Motor control system based on distributed control architecture

Abstract

The invention belongs to the technical field of motor distributed control, and particularly relates to a motor control system based on a distributed control architecture, the system comprises a demand analysis module, a routing module, a summary board control module and a control feedback module, and multi-motor cooperative control is realized through a five-level distributed node network; wherein the demand analysis module converts a process template demand into a three-dimensional process parameter space; the routing module realizes accurate instruction distribution through a tag hash mapping table; the summary board control module adopts a distributed instruction generation model, and combines fuzzy control and reinforcement learning algorithms to generate a motor control instruction; the control feedback module monitors the height deviation and torque distribution of the molding column in real time to realize closed-loop adjustment; through three-dimensional process parameter space analysis, a dynamic priority communication mechanism, a federal instruction generation model and a multi-stage real-time feedback closed loop, multi-column height synchronous error compensation adjustment is finally realized, and the precision and reliability of precision molding are remarkably improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of motor distributed control, and in particular relates to a motor control system based on a distributed control architecture. Background Art

[0002] Traditional multi-column molding control systems often rely on centralized architectures, suffering from core flaws such as inefficient 3D process parameter mapping, insufficient real-time node communication, and poor dynamic load adaptability. For example, in complex multi-column synchronous molding scenarios, existing technologies require manual point-by-point mapping of process template parameters to motor control commands, resulting in time-consuming trajectory planning and the introduction of cumulative errors. Furthermore, the traditional CAN bus has a fixed hierarchy, making it difficult to dynamically allocate communication resources. Large-scale node deployments are prone to data congestion, leading to loss of control over multi-column height and uneven pressure distribution during molding, seriously compromising mold accuracy. Summary of the Invention

[0003] In response to the shortcomings of the existing technology, the present invention proposes a motor control system based on a distributed control architecture, which includes a demand analysis module, a routing module, a summary board control module and a control feedback module, and realizes multi-motor collaborative control through a five-level distributed node network; the demand analysis module converts the process template requirements into a three-dimensional process parameter space; the routing module realizes accurate instruction distribution through a label hash mapping table; the summary board control module adopts a distributed instruction generation model, combines fuzzy control and reinforcement learning algorithms to generate motor control instructions; the control feedback module monitors the mold column height deviation and torque distribution in real time to achieve closed-loop adjustment; this application finally realizes multi-column height synchronization error compensation adjustment through three-dimensional process parameter space analysis, dynamic priority communication mechanism, federal instruction generation model and multi-level real-time feedback closed loop, significantly improving the accuracy and reliability of precision molds.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] A motor control system based on a distributed control architecture includes: a demand analysis module, a routing module, a summary board control module, and a control feedback module;

[0006] The demand analysis module is used to obtain a three-dimensional process parameter space based on the process template requirements input by the front end and the preset three-dimensional modeling analysis model;

[0007] Transmitting the three-dimensional process parameter space to the summary board control module through the primary connection and the secondary connection in the distributed control node network in the routing module;

[0008] The summary board control module obtains a distributed control instruction set corresponding to each summary board based on the three-dimensional process parameter space combined with a preset distributed instruction generation model and the parameter space of the motor drive board corresponding to each summary board;

[0009] Based on the distributed control instruction set corresponding to each aggregation board, the corresponding motor is driven and controlled through the three-level connection, the four-level connection and the motor drive board in the distributed control node network;

[0010] At the same time, the control feedback module monitors the corresponding multi-column molding results and motor operating parameters during the motor drive control process in real time and feeds them back to the distributed instruction generation model, and adjusts the operating control parameters of each motor until the multi-column molding deviation meets the process template requirements.

[0011] Specifically, the demand analysis module includes a process analysis unit and a three-dimensional modeling unit; the routing module includes a node network construction unit and a routing unit;

[0012] The process analysis unit is used to obtain the process template parameter space and the mold column distribution state space according to the input process template requirements and the preset analysis sub-model;

[0013] The three-dimensional modeling unit obtains a three-dimensional process model of the target to be molded based on the process template parameter space and the mold column distribution state space in combination with the three-dimensional coordinates with the center of gravity of the target to be molded as the origin, and marks the process template parameters at corresponding positions on the three-dimensional process model;

[0014] The node network construction unit is used to construct a distributed control node network through a graph algorithm and a distributed framework based on the front end, routing board, aggregation board, motor drive board, motor position information, operation information and signal transmission line information;

[0015] The routing unit is used to perform directed transmission of information instructions between nodes according to the distributed control node network.

[0016] Specifically, the summary board control module includes a parameter decomposition unit, a motion planning unit, and an instruction generation unit;

[0017] The parameter decomposition unit obtains a process parameter sequence corresponding to each motor under each motor drive board based on the three-dimensional process model and the marked process parameters combined with a preset processing mapping space;

[0018] The motion planning unit is used to obtain the height trajectory pulse sequence corresponding to the mold columns of all motors at each moment through a trajectory synchronization planning algorithm based on the three-dimensional process model and the marked process parameters, the process parameter sequence corresponding to each motor, and the speed and acceleration parameters of each motor for controlling the extension and contraction of the mold column;

[0019] The processing mapping space is constructed by the process parameter sequence corresponding to each motor and the motor label corresponding to each node in the collaborative processing subnet in the distributed control node network;

[0020] The collaborative processing subnet is constructed by combining all motors and the corresponding data interaction information between motors with a graph algorithm;

[0021] The instruction generation unit is used to obtain the telescopic control instruction corresponding to each motor according to the process parameter sequence corresponding to each motor and the height trajectory pulse sequence corresponding to the mold columns of all motors at each moment in combination with the distributed instruction generation model.

[0022] Specifically, the control feedback module includes a processing monitoring unit, a motor operation evaluation unit, and a monitoring feedback unit;

[0023] The processing monitoring unit is used to monitor in real time the height deviation of the shaping column controlled by each motor and the torque of the molding column corresponding to each position of the overall mold, and obtain a processing result deviation sequence;

[0024] The motor operation evaluation unit is used to monitor the interaction status between the operation data and the motor during the motor processing control process in real time according to a preset state monitoring node network, and to evaluate the motor operation status in real time through a built-in evaluation algorithm to obtain the motor operation state space;

[0025] The monitoring feedback unit is used to feed back the processing result deviation sequence, the torque distribution deviation corresponding to each motor and the motor operation state space to the distributed instruction generation model and the distributed control node network, so as to adjust the motor control process in real time and track and locate the abnormal state motor in real time.

[0026] Specifically, the connection process of the distributed control node network includes:

[0027] Set the front end as the first-level node, the signal routing board as the second-level node, the motor drive summary board as the third-level control node, the motor drive board as the fourth-level node, and the motor as the fifth-level node;

[0028] The first-level node is connected to at least one of the second-level nodes through at least one first-level connection established by the network port or RS232; each of the second-level nodes is connected to at least one third-level node through at least one second-level connection established by the CAN bus and the tag information of the corresponding third-level node is stored in the corresponding second-level node;

[0029] Each of the three-level nodes is connected to M four-level nodes through a three-level connection set constructed by the CAN bus and a 24V power supply, and the label information of the M four-level nodes is stored in the corresponding three-level node; each of the four-level nodes is connected to N five-level nodes through N four-level connections constructed by N motor lines, and the labels of the N five-level nodes are stored in the corresponding four-level nodes;

[0030] Each of the five-level nodes corresponds to each mold column one by one and all the five-level nodes are connected through an interactive control connection set constructed by data interaction information between the nodes;

[0031] A label hash mapping table is constructed based on the labels maintained by each level node and the connection relationship of the corresponding level in combination with a hash algorithm, and the constructed label hash mapping table is saved in the secondary node.

[0032] Specifically, the distributed instruction generation model is obtained by integrating the instruction generation sub-models with the same number of motors and a joint control sub-model through a federated algorithm. The instruction generation sub-model is used to obtain the corresponding expansion and contraction control instructions for each motor based on the process parameter sequence corresponding to each motor and the fuzzy control algorithm.

[0033] The joint control model is used to obtain the control deviation compensation corresponding to all motors based on the operating status of all motors, the transmission delay distribution of the command signal, the height trajectory of the mold columns corresponding to all motors at each moment, the overall mold deviation and the torque deviation corresponding to all motors combined with the reinforcement learning algorithm; the control deviation compensation corresponding to all motors includes the height and torque deviation corresponding to each mold column and the transmission signal delay deviation corresponding to each motor.

[0034] Specifically, the working process of the distributed control node network includes:

[0035] Based on the first-level node and the process template requirements, a process template parameter space and a mold column distribution state space are obtained, and the process template parameter space and the mold column distribution state space are transmitted to the second-level node through the first-level connection;

[0036] The second-level node decomposes and marks the process template parameter space and the mold column distribution state space according to the saved label hash mapping table, and at the same time transmits the marking results and all the third-level connections to the corresponding third-level nodes according to the marking. Each of the third-level nodes generates a telescopic control instruction set corresponding to all motors under each third-level node through a distributed instruction generation model combined with the process parameter sequence corresponding to each marked motor.

[0037] Specifically, the working process of the distributed control node network also includes:

[0038] The expansion and contraction control instruction sets corresponding to all motors under each third-level node are transmitted to the corresponding fourth-level nodes through the third-level connection relationship and the label in the label hash mapping table. Each of the fourth-level nodes converts the received expansion and contraction control instruction subset into a step pulse signal corresponding to each motor according to the stored motor label, and controls the operating status of each fifth-level node in real time.

[0039] The state monitoring node network has the same structure as the distributed control node network and the nodes correspond one to one. Each level corresponds to the working status of the node and the deviation between the motor operating parameters and the process parameters;

[0040] The status monitoring node network is used to monitor in real time the data transmission delay in each connection relationship within the distributed control node network, the corresponding operating status of each level node, the height deviation of the mold columns corresponding to all five-level nodes, the corresponding height trajectory of all mold columns at each moment, and the torque deviation of all motors.

[0041] Specifically, the working process of the distributed control node network also includes:

[0042] At the same time, the abnormality assessment algorithm configured for each monitoring node is used to perform abnormality assessment according to the operating status corresponding to each level node, and the process assessment algorithm configured for the collaborative processing subnet is used to perform real-time assessment of the mold column height deviation and torque deviation corresponding to each position point and the corresponding height trajectory of all mold columns at each moment, thereby obtaining abnormality assessment results and process assessment results respectively;

[0043] The process evaluation results are fed back to the distributed instruction generation model to obtain the process deviation adjustment parameter sequences corresponding to all motor controls to adjust all motor control states in real time. At the same time, the abnormality evaluation results are fed back to the state monitoring node network to perform real-time traceability and early warning of abnormal nodes in the distributed control node network.

[0044] Specifically, the construction process of the instruction generation sub-model and the joint control sub-model includes:

[0045] Constructing a single input state of a single instruction generation sub-model based on the process parameter sequence corresponding to each motor, the motor's historical operating parameters, and real-time feedback data, and constructing a global input state based on the global operating state of the distributed control node network, the molding process evaluation results, the global operating abnormality evaluation results, and the characteristics of the material to be molded;

[0046] Construct local execution action trigger information based on the single input state corresponding to each motor, including:

[0047] When the three-dimensional process model or the processing mapping space changes, the first execution action information is triggered, that is, the corresponding motor is controlled to control the corresponding mold column according to the input control instruction;

[0048] When the mold column height deviation exceeds a preset threshold or the corresponding torque distribution does not meet the process requirements, the second execution action information is triggered;

[0049] At the same time, based on the global input state, build global execution action trigger information, including:

[0050] When the second execution action information is triggered or the height trajectory deviation of all mold columns at each moment is greater than a preset trajectory deviation threshold, the global first execution action information is triggered; that is, the process deviation adjustment parameter sequence is generated through the joint control sub-model;

[0051] When the data transmission delay between nodes exceeds a threshold or an abnormality occurs in at least one node in the distributed control node network, the global second execution action information is triggered; that is, information transmission adjustment and abnormal node tracing are performed through the joint control sub-model.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] In response to the deficiencies in the prior art, the present invention maps process template parameters into a three-dimensional process parameter space through a three-dimensional modeling analytical model, and dynamically optimizes communication paths in combination with a five-level hierarchical architecture (front end → routing board → aggregation board → driver board → motor) of a distributed control node network and a label hash mapping table, thereby solving the problem of multi-node data transmission congestion. Fuzzy control and reinforcement learning algorithms are integrated based on a federated learning framework to achieve joint optimization of local instruction generation and global collaborative compensation, thereby improving the accuracy of multi-column height synchronization. The state monitoring node network is used to collect mold column height deviation, torque distribution and communication delay data in real time, and the abnormality assessment algorithm and process assessment algorithm are combined to dynamically adjust control parameters and trace abnormal nodes, thereby enhancing the system's adaptability to dynamic loads and external interference. Distributed redundant communication and adaptive current regulation technology further suppress the risk of cascading failures, thereby ensuring control stability and reliability in complex process scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a structural diagram of a motor control system based on a distributed control architecture according to Example 1 of the present invention;

[0055] Figure 2 This is a schematic diagram of a motor control system based on a distributed control architecture according to Example 1 of the present invention. DETAILED DESCRIPTION

[0056] Example 1

[0057] See also Figure 1, an embodiment provided by the present invention: a motor control system based on a distributed control architecture, comprising: a demand parsing module, a routing module, a summary board control module, and a control feedback module;

[0058] The demand analysis module is used to obtain a three-dimensional process parameter space based on the process template requirements input by the front end and the preset three-dimensional modeling analysis model;

[0059] Furthermore, the 3D modeling parsing model in this embodiment is constructed by integrating a Chinese pre-trained text parsing model with a 3D modeling algorithm;

[0060] The routing module is used for building a distributed control node network and transmitting signals;

[0061] The summary board control module is used to control the generation of control instructions and the drive control of the motor;

[0062] The control feedback module is used for real-time monitoring, evaluation and feedback adjustment of the motor control system;

[0063] Transmitting the three-dimensional process parameter space to the summary board control module through the primary connection and the secondary connection in the distributed control node network in the routing module;

[0064] The summary board control module obtains a distributed control instruction set corresponding to each summary board based on the three-dimensional process parameter space combined with a preset distributed instruction generation model and the parameter space of the motor drive board corresponding to each summary board;

[0065] Based on the distributed control instruction set corresponding to each aggregation board, the corresponding motor is driven and controlled through the three-level connection, the four-level connection and the motor drive board in the distributed control node network;

[0066] At the same time, the control feedback module monitors the corresponding multi-column molding results and motor operating parameters during the motor drive control process in real time, and feeds back to the distributed instruction generation model to adjust the operating control parameters of each motor until the multi-column molding deviation meets the process template requirements.

[0067] Furthermore, the demand analysis module includes a process analysis unit and a three-dimensional modeling unit; the routing module includes a node network construction unit and a routing unit;

[0068] The process analysis unit is used to obtain the process template parameter space and the mold column distribution state space according to the input process template requirements and the preset analysis sub-model;

[0069] The three-dimensional modeling unit obtains a three-dimensional process model of the target to be molded based on the process template parameter space and the mold column distribution state space in combination with the three-dimensional coordinates with the center of gravity of the target to be molded as the origin, and marks the process template parameters at corresponding positions on the three-dimensional process model;

[0070] The node network construction unit is used to construct a distributed control node network through a graph algorithm and a distributed framework based on the front end, routing board, aggregation board, motor drive board, motor position information, operation information and signal transmission line information;

[0071] The routing unit is used to perform directed transmission of information instructions between nodes according to the distributed control node network.

[0072] Furthermore, the summary board control module includes a parameter decomposition unit, a motion planning unit, and an instruction generation unit;

[0073] The parameter decomposition unit obtains a process parameter sequence corresponding to each motor under each motor drive board based on the three-dimensional process model and the marked process parameters combined with a preset processing mapping space;

[0074] The motion planning unit is used to obtain the height trajectory pulse sequence corresponding to the mold columns of all motors at each moment through a trajectory synchronization planning algorithm based on the three-dimensional process model and the marked process parameters, the process parameter sequence corresponding to each motor, and the speed and acceleration parameters of each motor for controlling the extension and contraction of the mold column;

[0075] The processing mapping space is constructed by the process parameter sequence corresponding to each motor and the motor label corresponding to each node in the collaborative processing subnet in the distributed control node network;

[0076] The collaborative processing subnet is constructed by combining all motors and the corresponding data interaction information between motors with a graph algorithm;

[0077] The instruction generation unit is used to obtain the telescopic control instruction corresponding to each motor according to the process parameter sequence corresponding to each motor and the height trajectory pulse sequence corresponding to the mold columns of all motors at each moment in combination with the distributed instruction generation model.

[0078] Furthermore, the control feedback module includes a processing monitoring unit, a motor operation evaluation unit, and a monitoring feedback unit;

[0079] The processing monitoring unit is used to monitor in real time the height deviation of the shaping column controlled by each motor and the torque of the molding column corresponding to each position of the overall mold, and obtain a processing result deviation sequence;

[0080] The motor operation evaluation unit is used to monitor the interaction status between the operation data and the motor during the motor processing control process in real time according to a preset state monitoring node network, and to evaluate the motor operation status in real time through a built-in evaluation algorithm to obtain the motor operation state space;

[0081] The monitoring feedback unit is used to feed back the processing result deviation sequence, the torque distribution deviation corresponding to each motor and the motor operation state space to the distributed instruction generation model and the distributed control node network, so as to adjust the motor control process in real time and track and locate the abnormal state motor in real time.

[0082] Furthermore, the distributed instruction generation model is obtained by integrating instruction generation sub-models with the same number of motors and a joint control sub-model through a federated algorithm; the instruction generation sub-model is used to obtain the corresponding expansion and contraction control instruction for each motor based on the process parameter sequence corresponding to each motor in combination with the fuzzy control algorithm;

[0083] The joint control model is used to obtain the control deviation compensation corresponding to all motors based on the operating status of all motors, the transmission delay distribution of the command signal, the height trajectory of the mold columns corresponding to all motors at each moment, the overall mold deviation and the torque deviation corresponding to all motors combined with the reinforcement learning algorithm; the control deviation compensation corresponding to all motors includes the height and torque deviation corresponding to each mold column and the transmission signal delay deviation corresponding to each motor.

[0084] Furthermore, the connection process of the distributed control node network includes:

[0085] Set the front end as the first-level node, the signal routing board as the second-level node, the motor drive summary board as the third-level control node, the motor drive board as the fourth-level node, and the motor as the fifth-level node;

[0086] The first-level node is connected to at least one of the second-level nodes through at least one first-level connection established by the network port or RS232; each of the second-level nodes is connected to at least one third-level node through at least one second-level connection established by the CAN bus and the tag information of the corresponding third-level node is stored in the corresponding second-level node;

[0087] Each of the three-level nodes is connected to M four-level nodes through a three-level connection set constructed by the CAN bus and a 24V power supply, and the label information of the M four-level nodes is stored in the corresponding three-level node; each of the four-level nodes is connected to N five-level nodes through N four-level connections constructed by N motor lines, and the labels of the N five-level nodes are stored in the corresponding four-level nodes;

[0088] Each of the five-level nodes corresponds to each mold column one by one and all the five-level nodes are connected through an interactive control connection set constructed by data interaction information between the nodes;

[0089] A label hash mapping table is constructed based on the labels maintained by each level node and the connection relationship of the corresponding level in combination with a hash algorithm, and the constructed label hash mapping table is saved in the secondary node.

[0090] Further, see Figure 2 , is a schematic diagram of a motor control system based on a distributed control architecture in this embodiment, specifically including:

[0091] The front end corresponds to a computer user parameter input processing, the computer is connected to the signal routing board through the network port or RS232 to send data and receive feedback data, each routing board can connect up to 100 motor drive summary boards through the CAN bus, and each motor drive summary board can be connected to 10 motor drive boards through the CAN bus +24V power supply (such as Figure 2 01 motor drive board to 10 motor drive board), each motor drive board controls 10 stepper motors through 10 motor lines. Further, the stepper motors in this embodiment are 10mm micro motors; each motor drive summary board can control the position of 100 stepper motors and feedback the status of each motor;

[0092] Furthermore, the working process of the distributed control node network includes:

[0093] Based on the first-level node and the process template requirements, a process template parameter space and a mold column distribution state space are obtained, and the process template parameter space and the mold column distribution state space are transmitted to the second-level node through the first-level connection;

[0094] The second-level node decomposes and marks the process template parameter space and the mold column distribution state space according to the saved label hash mapping table, and at the same time transmits the marking results and all the third-level connections to the corresponding third-level nodes according to the marking. Each of the third-level nodes generates a telescopic control instruction set corresponding to all motors under each third-level node through a distributed instruction generation model combined with the process parameter sequence corresponding to each marked motor.

[0095] The expansion and contraction control instruction sets corresponding to all motors under each third-level node are transmitted to the corresponding fourth-level nodes through the third-level connection relationship and the label in the label hash mapping table. Each of the fourth-level nodes converts the received expansion and contraction control instruction subset into a step pulse signal corresponding to each motor according to the stored motor label, and controls the operating status of each fifth-level node in real time.

[0096] The state monitoring node network has the same structure as the distributed control node network and the nodes correspond one to one. Each level corresponds to the working status of the node and the deviation between the motor operating parameters and the process parameters;

[0097] The status monitoring node network is used to monitor in real time the data transmission delay in each connection relationship within the distributed control node network, the corresponding operating status of each level node, the height deviation of the mold columns corresponding to all five-level nodes, the corresponding height trajectory of all mold columns at each moment, and the torque deviation of all motors.

[0098] At the same time, the abnormality assessment algorithm configured for each monitoring node is used to perform abnormality assessment according to the operating status corresponding to each level node, and the process assessment algorithm configured for the collaborative processing subnet is used to perform real-time assessment of the mold column height deviation and torque deviation corresponding to each position point and the corresponding height trajectory of all mold columns at each moment, thereby obtaining abnormality assessment results and process assessment results respectively;

[0099] The process evaluation results are fed back to the distributed instruction generation model to obtain the process deviation adjustment parameter sequence corresponding to all motor controls. While all motor control states are adjusted in real time, the abnormality evaluation results are fed back to the state monitoring node network to perform real-time traceability, positioning and early warning of abnormal nodes in the distributed control node network.

[0100] Furthermore, the construction process of the instruction generation sub-model and the joint control sub-model includes:

[0101] Constructing a single input state of a single instruction generation sub-model based on the process parameter sequence corresponding to each motor, the motor's historical operating parameters, and real-time feedback data, and constructing a global input state based on the global operating state of the distributed control node network, the molding process evaluation results, the global operating abnormality evaluation results, and the characteristics of the material to be molded;

[0102] Furthermore, in this embodiment, the process parameter sequence includes process requirements such as target mold column height, speed, and acceleration; historical motor operating parameters such as historical position deviation, current fluctuation, and temperature change data; real-time feedback data such as current motor position, torque output, and communication delay status; environmental parameters such as load changes, temperature, and vibration interference signals; global operating status such as the real-time position, torque, and communication delay data of all motors; molding process evaluation results such as the distribution of multi-column height deviations, surface deformation data, and torque balance; node network status such as the topological health of the distributed control node network and the distribution of data transmission delays; and properties of the material to be molded such as material property changes and mechanical vibration interference.

[0103] Construct local execution action trigger information based on the single input state corresponding to each motor, including:

[0104] When the three-dimensional process model or the processing mapping space changes, the first execution action information is triggered, that is, the corresponding motor is controlled to control the corresponding mold column according to the input control instruction;

[0105] When the mold column height deviation exceeds a preset threshold or the corresponding torque distribution does not meet the process requirements, the second execution action information is triggered;

[0106] At the same time, based on the global input state, build global execution action trigger information, including:

[0107] When the second execution action information is triggered or the height trajectory deviation of all mold columns at each moment is greater than a preset trajectory deviation threshold, the global first execution action information is triggered; that is, the process deviation adjustment parameter sequence is generated through the joint control sub-model;

[0108] When the data transmission delay between nodes exceeds a threshold or an abnormality occurs in at least one node in the distributed control node network, a global second execution action information is triggered; that is, information transmission adjustment and abnormal node tracing are performed through the joint control sub-model;

[0109] Constructing a comprehensive reward function based on the local execution action trigger information and the global execution action trigger information;

[0110] A command generation sub-model is constructed based on a fuzzy control algorithm, and a joint control sub-model is constructed using a reinforcement learning algorithm. The single input state and the corresponding local execution action trigger information corresponding to each motor are input into the corresponding command generation sub-model. At the same time, the global input state and the global execution action trigger information are input into the joint control sub-model. Integrated training is performed through a federated algorithm framework combined with a comprehensive reward function to obtain a trained distributed command generation model.

[0111] The trained distributed instruction generation model is configured to a distributed control node network to perform real-time control on the motor control system, so that the molding process meets the process template requirements in real time.

[0112] Furthermore, in order to illustrate the above motor control system in more detail, this embodiment takes the multi-column synchronous shaping control in injection molding as an example and provides the following examples, specifically including:

[0113] Application Scenario: A precision injection molding production line uses 1,000 stepper motors to control the height of mold columns, synchronously shaping complex three-dimensional structures. The traditional system, due to inefficient parameter mapping and high communication latency, resulted in multi-column height deviations of ≥0.1mm, uneven mold pressure distribution, and scrap rates as high as 15%. With this system, optimized distributed control architecture, multi-column synchronization errors of ≤±0.01mm were achieved, reducing scrap rates to below 1%.

[0114] Step 1: Process template input and three-dimensional parameter space generation:

[0115] Implementation details:

[0116] Process template input: The operator inputs the molding process template (such as the target structure CAD file) through the front-end computer, including the target height, speed, acceleration and other parameters of the molding column.

[0117] Requirements parsing module work:

[0118] Process analysis unit: Extracts the process template parameter space through the analysis sub-model (based on the rule engine), including the target height sequence and distribution status (such as column spacing and arrangement topology) of 1,000 mold columns.

[0119] 3D modeling unit: Establish a 3D coordinate system with the center of gravity of the mold target as the origin, map the process parameters into a 3D process model, and mark key position points (such as column top coordinates) on the model surface.

[0120] Output: Generates a 3D process parameter space (JSON format), including the ID, target height, motion constraints, etc. of each mold column.

[0121] Principle: Through the rule engine and coordinate transformation algorithm, the 2D / 3D process template is converted into a machine-parseable parameter space, avoiding the errors of manual point-by-point mapping.

[0122] Step 2: Distributed control node network construction and label mapping:

[0123] Implementation details:

[0124] Node network building unit:

[0125] Based on the physical locations of the front end, routing board, aggregation board, driver board, and motor (such as the production line layout) and the communication link (CAN bus topology), a five-level distributed node network is constructed:

[0126] Level 1 node: front-end computer (IP: 192.168.1.100);

[0127] Secondary node: 2 routing boards (CAN ID: R001, R002);

[0128] Level 3 node: 10 aggregation boards (CAN ID: S001-S010);

[0129] Level 4 node: 100 driver boards (CAN ID: D001-D100);

[0130] Level 5 nodes: 1000 stepper motors (labels: M0001-M1000).

[0131] Optimize the connection relationship between nodes through graph algorithms (such as Dijkstra shortest path) to ensure the shortest communication path.

[0132] Label hash map generation:

[0133] Perform a SHA-256 hash operation on each node label (such as M0001-M0010, D001) to generate a unique hash value (such as M0001→0x3a7b...).

[0134] Bind the hash value with the node hierarchy and connection relationship and save it to the Flash storage of the secondary node (routing board).

[0135] Principle: The hash algorithm ensures the uniqueness of node labels, and the distributed framework dynamically optimizes communication paths to avoid the fixed-level congestion of the traditional CAN bus.

[0136] Step 3: Distributed command generation and motion planning:

[0137] Implementation details:

[0138] Parameter decomposition unit:

[0139] Decompose the 3D process model according to the machining mapping space (preset motor-mold column correspondence table) to generate the process parameter sequence of 1000 motors (e.g. M001 corresponds to the target height of 50mm, acceleration of 2m / s 2 ).

[0140] Motion Planning Unit:

[0141] Using trajectory synchronization planning algorithm (based on B-spline interpolation):

[0142] Input: target height, speed, acceleration of each motor;

[0143] Output: Synchronous height trajectory pulse sequence of all mold columns (time stamp + pulse number) to ensure phase synchronization of multi-column motion.

[0144] For example, mold columns M0001-M1000 are started synchronously at t=0s and reach the target height at the same time at t=2s.

[0145] Instruction Generation Unit:

[0146] Instruction generation sub-model (fuzzy control):

[0147] Input: historical position deviation of M0001 (e.g. ±0.02mm), current load torque (e.g. 1.2N·m);

[0148] Output: Adjust the pulse frequency (e.g. from 200Hz to 210Hz) to compensate for the deviation.

[0149] Joint control sub-model (reinforcement learning):

[0150] Input: global height deviation (e.g. M0005 deviation + 0.03mm), communication delay distribution (average 0.5ms);

[0151] Output: Generates pulse number compensation (+3 pulses) for M0005 and assigns higher communication priority to routing board R001.

[0152] Principle: Fuzzy control solves local dynamic adjustment, reinforcement learning achieves global collaborative optimization, and the federated learning framework integrates the parameters of both.

[0153] Step 4: Real-time control and feedback adjustment:

[0154] Implementation details:

[0155] Instruction issuance and execution:

[0156] The second-level node (routing board R001) distributes the instruction set to the corresponding third-level nodes (aggregation boards S001-S010) according to the label hash mapping table;

[0157] The third-level node transmits the instructions to the fourth-level node (driver board D001-D100) through the CAN bus, and the driver board converts the pulse signal into motor action.

[0158] For example, the driver board D001 receives the instruction "M0001:50mm@200Hz" and drives the motor M0001 to move at a pulse frequency of 200Hz.

[0159] Condition Monitoring Node Network:

[0160] Process monitoring unit: The laser ranging sensor is used to collect the mold column height in real time (for example, the actual height of M0001 is 49.98 mm), and the torque sensor monitors the pressure distribution (for example, the output torque of M0001 is 1.1 N·m).

[0161] Motor operation evaluation unit: Based on the condition monitoring node network, a sudden increase in M0005 communication delay (1.5ms) was detected, triggering the abnormality evaluation algorithm.

[0162] Feedback Adjustment:

[0163] Monitoring feedback unit: Feedback the M0001 height deviation (-0.02mm) and M0005 delay data to the joint control sub-model;

[0164] Joint control sub-model: generates M0001 pulse compensation (+2 pulses) and M0005 communication path switching to the backup routing board R002;

[0165] Instruction generation sub-model: Dynamically adjust the fuzzy rule weights of M0001 and give priority to responding to altitude deviations.

[0166] Principle: Real-time data closed-loop feedback combined with dynamic path switching suppresses communication anomalies and mechanical disturbances.

[0167] Step 5: Abnormal tracing and process optimization:

[0168] Implementation details:

[0169] Anomaly Assessment Algorithm:

[0170] The communication of the fourth-level node D050 was detected to be interrupted (due to a loose CAN line), and the problem was traced back to the physical location (the 10th driver board in the 5th area of ​​the production line) through the tag hash mapping table, triggering an audible and visual alarm.

[0171] Process evaluation algorithm:

[0172] Analyzing the surface deformation data of the mold column revealed deviations in material shrinkage (e.g., the actual shrinkage of ABS material was 0.5% > the expected 0.3%).

[0173] After the digital twin model simulation, a process parameter correction table is generated (such as increasing the target height by 0.2 mm).

[0174] Federated learning parameter update:

[0175] Each instruction generates a sub-model to upload local parameters (such as the M0001 fuzzy rule weight);

[0176] The sub-model aggregation parameters are jointly controlled and the globally optimized instruction generation model is issued.

[0177] Principle: Digital twins and federated learning enable adaptive optimization of process parameters, and anomaly tracing algorithms accurately locate fault points.

[0178] In summary, this embodiment first converts process template parameters into a structured three-dimensional process parameter space at the parameter parsing and instruction mapping level through the combination of rule engine and three-dimensional modeling algorithm. The coordinate system design with the center of gravity of the mold as the origin ensures geometric consistency and avoids error accumulation and semantic ambiguity in manual mapping. The decomposition mechanism of the processing mapping space realizes the lossless transmission of "global goal-local instruction", which upgrades the parameter parsing efficiency and accuracy from the traditional mode relying on manual experience to a digital process that can be automatically parsed by the machine, laying a data foundation for the precise execution of subsequent control strategies. Secondly, the optimization of the communication architecture is the key support for improving system performance. The five-level distributed node DotNet builds the optimal communication topology through the Dijkstra algorithm, reduces the number of signal transmission hops, and combines the SHA-256 hash mapping table to achieve unique identification and rapid positioning of nodes, minimizing communication delays and label conflict risks; the layered architecture limits the size of the broadcast domain and suppresses CAN bus conflicts, while the dynamic path switching mechanism enhances the system's ability to resist single-point failures, thereby achieving a double improvement in bandwidth utilization and communication robustness, providing an underlying guarantee for the reliable transmission of real-time control instructions; thirdly, at the control strategy level, the introduction of the federated learning framework breaks through the bottleneck of traditional centralized control; the fuzzy control sub-model generates smooth local adjustment instructions based on real-time status , suppressing motor oscillation; the reinforcement learning sub-model optimizes the compensation strategy through the global reward function to solve the "combinatorial explosion" problem of multi-motor coordination; the federal mechanism aggregates local model parameters while protecting data privacy to achieve continuous iteration of the control strategy, so that the system has both local response speed and global coordination accuracy, and effectively copes with dynamic load changes under complex working conditions; Fourth, the data feedback and exception handling module builds an autonomous closed loop of "perception-decision-execution"; the multi-source data fusion of laser ranging and torque sensors provides millimeter-level precision state feedback, providing reliable input for the control algorithm; the anomaly assessment algorithm combines the hash mapping table to achieve rapid fault location and physical traceability, Shorten the maintenance cycle; the digital twin model reversely corrects the process parameters through material property simulation, solves the systematic deviations caused by external disturbances such as temperature, and significantly improves the system's dynamic adaptability to complex environments; Fifth, resource scheduling and redundant design further enhance the engineering practicality of the system; the dynamic priority allocation mechanism based on TSN technology ensures the deterministic transmission of key instructions, and the supercapacitor module smoothes power spikes, maintains power supply stability, and avoids hardware failures; the redundant circuits of the driver board and the wireless power supply module achieve seamless fault switching through multiplexing and magnetic resonance technology, dispersing the risk of cascading failures to independent units, greatly improving the system's fault tolerance and continuous operation reliability.In summary, this embodiment is not a breakthrough in a single technology, but a full-chain technology system from the bottom-level hardware to the upper-level algorithm is constructed through the digitization of parameter analysis, the intelligence of the communication network, the coordination of the control strategy, the real-time feedback mechanism, the dynamic resource scheduling and the redundancy of the hardware design; each module forms a closed-loop ecology of "precise mapping-efficient transmission-intelligent decision-making-real-time adjustment-reliable execution" through data interoperability and functional complementarity, and finally achieves a comprehensive improvement in control accuracy, communication efficiency (O(1) level node query), dynamic adaptability (complex environment disturbance response) and system robustness (redundant design fault tolerance), providing a scalable distributed control paradigm for high-complexity industrial molding scenarios.

[0179] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also change, modify, replace and modify the above-mentioned embodiments without departing from the scope of protection of the purpose of the present invention and the claims, and all of these are protected by the present invention.

Claims

1. A motor control system based on a distributed control architecture, characterized in that: include: Demand analysis module, routing module, summary board control module, control feedback module; The demand analysis module obtains a three-dimensional process parameter space based on the process template requirements input by the front end and the preset three-dimensional modeling analysis model; Transmitting the three-dimensional process parameter space to the summary board control module through the primary connection and the secondary connection in the distributed control node network in the routing module; The summary board control module obtains a distributed control instruction set corresponding to each summary board based on the three-dimensional process parameter space combined with a preset distributed instruction generation model and the parameter space of the motor drive board corresponding to each summary board; Based on the distributed control instruction set corresponding to each aggregation board, the corresponding motor is driven and controlled through the three-level connection, the four-level connection and the motor drive board in the distributed control node network; At the same time, the control feedback module monitors the corresponding multi-column molding results and motor operating parameters during the motor drive control process in real time and feeds them back to the distributed instruction generation model, and adjusts the operating control parameters of each motor until the multi-column molding deviation meets the process template requirements.

2. The motor control system based on a distributed control architecture according to claim 1, characterized in that: The demand analysis module includes a process analysis unit and a three-dimensional modeling unit; the routing module includes a node network construction unit and a routing unit; The process analysis unit obtains the process template parameter space and the mold column distribution state space according to the input process template requirements and the preset analysis sub-model; The three-dimensional modeling unit obtains a three-dimensional process model of the target to be molded based on the process template parameter space and the mold column distribution state space in combination with the three-dimensional coordinates with the center of gravity of the target to be molded as the origin, and marks the process template parameters at corresponding positions on the three-dimensional process model; The node network construction unit constructs a distributed control node network through a graph algorithm and a distributed framework based on the front end, routing board, aggregation board, motor drive board, motor position information, operation information and signal transmission line information; The routing unit performs directed transmission of information instructions between nodes according to the distributed control node network.

3. The motor control system based on a distributed control architecture according to claim 2, characterized in that: The summary board control module includes a parameter decomposition unit, a motion planning unit and an instruction generation unit; The parameter decomposition unit obtains a process parameter sequence corresponding to each motor under each motor drive board based on the three-dimensional process model and the marked process parameters combined with a preset processing mapping space; The motion planning unit obtains the height trajectory pulse sequence corresponding to the mold columns of all motors at each moment through a trajectory synchronization planning algorithm based on the three-dimensional process model and the marked process parameters, the process parameter sequence corresponding to each motor, and the speed and acceleration parameters of each motor controlling the extension and contraction of the mold column; The processing mapping space is constructed by the process parameter sequence corresponding to each motor and the motor label corresponding to each node in the collaborative processing subnet in the distributed control node network; The collaborative processing subnet is constructed by combining all motors and the corresponding data interaction information between motors with a graph algorithm; The instruction generation unit obtains the telescopic control instruction corresponding to each motor according to the process parameter sequence corresponding to each motor and the height trajectory pulse sequence corresponding to the mold columns of all motors at each moment in combination with the distributed instruction generation model.

4. The motor control system based on a distributed control architecture according to claim 3, characterized in that: The control feedback module includes a processing monitoring unit, a motor operation evaluation unit and a monitoring feedback unit; The processing monitoring unit is used to monitor in real time the height deviation of the shaping column controlled by each motor and the torque of the molding column corresponding to each position of the overall mold, and obtain a processing result deviation sequence; The motor operation evaluation unit monitors the interaction between the operation data and the motor during the motor processing control process in real time according to a preset state monitoring node network, and evaluates the motor operation state in real time through a built-in evaluation algorithm to obtain the motor operation state space; The monitoring feedback unit is used to feed back the processing result deviation sequence, the torque distribution deviation corresponding to each motor and the motor operation state space to the distributed instruction generation model and the distributed control node network, so as to adjust the motor control process in real time and track and locate the abnormal state motor in real time.

5. The motor control system based on a distributed control architecture according to claim 4, characterized in that: The connection process of the distributed control node network includes: Set the front end as the first-level node, the signal routing board as the second-level node, the motor drive summary board as the third-level node, the motor drive board as the fourth-level node, and the motor as the fifth-level node; The first-level node is connected to at least one of the second-level nodes through at least one first-level connection established by the network port or RS232; each of the second-level nodes is connected to at least one third-level node through at least one second-level connection established by the CAN bus and the tag information of the corresponding third-level node is stored in the corresponding second-level node; Each of the three-level nodes is connected to M four-level nodes through a three-level connection set constructed by the CAN bus and a 24V power supply, and the label information of the M four-level nodes is stored in the corresponding three-level node; each of the four-level nodes is connected to N five-level nodes through N four-level connections constructed by N motor lines, and the labels of the N five-level nodes are stored in the corresponding four-level nodes; Each of the five-level nodes corresponds to each mold column one by one and all the five-level nodes are connected through an interactive control connection set constructed by data interaction information between the nodes; A label hash mapping table is constructed based on the labels maintained by each level node and the connection relationship of the corresponding level in combination with a hash algorithm, and the constructed label hash mapping table is saved in the secondary node.

6. The motor control system based on a distributed control architecture according to claim 5, characterized in that: The distributed instruction generation model is obtained by integrating the instruction generation sub-models with the same number of motors and a joint control sub-model through a federated algorithm; the instruction generation sub-model is used to obtain the corresponding expansion and contraction control instructions for each motor based on the process parameter sequence corresponding to each motor and the fuzzy control algorithm; The joint control model is used to obtain the control deviation compensation corresponding to all motors based on the operating status of all motors, the transmission delay distribution of the command signal, the height trajectory of the mold columns corresponding to all motors at each moment, the overall mold deviation and the torque deviation corresponding to all motors combined with the reinforcement learning algorithm; the control deviation compensation corresponding to all motors includes the height and torque deviation corresponding to each mold column and the transmission signal delay deviation corresponding to each motor.

7. The motor control system based on a distributed control architecture according to claim 6, characterized in that: The working process of the distributed control node network includes: Based on the first-level node and the process template requirements, a process template parameter space and a mold column distribution state space are obtained, and the process template parameter space and the mold column distribution state space are transmitted to the second-level node through the first-level connection; The second-level node decomposes and marks the process template parameter space and the mold column distribution state space according to the saved label hash mapping table, and at the same time transmits the marking results and all the third-level connections to the corresponding third-level nodes according to the marking. Each of the third-level nodes generates a telescopic control instruction set corresponding to all motors under each third-level node through a distributed instruction generation model combined with the process parameter sequence corresponding to each marked motor.

8. The motor control system based on a distributed control architecture according to claim 7, characterized in that: The working process of the distributed control node network also includes: The expansion and contraction control instruction sets corresponding to all motors under each third-level node are transmitted to the corresponding fourth-level nodes through the third-level connection relationship and the label in the label hash mapping table. Each of the fourth-level nodes converts the received expansion and contraction control instruction subset into a step pulse signal corresponding to each motor according to the stored motor label, and controls the operating status of each fifth-level node in real time. The state monitoring node network has the same structure as the distributed control node network and the nodes correspond one to one. Each level corresponds to the working status of the node and the deviation between the motor operating parameters and the process parameters; The status monitoring node network is used to monitor in real time the data transmission delay in each connection relationship within the distributed control node network, the corresponding operating status of each level node, the height deviation of the mold columns corresponding to all five-level nodes, the corresponding height trajectory of all mold columns at each moment, and the torque deviation of all motors.

9. The motor control system based on a distributed control architecture according to claim 8, characterized in that: The working process of the distributed control node network also includes: At the same time, the abnormality assessment algorithm configured for each monitoring node is used to perform abnormality assessment according to the operating status corresponding to each level node, and the process assessment algorithm configured for the collaborative processing subnet is used to perform real-time assessment of the mold column height deviation and torque deviation corresponding to each position point and the corresponding height trajectory of all mold columns at each moment, thereby obtaining abnormality assessment results and process assessment results respectively; The process evaluation results are fed back to the distributed instruction generation model to obtain the process deviation adjustment parameter sequences corresponding to all motor controls to adjust all motor control states in real time. At the same time, the abnormality evaluation results are fed back to the state monitoring node network to perform real-time traceability and early warning of abnormal nodes in the distributed control node network.

10. The motor control system based on a distributed control architecture according to claim 9, characterized in that: The construction process of the instruction generation sub-model and the joint control sub-model includes: Constructing a single input state of a single instruction generation sub-model based on the process parameter sequence corresponding to each motor, the motor's historical operating parameters, and real-time feedback data, and constructing a global input state based on the global operating state of the distributed control node network, the molding process evaluation results, the global operating abnormality evaluation results, and the characteristics of the material to be molded; Construct local execution action trigger information based on the single input state corresponding to each motor, including: When the three-dimensional process model or the processing mapping space changes, the first execution action information is triggered, that is, the corresponding motor is controlled to control the corresponding mold column according to the input control instruction; When the mold column height deviation exceeds a preset threshold or the corresponding torque distribution does not meet the process requirements, the second execution action information is triggered; At the same time, based on the global input state, build global execution action trigger information, including: When the second execution action information is triggered or the height trajectory deviation of all mold columns at each moment is greater than a preset trajectory deviation threshold, the global first execution action information is triggered; that is, the process deviation adjustment parameter sequence is generated through the joint control sub-model; When the data transmission delay between nodes exceeds a threshold or an abnormality occurs in at least one node in the distributed control node network, the global second execution action information is triggered; that is, information transmission adjustment and abnormal node tracing are performed through the joint control sub-model.

Images