A direct current motor parameter optimization method based on digital twinning

Abstract

The application discloses a DC motor parameter optimization method based on digital twinning, and relates to the technical field of motor intelligent control, and comprises the following steps: S1, dividing into electrical excitation data, mechanical response data and thermal hysteresis data; S2, generating a misaligned track by fusion; S3, generating a drift anchor piece; S4, constructing a three-domain twinning model to obtain twinning prediction data and generate a three-domain deviation tensor; S5, constructing an attribution matrix to determine a parameter drift domain; S6, updating the three-domain twinning model to generate a control parameter vector; S7, optimizing the control parameter vector by adopting a May beetle optimization algorithm based on drift attribution to obtain a target parameter group. The application can improve the DC motor parameter drift identification accuracy, reduce the irrelevant control parameter search range and parameter out-of-bound risk, and improve the operation stability of the DC motor under the conditions of load change, temperature rise accumulation and mechanical response hysteresis.

Description

Technical Field

[0001] This invention relates to the field of intelligent motor control technology, and in particular to a method for optimizing DC motor parameters based on digital twins. Background Technology

[0002] With the increasing demands for efficient, stable, and precise control of DC motors from industrial automation equipment, electric actuators, intelligent manufacturing production lines, and new energy-related equipment, control technologies for DC motor operating status perception, model simulation, and parameter optimization have received widespread attention. Existing DC motor control systems typically rely on factory-calibrated parameters, manually tuned parameters, or offline simulation models for speed loop, current loop, and torque compensation control, and combine this with operating data such as voltage, current, speed, and temperature for status monitoring. However, the following problems commonly exist in practical applications: Under conditions of sudden load changes, frequent start-stop cycles, winding temperature rise, and long-term wear, the armature resistance, armature inductance, friction damping, load torque, and heat transfer parameters of DC motors are prone to change. Fixed control parameters or simple parameter table switching cannot adapt to the actual operating conditions in a timely manner, resulting in increased speed tracking error, increased current surge, and lag in dynamic response. Existing digital twin or simulation models often focus on state prediction, fault warning, or comparison of single physical quantities, making it difficult to distinguish the temporal correlation and coupling effects between electrical deviations, mechanical deviations, thermal deviations, and external load disturbances, which can easily lead to inaccurate identification of parameter drift sources. Existing optimization algorithms typically iterate with a unified search range and a general objective function when optimizing control parameters, without fully considering the impact of different drift sources of DC motors on the search direction and safety boundaries. This results in problems such as a large search range for irrelevant parameters, low convergence efficiency, and the possibility that candidate parameters may cause current overruns, excessive temperature rise, or speed fluctuations.

[0003] Therefore, how to provide a DC motor parameter optimization method based on digital twins is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] One objective of this invention is to propose a DC motor parameter optimization method based on digital twins. This invention aligns the real-time operating data of the DC motor to be optimized by time alignment and electromechanical / thermal attribute partitioning, forming electrical excitation data, mechanical response data, and thermal hysteresis data. Based on the response hysteresis and thermal accumulation hysteresis, misalignment trajectories and drift anchors are generated. A three-domain twin model is constructed, and a three-domain deviation tensor is generated. Based on the temporal and coupling states of the three-domain deviation tensor, an attribution matrix is ​​constructed, and the parameter drift domain is determined. A dung beetle optimization algorithm based on drift attribution is used to optimize the control parameter vector. The target parameter set is obtained by combining gated behavior allocation, asynchronous memory kernel, and barrier projection. This solves the problems in existing technologies where fixed control parameters are difficult to adapt to the electrical, mechanical, and thermal parameter drift of DC motors, digital twin models are difficult to accurately determine the source of parameter drift, and general optimization algorithms have problems such as large search ranges for irrelevant parameters and risks of candidate parameters going out of bounds. This invention achieves drift attribution, adaptive optimization, and stable optimization of DC motor parameters.

[0005] A method for optimizing DC motor parameters based on digital twins according to an embodiment of the present invention includes the following steps: S1. Time-align the real-time operating data of the DC motor to be optimized, and divide it into electrical excitation data, mechanical response data and thermal hysteresis data according to electromechanical and thermal properties; S2. Using electrical excitation data as the driving reference, determine the response hysteresis between mechanical response data and electrical excitation data. Using mechanical response data as the transmission reference, determine the thermal accumulation hysteresis between thermal hysteresis data and mechanical response data. Combine the response hysteresis and thermal accumulation hysteresis to generate a misalignment trajectory. S3. Generate drifting anchor plates based on misaligned feature points in the misaligned trajectory; S4. Construct a three-domain twin model, input the drifting anchor piece into the three-domain twin model to obtain twin prediction data, and compare the real-time running data corresponding to the drifting anchor piece with the twin prediction data to generate a three-domain deviation tensor. S5. Construct an attribution matrix based on the temporal and coupling states of the three-domain deviation tensor to determine the parameter drift domain; S6. Update the three-domain twin model based on the parameter drift domain to generate the control parameter vector; S7. The dung beetle optimization algorithm based on drift attribution is used to optimize the control parameter vector. The gating behavior of individual dung beetles is assigned based on the attribution matrix. An asynchronous memory kernel is constructed to update the position of individual dung beetles. The three-domain twin model is used to perform barrier projection on the position of individual dung beetles to obtain the target parameter set.

[0006] Optionally, S1 specifically includes: To optimize the real-time operating data of the DC motor, a unified sampling period and acquisition time identifier are configured; the real-time operating data includes armature voltage, armature current, excitation current, PWM duty cycle, current loop output, speed, torque, speed loop output, winding temperature, housing temperature, ambient temperature, and load disturbance. The real-time running data is resampled according to a unified sampling period. The resampled real-time running data is then arranged according to the collection time identifier to generate time-aligned data. Armature voltage, armature current, magnetizing current, PWM duty cycle, and current loop output in the time-aligned data are divided into electrical excitation data. The speed, torque, and speed loop outputs in the time-aligned data are divided into mechanical response data. The winding temperature, housing temperature, and ambient temperature in the time-aligned data are classified as thermal hysteresis data. Establish a correspondence between electrical excitation data, mechanical response data, and thermal hysteresis data according to the acquisition time identifier.

[0007] Optionally, S2 specifically includes: The electrical excitation data, mechanical response data, and thermal hysteresis data are arranged according to the time of acquisition. Amplitude normalization was performed on the electrical excitation data, mechanical response data, and thermal hysteresis data respectively. The response start time difference is determined based on the start time of the change in electrical excitation data and the start time of the change in mechanical response data; the response peak time difference is determined based on the peak time of electrical excitation data and the peak time of mechanical response data. The response hysteresis is generated based on the response start time difference and the response peak time difference; The thermal state initiation time difference is determined based on the start time of the change in mechanical response data and the start time of the change in thermal hysteresis data; the thermal state peak time difference is determined based on the peak time of mechanical response data and the peak time of thermal hysteresis data. The thermal accumulation hysteresis is generated based on the thermal start time difference and the thermal peak time difference. The response hysteresis and the thermal accumulation hysteresis are arranged and fused according to the acquisition time identifier to generate the misaligned trajectory.

[0008] Optionally, S3 specifically includes: The moment when the amplitude of the misalignment in the misalignment trajectory first exceeds the first preset threshold is identified as the misalignment start point. The acquisition time when the misalignment amplitude in the misalignment trajectory reaches a local maximum value is identified as the misalignment peak point; The moment when the amplitude of the misalignment in the misalignment trajectory decreases from the local maximum value to the range of the second preset threshold is identified as the misalignment fallback point. The last acquisition time marker in the misalignment trajectory where the change in the misalignment amplitude corresponding to a continuous preset number of acquisition time markers is less than the third preset threshold is determined as the misalignment stable point. The starting point of the misalignment, the peak point of the misalignment, the fallback point of the misalignment, and the stable point of the misalignment are denoted as the misalignment characteristic points; Based on the order of the misalignment feature points in the acquisition time markers, the corresponding data intervals are extracted from the electrical excitation data, mechanical response data, and thermal hysteresis data to generate drift anchor plates.

[0009] Optionally, the construction of the three-domain twin model involves inputting the drifting anchor piece into the three-domain twin model to obtain twin prediction data, specifically as follows: Establish electrical, mechanical, and thermal domains based on the rated parameters and historical operating data of the DC motor; The electrical domain is configured with armature resistance, armature inductance, back electromotive force coefficient, torque coefficient, and current response delay parameters. The mechanical domain is configured with rotational inertia, frictional damping, estimated load torque, and speed response delay parameters. The thermal domain configuration includes winding thermal capacity, shell thermal resistance, ambient heat transfer coefficient, temperature rise transfer delay parameter, and temperature compensation parameter. The electrical, mechanical, and thermal domains are coupled according to the relationships between electromagnetic torque, speed feedback, and temperature correction to generate a three-domain twin model; Input the electrical excitation data in the drifting anchor into the electrical domain to generate twin predicted current and twin predicted back electromotive force; The twin-predicted back EMF, twin-predicted current, and mechanical response data in the drifting anchor are input into the mechanical domain to generate twin-predicted speed and twin-predicted torque. The twin predicted current, twin predicted rotation speed, and thermal hysteresis data in the drifting anchor are input into the thermal domain to generate the twin predicted temperature. Twin prediction data is generated by combining twin prediction current, twin prediction speed, twin prediction torque, and twin prediction temperature.

[0010] Optionally, the step of comparing the real-time operating data corresponding to the drifting anchor piece with the twin prediction data to generate a three-domain bias tensor is as follows: The measured electrical quantities, measured mechanical quantities, and measured thermal quantities are extracted from the real-time operating data corresponding to the drifting anchor plate. The difference between the measured electrical quantities and the twin prediction current and twin prediction back electromotive force in the twin prediction data is calculated to generate electrical deviation components; The difference between the measured mechanical quantities and the twin predicted speed and twin predicted torque in the twin predicted data is calculated to generate mechanical deviation components; The difference between the measured thermal state quantity and the twin predicted temperature in the twin prediction data is calculated to generate a thermal deviation component; The electrical deviation component, mechanical deviation component, and thermal deviation component are combined according to the data acquisition time identifier to generate a three-domain deviation tensor.

[0011] Optionally, the construction of the attribution matrix based on the temporal state and coupling state of the three-domain deviation tensor specifically involves: According to the data acquisition time marker, the deviation start time, deviation peak time, deviation fall time, and deviation stabilization time of the electrical deviation component, mechanical deviation component, and thermal deviation component are extracted respectively. The timing state is generated based on the difference in the start time of the deviation, the difference in the peak time of the deviation, the difference in the duration of the deviation, and the difference in the duration of the deviation fall-back among the electrical deviation component, the mechanical deviation component, and the thermal deviation component. A coupling state is generated based on the synchronous change values ​​between the electrical deviation component and the mechanical deviation component, the synchronous change values ​​between the electrical deviation component and the thermal deviation component, and the synchronous change values ​​between the mechanical deviation component and the thermal deviation component. Arrange the timing and coupling states according to the electrical, mechanical, thermal, and external load disturbances to generate an attribution matrix.

[0012] Optionally, determining the parameter drift domain specifically involves: The attribution matrix describes the behavior of the electrical, mechanical, thermal, and external load disturbances. The columns of the attribution matrix are the deviation start time difference, deviation peak time difference, deviation duration difference, deviation fallback duration difference, synchronous change value between electrical deviation component and mechanical deviation component, synchronous change value between electrical deviation component and thermal deviation component, and synchronous change value between mechanical deviation component and thermal deviation component. Based on the elements and values ​​of the corresponding rows in the attribution matrix for the electrical, mechanical, thermal, and external load disturbances, attribution values ​​for the electrical, mechanical, thermal, and load disturbances are generated respectively. The row corresponding to the maximum value among the electrical domain attribution value, mechanical domain attribution value, thermal domain attribution value, and load disturbance attribution value is determined as the drift direction; The drift direction is determined based on the positive and negative directions of the corresponding electrical deviation component, mechanical deviation component, or thermal deviation component. The drift amplitude boundary is determined based on the peak deviation, duration of deviation, and duration of deviation fallback corresponding to the drift direction. The search contraction coefficient is determined based on the drift direction and drift amplitude boundaries; The drift direction, drift amplitude boundary, and search shrinkage coefficient are combined to generate the parameter drift domain.

[0013] Optionally, S6 specifically includes: If the drift direction is in the electrical domain, then update the electrical domain parameters according to the drift direction and drift amplitude boundaries; If the drift direction is to the mechanical domain, then update the mechanical domain parameters according to the drift direction and drift amplitude boundaries; If the drift direction is a hot zone, then update the hot zone parameters according to the drift direction and drift amplitude boundaries; If the drift direction is due to an external load disturbance, then update the load torque estimate according to the drift direction and drift amplitude boundaries; The range of control parameters is limited by the search contraction coefficient, and the control parameters within the limited range are combined to generate a control parameter vector.

[0014] Optionally, S7 specifically includes: The initial position of each dung beetle is generated based on the control parameter vector, and the search range of each dung beetle is limited according to the parameter drift domain. The drift dominance type is determined based on the electrical domain attribution value, mechanical domain attribution value, thermal domain attribution value, and load disturbance attribution value in the attribution matrix; Based on the drift dominance type, assign one of the following behaviors to dung beetle individuals: rolling ball behavior, reproductive behavior, foraging behavior, and stealing behavior, and generate gating behavior assignment results; Based on the electrical deviation component, mechanical deviation component, and thermal deviation component, the historical position increment of individual dung beetles is extracted respectively, and an asynchronous memory kernel is constructed; Based on the gating behavior allocation results and the asynchronous memory kernel, update the dung beetle individual position and generate candidate parameter positions; Input the candidate parameter positions into the three-domain twin model to obtain candidate prediction data; The barrier projection direction is generated based on the peak current, speed fluctuation, temperature peak, and PWM duty cycle change in the candidate prediction data. The candidate parameter positions are corrected according to the potential barrier projection direction to obtain safe candidate positions; The candidate safe locations are sorted by non-dominated order, and the control parameters corresponding to the candidate safe location with the largest crowding distance in the first non-dominated layer are selected to obtain the target parameter set.

[0015] The beneficial effects of this invention are: (1) This invention generates a misaligned trajectory by fusing response hysteresis and thermal accumulation hysteresis, and generates drift anchors based on the misaligned feature points in the misaligned trajectory. This makes the DC motor operating data no longer extracted according to time period or single working condition, but determined according to the time sequence deviation state between electrical excitation, mechanical response and thermal hysteresis. This can more accurately capture the occurrence, enhancement, fall back and stabilization process of parameter drift, and improve the pertinence of subsequent three-domain deviation tensor generation. (2) The attribution matrix is ​​constructed by the temporal state and coupling state of the three-domain deviation tensor, and the parameter drift domain is determined based on the attribution matrix. This enables the electrical domain, mechanical domain, thermal domain and external load disturbance to form a distinguishable drift source judgment result, avoiding the problem that the existing digital twin method can only compare the real data and the predicted data and cannot judge the source of deviation, reducing the situation of misjudging the load disturbance as the drift of the motor body parameters, and improving the accuracy of the three-domain twin model update; (3) The dung beetle optimization algorithm based on drift attribution is used to optimize the control parameter vector. The gating behavior of individual dung beetles is assigned through the attribution matrix, the position of individual dung beetles is updated through the asynchronous memory kernel, and the position of individual dung beetles is projected onto the position of individual dung beetles through the three-domain twin model. This makes the optimization process no longer use fixed behavior assignment and unified search range, but instead conducts directional search around the parameter drift domain, reducing the amount of irrelevant control parameter search, reducing the risk of current peak, speed fluctuation, temperature peak and PWM duty cycle change exceeding the limit caused by candidate parameters, and improving the matching degree between the target parameter group and the actual operating state of the DC motor. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a DC motor parameter optimization method based on digital twin proposed in this invention; Figure 2 This is a schematic diagram of the three-domain twin model proposed in this invention; Figure 3 This is a schematic diagram of the dung beetle optimization algorithm for drift attribution proposed in this invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0018] refer to Figures 1-3 A method for optimizing DC motor parameters based on digital twins includes the following steps: S1. Time-align the real-time operating data of the DC motor to be optimized, and divide it into electrical excitation data, mechanical response data and thermal hysteresis data according to electromechanical and thermal properties; S2. Using electrical excitation data as the driving reference, determine the response hysteresis between mechanical response data and electrical excitation data. Using mechanical response data as the transmission reference, determine the thermal accumulation hysteresis between thermal hysteresis data and mechanical response data. Combine the response hysteresis and thermal accumulation hysteresis to generate a misalignment trajectory. S3. Generate drifting anchor plates based on misaligned feature points in the misaligned trajectory; S4. Construct a three-domain twin model, input the drifting anchor piece into the three-domain twin model to obtain twin prediction data, and compare the real-time running data corresponding to the drifting anchor piece with the twin prediction data to generate a three-domain deviation tensor. S5. Construct an attribution matrix based on the temporal and coupling states of the three-domain deviation tensor to determine the parameter drift domain; S6. Update the three-domain twin model based on the parameter drift domain to generate the control parameter vector; S7. The dung beetle optimization algorithm based on drift attribution is used to optimize the control parameter vector. The gating behavior of individual dung beetles is assigned based on the attribution matrix. An asynchronous memory kernel is constructed to update the position of individual dung beetles. The three-domain twin model is used to perform barrier projection on the position of individual dung beetles to obtain the target parameter set.

[0019] In this embodiment, S1 specifically refers to: To optimize the real-time operating data of the DC motor, a unified sampling period and acquisition time identifier are configured. The unified sampling period is determined by the data period with the shortest and most stable continuous sampling among armature current, PWM duty cycle, and speed. The acquisition time identifier consists of the acquisition date, acquisition time, and cycle number. The real-time operating data includes armature voltage, armature current, excitation current, PWM duty cycle, current loop output, speed, torque, speed loop output, winding temperature, housing temperature, ambient temperature, and load disturbance. Real-time running data is resampled according to a unified sampling period. Data with a sampling period shorter than the unified sampling period is averaged according to the unified sampling period. Data with a sampling period longer than the unified sampling period is linearly interpolated according to adjacent collection times. The resampled real-time running data is arranged according to the acquisition time identifier to generate time-aligned data; Armature voltage, armature current, magnetizing current, PWM duty cycle, and current loop output in the time-aligned data are classified as electrical excitation data; speed, torque, and velocity loop output in the time-aligned data are classified as mechanical response data; and winding temperature, housing temperature, and ambient temperature in the time-aligned data are classified as thermal hysteresis data. Establish a correspondence between load disturbance values ​​and mechanical response data according to the acquisition time identifier; establish a one-to-one correspondence between electrical excitation data, mechanical response data, and thermal hysteresis data according to the same acquisition time identifier; delete time-aligned data with missing acquisition time identifiers; Time-aligned data with duplicate records under the same acquisition time identifier are averaged and merged; electrical excitation data, mechanical response data, and thermal hysteresis data are sorted in ascending order according to the acquisition time identifier.

[0020] In this embodiment, S2 specifically refers to: The electrical excitation data, mechanical response data, and thermal hysteresis data are arranged in ascending order according to the acquisition time identifier; The electrical excitation data, mechanical response data, and thermal hysteresis data are respectively normalized to obtain electrical normalized data, mechanical normalized data, and thermal normalized data. Based on the amplitude difference corresponding to the adjacent acquisition time markers in the electrical normalized data, mechanical normalized data, and thermal normalized data, the starting time of the change in electrical excitation data, mechanical response data, and thermal hysteresis data is determined respectively. Based on the maximum amplitude values ​​in the electrical normalized data, mechanical normalized data, and thermal normalized data, the peak times of the electrical excitation data, mechanical response data, and thermal hysteresis data are determined respectively. The time difference between the start time of the change in mechanical response data and the start time of the change in electrical excitation data is taken as the response start time difference, and the time difference between the peak time of mechanical response data and the peak time of electrical excitation data is taken as the response peak time difference. The response hysteresis is generated based on the response start time difference and the response peak time difference; The time difference between the start time of the change in thermal hysteresis data and the start time of the change in mechanical response data is taken as the thermal state start time difference, and the time difference between the peak time of thermal hysteresis data and the peak time of mechanical response data is taken as the thermal state peak time difference. The heat accumulation hysteresis is generated based on the thermal initiation time difference and the thermal peak time difference; The response hysteresis and thermal accumulation hysteresis are arranged according to the acquisition time marker, and the response hysteresis and thermal accumulation hysteresis under the same acquisition time marker are merged to generate a misaligned trajectory.

[0021] In this embodiment, S3 specifically refers to: Extract the misalignment amplitude and misalignment direction corresponding to each acquisition moment from the misalignment trajectory; The first preset threshold is determined by adding three times the standard deviation of the misalignment amplitude to the mean of the misalignment amplitude in the stable operation samples of the DC motor; the range of the second preset threshold is determined by the amplitude interval between the mean of the misalignment amplitude in the stable operation samples of the DC motor and the first preset threshold; the third preset threshold is determined by adding three times the standard deviation of the adjacent misalignment amplitude changes to the mean of the adjacent misalignment amplitude changes in the stable operation samples of the DC motor; the number of consecutive presets is determined by the number of sampling points per hundred milliseconds corresponding to the uniform sampling period, and the number of consecutive presets is not less than three; The acquisition time when the misalignment amplitude first exceeds the first preset threshold in the misalignment trajectory is identified as the misalignment start point; the acquisition time when the misalignment amplitude reaches a local maximum value after the misalignment start point is identified as the misalignment peak point; the acquisition time when the misalignment amplitude drops to the range of the second preset threshold after the misalignment peak point is identified as the misalignment fallback point; and the acquisition time when the change in misalignment amplitude corresponding to a preset number of consecutive acquisition time points after the misalignment fallback point is less than the third preset threshold is identified as the misalignment stabilization point. The starting point of the misalignment, the peak point of the misalignment, the fallback point of the misalignment, and the stable point of the misalignment are denoted as the misalignment characteristic points; Based on the acquisition time markers from the misalignment initiation point to the misalignment stabilization point, extract the electrical excitation data interval from the electrical excitation data; based on the acquisition time markers from the misalignment initiation point to the misalignment stabilization point, extract the mechanical response data interval from the mechanical response data; based on the acquisition time markers from the misalignment initiation point to the misalignment stabilization point, extract the thermal hysteresis data interval; establish a correspondence between the electrical excitation data interval, the mechanical response data interval, and the thermal hysteresis data interval based on the acquisition time markers. The electrical excitation data range, mechanical response data range, and thermal hysteresis data range, after establishing the correspondence, are combined to generate the drift anchor plate.

[0022] In this embodiment, a three-domain twin model is constructed, and the drifting anchor piece is input into the three-domain twin model to obtain twin prediction data, specifically: The electrical, mechanical, and thermal domains are established based on the rated parameters and historical operating data of the DC motor. The rated parameters of the DC motor include rated voltage, rated current, rated speed, rated torque, rated power, rated winding temperature, and rated housing temperature. The historical operating data includes historical electrical excitation data, historical mechanical response data, historical thermal hysteresis data, and historical load disturbance. Based on the armature voltage, armature current, magnetizing current, PWM duty cycle, and current loop output from historical electrical excitation data, determine the armature resistance, armature inductance, back EMF coefficient, torque coefficient, and current response delay parameters; based on the speed, torque, speed loop output, and load disturbance from historical mechanical response data, determine the moment of inertia, friction damping, estimated load torque, and speed response delay parameters; based on the winding temperature, shell temperature, and ambient temperature from historical thermal hysteresis data, determine the winding heat capacity, shell thermal resistance, ambient heat transfer coefficient, temperature rise propagation delay parameters, and temperature compensation parameters. Armature resistance, armature inductance, back EMF coefficient, torque coefficient, and current response delay parameter are configured in the electrical domain; moment of inertia, friction damping, load torque estimate, and speed response delay parameter are configured in the mechanical domain; winding heat capacity, housing thermal resistance, ambient heat transfer coefficient, temperature rise transfer delay parameter, and temperature compensation parameter are configured in the thermal domain; the twin predicted current output from the electrical domain is multiplied by the torque coefficient to generate electromagnetic torque; The electromagnetic torque is input into the mechanical domain, and the twin predicted speed and twin predicted torque are generated by combining the load torque estimate and friction damping; the twin predicted speed output from the mechanical domain is fed back to the electrical domain, and the twin predicted back EMF is generated by combining the back EMF coefficient; the twin predicted temperature output from the thermal domain is input into the electrical domain, and the armature resistance is corrected by combining the temperature compensation parameters. The electromagnetic torque, twin predicted speed feedback, and temperature correction relationship are configured between the electrical, mechanical, and thermal domains to generate a three-domain twin model; The electrical excitation data in the drift anchor is input into the electrical domain. The electrical domain generates twin predicted current and twin predicted back EMF based on armature voltage, armature current, excitation current, PWM duty cycle, current loop output, armature resistance, armature inductance, back EMF coefficient and current response delay parameters. The twin predicted current, twin predicted back electromotive force, and mechanical response data in the drifting anchor are input into the mechanical domain. The mechanical domain generates twin predicted speed and twin predicted torque based on the moment of inertia, friction damping, load torque estimate, and speed response delay parameter. The twin predicted current, twin predicted rotation speed, and thermal hysteresis data in the drift anchor are input into the thermal domain. The thermal domain generates the twin predicted temperature based on the winding thermal capacity, shell thermal resistance, ambient heat transfer coefficient, temperature rise transfer delay parameter, and temperature compensation parameter. The twin predicted current, twin predicted back electromotive force, twin predicted speed, twin predicted torque, and twin predicted temperature are arranged according to the data acquisition time marker. Establish a correspondence between twin predicted current, twin predicted back electromotive force, twin predicted speed, twin predicted torque, and twin predicted temperature under the same acquisition time identifier, and generate twin predicted data.

[0023] In this embodiment, the real-time operating data corresponding to the drifting anchor piece is compared with the twin prediction data to generate a three-domain bias tensor, specifically: Extract the real-time operational data corresponding to the drifting anchor piece according to the collection time identifier; Armature voltage, armature current and excitation current are extracted from the real-time operating data corresponding to the drifting anchor to generate measured electrical quantities; The measured back electromotive force is generated based on the armature voltage, armature current and rotational speed, and the armature current and measured back electromotive force are incorporated into the measured electrical quantities. From the real-time operating data corresponding to the drifting anchor, the rotational speed and torque are extracted to generate measured mechanical quantities; the winding temperature and shell temperature are extracted to generate measured thermal quantities. The armature current under the same acquisition time mark is subtracted from the twin predicted current to generate a current deviation term; the measured back EMF under the same acquisition time mark is subtracted from the twin predicted back EMF to generate a back EMF deviation term. Arrange the current deviation term and back electromotive force deviation term according to the data acquisition time identifier to generate electrical deviation components; The difference between the rotational speed at the same acquisition time and the twin predicted rotational speed is used to generate a rotational speed deviation term; the difference between the torque at the same acquisition time and the twin predicted torque is used to generate a torque deviation term. Arrange the speed deviation and torque deviation items according to the data acquisition time identifier to generate mechanical deviation components; The winding temperature deviation term is generated by subtracting the winding temperature under the same acquisition time marker from the twin predicted winding temperature in the twin predicted temperature; the shell temperature deviation term is generated by subtracting the shell temperature under the same acquisition time marker from the twin predicted shell temperature in the twin predicted temperature. Arrange the winding temperature deviation items and the shell temperature deviation items according to the data acquisition time markers to generate thermal deviation components; Establish a three-dimensional arrangement relationship based on the data acquisition time identifier, the category of deviation components, and the category of deviation items; The electrical deviation component, mechanical deviation component, and thermal deviation component are filled into a three-dimensional arrangement relationship to generate a three-domain deviation tensor. The first dimension of the three-domain deviation tensor is the acquisition time identifier, the second dimension is the electrical deviation component, mechanical deviation component, and thermal deviation component, and the third dimension is the deviation term contained in each deviation component.

[0024] In this embodiment, an attribution matrix is ​​constructed based on the temporal state and coupling state of the three-domain deviation tensor, specifically as follows: The electrical deviation component, mechanical deviation component, and thermal deviation component are extracted according to the data acquisition time marker. Calculate the deviation amplitude of the electrical deviation component, mechanical deviation component, and thermal deviation component at each acquisition time. Based on the sum of the mean deviation amplitude and three times the standard deviation of the deviation amplitude in the stable operation samples of the DC motor, the deviation judgment thresholds for the electrical deviation component, mechanical deviation component and thermal deviation component are determined respectively. The sampling time when the deviation amplitude first exceeds the corresponding deviation judgment threshold is identified as the deviation start time; the sampling time when the deviation amplitude reaches a local maximum value is identified as the deviation peak time; the sampling time when the deviation amplitude drops from the local maximum value to the range of the corresponding deviation judgment threshold is identified as the deviation fall-off time; and the sampling time when the change in deviation amplitude corresponding to a preset number of consecutive sampling time markers is less than the third preset threshold is identified as the deviation stabilization time. The timing state is generated based on the difference in the start time of the deviation, the difference in the peak time of the deviation, the difference in the duration of the deviation, and the difference in the duration of the deviation fall-back among the electrical deviation component, the mechanical deviation component, and the thermal deviation component. Calculate the deviation increments of the electrical deviation component, mechanical deviation component, and thermal deviation component between adjacent acquisition time markers, respectively; When the deviation increments of two types of deviation components are in the same direction at the same acquisition time, they are recorded as positive synchronization values; when the deviation increments of two types of deviation components are in opposite directions at the same acquisition time, they are recorded as negative synchronization values; and when any deviation increment is 0, it is recorded as zero synchronization value. Based on the positive synchronization value, the reverse synchronization value, and the zero synchronization value, the synchronization change values ​​between the electrical deviation component and the mechanical deviation component, the synchronization change values ​​between the electrical deviation component and the thermal deviation component, and the synchronization change values ​​between the mechanical deviation component and the thermal deviation component are generated respectively. A coupling state is generated based on the synchronous change values ​​between the electrical deviation component and the mechanical deviation component, the synchronous change values ​​between the electrical deviation component and the thermal deviation component, and the synchronous change values ​​between the mechanical deviation component and the thermal deviation component. The rows of the attribution matrix are set as electrical, mechanical, thermal, and external load disturbances; The columns of the attribution matrix are set as follows: deviation start time difference, deviation peak time difference, deviation duration difference, deviation fallback duration difference, synchronous change value between electrical deviation component and mechanical deviation component, synchronous change value between electrical deviation component and thermal deviation component, and synchronous change value between mechanical deviation component and thermal deviation component. The timing state and coupling state are filled in according to the rows and columns of the attribution matrix to generate the attribution matrix.

[0025] In this embodiment, the parameter drift domain is determined as follows: Read the elements of the corresponding rows of the electrical, mechanical, thermal, and external load disturbances in the attribution matrix; normalize the elements of each column of the attribution matrix according to the maximum absolute value in the same column, and record the column elements with the maximum absolute value of 0 as 0; The normalized elements of the corresponding rows in the electrical domain, mechanical domain, thermal domain, and external load disturbance are summed to generate the electrical domain attribution value, mechanical domain attribution value, thermal domain attribution value, and load disturbance attribution value. The row corresponding to the maximum value among the electrical domain attribution value, mechanical domain attribution value, thermal domain attribution value, and load disturbance attribution value is determined as the drift direction; if there are multiple maximum values, the row corresponding to the earliest deviation component at the peak time of the deviation is determined as the drift direction. If the drift direction is in the electrical domain, the drift direction is determined by the positive and negative directions of the electrical deviation component between the initial and peak values ​​of the deviation. If the drift direction is in the mechanical domain, the drift direction is determined by the positive and negative directions of the mechanical deviation component between the initial and peak values ​​of the deviation. If the drift direction is in the thermal domain, the drift direction is determined by the positive and negative directions of the thermal deviation component between the initial and peak values ​​of the deviation. If the drift direction is due to external load disturbance, the drift direction is determined by the changing directions of the mechanical deviation component and the load disturbance at the same acquisition time. The first boundary value is determined based on the peak amplitude of the deviation corresponding to the drift direction; the second boundary value is determined based on the ratio of the duration of the deviation corresponding to the drift direction to the duration of the drift anchor piece; the third boundary value is determined based on the ratio of the duration of the deviation fallback corresponding to the drift direction to the duration of the drift anchor piece. The maximum value among the first, second, and third boundary values ​​is used as the drift amplitude boundary; the drift amplitude boundary is multiplied by the attribution value corresponding to the drift direction to generate the search contraction coefficient; the drift direction, drift direction, drift amplitude boundary, and search contraction coefficient are combined to generate the parameter drift domain.

[0026] In this embodiment, S6 specifically refers to: The drift direction includes positive and negative; the parameter correction amount is generated by multiplying the drift amplitude boundary and the original parameter value corresponding to the drift direction. If the drift direction is in the electrical domain, update the electrical domain parameters according to the drift direction and parameter correction amount; if the drift direction is in the mechanical domain, update the mechanical domain parameters according to the drift direction and parameter correction amount; if the drift direction is in the thermal domain, update the thermal domain parameters according to the drift direction and parameter correction amount; if the drift direction is due to external load disturbance, update the load torque estimate according to the drift direction and parameter correction amount. The allowable parameter ranges for DC motor controllers include the speed loop proportional parameter range, speed loop integral parameter range, current loop proportional parameter range, current loop integral parameter range, PWM duty cycle limiting parameter range, torque feedforward compensation parameter range, speed error compensation parameter range, current safety limiting parameter range, temperature compensation parameter range, and current derating parameter range. Keep the center value of the allowable parameter range of the DC motor controller unchanged, and multiply the width of the allowable parameter range of the DC motor controller by the search contraction coefficient to generate the limited value range; The control parameter vector is generated by combining the speed loop proportional parameter, speed loop integral parameter, current loop proportional parameter, current loop integral parameter, PWM duty cycle limiting parameter, torque feedforward compensation parameter, speed error compensation parameter, current safety limiting parameter, temperature compensation parameter, and current derating parameter within the limited range of values.

[0027] In this embodiment, S7 specifically refers to: The position dimension of each dung beetle is established according to the control parameter dimension in the control parameter vector, and each position dimension of each dung beetle corresponds to one control parameter; The search range for individual dung beetles is limited according to the range of values ​​given by the parameter drift domain; the initial position of each dung beetle is generated within the search range. Read the electrical domain attribution value, mechanical domain attribution value, thermal domain attribution value and load disturbance attribution value in the attribution matrix, and determine the electrical domain, mechanical domain, thermal domain or external load disturbance corresponding to the maximum attribution value as the drift-dominant type; If the drift dominance type is electrical, then the dung beetle is assigned the ball-rolling behavior; if the drift dominance type is mechanical, then the dung beetle is assigned the foraging behavior; if the drift dominance type is thermal, then the dung beetle is assigned the reproductive behavior; if the drift dominance type is external load disturbance, then the dung beetle is assigned the theft behavior. The position difference of a dung beetle in adjacent iterations is recorded as the historical position increment; The electrical memory length is the number of sampling points corresponding to a uniform sampling period of 20 milliseconds; the mechanical memory length is the number of sampling points corresponding to the response hysteresis; and the thermal memory length is the number of sampling points corresponding to the thermal accumulation hysteresis. Based on the electrical memory length, historical position increments are extracted from the iterations corresponding to the electrical deviation components to generate an electrical memory sequence; based on the mechanical memory length, historical position increments are extracted from the iterations corresponding to the mechanical deviation components to generate a mechanical memory sequence; based on the thermal memory length, historical position increments are extracted from the iterations corresponding to the thermal deviation components to generate a thermal memory sequence. An asynchronous memory core is generated by combining electrical memory sequences, mechanical memory sequences, and thermal memory sequences. If an individual dung beetle is assigned to the rolling ball behavior, then the position of the individual dung beetle is updated based on the position difference between the group's optimal position and the individual dung beetle's position, as well as the electrical memory sequence. If dung beetle individuals are assigned foraging behaviors, then the dung beetle individual positions are updated based on the positional difference between the random individual positions and the dung beetle individual positions, as well as the mechanical memory sequence. If a dung beetle is assigned reproductive behavior, then the dung beetle's position is updated based on the neighborhood boundary of the individual's location and the thermal memory sequence. If an individual dung beetle is assigned the act of stealing, then the position of the individual dung beetle is updated based on the position difference between the group's optimal position, the random individual position, and the individual dung beetle position. Record the updated dung beetle individual position as the candidate parameter position; The candidate parameter positions are input into the three-domain twin model to generate candidate prediction data; the peak current, speed fluctuation, temperature peak and PWM duty cycle change are extracted from the candidate prediction data. The current safety boundary is determined by the product of the rated current and the driver's allowable overload ratio; the speed safety boundary is determined by the product of the rated speed and the allowable overspeed coefficient; the temperature safety boundary is determined by the smaller of the rated winding temperature and the allowable temperature of the insulation class; the PWM duty cycle variation boundary is determined by the allowable parameter range of the DC motor controller. If any of the values ​​of peak current, speed fluctuation, peak temperature, and PWM duty cycle change exceeds the corresponding safety boundary, the dimension of the position corresponding to the value exceeding the safety boundary will be shrunk towards the center of the value range, and the direction of the shrunk direction will be recorded as the barrier projection direction. If the peak current, speed fluctuation, temperature peak, and PWM duty cycle change do not exceed the corresponding safety boundary, then the zero vector is recorded as the barrier projection direction. The candidate parameter positions are corrected according to the potential barrier projection direction to obtain safe candidate positions; evaluation parameters are established based on the three-domain deviation amplitude, current peak, speed fluctuation, temperature peak and PWM duty cycle change corresponding to the safe candidate positions. If the total evaluation value of one safe candidate position is not greater than the total evaluation value of another safe candidate position, and at least one evaluation value is less than the corresponding evaluation value of another safe candidate position, then it is determined that the former safe candidate position dominates the latter safe candidate position. Security candidate positions that are not dominated by any security candidate position are assigned to the first non-dominated layer. The safe candidate positions in the first non-dominated layer are sorted in ascending order according to each evaluation value; For each safe candidate location, the sum of the differences between adjacent safe candidate locations on each evaluation metric is calculated to generate the congestion distance; The control parameters corresponding to the safe candidate position with the largest congestion distance in the first non-dominated layer are selected to obtain the target parameter set.

[0028] In this embodiment, the dung beetle optimization algorithm based on drift attribution inherits the group search framework, dung beetle individual position representation method, search boundary limitation method, individual fitness evaluation method, and position update ideas corresponding to rolling ball behavior, reproductive behavior, foraging behavior, and stealing behavior from the dung beetle optimization algorithm. It uses the control parameter vector as the dung beetle individual position, the range of control parameter values ​​as the dung beetle individual search boundary, and obtains the target parameter set through multiple rounds of iterative search. Building upon this foundation, the dung beetle optimization algorithm based on drift attribution does not employ the random or fixed-proportion allocation of dung beetle behavior as in the original algorithm. Instead, it determines the drift dominance type based on the electrical, mechanical, thermal, and load disturbance attribution values ​​in the attribution matrix, and then assigns gated behavior to individual dung beetles according to this dominance type. Furthermore, instead of using a single historical position increment for individual position updates, it constructs asynchronous memory kernels based on the electrical, mechanical, and thermal deviation components, ensuring that the dung beetle position update simultaneously preserves the variation characteristics of electrical response, mechanical response, and thermal hysteresis at different time scales. Further, after updating the individual dung beetle position, the candidate parameter positions are input into a three-domain twin model. Based on the current peak, speed fluctuation, temperature peak, and PWM duty cycle change in the candidate prediction data, a barrier projection direction is generated, and the candidate parameter positions are corrected according to this barrier projection direction. Through the above improvements, the parameter optimization process can search around the drift sources corresponding to the electrical domain, mechanical domain, thermal domain, or external load disturbances, reducing the large-scale search of irrelevant control parameters, reducing the risk of candidate parameters causing current out-of-bounds, temperature out-of-bounds, and speed fluctuation out-of-bounds, and improving the matching degree between the target parameter set and the actual operating state of the DC motor.

[0029] Example 1: To verify the feasibility of this invention in practice, it was applied to a DC motor drive unit on an automotive parts assembly line. This drive unit uses a brushed DC motor with a rated power of 2.2kW, a rated voltage of 220V, and a rated speed of 1500r / min, primarily responsible for the reciprocating movement of the feeding slide. During production, the feeding slide needs to switch between no-load, half-load, full-load, and frequent start-stop states. The original control scheme used fixed speed loop parameters and fixed current loop parameters. After continuous operation, due to increased slide rail friction, load fluctuations, and the combined effect of winding temperature rise, the motor experienced slow speed recovery, high armature current peaks, and continuously rising housing temperature during full-load start-up and emergency stop reset phases, leading to unstable feeding cycle time.

[0030] When applying this invention, armature voltage, armature current, excitation current, PWM duty cycle, current loop output, speed, torque, speed loop output, winding temperature, housing temperature, ambient temperature, and load disturbance are time-aligned and divided into electrical excitation data, mechanical response data, and thermal hysteresis data according to electromechanical and thermal properties. After generating a misalignment trajectory based on response hysteresis and thermal accumulation hysteresis, misalignment feature points are extracted from the misalignment trajectory to generate drift anchors. A three-domain twin model predicts the drift anchors, obtaining twin prediction data, which is then compared with the corresponding real-time operating data to generate a three-domain deviation tensor. The attribution matrix shows that the mechanical domain attribution value is highest during the full-load start-up phase, and the thermal domain attribution value increases after 45 minutes of continuous operation, indicating that the control parameter mismatch is not a single current loop problem, but is caused by both mechanical response hysteresis and thermal drift. After updating the three-domain twin model according to the parameter drift domain, a dung beetle optimization algorithm for drift attribution is used to generate the target parameter set.

[0031] Table 1 Comparison of DC motor operating performance before and after optimization

[0032] As shown in Table 1, under all four typical operating conditions, this invention significantly reduces the maximum speed error and recovery time. Under full-load start-up conditions, the maximum speed error decreased from 118.4 r / min to 49.7 r / min, and the recovery time decreased from 930 ms to 548 ms, indicating that the analysis interval extracted by the misalignment trajectory and drift anchor can accurately reflect the mechanical response hysteresis. Under frequent start-stop conditions, the winding temperature rise decreased from 26.4℃ to 18.2℃, and the peak current decreased from 33.7A to 26.6A, indicating that the three-domain deviation tensor and attribution matrix can distinguish between thermal domain drift and mechanical domain drift, avoiding coarse compensation by simply increasing the current loop response speed, thereby reducing current surges and heat accumulation.

[0033] Table 2 Comparison of Parameter Drift Attribution and Optimization Process

[0034] Table 2 shows that the parameter drift domain can provide different drift directions depending on the operating conditions. During the cold start-up phase, the main deviation is in the electrical domain; during the full-load start-up phase, the mechanical domain attribution value reaches 0.81; and after 45 minutes of continuous operation, the thermal domain attribution value reaches 0.79, consistent with the change pattern of the motor in the field from electrical response to mechanical hysteresis and then to thermal accumulation. The search contraction coefficient is 0.46 and 0.49 during the full-load start-up and hot-state operation phases, respectively, which narrows the optimization range of the control parameter vector around the main drift source, reducing the search for irrelevant parameters. Barrier projection processing corrects out-of-bounds candidate positions to safe candidate positions, maintaining the target parameter group verification pass rate above 92.6%, proving that the dung beetle optimization algorithm for drift attribution can improve optimization efficiency and safety.

[0035] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for optimizing DC motor parameters based on digital twins, characterized in that, Includes the following steps: S1. Time-align the real-time operating data of the DC motor to be optimized, and divide it into electrical excitation data, mechanical response data and thermal hysteresis data according to electromechanical and thermal properties; S2. Using electrical excitation data as the driving reference, determine the response hysteresis between mechanical response data and electrical excitation data. Using mechanical response data as the transmission reference, determine the thermal accumulation hysteresis between thermal hysteresis data and mechanical response data. Combine the response hysteresis and thermal accumulation hysteresis to generate a misalignment trajectory. S3. Generate drifting anchor plates based on misaligned feature points in the misaligned trajectory; S4. Construct a three-domain twin model, input the drifting anchor piece into the three-domain twin model to obtain twin prediction data, and compare the real-time running data corresponding to the drifting anchor piece with the twin prediction data to generate a three-domain deviation tensor. S5. Construct an attribution matrix based on the temporal and coupling states of the three-domain deviation tensor to determine the parameter drift domain; S6. Update the three-domain twin model based on the parameter drift domain to generate the control parameter vector; S7. The dung beetle optimization algorithm based on drift attribution is used to optimize the control parameter vector. The gating behavior of individual dung beetles is assigned based on the attribution matrix. An asynchronous memory kernel is constructed to update the position of individual dung beetles. The three-domain twin model is used to perform barrier projection on the position of individual dung beetles to obtain the target parameter set.

2. The DC motor parameter optimization method based on digital twin according to claim 1, characterized in that, Specifically, S1 is: To optimize the real-time operating data of the DC motor, a unified sampling period and acquisition time identifier are configured; the real-time operating data includes armature voltage, armature current, excitation current, PWM duty cycle, current loop output, speed, torque, speed loop output, winding temperature, housing temperature, ambient temperature, and load disturbance. The real-time running data is resampled according to a unified sampling period. The resampled real-time running data is then arranged according to the collection time identifier to generate time-aligned data. Armature voltage, armature current, magnetizing current, PWM duty cycle, and current loop output in the time-aligned data are divided into electrical excitation data. The speed, torque, and speed loop outputs in the time-aligned data are divided into mechanical response data. The winding temperature, housing temperature, and ambient temperature in the time-aligned data are classified as thermal hysteresis data. Establish a correspondence between electrical excitation data, mechanical response data, and thermal hysteresis data according to the acquisition time identifier.

3. The DC motor parameter optimization method based on digital twin according to claim 1, characterized in that, Specifically, S2 is: The electrical excitation data, mechanical response data, and thermal hysteresis data are arranged according to the time of acquisition. Amplitude normalization was performed on the electrical excitation data, mechanical response data, and thermal hysteresis data respectively. The response start time difference is determined based on the start time of the change in electrical excitation data and the start time of the change in mechanical response data; The peak response time difference is determined based on the peak times of the electrical excitation data and the mechanical response data. The response hysteresis is generated based on the response start time difference and the response peak time difference; The thermal initiation time difference is determined based on the start time of the change in mechanical response data and the start time of the change in thermal hysteresis data. The time difference between the thermal peak values ​​is determined based on the peak times of the mechanical response data and the thermal hysteresis data. The thermal accumulation hysteresis is generated based on the thermal start time difference and the thermal peak time difference. The response hysteresis and the thermal accumulation hysteresis are arranged and fused according to the acquisition time identifier to generate the misaligned trajectory.

4. The DC motor parameter optimization method based on digital twin according to claim 1, characterized in that, Specifically, S3 is: The moment when the amplitude of the misalignment in the misalignment trajectory first exceeds the first preset threshold is identified as the misalignment start point. The acquisition time when the misalignment amplitude in the misalignment trajectory reaches a local maximum value is identified as the misalignment peak point; The moment when the amplitude of the misalignment in the misalignment trajectory decreases from the local maximum value to the range of the second preset threshold is identified as the misalignment fallback point. The last acquisition time marker in the misalignment trajectory where the change in the misalignment amplitude corresponding to a continuous preset number of acquisition time markers is less than the third preset threshold is determined as the misalignment stable point. The starting point of the misalignment, the peak point of the misalignment, the fallback point of the misalignment, and the stable point of the misalignment are denoted as the misalignment characteristic points; Based on the order of the misalignment feature points in the acquisition time markers, the corresponding data intervals are extracted from the electrical excitation data, mechanical response data, and thermal hysteresis data to generate drift anchor plates.

5. The DC motor parameter optimization method based on digital twin according to claim 1, characterized in that, The construction of the three-domain twin model involves inputting the drifting anchor piece into the three-domain twin model to obtain twin prediction data, specifically as follows: Establish electrical, mechanical, and thermal domains based on the rated parameters and historical operating data of the DC motor; The electrical domain is configured with armature resistance, armature inductance, back electromotive force coefficient, torque coefficient, and current response delay parameters. The mechanical domain is configured with rotational inertia, frictional damping, estimated load torque, and speed response delay parameters. The thermal domain configuration includes winding thermal capacity, shell thermal resistance, ambient heat transfer coefficient, temperature rise transfer delay parameter, and temperature compensation parameter. The electrical, mechanical, and thermal domains are coupled according to the relationships between electromagnetic torque, speed feedback, and temperature correction to generate a three-domain twin model; Input the electrical excitation data in the drifting anchor into the electrical domain to generate twin predicted current and twin predicted back electromotive force; The twin-predicted back EMF, twin-predicted current, and mechanical response data in the drifting anchor are input into the mechanical domain to generate twin-predicted speed and twin-predicted torque. The twin predicted current, twin predicted rotation speed, and thermal hysteresis data in the drifting anchor are input into the thermal domain to generate the twin predicted temperature. Twin prediction data is generated by combining twin prediction current, twin prediction speed, twin prediction torque, and twin prediction temperature.

6. The DC motor parameter optimization method based on digital twin according to claim 1, characterized in that, The process of comparing the real-time operational data corresponding to the drifting anchor piece with the twin prediction data to generate a three-domain bias tensor is as follows: The measured electrical quantities, measured mechanical quantities, and measured thermal quantities are extracted from the real-time operating data corresponding to the drifting anchor plate. The difference between the measured electrical quantities and the twin prediction current and twin prediction back electromotive force in the twin prediction data is calculated to generate electrical deviation components; The difference between the measured mechanical quantities and the twin predicted speed and twin predicted torque in the twin predicted data is calculated to generate mechanical deviation components; The difference between the measured thermal state quantity and the twin prediction temperature in the twin prediction data is calculated to generate a thermal deviation component; The electrical deviation component, mechanical deviation component, and thermal deviation component are combined according to the data acquisition time identifier to generate a three-domain deviation tensor.

7. The DC motor parameter optimization method based on digital twin according to claim 1, characterized in that, The attribution matrix is ​​constructed based on the temporal state and coupling state of the three-domain deviation tensor, specifically as follows: According to the data acquisition time marker, the deviation start time, deviation peak time, deviation fall time, and deviation stabilization time of the electrical deviation component, mechanical deviation component, and thermal deviation component are extracted respectively. The timing state is generated based on the difference in the start time of the deviation, the difference in the peak time of the deviation, the difference in the duration of the deviation, and the difference in the duration of the deviation fall-back among the electrical deviation component, the mechanical deviation component, and the thermal deviation component. A coupling state is generated based on the synchronous change values ​​between the electrical deviation component and the mechanical deviation component, the synchronous change values ​​between the electrical deviation component and the thermal deviation component, and the synchronous change values ​​between the mechanical deviation component and the thermal deviation component. Arrange the timing and coupling states according to the electrical, mechanical, thermal, and external load disturbances to generate an attribution matrix.

8. The DC motor parameter optimization method based on digital twin according to claim 1, characterized in that, The determination of the parameter drift domain specifically refers to: The attribution matrix describes the behavior of the electrical, mechanical, thermal, and external load disturbances. The columns of the attribution matrix are the deviation start time difference, deviation peak time difference, deviation duration difference, deviation fallback duration difference, synchronous change value between electrical deviation component and mechanical deviation component, synchronous change value between electrical deviation component and thermal deviation component, and synchronous change value between mechanical deviation component and thermal deviation component. Based on the elements and values ​​of the corresponding rows in the attribution matrix for the electrical, mechanical, thermal, and external load disturbances, attribution values ​​for the electrical, mechanical, thermal, and load disturbances are generated respectively. The row corresponding to the maximum value among the electrical domain attribution value, mechanical domain attribution value, thermal domain attribution value, and load disturbance attribution value is determined as the drift direction; The drift direction is determined based on the positive and negative directions of the corresponding electrical deviation component, mechanical deviation component, or thermal deviation component. The drift amplitude boundary is determined based on the peak deviation, duration of deviation, and duration of deviation fallback corresponding to the drift direction. The search contraction coefficient is determined based on the drift direction and drift amplitude boundaries; The drift direction, drift amplitude boundary, and search shrinkage coefficient are combined to generate the parameter drift domain.

9. The DC motor parameter optimization method based on digital twin according to claim 1, characterized in that, Specifically, S6 is: If the drift direction is in the electrical domain, then update the electrical domain parameters according to the drift direction and drift amplitude boundaries; If the drift direction is to the mechanical domain, then update the mechanical domain parameters according to the drift direction and drift amplitude boundaries; If the drift direction is a hot zone, then update the hot zone parameters according to the drift direction and drift amplitude boundaries; If the drift direction is due to an external load disturbance, then update the load torque estimate according to the drift direction and drift amplitude boundaries; The range of control parameters is limited by the search contraction coefficient, and the control parameters within the limited range are combined to generate a control parameter vector.

10. The DC motor parameter optimization method based on digital twin according to claim 1, characterized in that, Specifically, S7 is: The initial position of each dung beetle is generated based on the control parameter vector, and the search range of each dung beetle is limited according to the parameter drift domain. The drift dominance type is determined based on the electrical domain attribution value, mechanical domain attribution value, thermal domain attribution value, and load disturbance attribution value in the attribution matrix; Based on the drift dominance type, assign one of the following behaviors to dung beetle individuals: rolling ball behavior, reproductive behavior, foraging behavior, and stealing behavior, and generate gating behavior assignment results; Based on the electrical deviation component, mechanical deviation component, and thermal deviation component, the historical position increment of individual dung beetles is extracted respectively, and an asynchronous memory kernel is constructed; Based on the gating behavior allocation results and the asynchronous memory kernel, update the dung beetle individual position and generate candidate parameter positions; Input the candidate parameter positions into the three-domain twin model to obtain candidate prediction data; The barrier projection direction is generated based on the peak current, speed fluctuation, temperature peak, and PWM duty cycle change in the candidate prediction data. The candidate parameter positions are corrected according to the potential barrier projection direction to obtain safe candidate positions; The candidate safe locations are sorted by non-dominated order, and the control parameters corresponding to the candidate safe location with the largest crowding distance in the first non-dominated layer are selected to obtain the target parameter set.

Images