Multi-protocol conversion electric actuating mechanism control method

Abstract

The invention discloses a multi-protocol conversion electric actuating mechanism control method, and particularly relates to the field of industrial automation control, and the method comprises the steps: firstly, carrying out system initialization, including hardware self-inspection, protocol automatic identification and parameter configuration loading, then converting different protocol signals into target formats through a master-slave protocol conversion module, and finally, carrying out multi-protocol conversion; data are transmitted to a cloud end through a 5G communication link, cloud end interaction and local edge calculation preprocessing are carried out at the same time, storage management and remote configuration management are carried out on the data at the cloud end, faults are analyzed and predicted through AI early warning, and during fault early warning, local alarm is carried out and fault information is uploaded. According to the method, the problem of multi-protocol communication of the electric actuating mechanism is solved, reliable transmission of data is guaranteed, accurate fault prediction and processing are realized, convenient configuration management is supported, and the operation stability and the intelligent level of the system are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial automation control, and more particularly to a multi-protocol conversion electric actuator control method. Background Art

[0002] In the process of industrial automation, electric actuators, as key equipment, play a vital role. They can accurately convert electrical signals into mechanical motion and are widely used in many industries such as petrochemicals, electric power, metallurgy, and water conservancy. They are used to drive valves, dampers and other devices to achieve precise control of key process parameters such as fluid flow, pressure, and temperature. They are core components to ensure safe and stable production and improve product quality.

[0003] Currently, the Industrial Internet of Things and intelligent manufacturing are booming, the industrial environment is becoming increasingly complex, and the demand for equipment interconnection is growing rapidly. This has prompted the emergence of a large number of communication protocols in industrial sites. Common ones include Profibus-DP, which is suitable for industrial fieldbuses and has high-speed and reliable characteristics but a complex protocol structure; MQTT, which is lightweight, focuses on communication for resource-constrained devices, has certain requirements for network stability, and has a simple data format; and Modbus RTU, HART and other protocols. These protocols vary significantly in data format, transmission rate, communication mechanism, etc., bringing many difficulties to the operation of electric actuators. On the one hand, traditional electric actuators usually only support a single or a few protocols, which makes it difficult to meet diverse communication needs. As a result, additional conversion equipment and extensive debugging are required for device integration, which greatly increases system cost and maintenance difficulty. For example, in an industrial automation production line containing equipment from multiple brands, some equipment uses the Profibus-DP protocol, while others use the Modbus RTU protocol. It is difficult for traditional electric actuators to directly achieve seamless communication with devices with different protocols.

[0004] On the other hand, the industrial field environment is harsh, with high temperature, high humidity, strong electromagnetic interference and other conditions occurring frequently, which seriously threaten the stability and accuracy of data transmission. Even if the electric actuator can recognize multiple protocols, it may still be due to data transmission errors or losses, resulting in the inability to accurately convey the control instructions, causing the actuator to operate abnormally, affecting the normal production. For example, in a steel smelting workshop, strong electromagnetic interference may cause data transmission to be lost, resulting in errors in the control signal received by the electric actuator, and then making the valve opening control inaccurate, affecting the transportation and reaction of materials in the furnace. From the perspective of equipment management and maintenance, with the expansion of industrial production scale and the increase in the number of equipment, The demand for real-time monitoring, fault prediction and remote management of electric actuators is becoming more and more urgent. However, existing control methods have obvious shortcomings in these aspects. Traditional methods are difficult to obtain equipment operation status data in real time and accurately, and cannot provide early warning and diagnosis of potential faults. Moreover, when the equipment needs parameter adjustment or firmware upgrade, most of them rely on manual on-site operation, which is inefficient and prone to errors. For example, in large chemical companies, a large number of electric actuators distributed in different areas, if all need to be manually adjusted on-site, will not only consume a lot of manpower and material resources, but may also cause new problems due to improper operation, seriously affecting the company's production efficiency and economic benefits. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a multi-protocol conversion electric actuator control method, which solves the problems proposed in the above background technology through the following scheme: the traditional electric actuator usually only supports a single or a few protocols, is difficult to meet diversified communication needs, has poor stability and accuracy of data transmission, is difficult to obtain equipment operation status data in real time and accurately by traditional methods, cannot provide early warning and diagnosis of potential faults, and when the equipment requires parameter adjustment or firmware upgrade, mostly relies on manual on-site operation, which is inefficient and prone to errors.

[0006] To achieve the above-mentioned object, the present invention provides the following technical solution: a multi-protocol conversion electric actuator control method, comprising: S1: system initialization: performing hardware self-test, automatic protocol identification, and parameter configuration loading to put the system in a state where it can work normally and accurately respond to input signals;

[0007] S2: Protocol conversion execution: Through the collaborative work of the master protocol conversion module and the slave protocol conversion module, the output and input signals of different protocols are uniformly converted into the target protocol format, providing a standardized data interface for subsequent data transmission and processing;

[0008] S3: Data transmission and processing: After the protocol conversion is completed, the converted data is transmitted to the cloud through the 5G communication link, while cloud data interaction is carried out and edge computing pre-processing is performed locally;

[0009] S4: Cloud service processing: After data is transmitted to the cloud via a 5G communication link and initial interaction and preprocessing are completed, data storage management and remote configuration management are performed on the transmitted data, and fault prediction of electric actuators is performed through AI early warning analysis;

[0010] S5: Fault handling: When a fault warning occurs, the local buzzer alarm is triggered and the fault information is uploaded to the cloud to notify the operation and maintenance personnel. When a red warning is triggered, the thyristor fast shutdown method is immediately used to cut off the power supply of the actuator motor. At the same time, an emergency shutdown notification containing the fault code is sent to the cloud through the UDP protocol.

[0011] Technical effects and advantages of the present invention:

[0012] 1. The present invention achieves compatibility with multiple protocols through steps such as hardware self-test, automatic protocol identification, and parameter configuration loading. The protocol detection chip can perform baud rate adaptive detection within the range of 300-3000Kbps and match it with the built-in feature library to accurately identify the protocol type of the input signal. The main protocol conversion module and the slave protocol conversion module work together to uniformly convert the output and input signals of different protocols into the target protocol format, providing a standardized data interface for subsequent data transmission and processing, effectively solving the communication problem between the electric actuator and various protocol devices, and reducing the difficulty of system integration.

[0013] 2. In terms of data transmission, the present invention adopts a variety of technologies to ensure reliable data transmission. During the establishment of the 5G communication link, the Internet of Things card authentication, NSA / SA dual-mode automatic attachment to the base station, the setting of a timeout retransmission mechanism, and the use of the AES-256 encryption algorithm ensure the safe and stable transmission of data in complex industrial environments. When interacting with the cloud, the device and the cloud interact with data at a fixed period, abnormal data is reported in a timely manner, and an ACK response timeout retransmission mechanism is set to ensure that the instructions are correctly received and executed by the device.

[0014] 3. The electric actuator health model established by the present invention is based on the LSTM neural network and is trained using a large amount of historical data, which can accurately predict the failure probability of the equipment. By collecting key operating parameters such as motor current, valve position feedback, gearbox temperature, etc. in real time, and analyzing the historical data of equipment performance indicators with a sliding window regression model, it is possible to predict the performance degradation trend, divide the warning level into yellow warning and red warning, and use different notification methods to remind users and operation and maintenance personnel. In terms of fault handling, when a fault warning occurs, the local buzzer alarm is triggered and the fault information is uploaded; when a red warning is triggered, the motor power is immediately cut off and an emergency shutdown notification is sent to the cloud, effectively ensuring equipment and production safety.

[0015] 4. The present invention supports two parameter configuration loading methods: local and cloud. The local device reads the XML format configuration file through the USB interface, and the cloud device obtains dynamic configuration information through the MQTT protocol, which is convenient for users to configure according to actual needs. It supports differential upgrade algorithm for firmware upgrade. When the OTA upgrade is interrupted, the progress can be recorded and the upgrade can be continued from the breakpoint, reducing the upgrade time and data transmission volume. Users can dynamically modify the control parameters through the cloud platform. After receiving the new parameters, the device performs a legitimacy check and updates the local configuration. At the same time, all configuration change operations are recorded in the audit log, which is convenient for query and management, and improves the efficiency and accuracy of configuration management. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the overall structure of the present invention.

[0017] Figure 2 It is a schematic diagram of the protocol conversion process of the present invention. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] As attached Figure 1 A multi-protocol conversion electric actuator control method shown includes:

[0020] S1: System initialization: Perform hardware self-test, automatic protocol identification, and parameter configuration loading to ensure that the system can work normally and accurately respond to input signals;

[0021] The hardware self-test refers to a comprehensive self-test of the motor drive module, gearbox position sensor, protocol conversion module, and 5G gateway after the actuator body is started. The motor drive module completes an overload test within 100ms to ensure its reliable power output during subsequent operation and avoid equipment damage and system failures caused by overload. The gearbox position sensor is calibrated to control the error within ±0.5% to ensure accurate acquisition of real-time gearbox position information, providing basic data support for precise motion control of the actuator. The protocol conversion module performs a master-slave processor communication handshake. Through specific communication protocols and handshake signals, it confirms that the data transmission channel between the master and slave processors is unobstructed, ensuring the smooth progress of the subsequent protocol conversion process. The 5G gateway detects the antenna signal strength, requiring the signal strength to be ≥-90dBm to ensure stable data transmission between the subsequent device and the cloud. At the same time, the sensor module reads the baseline values of operating parameters. These baseline values will serve as an important reference for parameter monitoring, comparison, and fault diagnosis during subsequent system operation.

[0022] The steps for automatic protocol recognition are as follows:

[0023] After the input signal enters the protocol detection chip, the chip first performs baud rate adaptive detection in the detection range of 300-3000Kbps to determine the data transmission rate of the input signal. Different industrial protocols and network protocols may use different baud rates for data transmission. Through adaptive detection, the baud rate of the input signal can be accurately matched, providing a basis for subsequent protocol identification. Then, the chip matches the input signal with the built-in protocol feature library. Each protocol has a unique feature code and data format. The protocol type followed by the input signal is preliminarily determined through feature matching. Finally, the data frame format is verified and the parity bit is automatically matched. The parity bit is used for error detection during data transmission to ensure the accuracy and integrity of the data. The entire protocol identification process is completed within 50ms, and the identification result is output to the main controller to provide a decision basis for the subsequent protocol conversion and data processing of the main controller.

[0024] The parameter configuration loading refers to:

[0025] The system loads the necessary configuration information and supports both local and cloud-based parameter configuration loading methods. Local configuration reads XML-formatted configuration files through a USB interface. XML, as a universal data markup language, has good data structure and readability, and can easily store and transmit various parameters required for system operation. Cloud configuration obtains dynamic configuration information from the server via the MQTT protocol. MQTT is a lightweight message transmission protocol suitable for communication between IoT devices and servers, enabling real-time parameter updates and remote configuration. Configuration content includes the target protocol type. The system supports simultaneous configuration of two output protocols to meet protocol conversion requirements in different application scenarios; a data mapping table that defines the correspondence between input / output signals and register addresses, providing mapping rules for accurate data transmission and processing; and a cloud server address and authentication key used to establish a secure connection between the device and the cloud server, enabling reliable data transmission and authentication.

[0026] S2: Protocol conversion execution: Through the collaborative work of the master protocol conversion module and the slave protocol conversion module, the output and input signals of different protocols are uniformly converted into the target protocol format, providing a standardized data interface for subsequent data transmission and processing;

[0027] It should be specifically noted that the main protocol conversion module adopts different parsing and conversion methods according to the protocol type of the input signal. For industrial protocols, the PDU protocol data unit data is parsed through the DP slave chip. For network protocols, the JSON format payload data is decoded and all input signals are uniformly converted to Modbus RTU format. The converted data is stored in a double-buffered data area with a capacity of 2KB.

[0028] The double buffer data area adopts a double buffer alternate working mode, which can improve the efficiency of data processing and avoid data loss and data processing delay;

[0029] It should be further explained that PDU is the basic unit of data transmission in the protocol. By parsing PDU data, the specific content of the input signal can be obtained. JSON is a lightweight data exchange format commonly used for data transmission and storage in network communications. Modbus RTU is a communication protocol widely used in the field of industrial automation, with the characteristics of versatility and strong compatibility.

[0030] It should be specifically noted that the slave protocol conversion module further converts the uniformly converted ModbusRTU format data into the target protocol TCP according to the multi-protocol concurrent requirement ModbusRTU+TCP preset by the user. For industrial protocol conversion, a dedicated protocol stack such as Profibus-DPV1 protocol stack is used. For network protocol conversion, TCP / IP encapsulation is implemented through the LwIP protocol stack. The converted data is subjected to CRC check for the Modbus protocol or FCS frame check sequence check for the Profibus protocol, and data integrity check is performed.

[0031] The data integrity check refers to byte alignment detection to ensure that no errors occur during data transmission and processing, thereby ensuring the accuracy and integrity of the data;

[0032] It should be further explained that a protocol stack is a series of software programs and algorithms that implement a specific protocol. It can perform functions such as protocol encapsulation, parsing, and error handling. LwIP is a lightweight TCP / IP protocol stack suitable for network communications in embedded systems. CRC refers to cyclic redundancy check.

[0033] The above protocol conversion process can be found in Figure 2 As shown;

[0034] S3: Data transmission and processing: After the protocol conversion is completed, the converted data is transmitted to the cloud through the 5G communication link, while cloud data interaction is carried out and edge computing preprocessing is performed locally.

[0035] It should be specifically noted that the establishment of the 5G communication link refers to authentication through the Internet of Things card, supporting the eSIM air card writing function, the 5G gateway adopts NSA / SA dual-mode automatic attachment to the base station, selects the networking mode according to the actual network environment and signal strength, ensures a stable connection with the base station, and sets a timeout retransmission mechanism when establishing a TCP connection, with a maximum of 3 retries, and data transmission adopts the AES-256 encryption algorithm, namely CBC mode.

[0036] It should be further explained that eSIM can realize remote configuration and management, improving the flexibility of device use. TCP is a connection-oriented transmission protocol. Timeout retransmission can ensure reliable data transmission. AES-256 encryption has high strength. CBC mode increases the randomness and security of encryption, protecting data from being stolen or tampered with during transmission.

[0037] The cloud data interaction refers to the data interaction between the device and the cloud at a fixed period, sending device status data including operating parameters to the cloud every 500ms. This data can fully reflect the real-time operating status of the device. Once abnormal data is detected, the system reports it within ≤200ms so that the cloud can respond in time. The control instructions issued by the cloud enter the preset priority queue, and emergency instructions are processed first. Priority management ensures that important instructions are executed in time. At the same time, an ACK response timeout retransmission mechanism is set to ensure that the instructions are correctly received and executed by the device.

[0038] The edge computing preprocessing includes:

[0039] A1: Real-time data acquisition and caching: Utilizing various local sensors on the device, key operating parameters of the electric actuator are continuously collected 10 times per second, including motor current, valve position feedback, and gearbox temperature. This data directly reflects the device's immediate operating status and serves as raw material for subsequent analysis.

[0040] High-speed flash memory chips are used locally to cache the latest 1,000 historical data items, and the cache is managed using the FIFO principle. As new data is continuously collected, the oldest data is squeezed out of the cache in turn, ensuring that the latest historical data is always retained in the cache, providing timeliness for subsequent analysis. At the same time, by establishing a timestamp index table, data reading and writing efficiency is greatly improved, making it easier to quickly locate and retrieve data at specific time points.

[0041] A2: Data fluctuation monitoring and preliminary screening: Calculate the parameter fluctuation range in real time, set a dynamic threshold of ±5% for each parameter, and use a sliding window algorithm with a sliding window of 10 seconds to process the data within the window. In each window, calculate the mean and standard deviation of the data. The mean reflects the average level of the data, and the standard deviation reflects the degree of dispersion of the data. These two statistics are used to further determine whether the parameter is within the normal fluctuation range. If the mean or standard deviation exceeds the preset range based on historical data statistics, the data in the window is also marked as abnormal. For example, for motor current parameters, when the real-time current value fluctuates by more than ±5% of its current mean within 1 second, the data is marked as potentially abnormal data.

[0042] A3: Key Feature Extraction and Data Output: Key features are extracted from the initially screened data. In addition to the mean and standard deviation, the rate of change (the ratio of the difference between two adjacent sampling points to the time interval) is calculated. These features can more comprehensively reflect the dynamic changes in equipment operating parameters.

[0043] The extracted key features and corresponding timestamps and other information are packaged and organized in JSON format for AI early warning analysis. The output data undergoes preliminary screening and feature extraction to remove redundant data, reducing the data processing volume of the AI early warning analysis module and providing a high-quality data foundation for the efficient operation of subsequent complex models.

[0044] S4: Cloud service processing: After data is transmitted to the cloud via a 5G communication link and initial interaction and preprocessing are completed, data storage management and remote configuration management are performed on the transmitted data, and fault prediction of electric actuators is performed through AI early warning analysis;

[0045] It should be noted that the data storage management uses the TimescaleDB time series database to store device data. When entering the data into the database, the data is partitioned and stored according to the device ID and timestamp to improve storage and query efficiency. At the same time, the original data is losslessly compressed, and the compression rate is controlled at ≥80%;

[0046] It should be further explained that the TimescaleDB time series database supports multi-dimensional data retrieval based on time, device ID, parameter type, etc. Users enter conditions through the cloud platform query interface. The system displays the results in the form of tables or charts after retrieval. The original data is stored for one year, the aggregated data is stored permanently, expired data is cleaned up regularly, and data is backed up off-site to ensure data security and reliability.

[0047] The AI early warning analysis refers to establishing an electric actuator health model based on the LSTM neural network, using historical data to train the model, taking the electric actuator operating parameters as input and the fault status as output, iteratively adjusting the model weights and biases, inputting the real-time collected operating parameters into the model, outputting the fault probability, and issuing an early warning when the preset threshold is exceeded. The sliding window regression model is used to analyze the historical data of equipment performance indicators, and a regression model of performance and time is established to predict the performance degradation trend. The early warning levels are divided into yellow warnings and red warnings, and different notification methods are used to remind users and operation and maintenance personnel.

[0048] Specifically, the process of establishing the electric actuator health model is as follows:

[0049] B1: Historical data collection: For electric actuators, focus on collecting historical data of three key operating parameters: motor current I, valve position feedback P, and gearbox temperature T. Assume that m samples are collected, and each sample corresponds to time t i (i=1,2,…,m), then the sample data can be expressed as {(I ti ,P ti ,T ti ,y ti )}, where y ti∈{0,1}, 0 represents normal operation state, and 1 represents fault state.

[0050] B2: Data cleaning: For the missing values in motor current I, valve position feedback P, and gearbox temperature T, the mean filling method is used. Taking motor current as an example, its mean k is the number of missing values of motor current, and the missing values are filled with I. In terms of outlier processing, the values with a difference from the mean greater than three times the standard deviation are regarded as outliers and replaced by the neighborhood mean. Taking the motor current I as an example, let its mean be μ I , with a standard deviation of σ I , when |I ti -μ I ∣>3σ I When it is considered as an outlier, the neighborhood mean is used Instead, when data is normalized, the motor current, valve position feedback, and gearbox temperature are Z-score normalized. For example, the motor current is normalized to where μ I is the mean motor current, σ I is the standard deviation.

[0051] B3: Dataset division: Divide the cleaned data into training set D according to the ratio of 7:2:1 train , validation set D val and the test set D test , the number of training set samples m1 = 0.7m, used for model training, the number of validation set samples m2 = 0.2m, to assist in adjusting model hyperparameters, and the number of test set samples m3 = 0.1m, to evaluate model performance.

[0052] B4: LSTM model architecture design: Construct an LSTM model with an input layer dimension of 3 corresponding to the motor current, valve position feedback, and gearbox temperature. The hidden layer has h LSTM units, divided into l layers. The output layer dimension is 1, which is the fault probability. The forget gate Input Gate Output Gate and candidate cell states Calculated by the following formulas:

[0053]

[0054] Current cell status and hidden state for:

[0055]

[0056] The final output failure probability prediction value

[0057]

[0058] B4: Model training: The binary cross entropy loss function is used to calculate the loss value of the entire training batch. The binary cross entropy loss function is as follows:

[0059]

[0060] For each sample i, is the true fault state of sample i, which can be 0 or 1. is the failure probability predicted by the model, and is calculated separately. and The values of the two items are then added together, the sum of the above values of all m1 samples is accumulated, and then divided by the number of samples m1, and the negative sign is taken to obtain the loss value L of the training batch. The loss value L reflects the degree of difference between the current predicted value of the model and the true value. The smaller L is, the closer the predicted value is to the true value. Using the gradient descent method, the learning rate is set to α. In each training cycle, the gradient of the loss function with respect to the weight and bias is calculated. And update the weights and biases;

[0061] Use the chain rule to find the partial derivative of the loss function L with respect to the weights and biases in the model to get the gradient Take the weight matrix W of the fully connected layer out For example, assume that the input of the fully connected layer is That is, the hidden state of the last time step of the LSTM layer, the output is First of all, About W out Find the partial derivative and combine it with the loss function The partial derivative of the loss function is obtained according to the chain rule. out The partial derivative of is the gradient;

[0062] For weight W, the update formula is Where α is the learning rate, which controls the step size of each update, is the gradient of the loss function with respect to the weight W, which is obtained by subtracting the learning rate α from the weight W and the gradient The product of , to achieve weight update;

[0063] For bias b, the update formula is The principle is similar to weight update, and the bias value is updated by subtracting the product of the learning rate and the bias gradient.

[0064] B5: Model evaluation: On the test set D test Evaluate the model performance and calculate the accuracy Acc, recall Rec and F1 value F1;

[0065] The calculation method of the accuracy Acc is as follows:

[0066]

[0067] The calculation method of the recall rate Rec is as follows:

[0068]

[0069] The calculation method of the F1 value is as follows:

[0070]

[0071] B6: Real-time fault warning: Real-time acquisition of the motor current I of the electric actuator real , valve position feedback P real , gearbox temperature T real , using the same data cleaning and standardization methods as historical data, the standardized real-time parameter I real ′、P real ′、T real ′, input the standardized real-time parameters into the trained LSTM model to obtain the fault probability prediction value like θ refers to the preset threshold, which determines that the equipment may fail and triggers a fault warning.

[0072] B7: Performance degradation trend analysis based on sliding window regression model: Select the sliding window size as s, for the time series data of motor current I, valve position feedback P, and gearbox temperature T, starting from t = 1, take the window [I t ,I t+1 ,…,I t+s-1 ]、[P t ,P t+1 ,…,P t+s-1 ]、[T t ,T t+1 ,…,T t+s-1 ], each time the window slides one time step, for the data in each sliding window, taking the motor current as an example, the linear regression model y t =β0+β1t+∈, use the least squares method to estimate the regression coefficients β0 and β1, the goal is to minimize the residual sum of squares Based on the established regression model, the motor current, valve position feedback, and gearbox temperature values for the next p time steps are predicted. If the predicted values show that the performance indicators continue to decline and the decline exceeds the preset range, it is considered that there is a risk of performance degradation of the equipment, and maintenance measures need to be taken in advance.

[0073] B8: Warning level classification: Through statistical analysis of the historical normal operation data of the equipment, the average value of the motor current, the average value of the valve position feedback, and the average value of the gearbox temperature are determined, and the motor current reference value μ is obtained based on the average value. I, valve position feedback reference value μ P , gearbox temperature reference value μ T , when the motor current I real satisfy Or valve position feedback P real satisfy Or the gearbox temperature Treal satisfies When the yellow warning is triggered, the motor current I real Exceeding the preset safety threshold I th , or valve position feedback P real Exceeding the preset safety threshold P th , or gearbox temperature T real Exceeding the preset safety threshold T th A red alert is triggered.

[0074] The remote configuration management includes upgrading the electric actuator’s firmware through a differential upgrade algorithm. The system compares the current firmware version of the device with the latest firmware version on the cloud, calculates the difference and transmits it.

[0075] It should be further explained that during the OTA upgrade process, if an interruption occurs, the device will record the upgrade progress and continue the upgrade from the breakpoint after the network is restored. Users can dynamically modify control parameters such as dead zone settings and response time through the cloud platform. After receiving the new parameters, the device will perform a legitimacy check. If the check passes, the local configuration will be updated. In addition, all configuration change operations will be recorded in the audit log. The log content includes operation time, operator, change content, etc., which is convenient for users to query and manage.

[0076] S5: Fault handling: When a fault warning occurs, the local buzzer alarm is triggered and the fault information is uploaded to the cloud to notify the operation and maintenance personnel. When a red warning is triggered, the thyristor fast shutdown method is immediately used to cut off the power supply of the actuator motor. At the same time, an emergency shutdown notice containing the fault code is sent to the cloud via the UDP protocol;

[0077] It should be further explained that the use of predefined fault code tables reduces the amount of data transmission, making it easier for remote monitoring personnel to understand and handle fault conditions in a timely manner.

[0078] Secondly: The drawings of the embodiments disclosed in the present invention only involve structures related to the embodiments disclosed in the present invention. Other structures may refer to conventional designs. The same embodiment and different embodiments of the present invention may be combined with each other without conflict.

[0079] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A multi-protocol conversion electric actuator control method, characterized in that: include: S1: System initialization: Perform hardware self-test, automatic protocol identification, and parameter configuration loading to ensure that the system can work normally and accurately respond to input signals; S2: Protocol conversion execution: Through the collaborative work of the master protocol conversion module and the slave protocol conversion module, the output and input signals of different protocols are uniformly converted into the target protocol format, providing a standardized data interface for subsequent data transmission and processing; S3: Data transmission and processing: After the protocol conversion is completed, the converted data is transmitted to the cloud through the 5G communication link, while cloud data interaction is carried out and edge computing pre-processing is performed locally; S4: Cloud service processing: After data is transmitted to the cloud via a 5G communication link and initial interaction and preprocessing are completed, data storage management and remote configuration management are performed on the transmitted data, and fault prediction of electric actuators is performed through AI early warning analysis; S5: Fault handling: When a fault warning occurs, the local buzzer alarm is triggered and the fault information is uploaded to the cloud to notify the operation and maintenance personnel. When a red warning is triggered, the thyristor fast shutdown method is immediately used to cut off the power supply of the actuator motor. At the same time, an emergency shutdown notification containing the fault code is sent to the cloud through the UDP protocol.

2. The multi-protocol conversion electric actuator control method according to claim 1, characterized in that: The hardware self-test refers to a comprehensive self-test of the motor drive module, gearbox position sensor, protocol conversion module, and 5G gateway after the actuator body is started. The motor drive module completes the overload test within 100ms, the gearbox position sensor is calibrated to control the error within ±0.5%, the protocol conversion module performs a master-slave processor communication handshake, and the 5G gateway detects the antenna signal strength, requiring the signal strength to be ≥-90dBm. At the same time, the sensor module reads the baseline value of the operating parameter; The steps for automatic protocol recognition are as follows: After the input signal enters the protocol detection chip, the chip first performs baud rate adaptive detection within the detection range of 300-3000Kbps. Then, the chip matches the input signal with the built-in protocol feature library, and preliminarily determines the protocol type followed by the input signal through feature matching. Finally, the data frame format is verified and the parity bit is automatically matched. The recognition result is output to the main controller, providing a decision basis for the main controller's subsequent protocol conversion and data processing. The parameter configuration loading refers to: The system loads the necessary configuration information and supports both local and cloud-based parameter configuration loading methods. Local configuration reads XML-formatted configuration files through the USB interface, while cloud configuration obtains dynamic configuration information from the server through the MQTT protocol. The configuration content includes the target protocol type, data mapping table, cloud server address, and authentication key, which are used to establish a secure connection between the device and the cloud server, achieving reliable data transmission and authentication.

3. The multi-protocol conversion electric actuator control method according to claim 1, characterized in that: The main protocol conversion module adopts different parsing and conversion methods according to the protocol type of the input signal. For industrial protocols, the PDU protocol data unit data is parsed by the DP slave chip. For network protocols, the load data in JSON format is decoded and all input signals are uniformly converted to Modbus RTU format. The converted data is stored in a double-buffered data area with a capacity of 2KB. The slave protocol conversion module further converts the uniformly converted ModbusRTU format data into the target protocol TCP according to the multi-protocol concurrent requirement ModbusRTU+TCP pre-set by the user. For industrial protocol conversion, a dedicated protocol stack such as Profibus-DPV1 protocol stack is used. For network protocol conversion, TCP / IP encapsulation is implemented through the LwIP protocol stack. The converted data is subjected to CRC check for the Modbus protocol or FCS frame check sequence check for the Profibus protocol, and data integrity check is performed.

4. The multi-protocol conversion electric actuator control method according to claim 1, characterized in that: The establishment of the 5G communication link refers to authentication through the Internet of Things card. The 5G gateway adopts NSA / SA dual-mode automatic attachment to the base station, selects the networking mode according to the actual network environment and signal strength, sets a timeout retransmission mechanism when establishing a TCP connection, and uses the AES-256 encryption algorithm, namely CBC mode, for data transmission; The cloud data interaction refers to the data interaction between the device and the cloud at a fixed period, sending device status data including operating parameters to the cloud every 500ms; The edge computing preprocessing includes: A1: Real-time data acquisition and caching: Utilizing various local sensors, the device continuously collects key operating parameters of the electric actuator, including motor current, valve position feedback, and gearbox temperature, at a frequency of 10 times per second. High-speed flash memory chips are used to cache the most recent 1,000 historical data items. The cache is managed using the FIFO (first-in, first-out) principle. Furthermore, a timestamp index table is established. A2: Data Fluctuation Monitoring and Preliminary Screening: Parameter fluctuation ranges are calculated in real time. A dynamic threshold of ±5% is set for each parameter. A sliding window algorithm is used to process data within a 10-second sliding window. Within each window, the mean and standard deviation are calculated. The mean reflects the average level of the data, while the standard deviation reflects the degree of dispersion. These two statistics are used to further determine whether the parameter is within the normal fluctuation range. If the mean or standard deviation exceeds the preset range, the data within that window is also marked as abnormal. A3: Key feature extraction and data output: For the data that has undergone preliminary screening, key features are extracted. In addition to the mean and standard deviation, the data change rate is calculated, that is, the ratio of the change difference between two adjacent sampling points to the time interval. The extracted key features and corresponding timestamps and other information are packaged and organized in JSON format for AI early warning analysis.

5. The multi-protocol conversion electric actuator control method according to claim 1, characterized in that: The data storage management uses the TimescaleDB time series database to store device data. When entering the database, it is partitioned and stored according to the device ID and timestamp, and the original data is losslessly compressed. The AI early warning analysis refers to establishing an electric actuator health model based on the LSTM neural network, using historical data to train the model, taking the electric actuator operating parameters as input and the fault status as output, iteratively adjusting the model weights and biases, inputting the real-time collected operating parameters into the model, outputting the fault probability, and issuing an early warning when the preset threshold is exceeded. The sliding window regression model is used to analyze the historical data of equipment performance indicators, and a regression model of performance and time is established to predict the performance degradation trend. The early warning levels are divided into yellow warnings and red warnings, and different notification methods are used to remind users and operation and maintenance personnel.

6. The multi-protocol conversion electric actuator control method according to claim 5, characterized in that: The establishment process of the electric actuator health model is as follows: B1: Historical data collection: For electric actuators, focus on collecting historical data of three key operating parameters: motor current I, valve position feedback P, and gearbox temperature T. Assume that m samples are collected, and each sample corresponds to time t i (i=1,2,…,m), then the sample data can be expressed as {(I ti ,P ti ,T ti ,y ti )}, where y ti ∈{0,1}, 0 represents normal operation and 1 represents fault state; B2: Data cleaning: For missing values in motor current I, valve position feedback P, and gearbox temperature T, if the mean filling method is used, values with a difference from the mean greater than three times the standard deviation are considered outliers and replaced with the neighborhood mean. When normalizing the data, Z-score normalization is performed on the motor current, valve position feedback, and gearbox temperature respectively. B3: Dataset division: Divide the cleaned data into training set D according to the ratio of 7:2:1 train , validation set D val and the test set D test , the number of training set samples m1 = 0.7m, used for model training, the number of validation set samples m2 = 0.2m, to assist in adjusting model hyperparameters, and the number of test set samples m3 = 0.1m, to evaluate model performance; B4: LSTM model architecture design: Construct an LSTM model with an input layer dimension of 3, corresponding to the three parameters of motor current, valve position feedback, and gearbox temperature. The hidden layer has h LSTM units, divided into l layers. The output layer dimension is 1, which is the fault probability. The forget gate Input Gate Output Gate and candidate cell states Calculated by the following formulas: Current cell status and hidden state for: The final output failure probability prediction value B4: Model training: The binary cross entropy loss function is used to calculate the loss value of the entire training batch. The binary cross entropy loss function is as follows: For each sample i, is the true fault state of sample i, which can be 0 or 1. is the failure probability predicted by the model, and is calculated separately. and The values of the two items are then added together, the sum of the above values of all m1 samples is accumulated, and then divided by the number of samples m1 and minus the sign to get the loss value L of the training batch. Using the gradient descent method, the learning rate is set to α. In each training cycle, the gradient of the loss function with respect to the weight and bias is calculated. And update the weights and biases; Use the chain rule to find the partial derivative of the loss function L with respect to the weights and biases in the model to get the gradient For weight W, the update formula is Where α is the learning rate, which controls the step size of each update, is the gradient of the loss function with respect to the weight W, which is obtained by subtracting the learning rate α from the weight W and the gradient The product of , to achieve weight update; For bias b, the update formula is The principle is similar to weight update, and the bias value is updated by subtracting the product of the learning rate and the bias gradient; B5: Model evaluation: On the test set D test Evaluate the model performance and calculate the accuracy Acc, recall Rec and F1 value F1; The calculation method of the accuracy Acc is as follows: The calculation method of the recall rate Rec is as follows: The calculation method of the F1 value is as follows: B6: Real-time fault warning: Real-time acquisition of the motor current I of the electric actuator real , valve position feedback P real , gearbox temperature T real , using the same data cleaning and standardization methods as historical data, the standardized real-time parameter I real ′、P real ′、T real ′, input the standardized real-time parameters into the trained LSTM model to obtain the fault probability prediction value like θ refers to the preset threshold, which determines that the equipment may fail and triggers a fault warning; B7: Performance degradation trend analysis based on sliding window regression model: Select the sliding window size as s, for the time series data of motor current I, valve position feedback P, and gearbox temperature T, starting from t = 1, take the window [I t ,I t+1 ,…,I t+s-1 ]、[P t ,P t+1 ,…,P t+s-1 ]、[T t ,T t+1 ,…,T t+s-1 ], each time the window slides one time step, for the data in each sliding window, taking the motor current as an example, the linear regression model y t =β0+β1t+∈, use the least squares method to estimate the regression coefficients β0 and β1, the goal is to minimize the residual sum of squares Based on the established regression model, the motor current, valve position feedback, and gearbox temperature values for the next p time steps are predicted. If the predicted values show a continuous decline in performance indicators and the decline exceeds the preset range, the equipment is considered to be at risk of performance degradation and maintenance measures must be taken in advance. B8: Warning level classification: Through statistical analysis of the historical normal operation data of the equipment, the average value of the motor current, the average value of the valve position feedback, and the average value of the gearbox temperature are determined, and the motor current reference value μ is obtained based on the average value. I , valve position feedback reference value μ P , gearbox temperature reference value μ T , when the motor current I real satisfy Or valve position feedback P real satisfy Or the gearbox temperature Treal meets When the yellow warning is triggered, the motor current I real Exceeding the preset safety threshold I th , or valve position feedback P real Exceeding the preset safety threshold P th , or gearbox temperature T real Exceeding the preset safety threshold T th A red alert is triggered.

Images