A control method and system of a distributed motion controller

Abstract

This invention discloses a control method and system for a distributed motion controller, relating to the field of motor control technology. The method includes: acquiring operational data from the motion controllers of each production line via real-time Ethernet and processing the acquired data; performing fuzzy inference based on the underlying single-axis servo fuzzy controller of the motion controller to obtain position feedforward compensation time and distributed clock phase adjustment; performing fuzzy inference based on the upper-level production line collaborative fuzzy scheduler of the main controller to obtain speed multipliers and global command timestamp offsets; generating motor control commands conforming to the real-time Ethernet protocol; and using a distributed clock synchronization mechanism to synchronously adjust the operating parameters of the servo motors of each production line. By real-time monitoring and scheduling of the motion controllers of multiple production lines, multiple production lines can cooperate with each other, achieving intelligent dynamic adjustment of the production line system, thereby improving the processing efficiency of the production line system.

Description

Technical Field

[0001] This invention relates to the field of motor control technology, and specifically to a control method and system for a distributed motion controller. Background Technology

[0002] In existing machining production line systems, multiple production lines need to work collaboratively. Each production line is equipped with a motion controller, a servo driver controlled by the motion controller, and a servo motor driven by the servo driver. The servo motor drives the corresponding equipment within the production line. Current production line systems primarily achieve this by setting fixed control parameters for each production line and integrating buffer units between them to buffer semi-finished products. By controlling product transport to adapt to differences in production rhythm between different production lines, they integrate multiple processing steps for the product. However, this collaborative working method of production line systems suffers from the problem of semi-finished product accumulation. The independent operation of each production line, lacking coordinated adjustment capabilities, leads to material jams or accumulation of semi-finished workpieces, resulting in overall production efficiency that needs improvement. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a control method and system for a distributed motion controller. By real-time monitoring and scheduling of the motion controllers of each production line, the collaborative control and management of multiple production lines can be realized, and intelligent dynamic adjustment of the multi-production line system can be achieved, thereby improving the processing and production efficiency of the production line system.

[0004] This invention provides a control method for a distributed motion controller, comprising the following steps:

[0005] Step S1: Transmit the running data of the motion controller of each production line to the main controller through the real-time Ethernet communication network. Process the running data of the motion controller of each production line to obtain servo following delay, multi-axis synchronization phase difference, inter-production line transmission delay deviation, processing cycle residual, final assembly line buffer queue length deviation and processing cycle residual change.

[0006] Step S2: Input the servo follow delay and the multi-axis synchronization phase difference into the underlying single-axis servo fuzzy controller of the motion controller of each production line for fuzzy inference to obtain the position feedforward compensation time and the distributed clock phase adjustment amount;

[0007] Step S3: Input the inter-production line transmission delay deviation, the processing cycle time residual, the final assembly line buffer queue length deviation, and the processing cycle time residual change into the upper-level production line collaborative fuzzy scheduler running on the main controller for fuzzy inference to obtain the speed multiplier and global instruction timestamp offset.

[0008] Step S4: Convert the position feedforward compensation time, the distributed clock phase adjustment amount, the speed multiplier, and the global command timestamp offset into motor control commands that conform to the real-time Ethernet protocol.

[0009] Step S5: Using a distributed clock synchronization mechanism, each production line's motion controller executes the instruction corresponding to the position feedforward compensation time, the instruction corresponding to the distributed clock phase adjustment amount, the instruction corresponding to the speed multiplier, and the instruction corresponding to the global instruction timestamp offset at a unified time to synchronize and adjust the operating parameters of each production line's servo motor.

[0010] Furthermore, step S1 includes:

[0011] The process data object via real-time Ethernet reads the servo axis command position register and actual position register in each control cycle. The read command position sequence and actual position sequence are time-domain shift matched. The shift amount when the correlation between the two sequences is the highest is detected. The servo following delay is determined based on the shift amount and the control cycle.

[0012] Record the local timestamp of each motion controller at the moment of triggering the distributed clock hardware synchronization signal, compare the local timestamps of different production lines, and obtain the multi-axis synchronization phase difference;

[0013] At the workpiece exit of the production line and the entrance of the final assembly line, a distributed clock timestamp of the workpiece's passage time is latched by a digital input module connected to a real-time Ethernet. The difference between the latched departure time and arrival time is calculated to obtain the actual transmission time. The actual transmission time is compared with a preset reference transmission time to obtain the transmission delay deviation between the production lines.

[0014] Furthermore, step S1 also includes:

[0015] Record the start and end times of processing each workpiece and calculate the actual processing time. Compare the actual processing time with the predicted processing time output by the pre-trained time series prediction model to obtain the processing cycle residual.

[0016] The actual queue length of the buffer is obtained by reading the digital input module counts of the inlet and outlet of the final assembly line buffer through real-time Ethernet, calculating the difference between the inlet count and the outlet count, and comparing the actual queue length of the buffer with the preset target queue length to obtain the queue length deviation of the final assembly line buffer.

[0017] The difference between the machining cycle time residuals of two consecutive workpieces is calculated to obtain the change in the machining cycle time residual.

[0018] Furthermore, step S2 includes:

[0019] The underlying single-axis servo fuzzy controller inputs the servo follow delay and the multi-axis synchronization phase difference into the motion controller of each production line.

[0020] The servo following delay is fuzzified based on the triangular membership function built into the underlying single-axis servo fuzzy controller to obtain the first fuzzy distribution result corresponding to the servo following delay;

[0021] The multi-axis synchronous phase difference is fuzzified based on the triangular membership function built into the underlying single-axis servo fuzzy controller to obtain the second fuzzy distribution result corresponding to the multi-axis synchronous phase difference;

[0022] When the second fuzzy distribution result indicates that the local axis is severely lagging, and the first fuzzy distribution result indicates that the servo following delay is less than or equal to the normal range, the first phase compensation trend and the first feedforward compensation trend are inferred.

[0023] When the second fuzzy distribution result indicates that the local axis is severely ahead, and the first fuzzy distribution result indicates that the servo following delay is higher than the normal range, the second phase compensation trend and the second feedforward compensation trend are inferred.

[0024] The first feedforward compensation trend is deblurred to obtain the position feedforward compensation time; the first phase compensation trend is deblurred to obtain the distributed clock phase adjustment amount; or

[0025] The second feedforward compensation trend is defuzzified to obtain the position feedforward compensation time, and the second phase compensation trend is defuzzified to obtain the distributed clock phase adjustment amount.

[0026] Furthermore, step S3 includes:

[0027] The processing cycle time residual and the length deviation of the final assembly line buffer queue are input into the upper-level production line collaborative fuzzy scheduler and processed based on the first fuzzy inference module built into the upper-level production line collaborative fuzzy scheduler to obtain the temporary speed multiplier.

[0028] The inter-production line transmission delay deviation and the change in processing cycle time residual are input into the upper-level production line collaborative fuzzy scheduler. The upper-level production line collaborative fuzzy scheduler is processed based on the second fuzzy inference module built into it to obtain the speed ratio correction value and the global instruction timestamp offset.

[0029] The temporary speed multiplier is added to the speed multiplier correction value, and the addition result is subjected to amplitude limiting to obtain the speed multiplier.

[0030] Furthermore, the processing logic of the first fuzzy inference module is as follows: when the processing cycle residual indicates that the processing is slower than expected and the deviation of the final assembly line buffer queue length indicates that the buffer is missing parts, an increased temporary speed multiplier is generated;

[0031] The processing logic of the second fuzzy inference module is as follows: when the inter-line transmission delay deviation indicates that the material arrives late and the change in the processing cycle residual indicates that the delay trend is aggravated, a positive global instruction timestamp offset and a positive speed ratio correction value are generated.

[0032] Furthermore, step S4 includes:

[0033] The position feedforward compensation time is written into the interpolator feedforward parameter table of the motion controller, so that the subsequently generated command position sequence is shifted forward on the time axis by the position feedforward compensation time, and the command corresponding to the position feedforward compensation time is obtained.

[0034] The distributed clock phase adjustment amount is written into the distributed clock system time offset register of the corresponding slave controller through the service data object of real-time Ethernet, so as to obtain the instruction corresponding to the distributed clock phase adjustment amount.

[0035] The speed multiplier is encoded into a real-time Ethernet process data object data frame, and a first effective timestamp is appended to the data frame to obtain the instruction corresponding to the speed multiplier;

[0036] After appending a second effective timestamp to the global command timestamp offset, the command is broadcast to the motion controllers of each production line via real-time Ethernet to obtain the command corresponding to the global command timestamp offset.

[0037] Furthermore, step S5 includes:

[0038] When the local distributed clock reaches the first effective timestamp, each motion controller writes the speed multiplier to the servo driver within the next distributed clock hardware synchronization signal interrupt;

[0039] Each motion controller uniformly applies the global command timestamp offset carrying the second effective timestamp in its local command queue.

[0040] Furthermore, the control method further includes: before executing step S1, performing a global scheduling optimization step at the non-real-time task layer, the global scheduling optimization step including:

[0041] The scheduling engine matches order parameters with preset scheduling templates to generate scheduling constraints.

[0042] During online operation, a global scheduling scheme containing a base speed ratio and base timing parameters is generated by the scheduling policy network based on the scheduling constraints.

[0043] In a digital twin environment, the global scheduling scheme is pre-validated at a speed faster than real-time simulation. The pre-validated global scheduling scheme is then sent to the main controller to update the baseline parameters or constraint range of the upper-layer production line collaborative fuzzy scheduler.

[0044] The present invention also provides a control system for a distributed motion controller, the control system comprising:

[0045] Data processing module: Used to transmit the running data of each production line's motion controller to the main controller via a real-time Ethernet communication network, and to process the running data of each production line's motion controller to obtain servo following delay, multi-axis synchronization phase difference, inter-production line transmission delay deviation, processing cycle residual, final assembly line buffer queue length deviation, and processing cycle residual change.

[0046] The pre-compensation module is used to perform fuzzy inference on the underlying single-axis servo fuzzy controller of the motion controller of each production line by inputting the servo following delay and the multi-axis synchronous phase difference into the motion controller of each production line, so as to obtain the position feedforward compensation time and the distributed clock phase adjustment amount.

[0047] Fuzzy calculation module: used to input the inter-production line transmission delay deviation, the processing cycle time residual, the final assembly line buffer queue length deviation and the processing cycle time residual change into the upper-level production line collaborative fuzzy scheduler running on the main controller for fuzzy inference to obtain the speed multiplier and global instruction timestamp offset;

[0048] Command generation module: used to convert the position feedforward compensation time, the distributed clock phase adjustment amount, the speed multiplier, and the global command timestamp offset into motor control commands that conform to the real-time Ethernet protocol;

[0049] Global Adjustment Module: This module utilizes a distributed clock synchronization mechanism to enable the motion controllers of each production line to execute the instructions corresponding to the position feedforward compensation time, the distributed clock phase adjustment amount, the speed multiplier, and the global instruction timestamp offset at a unified time, thereby synchronizing the operating parameters of the servo motors of each production line.

[0050] This invention provides a control method and system for a distributed motion controller. By real-time monitoring and scheduling of motion controllers on multiple production lines, multiple production lines can cooperate with each other, realizing intelligent dynamic adjustment of the production line system, thereby improving the processing and production efficiency of the production line system. Attached Figure Description

[0051] Figure 1 This is a flowchart of the control method of the distributed motion controller in an embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram of the production line system in an embodiment of the present invention;

[0053] Figure 3 This is a schematic diagram of the control system structure of the distributed motion controller in an embodiment of the present invention. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] In a machining production line system, multiple production lines need to work together. Each production line is equipped with a motion controller, a servo driver controlled by the motion controller, and a servo motor driven by the servo driver. The operation of the servo motor drives the corresponding equipment in the production line.

[0056] This invention provides a control method and system for a distributed motion controller. By real-time monitoring and scheduling of the motion controllers corresponding to each production line, multiple production lines can cooperate with each other, realizing intelligent dynamic adjustment of the production line system, thereby improving the processing and production efficiency of the production line system.

[0057] Please refer to Figures 1 to 3 This invention provides a control method for a distributed motion controller, comprising the following steps:

[0058] S1. The running data of the motion controller of each production line is transmitted to the main controller through the real-time Ethernet communication network. The running data of the motion controller of each production line is processed to obtain the servo following delay, multi-axis synchronization phase difference, inter-production line transmission delay deviation, processing cycle residual, final assembly line buffer queue length deviation and processing cycle residual change.

[0059] Specifically, step S1 includes:

[0060] The process data object via real-time Ethernet reads the servo axis command position register and actual position register in each control cycle. The read command position sequence and actual position sequence are time-domain shift matched, and the shift amount when the correlation between the two sequences is the highest is detected. The servo following delay is determined based on the shift amount and the control cycle.

[0061] Specifically, the real-time Ethernet mentioned is an industrial Ethernet technology that supports deterministic data transmission, ensuring low latency and high synchronization of data transmission. Process Data Objects (PDOs) are a mechanism in the real-time Ethernet protocol used for the periodic, real-time exchange of process data. They allow the controller to efficiently acquire the commanded position (i.e., the position the controller expects the servo axis to reach) and the actual position (i.e., the current position fed back by the servo axis encoder) of the servo axis within each fixed control cycle. This reading method ensures the real-time and synchronous nature of data acquisition, providing a reliable data foundation for subsequent accurate analysis.

[0062] Record the local timestamp of each motion controller at the moment of triggering the distributed clock hardware synchronization signal, compare the local timestamps of different production lines, and obtain the multi-axis synchronization phase difference;

[0063] A time-domain shift matching process is performed on the read instruction position sequence and the actual position sequence to detect the shift amount at which the correlation between the two sequences is highest. Time-domain shift matching aims to find the optimal correspondence between the two sequences by aligning them on the time axis. This can be achieved by calculating the cross-correlation function of the two sequences, which quantifies the similarity between the two signals at different time delays. When the cross-correlation function reaches its global maximum, the corresponding shift amount represents the shift amount at which the correlation between the two sequences is highest, reflecting the time delay between the instruction and the actual response.

[0064] By determining the shift amount with the highest correlation between the command position sequence and the actual position sequence, it can be converted into a specific time delay. If the shift amount is expressed as an integer multiple of the control cycle, the servo follow delay can be directly calculated by multiplying the shift amount by the control cycle.

[0065] At the workpiece exit of the production line and the entrance of the final assembly line, a distributed clock timestamp of the workpiece's passage time is latched by a digital input module connected to a real-time Ethernet. The difference between the latched departure time and arrival time is calculated to obtain the actual transmission time. The actual transmission time is compared with a preset reference transmission time to obtain the transmission delay deviation between the production lines.

[0066] Furthermore, the preset reference transmission time is a theoretical transmission time calculated in advance based on the physical layout of the production line (such as the distance between two workstations) and the designed standard transmission speed (such as the rated speed of the conveyor belt). The preset reference transmission time is a benchmark value set by the system to compare with the actual measured value to evaluate the transmission delay.

[0067] Furthermore, digital input modules are used to detect physical switching signals, such as the signal from a photoelectric sensor detecting the passage of a workpiece. By connecting these digital input modules to a real-time Ethernet network and utilizing their timestamp capture function, the timestamp of the current distributed clock can be precisely latched the instant the workpiece passes through the sensor-triggered signal.

[0068] In an automated production line using EtherCAT real-time Ethernet, to accurately measure servo following delay, a 1-millisecond control cycle is set to read data from the command position register and the actual position register from the servo driver, and store the read data as a command position sequence and an actual position sequence. Subsequently, the main controller 10 determines the time-domain shift amount of the command position sequence and the actual position sequence when the cross-correlation coefficient is maximized.

[0069] Furthermore, in this embodiment, the optimal shift amount for the command position sequence and the real-time position sequence is 5 control cycles, so the servo following delay is determined to be 5 milliseconds. To measure the multi-axis synchronization phase difference, each EtherCAT slave (motion controller) records its internal local timestamp upon receiving the hardware synchronization signal from the EtherCAT distributed clock (DC). The master controller 10 periodically collects these timestamps and calculates the difference between the timestamps of different slaves. For measuring the inter-line transmission delay deviation, photoelectric sensors are installed at the workpiece exit and the final assembly line entrance, respectively. These sensors are connected to the EtherCAT digital input module. When a workpiece passes through the sensor, the digital input module triggers a latching function, accurately recording the current EtherCAT distributed clock timestamp. Assuming the latched timestamp of the workpiece at the exit is... The latch timestamp at the entrance is The actual transmission time is - By comparing this actual transmission time with a preset reference transmission time (e.g., 2.5 seconds calculated based on the physical distance to the production line and the design speed), the transmission delay deviation between production lines can be obtained. For example, if the actual transmission time is 2.7 seconds, the transmission delay deviation is 0.2 seconds.

[0070] Specifically, step S1 also includes:

[0071] Record the start and end times of processing each workpiece and calculate the actual processing time. Compare the actual processing time with the predicted processing time output by the pre-trained time series prediction model to obtain the processing cycle residual.

[0072] The actual queue length of the buffer is obtained by reading the digital input module counts of the inlet and outlet of the final assembly line buffer through real-time Ethernet, calculating the difference between the inlet count and the outlet count, and comparing the actual queue length of the buffer with the preset target queue length to obtain the queue length deviation of the final assembly line buffer.

[0073] The difference between the machining cycle time residuals of two consecutive workpieces is calculated to obtain the change in the machining cycle time residual.

[0074] At each machining station, a pair of photoelectric sensors can be installed, one at the starting position when the workpiece enters the station and the other at the ending position when the workpiece leaves the station. When the workpiece blocks the starting sensor, the motion controller records a distributed clock timestamp. When the workpiece occlusion ends at the sensor, the motion controller records a timestamp. The actual processing time is... - A lightweight LSTM model can be used as a pre-trained time-series prediction model. This model is trained offline using historical production data (such as workpiece type, material, equipment load, and ambient temperature) to learn the temporal patterns of processing time. During online operation, the model receives the feature parameters of the current workpiece as input and outputs the predicted processing time. The actual processing time is subtracted from the predicted processing time output by the LSTM model to obtain the processing cycle time residual.

[0075] Furthermore, the LSTM model is trained as follows: During the offline training phase, the actual processing time of each workpiece (obtained from the difference between the timestamps of the start and end photoelectric sensors) and its corresponding multidimensional feature parameters are extracted from the historical production database. These multidimensional feature parameters include: workpiece type code, material hardness grade, geometric complexity index, current spindle load rate, cumulative tool usage time, coolant temperature, previous actual processing time, workshop ambient temperature and humidity, and shift identifier, etc.

[0076] To capture the temporal dependency of processing time on the order of workpieces, the system constructs a sliding window according to the processing order: the feature vector sequence of T consecutive workpieces is used as the input sample of the LSTM model, and the corresponding label is the actual processing time of the next workpiece. All numerical features are standardized by Z-score, and the data is divided into training set, validation set and test set according to time order.

[0077] The model employs a lightweight single-layer LSTM structure with 32–64 hidden units, followed by a Dropout layer and a fully connected layer. It outputs a numerical value representing the prediction processing time. Mean absolute error (MAE) is used as the loss function during offline training. Mini-batch gradient descent (batch size 32–64) is used during training. The loss is evaluated on the validation set after each epoch, and training is complete when the loss reaches its minimum value.

[0078] After training, the model's final performance is evaluated on the test set, and the statistical distribution of the prediction residuals (such as mean and standard deviation) is calculated as the threshold for online anomaly detection. The entire offline process includes data cleaning, sequence sample construction, normalized parameter saving, model hyperparameter tuning, and final model solidification (exporting to a lightweight format for edge controller loading). This allows for direct application of the same transformation to the real-time input using the normalized parameters during online runtime, and the predicted processing time is obtained by subtracting the actual processing time measured by the sensor.

[0079] If the actual processing time is 10.2 seconds and the predicted processing time is 10.0 seconds, the residual is 0.2 seconds. For the residual sequence of processing cycles for 10 consecutive workpieces, a simple moving average filter with a sliding window of 5 can be used to smooth the data, eliminate occasional fluctuations, and obtain the systematic cycle offset. At the entrance and exit of the assembly line buffer, a digital input module supporting the EtherCAT protocol is installed. Each module is connected to a counting sensor (e.g., a laser beam sensor). When a workpiece passes through, the sensor triggers the internal counter of the module to increment by 1. The main controller 10 reads the counter values ​​of these two modules in each control cycle through the EtherCAT Process Data Object (PDO) mechanism. The main controller 10 subtracts the exit count from the read entrance count to obtain the actual queue length of the buffer. For example, if the entrance count is 100 and the exit count is 90, the actual queue length is 10. The preset target queue length can be set to 8. Comparing the actual queue length of 10 with the target queue length of 8 yields a queue length deviation of 2 for the assembly line buffer. If the current workpiece's machining cycle time residual is 0.2 seconds and the previous workpiece's machining cycle time residual is 0.1 seconds, then the change in machining cycle time residual is 0.2 - 0.1 = 0.1 seconds.

[0080] It can provide highly accurate and real-time data on machining cycle time residuals, assembly line buffer queue length deviations, and machining cycle time residual changes. These parameters serve as inputs to the upper-level production line collaborative fuzzy scheduler, enabling the scheduler to perform fuzzy inference based on more reliable and forward-looking information. When machining cycle time fluctuates or buffer load changes, the system can quickly sense and quantify these changes. Through precise residual and rate-of-change information, the scheduler can adjust production line operation strategies in a timely manner, such as accelerating or decelerating specific production lines or adjusting workpiece transfer timing. This effectively avoids the accumulation of semi-finished workpieces between production lines, optimizes material flow between production lines, and thus significantly improves the collaborative efficiency and production stability of the entire distributed motion control system.

[0081] S2. The servo follow delay and the multi-axis synchronous phase difference are input into the underlying single-axis servo fuzzy controller of each motion controller 20 for fuzzy inference to obtain the position feedforward compensation time and the distributed clock phase adjustment amount.

[0082] Specifically, step S2 includes:

[0083] The servo following delay and the multi-axis synchronous phase difference are input into the underlying single-axis servo fuzzy controller of the motion controller in each production line. The servo following delay is fuzzified based on the triangular membership function built into the underlying single-axis servo fuzzy controller to obtain a first fuzzy distribution result corresponding to the servo following delay. The multi-axis synchronous phase difference is fuzzified based on the triangular membership function built into the underlying single-axis servo fuzzy controller to obtain a second fuzzy distribution result corresponding to the multi-axis synchronous phase difference. The servo following delay and the multi-axis synchronous phase difference are respectively input into a preset membership distribution for fuzzification to obtain the first fuzzy distribution result corresponding to the servo following delay and the second fuzzy distribution result corresponding to the multi-axis synchronous phase difference. The servo following delay refers to the time difference or spatial difference between the actual position and the commanded position of the servo motor, reflecting the response speed and tracking accuracy of the servo system.

[0084] When the second fuzzy distribution result indicates that the local axis is severely lagging, and the first fuzzy distribution result indicates that the servo following delay is small or normal, the first phase compensation trend and the first feedforward compensation trend are inferred.

[0085] When the second fuzzy distribution result indicates that the local axis is significantly ahead, and the first fuzzy distribution result indicates that the servo following delay is large, the second phase compensation trend and the second feedforward compensation trend are inferred. The first feedforward compensation trend is defuzzified to obtain the position feedforward compensation time. The first phase compensation trend is defuzzified to obtain the distributed clock phase adjustment amount. Alternatively, the second feedforward compensation trend is defuzzified to obtain the position feedforward compensation time. The second phase compensation trend is defuzzified to obtain the distributed clock phase adjustment amount.

[0086] In a distributed motion control system, the underlying single-axis servo fuzzy controller can employ a fuzzy inference system based on the Mamdani model. For servo following delay, three fuzzy sets can be defined, such as {"small", "normal", "large"}, and fuzzified using triangular membership functions. Similarly, for multi-axis synchronization phase difference, three fuzzy sets can be defined, such as {"severe lag", "normal", "severe lead"}, and also fuzzified using triangular membership functions. The fuzzy inference rules can be specifically set as follows: if the second fuzzy distribution result is "severe lag" and the first fuzzy distribution result is "small" or "normal", then the first phase compensation trend is "moderate increase" and the first feedforward compensation trend is "slight increase"; if the second fuzzy distribution result is "severe lead" and the first fuzzy distribution result is "large", then the second phase compensation trend is "moderate decrease" and the second feedforward compensation trend is "slight decrease". In actual operation, for example, if the servo follow delay is 2ms and the multi-axis synchronization phase difference is -50 clock cycles, after fuzzification, the 2ms servo follow delay might have a membership degree of 0.6 for "small" and 0.4 for "normal"; the -50 clock cycle multi-axis synchronization phase difference might have a membership degree of 0.5 for "severe lag" and 0.5 for "normal". According to the above rules, since the multi-axis synchronization phase difference has a membership degree of "severe lag" and the servo follow delay has a membership degree of "small" or "normal", the truth value of the rule antecedent (using the min operator) is min(0.5, max(0.6, 0.4)) = 0.5. At this point, the inferred first phase compensation trend (e.g., "moderate increase") and first feedforward compensation trend (e.g., "small increase") will be activated with a strength of 0.5. Finally, the fuzzy output is converted into precise values ​​using defuzzification methods such as the center of gravity method. For example, a specific position feedforward compensation time (e.g., +0.5ms) and a distributed clock phase adjustment amount (e.g., +20 clock cycles) are obtained. These values ​​are then converted into motor control commands to adjust the operating parameters of the servo motor.

[0087] Through the above technical solution, this invention clarifies the specific rules for fuzzy inference in the underlying single-axis servo fuzzy controller, solving the problem of inaccurate compensation trend inference caused by unclear fuzzy inference rules in traditional solutions. By fuzzifying the servo following delay and multi-axis synchronization phase difference, and performing targeted conditional inference based on their fuzzy distribution results, the system can more accurately determine the current operating state and synchronization deviation type of the servo axis. This refined and adaptive compensation strategy enables the distributed motion control system to significantly improve the accuracy and efficiency of multi-axis synchronization when facing complex dynamic conditions, effectively avoiding overshoot, undershoot, or oscillation caused by improper compensation, thereby improving the stability and production performance of the entire production line system.

[0088] S3. Input the inter-production line transmission delay deviation, the processing cycle time residual, the final assembly line buffer queue length deviation, and the processing cycle time residual change into the upper-level production line collaborative fuzzy scheduler running on the main controller 10 for fuzzy inference to obtain the speed multiplier and global instruction timestamp offset.

[0089] Specifically, step S3 includes:

[0090] The processing cycle time residual and the length deviation of the final assembly line buffer queue are input into the upper-level production line collaborative fuzzy scheduler and processed based on the first fuzzy inference module built into the upper-level production line collaborative fuzzy scheduler to obtain the temporary speed multiplier.

[0091] The inter-production line transmission delay deviation and the change in processing cycle time residual are input into the upper-level production line collaborative fuzzy scheduler. The upper-level production line collaborative fuzzy scheduler is processed based on the second fuzzy inference module built into it to obtain the speed ratio correction value and the global instruction timestamp offset.

[0092] The temporary speed multiplier is added to the speed multiplier correction value, and the addition result is subjected to amplitude limiting to obtain the speed multiplier.

[0093] Among them, the processing cycle residual refers to the difference between the actual processing time and the predicted processing time output by the pre-trained time series prediction model. It reflects the speed of the current production line processing relative to the expected baseline and is a key indicator for evaluating the production efficiency of the production line.

[0094] The deviation of the queue length in the final assembly line buffer refers to the difference between the actual queue length and the preset target queue length in the final assembly line buffer. It reflects the degree of material accumulation or shortage in the buffer and is an important parameter for measuring the balance of material flow between production lines.

[0095] Specifically, the processing logic of the first fuzzy inference module is as follows: when the processing cycle residual indicates that the processing is slower than expected and the deviation of the assembly line buffer queue length indicates that the buffer is missing parts, an increased temporary speed multiplier is generated;

[0096] The processing logic of the second fuzzy inference module is as follows: when the inter-line transmission delay deviation indicates that the material arrives late and the change in the processing cycle residual indicates that the delay trend is aggravated, a positive global instruction timestamp offset and a positive speed ratio correction value are generated.

[0097] By employing refined fuzzy inference logic, precise optimization of production line collaborative control is achieved. The first fuzzy inference module generates an increased temporary speed multiplier when it detects a combination of conditions: processing cycle time residual indicating slower-than-expected processing and a buffer queue length deviation indicating a shortage of parts in the buffer. This logic, by simultaneously considering processing delay and buffer shortage status, avoids inaccurate adjustments caused by relying on a single deviation, thus directly accelerating processing to alleviate the shortage problem. Simultaneously, the second fuzzy inference module generates a positive global instruction timestamp offset and a positive speed multiplier correction value when it detects a combination of conditions: inter-line transmission delay deviation indicating delayed material arrival and changes in processing cycle time residual indicating an worsening delay trend. This logic integrates transmission delay and delay deterioration trends, dynamically adjusting global synchronization and speed correction to avoid delay accumulation. The outputs of the two fuzzy inference modules—the temporary speed multiplier, the speed multiplier correction value, and the global command timestamp offset—are then integrated into the upper-level production line collaborative fuzzy scheduler. The temporary speed multiplier and the speed multiplier correction value are added and limited to form the final speed multiplier, which, along with the global command timestamp offset, is converted into motor control commands. This collaborative effect enables the control system to perform layered and interconnected optimization for different levels of problems (single production line efficiency and inter-production line collaboration), ensuring that the fuzzy inference output more closely matches actual operating conditions and enhancing the adaptability and efficiency of the control system.

[0098] S4. Convert the position feedforward compensation time, the distributed clock phase adjustment amount, the speed multiplier, and the global command timestamp offset into motor control commands that conform to the real-time Ethernet protocol.

[0099] Step S4 includes:

[0100] The position feedforward compensation time is written into the interpolator feedforward parameter table of the motion controller, so that the subsequently generated command position sequence is shifted forward on the time axis by the position feedforward compensation time, and the command corresponding to the position feedforward compensation time is obtained.

[0101] The distributed clock phase adjustment amount is written into the distributed clock system time offset register of the corresponding slave controller through the service data object of real-time Ethernet, so as to obtain the instruction corresponding to the distributed clock phase adjustment amount.

[0102] The speed multiplier is encoded into a real-time Ethernet process data object data frame, and a first effective timestamp is appended to the data frame to obtain the instruction corresponding to the speed multiplier;

[0103] After appending the second effective timestamp to the global command timestamp offset, the command is broadcast to the motion controllers 20 of each production line via real-time Ethernet to obtain the command corresponding to the global command timestamp offset.

[0104] S5. Using a distributed clock synchronization mechanism, each production line's motion controller executes the instruction corresponding to the position feedforward compensation time, the instruction corresponding to the distributed clock phase adjustment amount, the instruction corresponding to the speed multiplier, and the instruction corresponding to the global instruction timestamp offset at a unified time, thereby synchronizing the operating parameters of each production line's servo motor.

[0105] Specifically, step S5 includes:

[0106] When the local distributed clock reaches the first effective timestamp, each motion controller 20 writes the speed multiplier to the servo driver within the next distributed clock hardware synchronization signal interrupt;

[0107] Each motion controller 20 uniformly applies the global instruction timestamp offset carrying the second effective timestamp in its local instruction queue.

[0108] When converting the position feedforward compensation time into commands, the motion controller can employ a CoE (CANopen over EtherCAT) object dictionary mechanism based on the EtherCAT protocol. The interpolator feedforward parameter table within the motion controller can be mapped to a specific index in the CoE object dictionary, such as 0x60F0:01. After the underlying single-axis servo fuzzy controller calculates the position feedforward compensation time, the master controller 10 can send a CoE write request to the corresponding slave controller via the EtherCAT master station, writing the time value to the 0x60F0:01 object. Once the write is successful, the interpolator within the motion controller will automatically adjust the generation of subsequent command position sequences based on this new feedforward parameter, shifting it forward by the corresponding time amount on the time axis.

[0109] For distributed clock phase adjustment, the master controller 10 can use SDO communication via the EtherCAT master station to write the calculated distributed clock phase adjustment into the DC system time offset register inside the slave controller. For example, this register may correspond to a specific address or CoE object in the EtherCAT slave controller chip, such as 0x0F80:01. Through SDO writing, precise fine-tuning of the slave DC clock phase can be achieved to eliminate clock drift or initial phase difference between different slaves.

[0110] Furthermore, the main controller 10 can encapsulate the speed multiplier value in a specific PDO message (such as RxPDO) and configure this PDO message to be sent at a fixed period (e.g., 1 millisecond). To achieve synchronous activation, a first activation timestamp can be appended to this PDO message, which is based on a future time based on the EtherCAT distributed clock. When the motion controller 20 receives this PDO message, it caches the speed multiplier value and only applies the speed multiplier to the servo drive when its local distributed clock reaches the first activation timestamp, thereby ensuring that all motion controllers 20 change their operating speed at the same time.

[0111] The main controller 10 can attach a second effective timestamp and send it to all motion controllers 20 via EtherCAT broadcast message. After receiving the broadcast message, all motion controllers 20 will adjust the timestamps of all instructions in their local instruction queues according to the second effective timestamp at a unified time to achieve global cycle time coordination between production lines.

[0112] Specifically, the control method further includes a global scheduling optimization step performed at the non-real-time task layer:

[0113] The scheduling engine matches order parameters with preset scheduling templates to generate scheduling constraints. Order parameters are a set of data describing the specific requirements of customer orders or production tasks, such as product type, quantity, delivery date, and special process requirements. They serve as input to the scheduling engine and guide the generation of scheduling schemes.

[0114] During online runtime, a global scheduling scheme containing a base speed ratio and base timing parameters is generated based on the scheduling policy network according to the aforementioned scheduling constraints.

[0115] In a digital twin environment, the global scheduling scheme is pre-validated at a speed faster than real-time simulation. The pre-validated global scheduling scheme is then sent to the main controller 10 to update the baseline parameters or constraint range of the upper-layer production line collaborative fuzzy scheduler.

[0116] Furthermore, the global scheduling optimization step executed at the non-real-time task layer refers to the optimization decision-making process performed at an independent software layer, typically with a longer cycle or event-driven, during system operation, without directly affecting the real-time control loop. The purpose of this step is to avoid interference from real-time control tasks, ensure the optimization process is independent and efficient, and provide macro-level guidance for real-time control.

[0117] The global scheduling optimization process can be deployed on a standalone industrial PC that communicates with the main controller 10 via Ethernet. The scheduling engine can employ an open-source constraint programming solver, such as OptaPlanner, or be built using the Python-based PuLP library. A specific order parameter can include "Product A, quantity 1000 pieces, delivery time 3 days, special process requirement: surface polishing". The preset scheduling template can be an XML file defining the production process routes, required equipment, standard processing time, and priorities for different product types (e.g., Product A, Product B). Specific scheduling constraints can be: maximum daily capacity of the polishing equipment is 500 pieces; Product A must undergo a cutting process before polishing; maximum buffer capacity is 200 pieces. During online operation, a deep reinforcement learning model trained using TensorFlow can be used as the scheduling policy network. This model, through millions of simulations of production scenarios offline, has learned how to dynamically adjust the production cycle time and material flow strategies of each production line based on the current order and equipment status. The resulting global scheduling scheme, for example, can assign a base speed multiplier of 1.05 to the first production line (cutting) and 0.98 to the second production line (polishing) for product A, and set the base transfer time window from the exit of the first production line to the entrance of the second production line to be 10 seconds ± 1 second. In a digital twin environment, a virtual factory model can be built using Siemens Plant Simulation software to accurately simulate the processing time, transfer delay, and buffer dynamics of each device. In this digital twin environment, the simulation speed can be set to 100 times the actual time, meaning 1 second of simulation time corresponds to 100 seconds of actual production time, thus achieving speed faster than real-time simulation. During the pre-validation phase, the above global scheduling scheme can be run in the digital twin environment to observe whether problems such as buffer overflows or gaps, equipment idleness, or delivery delays occur during the simulated production process. If the simulation results show that all indicators are within acceptable ranges, the pre-validation is considered successful.

[0118] Subsequently, the pre-verified global scheduling scheme (e.g., a data packet containing a baseline speed multiplier and baseline timing parameters) is sent to the main controller 10 via the EtherCAT protocol. Upon receiving the global scheduling scheme, the main controller 10 uses the baseline speed multiplier of product A (1.05) as the target center value for the upper-level production line collaborative fuzzy scheduler when calculating the speed multiplier, and adjusts the membership functions of its fuzzy rules regarding "processing cycle time residual" and "assembly line buffer queue length deviation" to make it more inclined to fine-tune around 1.05. Simultaneously, the baseline transmission time window is used as a reference benchmark for inter-production line transmission delay deviation.

[0119] This invention also proposes a control system for a distributed motion controller, comprising:

[0120] Data processing module 100: Used to transmit the running data of the motion controller of each production line to the main controller through a real-time Ethernet communication network, process the running data of the motion controller of each production line, and obtain servo following delay, multi-axis synchronization phase difference, inter-production line transmission delay deviation, processing cycle residual, final assembly line buffer queue length deviation, and processing cycle residual change.

[0121] Pre-compensation module 200: used to perform fuzzy inference on the underlying single-axis servo fuzzy controller of the motion controller of each production line by inputting the servo following delay and the multi-axis synchronous phase difference into the motion controller of each production line, so as to obtain the position feedforward compensation time and the distributed clock phase adjustment amount;

[0122] Fuzzy calculation module 300: is used to input the inter-production line transmission delay deviation, the processing cycle time residual, the final assembly line buffer queue length deviation and the processing cycle time residual change into the upper-level production line collaborative fuzzy scheduler running on the main controller for fuzzy inference to obtain the speed multiplier and global instruction timestamp offset;

[0123] Command generation module 400: used to convert the position feedforward compensation time, the distributed clock phase adjustment amount, the speed multiplier and the global command timestamp offset into motor control commands that conform to the real-time Ethernet protocol;

[0124] Global Adjustment Module 500: Used to utilize a distributed clock synchronization mechanism to enable the motion controllers of each production line to execute the instructions corresponding to the position feedforward compensation time, the distributed clock phase adjustment amount, the speed multiplier, and the global instruction timestamp offset at a unified time, thereby synchronizing the operating parameters of the servo motors of each production line.

[0125] The production line system includes several motion controllers 20, each corresponding to one of the several production lines, and a main controller 10 electrically connected to each motion controller 20. The motion controller 20 is the execution unit in the distributed motion control system. Its main function is to receive instructions from the main controller 10 and, based on local sensor data and preset logic, to control the servo motors or other actuators on its assigned production line in real time and with precision. The motion controller 20 typically possesses powerful real-time processing capabilities and abundant I / O interfaces, enabling direct communication with field devices such as servo drives, encoders, and limit switches.

[0126] The main controller 10 is the coordination and decision-making center in the distributed motion control system. Its main responsibility is to perform global scheduling, optimization, and management. It is responsible for collecting the operating data reported by each motion controller 20, executing complex fuzzy inference and scheduling algorithms at higher levels, and issuing global control commands and parameters to each motion controller 20. The main controller 10 typically has good computing power and richer communication interfaces, enabling it to process large amounts of data and perform complex logical operations.

[0127] The distributed motion controller system of this invention achieves precise control and optimized scheduling of motion across multiple production lines through the collaborative work of the main controller 10 and motion controllers 20. The main controller 10, as the system's decision-making layer, is responsible for global data analysis, fuzzy inference, and scheduling optimization. The main controller 10 periodically collects operational data from each motion controller 20 via real-time Ethernet and performs fuzzy inference in the upper-level production line collaborative fuzzy scheduler based on this data (such as inter-production line transmission delay deviations, processing cycle time residuals, etc.) to generate global speed ratios and instruction timestamp offsets.

[0128] Global commands are then sent to each motion controller 20 via real-time Ethernet. Each motion controller 20, acting as the system's execution layer, is deployed on a specific production line and is responsible for the specific motion control tasks of its assigned line. Each motion controller 20 receives global commands from the main controller 10 and, in conjunction with locally acquired operational data (such as servo following delay and multi-axis synchronization phase difference), performs fuzzy inference in the underlying single-axis servo fuzzy controller to generate position feedforward compensation time and distributed clock phase adjustment. These local and global adjustments are then converted into motor control commands conforming to the real-time Ethernet protocol.

[0129] Ultimately, utilizing a distributed clock synchronization mechanism, each motion controller 20 executes these instructions synchronously at a unified time, precisely adjusting the operating parameters of its respective production line servo motors. This master-slave distributed architecture enables low-level control tasks to be executed in real-time and in parallel on the motion controllers 20, significantly reducing control latency and improving response speed. Simultaneously, the master controller 10 can perform higher-level global optimization and scheduling, ensuring highly coordinated material flow and cycle time between different production lines. This effectively avoids the problem of semi-finished workpiece accumulation that may occur in traditional centralized control, and improves the operating efficiency and stability of the entire production line system. Through this hierarchical, distributed control strategy, this invention enables refined management and dynamic optimization of complex multi-production line systems.

[0130] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, which may include: read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, etc.

[0131] Furthermore, the embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A control method for a distributed motion controller, characterized in that, include: Step S1: Transmit the running data of the motion controller of each production line to the main controller through the real-time Ethernet communication network. Process the running data of the motion controller of each production line to obtain servo following delay, multi-axis synchronization phase difference, inter-production line transmission delay deviation, processing cycle residual, final assembly line buffer queue length deviation and processing cycle residual change. Step S2: Input the servo follow delay and the multi-axis synchronization phase difference into the underlying single-axis servo fuzzy controller of the motion controller of each production line for fuzzy inference to obtain the position feedforward compensation time and the distributed clock phase adjustment amount; Step S3: Input the inter-production line transmission delay deviation, the processing cycle time residual, the final assembly line buffer queue length deviation, and the processing cycle time residual change into the upper-level production line collaborative fuzzy scheduler running on the main controller for fuzzy inference to obtain the speed multiplier and global instruction timestamp offset. Step S4: Convert the position feedforward compensation time, the distributed clock phase adjustment amount, the speed multiplier, and the global command timestamp offset into motor control commands that conform to the real-time Ethernet protocol. Step S5: Using a distributed clock synchronization mechanism, each production line's motion controller executes the instruction corresponding to the position feedforward compensation time, the instruction corresponding to the distributed clock phase adjustment amount, the instruction corresponding to the speed multiplier, and the instruction corresponding to the global instruction timestamp offset at a unified time to synchronize and adjust the operating parameters of each production line's servo motor.

2. The control method according to claim 1, characterized in that, Step S1 includes: The process data object via real-time Ethernet reads the servo axis command position register and actual position register in each control cycle. The read command position sequence and actual position sequence are time-domain shift matched. The shift amount when the correlation between the two sequences is the highest is detected. The servo following delay is determined based on the shift amount and the control cycle. Record the local timestamp of each motion controller at the moment of triggering the distributed clock hardware synchronization signal, compare the local timestamps of different production lines, and obtain the multi-axis synchronization phase difference; At the workpiece exit of the production line and the entrance of the final assembly line, a distributed clock timestamp of the workpiece's passage time is latched by a digital input module connected to a real-time Ethernet. The difference between the latched departure time and arrival time is calculated to obtain the actual transmission time. The actual transmission time is compared with a preset reference transmission time to obtain the transmission delay deviation between the production lines.

3. The control method according to claim 1, characterized in that, Step S1 also includes: Record the start and end times of processing each workpiece and calculate the actual processing time. Compare the actual processing time with the predicted processing time output by the pre-trained time series prediction model to obtain the processing cycle residual. The actual queue length of the buffer is obtained by reading the digital input module counts of the inlet and outlet of the final assembly line buffer through real-time Ethernet, calculating the difference between the inlet count and the outlet count, and comparing the actual queue length of the buffer with the preset target queue length to obtain the queue length deviation of the final assembly line buffer. The difference between the machining cycle time residuals of two consecutive workpieces is calculated to obtain the change in the machining cycle time residual.

4. The control method according to claim 1, characterized in that, Step S2 includes: The underlying single-axis servo fuzzy controller inputs the servo follow delay and the multi-axis synchronization phase difference into the motion controller of each production line. The servo following delay is fuzzified based on the triangular membership function built into the underlying single-axis servo fuzzy controller to obtain the first fuzzy distribution result corresponding to the servo following delay; The multi-axis synchronous phase difference is fuzzified based on the triangular membership function built into the underlying single-axis servo fuzzy controller to obtain the second fuzzy distribution result corresponding to the multi-axis synchronous phase difference; When the second fuzzy distribution result indicates that the local axis is severely lagging, and the first fuzzy distribution result indicates that the servo following delay is less than or equal to the normal range, the first phase compensation trend and the first feedforward compensation trend are inferred. When the second fuzzy distribution result indicates that the local axis is severely ahead, and the first fuzzy distribution result indicates that the servo following delay is higher than the normal range, the second phase compensation trend and the second feedforward compensation trend are inferred. The first feedforward compensation trend is deblurred to obtain the position feedforward compensation time; the first phase compensation trend is deblurred to obtain the distributed clock phase adjustment amount; or The second feedforward compensation trend is defuzzified to obtain the position feedforward compensation time, and the second phase compensation trend is defuzzified to obtain the distributed clock phase adjustment amount.

5. The method according to claim 1, characterized in that, Step S3 includes: The processing cycle time residual and the length deviation of the final assembly line buffer queue are input into the upper-level production line collaborative fuzzy scheduler and processed based on the first fuzzy inference module built into the upper-level production line collaborative fuzzy scheduler to obtain the temporary speed multiplier. The inter-production line transmission delay deviation and the change in processing cycle time residual are input into the upper-level production line collaborative fuzzy scheduler. The upper-level production line collaborative fuzzy scheduler is processed based on the second fuzzy inference module built into it to obtain the speed ratio correction value and the global instruction timestamp offset. The temporary speed multiplier is added to the speed multiplier correction value, and the addition result is subjected to amplitude limiting to obtain the speed multiplier.

6. The control method according to claim 5, characterized in that, The processing logic of the first fuzzy inference module is as follows: when the processing cycle residual indicates that the processing is slower than expected and the deviation of the assembly line buffer queue length indicates that the buffer is missing parts, an increased temporary speed multiplier is generated; The processing logic of the second fuzzy inference module is as follows: when the inter-line transmission delay deviation indicates that the material arrives late and the change in the processing cycle residual indicates that the delay trend is aggravated, a positive global instruction timestamp offset and a positive speed ratio correction value are generated.

7. The control method according to claim 1, characterized in that, Step S4 includes: The position feedforward compensation time is written into the interpolator feedforward parameter table of the motion controller, so that the subsequently generated command position sequence is shifted forward on the time axis by the position feedforward compensation time, and the command corresponding to the position feedforward compensation time is obtained. The distributed clock phase adjustment amount is written into the distributed clock system time offset register of the corresponding slave controller through the service data object of real-time Ethernet, so as to obtain the instruction corresponding to the distributed clock phase adjustment amount. The speed multiplier is encoded into a real-time Ethernet process data object data frame, and a first effective timestamp is appended to the data frame to obtain the instruction corresponding to the speed multiplier; After appending a second effective timestamp to the global command timestamp offset, the command is broadcast to the motion controllers of each production line via real-time Ethernet to obtain the command corresponding to the global command timestamp offset.

8. The control method according to claim 7, characterized in that, Step S5 includes: When the local distributed clock reaches the first effective timestamp, each motion controller writes the speed multiplier to the servo driver within the next distributed clock hardware synchronization signal interrupt; Each motion controller uniformly applies the global command timestamp offset carrying the second effective timestamp in its local command queue.

9. The control method according to claim 1, characterized in that, The control method further includes: before executing step S1, performing a global scheduling optimization step at the non-real-time task layer, the global scheduling optimization step including: The scheduling engine matches order parameters with preset scheduling templates to generate scheduling constraints. During online operation, a global scheduling scheme containing a base speed ratio and base timing parameters is generated by the scheduling policy network based on the scheduling constraints. In a digital twin environment, the global scheduling scheme is pre-validated at a speed faster than real-time simulation. The pre-validated global scheduling scheme is then sent to the main controller to update the baseline parameters or constraint range of the upper-layer production line collaborative fuzzy scheduler.

10. A control system for a distributed motion controller, characterized in that, The control system includes: Data processing module: Used to transmit the running data of each production line's motion controller to the main controller via a real-time Ethernet communication network, and to process the running data of each production line's motion controller to obtain servo following delay, multi-axis synchronization phase difference, inter-production line transmission delay deviation, processing cycle residual, final assembly line buffer queue length deviation, and processing cycle residual change. The pre-compensation module is used to perform fuzzy inference on the underlying single-axis servo fuzzy controller of the motion controller of each production line by inputting the servo following delay and the multi-axis synchronous phase difference into the motion controller of each production line, so as to obtain the position feedforward compensation time and the distributed clock phase adjustment amount. Fuzzy calculation module: used to input the inter-production line transmission delay deviation, the processing cycle time residual, the final assembly line buffer queue length deviation and the processing cycle time residual change into the upper-level production line collaborative fuzzy scheduler running on the main controller for fuzzy inference to obtain the speed multiplier and global instruction timestamp offset; Command generation module: used to convert the position feedforward compensation time, the distributed clock phase adjustment amount, the speed multiplier, and the global command timestamp offset into motor control commands that conform to the real-time Ethernet protocol; Global Adjustment Module: This module utilizes a distributed clock synchronization mechanism to enable the motion controllers of each production line to execute the instructions corresponding to the position feedforward compensation time, the distributed clock phase adjustment amount, the speed multiplier, and the global instruction timestamp offset at a unified time, thereby synchronizing the operating parameters of the servo motors of each production line.

Images