Intelligent actuator adaptive control method and system based on Internet of Things

Abstract

The invention relates to an intelligent actuator self-adaptive control method and system based on the Internet of Things, and belongs to the technical field of automation and intelligent control. The method comprises the following steps: integrating an actuating mechanism driven by a servo motor, a multi-mode sensor group, a microprocessor and a communication unit through an intelligent actuator designed in a mechanical-electrical integration manner; executing parameters are collected in real time and fed back to the microprocessor; the operation state parameter and the control instruction are encrypted and transmitted to the cloud server through the communication unit; the cloud server analyzes the system log through a mode clustering algorithm, identifies an abnormal working condition deviating from a normal operation sequence, and transmits a detection result to the terminal application; a terminal application receives multi-mode instruction input through a touch panel, maps the input into an operation instruction based on a rule engine, generates a control strategy and uploads the control strategy to a cloud server. The self-adaptive control of the intelligent actuator is realized, the working efficiency and precision of the actuator are improved, and remote monitoring and intelligent management are realized.

Description

Technical Field

[0001] The present invention belongs to the field of automation and intelligent control technology, and specifically relates to an adaptive control method and system for an intelligent actuator based on the Internet of Things. Background Art

[0002] Traditional actuators rely on local controllers based on PLCs or microcontrollers, and are controlled by manual adjustment or preset parameters. This model leads to the following problems: Lack of remote monitoring: Operational data (such as location, temperature, and current) cannot be obtained in real time, requiring reliance on on-site inspections, resulting in significant response delays. Data island effect: The local control system lacks a communication interface with the upper management platform (e.g., insufficient support for Modbus and Profibus protocols), resulting in information discontinuity. Limited flexibility: Adjusting parameters requires physical contact with the equipment, which cannot adapt to dynamic working conditions.

[0003] Traditional methods simplify the actuator into a linear system (amplifier + integrator), but actual parameters (such as friction coefficient and load inertia) vary significantly during operation. Relying on trial-and-error to adjust PID parameters is time-consuming and difficult to adapt to sudden environmental changes. Under complex operating conditions, the system exhibits strong nonlinearity (such as valve core vibration) and coupling characteristics (such as multi-degree-of-freedom robotic arms), rendering traditional linear control strategies ineffective. Furthermore, early processors (such as 8-bit microcontrollers) struggled to run advanced algorithms (such as model predictive control), limiting real-time performance. Mechanical transmission backlash and motor inertia led to delayed command response, exacerbating overshoot and oscillation.

[0004] Existing actuators mostly use analog signals (4-20mA) or single bus protocols (such as RS485), which makes it difficult to seamlessly connect with industrial Internet platforms (supporting OPC UA and MQTT); high-precision sensor data (such as 1 / 250 resolution positioning signals) is sensitive to network latency, and local devices lack data cleaning and feature extraction capabilities, and need to rely on cloud processing, increasing response delays; while intelligent algorithms require a large amount of historical data for training, traditional systems lack a data collection mechanism, and fixed-parameter PID cannot adapt to sudden load changes. Summary of the Invention

[0005] To solve the above problems existing in the prior art, the present invention provides an adaptive control method and system for intelligent actuators based on the Internet of Things. The objectives of the present invention can be achieved through the following technical solutions: An adaptive control method for an intelligent actuator based on the Internet of Things, comprising: S1: An intelligent actuator with mechatronic design integrates a servo motor-driven actuator, a multimodal sensor group, a microprocessor, and a communication unit. The multimodal sensor group collects execution parameters in real time and feeds them back to the microprocessor, which generates control instructions based on the motor servo control model. The motor servo control model uses the recursive least squares method to identify the electromagnetic torque and mechanical motion parameters offline, and adopts model reference adaptive control to achieve online parameter dynamic update; S2: Encrypting and transmitting the operating status parameters and control instructions to the cloud server via the communication unit; the cloud server uses a time series database to store data, analyzes system logs using a pattern clustering algorithm, identifies abnormal operating conditions that deviate from the normal operating sequence, and transmits the detection results to the terminal application; S3: The terminal application receives multimodal command input via the touch panel, maps the input into operation instructions based on the rule engine, generates a control strategy, and uploads it to the cloud server; the cloud server integrates real-time status parameters and historical operation data, optimizes the control strategy by integrating dynamic feedforward compensation and feedback correction compensation, and sends it to the microprocessor via the MQTT protocol; S4: The microprocessor generates a PWM modulation instruction according to the optimized control strategy to adjust the torque output characteristics of the servo motor, wherein the feedback correction compensation amount generates an optimal compensation voltage through the extended state observer to shorten the load step response time.

[0006] Specifically, the multimodal sensor group in S1 includes: A high-precision encoder provides real-time feedback of speed and position signals; a thermocouple array detects temperature rise signals of the motor and transmission components; the signals are amplified and filtered before being input into the microprocessor.

[0007] Specifically, the construction of the motor servo control model includes: Construct multi-source disturbance terms, including load mutation disturbance terms and parameter mismatch disturbance terms; Adopt dynamic compensation strategy to dynamically adjust proportional / integral coefficients by integrating incremental PID and fuzzy kernel; Among them, the execution of the dynamic compensation strategy satisfies: The feedforward compensation is adjusted in real time by the control strategy issued by the cloud server to pre-compensate for known disturbances; The feedback correction amount dynamically adjusts the uncompensated disturbance based on real-time feedback of the execution parameters.

[0008] Specifically, the feedback correction amount is generated as follows: The unmodeled dynamics are treated as sum disturbances by using an extended state observer, and a state space model is established to estimate the disturbance value in real time. Based on the disturbance estimation value, the model predictive control rolling optimization is adopted to generate the optimal compensation voltage sequence. The multi-source disturbance terms specifically include the load mutation disturbance term and the parameter mismatch disturbance term.

[0009] Specifically, the abnormal operating condition detection in S2 includes: Use the sliding window algorithm to align the real-time data in segments; When there is a time lag between real-time data and historical operating condition data, the missing points are filled in through interpolation algorithms; The multi-dimensional dataset is input into the LSTM network to predict the disturbance type and trigger feedforward compensation.

[0010] Specifically, the execution of the rule engine in S3 includes: Convert multimodal command input into structured data through a digital signal parser; A three-level rule base architecture is used for business logic management: Basic rule base: stores parameter threshold rules and state machine-driven mode switching logic; Scenario rule library: binds rule combinations for specific working conditions and sets fast response modes; Dynamic rule base: Achieve online rule optimization and iteration through a dual-engine architecture.

[0011] Specifically, the optimization of the control strategy in S3 includes: The steps for generating dynamic feedforward compensation include: Compare the deviation between the actual output and the reference model output; Combined with the disturbance type predicted by the LSTM network, the feedforward gain matrix is ​​dynamically adjusted through the gradient descent method; Steps for generating feedback correction compensation: Based on the online parameter identification results, interpolation optimization is performed through the compensation lookup table; The control strategy fusion steps include: Assign weights to the feedforward compensation and feedback correction compensation according to the confidence level of the working condition; The Lagrange multiplier method is used to ensure that the control quantity sequence meets the physical constraints of the actuator.

[0012] Specifically, the weight distribution satisfies: the feedback correction compensation weight is greater than 70% during steady-state operation, and the feedforward compensation weight is greater than 50% during sudden load changes.

[0013] An intelligent actuator adaptive control system based on the Internet of Things, comprising: Edge execution module: Integrates servo motors, multimodal sensor groups, and microprocessors for real-time data acquisition and local control command generation; Cloud collaboration module: This includes a time series database and anomaly detection unit, and uses a pattern clustering algorithm to align historical operating data with real-time data. Strategy optimization module: deployed on the cloud server, configured with a dynamic feedforward compensator and an extended state observer, and outputs a control strategy for the fusion compensation amount; Human-computer interaction module: includes a touch panel and a rule engine, mapping multimodal inputs into operational instructions; Among them, the edge execution module and the cloud collaboration module communicate through the encrypted MQTT protocol, and the output of the strategy optimization module directly drives the microprocessor to adjust the PWM modulation instructions.

[0014] Specifically, the microprocessor runs a real-time operating system and supports a fusion control algorithm of incremental PID and fuzzy kernel; The extended state observer is coupled with the model predictive control, and the generated optimal compensation voltage sequence is directly input into the motor drive circuit.

[0015] The beneficial effects of the present invention are: Through the adaptive control method of intelligent actuator based on the Internet of Things provided by the present invention, precise control of the actuator is achieved, and the stability and response speed of the system are significantly improved. First, the integration of the multimodal sensor group enables the system to collect and feedback execution parameters in real time, providing a reliable data basis for real-time analysis of the microprocessor. Secondly, the introduction of the cloud server not only realizes the remote storage of data and abnormal working condition detection, but also can optimize the control strategy in combination with real-time state parameters, further improving the intelligence level of the system. Finally, through PWM modulation and adjustment of the torque output characteristics of the servo motor, precise regulation of the operating state of the actuator is achieved, meeting the control requirements under various complex working conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0017] Figure 1 This is a structural diagram of an intelligent actuator adaptive control system based on the Internet of Things of the present invention. DETAILED DESCRIPTION

[0018] In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0019] See also Figure 1 , an adaptive control method for an intelligent actuator based on the Internet of Things, comprising: S1: An intelligent actuator with mechatronic design integrates a servo motor-driven actuator, a multimodal sensor group, a microprocessor, and a communication unit. The multimodal sensor group collects execution parameters in real time and feeds them back to the microprocessor, which generates control instructions based on the motor servo control model. The motor servo control model uses the recursive least squares method to identify the electromagnetic torque and mechanical motion parameters offline, and adopts model reference adaptive control to achieve online parameter dynamic update; S2: Encrypting and transmitting the operating status parameters and control instructions to the cloud server via the communication unit; the cloud server uses a time series database to store data, analyzes system logs using a pattern clustering algorithm, identifies abnormal operating conditions that deviate from the normal operating sequence, and transmits the detection results to the terminal application; S3: The terminal application receives multimodal command input via the touch panel, maps the input into operation instructions based on the rule engine, generates a control strategy, and uploads it to the cloud server; the cloud server integrates real-time status parameters and historical operation data, optimizes the control strategy by integrating dynamic feedforward compensation and feedback correction compensation, and sends it to the microprocessor via the MQTT protocol; S4: The microprocessor generates a PWM modulation instruction according to the optimized control strategy to adjust the torque output characteristics of the servo motor, wherein the feedback correction compensation amount generates an optimal compensation voltage through the extended state observer to shorten the load step response time.

[0020] Specifically, the multimodal sensor group in S1 includes a high-precision encoder for real-time feedback of rotational speed and position and an integrated thermocouple for detecting motor overheating or abnormal temperature rise of transmission components. The signal is input into the microprocessor after amplification and filtering.

[0021] In this embodiment, the servo motor drive uses a high-precision electrical actuator with fast dynamic response (microsecond adjustment) and adjustable torque output. The multimodal sensing network integrates a high-precision encoder and a thermocouple array to realize multi-dimensional parameter acquisition such as position, speed, and temperature. The microprocessor runs a real-time operating system (RTOS) and supports the integrated control of PID algorithm and load compensation.

[0022] Specifically, the motor servo control model in S1 performs offline parameter identification based on electromagnetic torque and mechanical motion through the recursive least squares method, and realizes dynamic update of online parameters in combination with model reference adaptive control; and constructs multi-source disturbance terms; eliminates the multi-source disturbance terms through the dynamic compensation strategy, and dynamically adjusts the proportional / integral coefficients through the fusion of incremental PID and fuzzy kernel to eliminate steady-state errors.

[0023] In this embodiment, based on the static characteristics of the motor, the electromagnetic torque equation is simplified, and only core parameters such as the stator resistance, rotor resistance, and stator / rotor inductance are retained to reduce the computational complexity. A pseudo-random M sequence (such as that generated by a linear feedback shift register) is used as the excitation signal, and the dynamic response of the motor is stimulated by step voltage injection to avoid voltage errors caused by dead time.

[0024] The offline identification objective function is: , in, , R s is the stator resistance, L s is the stator inductance, R r is the rotor resistance, L r is the rotor inductance; For input voltage, current and other observation data vectors, the Butterworth filter equation is solved by improving the Euler numerical solution to directly obtain the signal derivative information and reduce the discretization error.

[0025] The online parameter dynamic update method is as follows: an ideal motor model is constructed based on the offline identification results to output the reference speed; the actual motor model parameters are updated through online RLS to output the actual speed; the parameters are dynamically adjusted according to the error between the reference speed and the actual speed.

[0026] Through incremental PID dynamic parameter adjustment, the discretization equation is: , The conditional integration method is used to freeze the integral term when the error exceeds a threshold to prevent the actuator from saturating.

[0027] Specifically, the multi-source disturbance terms include load mutation disturbance terms and parameter mismatch disturbance terms.

[0028] Specifically, the dynamic compensation strategy adopts a method that combines feedforward compensation and feedback correction according to the characteristics of the multi-source disturbance terms; the feedforward compensation part is adjusted in real time by the control strategy sent by the cloud server received by the microprocessor to pre-compensate for known disturbance terms; the feedback correction part dynamically adjusts the compensation amount for the disturbance terms that are not fully compensated based on the real-time feedback of the execution parameters.

[0029] Specifically, the compensation amount of the feedback correction generates an optimal compensation voltage through an extended state observer rolling optimization, thereby shortening the load step response time.

[0030] Specifically, the abnormal operating condition detection method in S2 clusters system logs according to patterns and identifies abnormal operating states that deviate from normal operating sequences.

[0031] Specifically, the terminal application obtains multimodal command input through the touch panel parameter setting interface, including manually inputting control parameters and selecting a preset control mode, and performs condition-action mapping on the multimodal command input based on the rule engine to obtain an operation instruction; The rule engine uses a rule base to store predefined business logic rules, and converts the multimodal instruction input into structured data through a digital signal parser. The business logic rules include threshold rules for manually input control parameters and switching logic for preset control modes; the business logic rules in the rule base are matched according to the structured data, and the matched rules are converted into specific control strategies.

[0032] In this embodiment, a three-level rule base architecture is adopted to realize modular management of business logic. The basic rule base is used to store core business logic rules, define parameter boundary conditions through threshold rules, and set mode switching rules based on conditional judgment of the state machine. The scenario rule base is a combination of rules for specific working conditions and sets a rapid response mode; the dynamic rule base realizes online optimization of rules and implements rule iteration through a dual-engine architecture.

[0033] Specifically, the optimization method of the cloud server in S3 to optimize the control strategy in combination with the real-time state parameters of the actuator is: The cloud server integrates the real-time status parameters of the actuator and the historical operating data, aligns the real-time data of the multimodal sensor group with the historical operating condition data in the cloud time series database, compares the difference between the actual output and the reference model output through the cloud server, and dynamically adjusts the feedforward compensation to generate a dynamic feedforward compensation; the cloud uses a compensation lookup table and combines the online parameter identification results to perform interpolation optimization to obtain the feedback correction compensation; the control strategy sent by the terminal application is integrated with the feedforward compensation and the feedback correction compensation to obtain the optimal control quantity sequence, and sends it to the edge microprocessor through the MQTT protocol.

[0034] In this embodiment, the cloud server uses timestamp synchronization technology to align the real-time data from a multimodal sensor suite (encoders, thermocouples, etc.) with historical operating condition data in a time series database. A sliding window algorithm is used to segment the real-time data, ensuring consistency in dimensions such as sampling frequency and timestamp offset. For example, if there is a time lag between real-time speed data and historical temperature data, an interpolation algorithm is used to fill in the missing points, creating a time-aligned multidimensional dataset.

[0035] Dynamic feedforward compensation is implemented based on the following steps: Reference model comparison: The cloud server has a built-in ideal motor model (electromagnetic torque equation for offline parameter identification), which compares the actual output (speed, torque) with the reference model output in real time and calculates the deviation.

[0036] Disturbance prediction: Combined with historical operating condition data (load mutation pattern library), the LSTM network is used to predict the types of disturbances that may occur in the future. If historical data shows a sudden temperature rise before a load mutation, feedforward compensation is triggered in advance.

[0037] Compensation Generation: Based on the deviation and predicted disturbance, the feedforward gain matrix is ​​dynamically adjusted using a gradient descent method to generate feedforward compensation tailored to the current operating conditions. This process supports multivariable coupled compensation, such as simultaneous compensation for torque fluctuations and mechanical backlash changes caused by thermal expansion.

[0038] Interpolation optimization of feedback correction compensation: The extended state observer treats unmodeled dynamics (such as sudden changes in drive shaft friction) as a "sum disturbance" and estimates the disturbance value in real time using the state space model; Based on disturbance estimates, Model Predictive Control (MPC) is used to continuously optimize the compensation voltage sequence, shortening load step response time. For example, during motor acceleration, the ESO predicts the risk of current overshoot and injects a reverse compensation voltage in advance to suppress oscillations.

[0039] In this embodiment, the cloud server integrates the original control strategy (speed setting value), dynamic feedforward compensation, and feedback correction compensation issued by the terminal application: Weights are assigned based on the confidence level of different operating conditions. During steady-state operation, feedback correction is primarily used (weight >70%), while feedforward compensation is increased to over 50% during sudden load changes. Lagrange multipliers are used to ensure that the fused control sequence meets the physical limitations of the actuator (maximum current, temperature threshold). During the offline phase, a pseudo-random M sequence is used to excite the motor and identify core parameters. During the online phase, model reference adaptive control (MRAC) is combined to dynamically update parameters to avoid parameter drift caused by temperature rise.

[0040] An intelligent actuator adaptive control system based on the Internet of Things, comprising: Edge execution module: Integrates servo motors, multimodal sensor groups, and microprocessors for real-time data acquisition and local control command generation; Cloud collaboration module: This includes a time series database and anomaly detection unit, and uses a pattern clustering algorithm to align historical operating data with real-time data. Strategy optimization module: deployed on the cloud server, configured with a dynamic feedforward compensator and an extended state observer, and outputs a control strategy for the fusion compensation amount; Human-computer interaction module: includes a touch panel and a rule engine, mapping multimodal inputs into operational instructions; Among them, the edge execution module and the cloud collaboration module communicate through the encrypted MQTT protocol, and the output of the strategy optimization module directly drives the microprocessor to adjust the PWM modulation instructions.

[0041] Specifically, the microprocessor runs a real-time operating system and supports a fusion control algorithm of incremental PID and fuzzy kernel; The extended state observer is coupled with the model predictive control, and the generated optimal compensation voltage sequence is directly input into the motor drive circuit.

[0042] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as a preferred embodiment as above, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments using the technical contents disclosed above without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An adaptive control method for intelligent actuators based on the Internet of Things, characterized in that: include: S1: An intelligent actuator with mechatronic design integrates a servo motor-driven actuator, a multimodal sensor group, a microprocessor, and a communication unit. The multimodal sensor group collects execution parameters in real time and feeds them back to the microprocessor, which generates control instructions based on the motor servo control model. The motor servo control model uses the recursive least squares method to identify the electromagnetic torque and mechanical motion parameters offline, and adopts model reference adaptive control to achieve online parameter dynamic update; S2: Encrypting and transmitting the operating status parameters and control instructions to the cloud server via the communication unit; the cloud server uses a time series database to store data, analyzes system logs using a pattern clustering algorithm, identifies abnormal operating conditions that deviate from the normal operating sequence, and transmits the detection results to the terminal application; S3: The terminal application receives multimodal command input via the touch panel, maps the input into operation instructions based on the rule engine, generates a control strategy, and uploads it to the cloud server; the cloud server integrates real-time status parameters and historical operation data, optimizes the control strategy by integrating dynamic feedforward compensation and feedback correction compensation, and sends it to the microprocessor via the MQTT protocol; S4: The microprocessor generates a PWM modulation instruction according to the optimized control strategy to adjust the torque output characteristics of the servo motor, wherein the feedback correction compensation amount generates an optimal compensation voltage through the extended state observer to shorten the load step response time.

2. The method according to claim 1, characterized in that The multimodal sensor group in S1 includes: A high-precision encoder provides real-time feedback of speed and position signals; a thermocouple array detects temperature rise signals of the motor and transmission components; the signals are amplified and filtered before being input into the microprocessor.

3. The method according to claim 1, characterized in that The construction of the motor servo control model includes: Construct multi-source disturbance terms, including load mutation disturbance terms and parameter mismatch disturbance terms; Adopt dynamic compensation strategy to dynamically adjust proportional / integral coefficients by integrating incremental PID and fuzzy kernel; Among them, the execution of the dynamic compensation strategy satisfies: The feedforward compensation is adjusted in real time by the control strategy issued by the cloud server to pre-compensate for known disturbances; The feedback correction amount dynamically adjusts the uncompensated disturbance based on real-time feedback of the execution parameters.

4. The method according to claim 3, characterized in that The generation of the feedback correction amount is specifically as follows: The unmodeled dynamics are treated as sum disturbances by using an extended state observer, and a state space model is established to estimate the disturbance value in real time. Based on the disturbance estimation value, the model predictive control rolling optimization is adopted to generate the optimal compensation voltage sequence. The multi-source disturbance terms specifically include the load mutation disturbance term and the parameter mismatch disturbance term.

5. The method according to claim 3, characterized in that The abnormal operating condition detection in S2 includes: Use the sliding window algorithm to align the real-time data in segments; When there is a time lag between real-time data and historical operating condition data, the missing points are filled in through interpolation algorithms; The multi-dimensional dataset is input into the LSTM network to predict the disturbance type and trigger feedforward compensation.

6. The method according to claim 3, characterized in that The execution of the rule engine described in S3 includes: Convert multimodal command input into structured data through a digital signal parser; A three-level rule base architecture is used for business logic management: Basic rule base: stores parameter threshold rules and state machine-driven mode switching logic; Scenario rule library: binds rule combinations for specific working conditions and sets fast response modes; Dynamic rule base: Achieve online rule optimization and iteration through a dual-engine architecture.

7. The method according to claim 1, characterized in that The optimization of the control strategy described in S3 includes: The steps for generating dynamic feedforward compensation include: Compare the deviation between the actual output and the reference model output; Combined with the disturbance type predicted by the LSTM network, the feedforward gain matrix is ​​dynamically adjusted through the gradient descent method; Steps for generating feedback correction compensation: Based on the online parameter identification results, interpolation optimization is performed through the compensation lookup table; The control strategy fusion steps include: Assign weights to the feedforward compensation and feedback correction compensation according to the confidence level of the working condition; The Lagrange multiplier method is used to ensure that the control quantity sequence meets the physical constraints of the actuator.

8. The method according to claim 1, characterized in that The weight distribution satisfies: the feedback correction compensation weight is greater than 70% during steady-state operation, and the feedforward compensation weight is greater than 50% during sudden load changes.

9. The method according to claim 1, characterized in that An Internet of Things-based intelligent actuator adaptive control system, used to execute the method according to any one of claims 1 to 8, characterized in that it includes: Edge execution module: Integrates servo motors, multimodal sensor groups, and microprocessors for real-time data acquisition and local control command generation; Cloud collaboration module: This includes a time series database and anomaly detection unit, and uses a pattern clustering algorithm to align historical operating data with real-time data. Strategy optimization module: deployed on the cloud server, configured with a dynamic feedforward compensator and an extended state observer, and outputs a control strategy for the fusion compensation amount; Human-computer interaction module: includes a touch panel and a rule engine, mapping multimodal inputs into operational instructions; Among them, the edge execution module and the cloud collaboration module communicate through the encrypted MQTT protocol, and the output of the strategy optimization module directly drives the microprocessor to adjust the PWM modulation instructions.

10. The system according to claim 9, characterized in that The microprocessor runs a real-time operating system and supports a fusion control algorithm of incremental PID and fuzzy kernel; The extended state observer is coupled with the model predictive control, and the generated optimal compensation voltage sequence is directly input into the motor drive circuit.

Images