, the data forwarding module 60 does not perform a simple sequential reading of the queue, but instead iterates through each data packet in the queue to be sent. Calculate its transmission priority score The calculation formula is as follows:

[0100] ;

[0101] In the formula, For the first The semantic entropy value corresponding to each data packet reflects the fault correlation or information content of that data. The current system time; The time when the data packet was generated; To prevent the time constant from having a denominator of zero; This is a balancing coefficient, with a value between 0 and 1, used to adjust the weight of data value and data freshness in the ranking process. Score the transmission priority.

[0102] Step S603: Rearrange and send.

[0103] Data forwarding module 60 calculates... The pending data packets are reordered in descending order. Packets with higher scores are moved to the head of the queue.

[0104] Step S604: Flow control and confirmation.

[0105] The system retrieves and sends data packets in the rearranged order. After each data packet is sent, the system waits for an acknowledgment (ACK) from the receiver. If no acknowledgment is received within the timeout period, the system executes an exponential backoff retransmission strategy.

[0106] Through this mechanism, when the network fault is recovered, "high-value data" containing information about device mutations and high semantic entropy will receive the highest transmission priority, even if they were generated earlier; while "low-value data" that merely records stable operating conditions will be sent later. This strategy ensures that, with limited bandwidth resources and backlogged data, the operations and maintenance center can obtain critical anomaly information from the production site as quickly as possible, thereby shortening fault response time. For the specific handshake process of the MQTT protocol and the standard implementation of the AES encryption algorithm, those skilled in the art can refer to relevant international standard documents, which will not be elaborated here.

[0107] The security management module 70 and the dashboard display module 80 together construct the system's operation and maintenance interaction and security protection barrier.

[0108] The security management module 70 implements a multi-level permission management system based on a role-based access control (RBAC) model. The system pre-defines three standard roles: system administrator, operations engineer, and operator. The system administrator has the highest privileges, enabling them to manage user accounts, upgrade system firmware, and adjust core algorithm parameters; operations engineers are authorized to configure communication drivers, modify point table mappings, and view logs; operators only have read-only browsing permissions for dashboard data and basic report export permissions. When a user logs into the system, the security management module 70 verifies the user's credentials and generates a JSON Web Token (JWT) containing the role declaration. All subsequent API requests must include this token in the HTTP header, and the server verifies the validity of the token signature to determine whether to allow the operation request.

[0109] In addition, the security management module 70 integrates a full-process log auditing service. This service monitors the system's internal message bus, capturing all critical events, including user login / logout, configuration parameter changes, driver loading failures, and data transmission interruptions. The system encapsulates the event's trigger time, operator, IP address, and operation content into structured log records and stores them in a local encrypted database. For high-risk operations involving modifications to core configurations, the system also sends real-time alerts to administrators via a pre-defined SMTP email service or instant messaging software interface, ensuring that any changes to the system are traceable.

[0110] The Kanban module 80 is built on the Vue 3.0 progressive JavaScript framework, creating a front-end single-page application (SPA). To achieve low-latency visualization of industrial data, the Kanban module 80 uses the WebSocket full-duplex communication protocol to maintain a persistent connection with the back-end service. The front-end page uses the ECharts open-source charting library to render various visualization components.

[0111] Specifically, the Kanban display module 80 provides a device topology overview function, displaying the connection status of the edge gateway device and all its connected industrial field devices in vector format. Online devices are displayed in green, while offline or faulty devices are displayed in red, with the last communication time indicated. Simultaneously, this module features real-time data stream monitoring, using dynamic line graphs to display the real-time trends of core process parameters such as temperature, pressure, and speed. The line graphs support zooming and panning, allowing maintenance personnel to review historical fluctuations. Furthermore, the Kanban display module 80 is equipped with a system resource dashboard to display the edge gateway device's CPU utilization, memory usage, internal core temperature, and remaining SSD storage space in real time. In terms of report management, this module provides a multi-dimensional historical data query interface. Users can select start and end times, target devices, and variable names. The system backend retrieves data from the SQLite database based on the query conditions and generates a preview table on the front end. Using front-end libraries such as SheetJS, the query results can be directly exported to Excel or CSV files.

[0112] The Kanban display module 80 supports both local area network (LAN) and remote wide area network (WAN) access modes. In local mode, maintenance personnel can access the gateway directly via a laptop connected to the gateway's LAN port. In remote mode, the gateway maps the web service port to the public network using intranet penetration technology or VPN service, allowing maintenance personnel to view the site status anytime, anywhere via a mobile browser. The specific component lifecycles of the Vue framework and the details of the WebSocket handshake protocol are standard technologies in front-end development and will not be elaborated upon here.

Claims

1. A multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios, characterized in that: include: The configuration management module (10) is used to parse the encrypted configuration file and dynamically load the communication protocol driver using a reflection mechanism; The data acquisition module (20) is used to establish a connection with industrial field equipment and acquire raw data using the loaded communication protocol driver; The data processing module (30) is used to calculate the semantic entropy value of the raw data within the time window and to mark the data value level according to the semantic entropy value; The calculation of the semantic entropy value of the original data within the time window specifically includes the following steps: The probability distribution of the data sequence falling within the discrete value interval within the time window is statistically analyzed. Calculate the negative sum of the products of the probability of occurrence of each discrete value interval and the natural logarithm of the probability of occurrence, and use it as the basic information entropy; Calculate the absolute value of the difference between the current data and the previous data, and use the ratio of this difference to the standard deviation of the data sequence as the normalization mutation term; The semantic entropy value is obtained by adding the basic information entropy to the weighted normalized mutation term. The semantic entropy value is calculated using the following formula: ； In the formula, For the first The probability of occurrence of a discrete range of values; The data sample value at the current moment; This is the data sample value from the previous moment; This is the standard deviation of the data sequence within the current time window, used to normalize the magnitude of change. It is a very small positive number, used to prevent the denominator from being zero; The gain coefficient is a preset constant used to adjust the contribution weight of transient changes to the overall entropy value. Data storage module (40) is used to perform differentiated storage operations on the data according to the data value level; The memory management module (50) is used to monitor memory resources, calculate the eviction priority of cached data based on the semantic entropy value, and then perform memory cleanup. The data forwarding module (60) is used to send data to the upper-layer management system based on the transmission priority determined by the semantic entropy value.

2. The multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios according to claim 1, characterized in that, The process of parsing the encrypted configuration file and dynamically loading the communication protocol driver using a reflection mechanism specifically includes the following steps: Upon detecting external storage media access, the encrypted configuration file is read and decrypted to obtain the protocol type identifier; Retrieve locally stored dynamic link library files based on the protocol type identifier; Read the assembly metadata of the dynamic link library file, locate and instantiate the driver class that implements the standard driver interface; The instantiated driver class object is assigned to the driver interface variable to complete the loading of the communication protocol driver program.

3. The multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios according to claim 1, characterized in that, When acquiring raw data, the data acquisition module (20) is also used to execute adaptive drive correction logic, specifically including the following steps: When data acquisition fails using the communication protocol driver, the physical signal waveform sequence of the communication port is acquired at a sampling frequency that is higher than a preset multiple of the communication baud rate. Features of the physical signal waveform sequence are extracted to construct a signal feature vector, and the signal feature vector is matched with a preset fingerprint database template; Based on the matching results, the communication deviation mode is determined and driver patch parameters are generated. The driver patch parameters are then used to correct the protocol parsing logic in order to reacquire data. The communication deviation modes include byte order reversal, non-standard baud rate offset, or frame header verification rules.

4. The multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios according to claim 3, characterized in that, The matching of the signal feature vector with the preset fingerprint database template specifically includes the following steps: Calculate the square of the difference between the currently acquired physical feature component and the standard feature value in the fingerprint database template; The square roots of the weighted sums of the squares of the differences in each dimension are then taken. The reciprocal of the square root result plus one is taken as the similarity confidence score, and the template with the highest similarity confidence score is selected as the matching result.

5. The multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios according to claim 1, characterized in that, The process of marking data value levels based on semantic entropy values ​​specifically includes the following steps: The calculated semantic entropy value is compared with a preset high entropy threshold. If the semantic entropy value is greater than or equal to the high entropy threshold, the data is marked as high-value data; If the semantic entropy value is less than the high entropy threshold, the data is marked as low-value data.

6. The multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios according to claim 5, characterized in that, The differential storage operation on the data specifically includes the following steps: The data marked as high-value data is written into the core table of the database using a lossless compression algorithm; The data marked as low-value data is statistically reduced and then written into the circular buffer table of the database; When the circular buffer table is full, the earliest written data record is overwritten.

7. The multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios according to claim 1, characterized in that, The calculation of cache data eviction priority based on the semantic entropy value specifically includes the following steps: Get the difference between the current system time and the time when the data cache block was last accessed, and use it as the idle time. Obtain the semantic entropy value corresponding to the data cache block; Calculate the ratio of the idle time to the sum of the semantic entropy value and the preset adjustment constant, and use the ratio as the elimination score; The order in which the cached data is cleared is determined according to the elimination score from highest to lowest.

8. The multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios according to claim 1, characterized in that, The transmission priority determined by the semantic entropy value sends the data to the upper-layer management system, specifically including the following steps: After the network connection is restored, iterate through the data packets in the pending queue; A transmission priority score is calculated by weighting and summing the semantic entropy value of each data packet with the data packet's freshness, wherein the freshness of the data packet is inversely proportional to the elapsed time since the data packet was generated; The transmission priority score is used to reorder the queue of data packets to be sent in descending order, and the data packets at the head of the queue are sent first.

9. The multi-protocol compatible IoT data acquisition and intelligent operation and maintenance system for industrial scenarios according to claim 1, characterized in that, The system also includes: The security management module (70) is used to enforce role-based access control policies to verify user operation permissions and monitor configuration changes, driver loading and data transmission behavior to generate traceable system audit logs; The Kanban display module (80) is used to establish a full-duplex communication connection between the front end and the back end, render the connection topology and core process parameter flow of industrial field equipment in real time in the form of visual charts, and provide interfaces for querying historical data and exporting reports.

Images