A visual control fusion electrical drive method and system executed by an industrial intelligent gateway

By using the cross-physical field mapping technology of processing visual data and current signal flow through industrial intelligent gateways, the problem of electrical drive systems being unable to predict load disturbances during high-speed processing is solved, and the forward-looking compensation and control performance of the drive system are improved.

CN123122382APending Publication Date: 2026-11-17SHANDONG GUANFEI ENERGY EQUIPMENT MANUFACTURING CO LTD +1
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202611232275.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing electric drive systems cannot predict and actively suppress transient load disturbances caused by sudden changes in workpiece geometry or transmission characteristics during high-speed machining, resulting in dynamic errors such as trajectory overshoot and resonance, which limits control accuracy and stability.

Method used

By simultaneously processing visual data streams and current signal streams through an industrial smart gateway, cross-physical field mapping and forward-looking compensation technology are generated to reshape the drive current waveform to actively suppress load disturbances and improve control performance.

Benefits of technology

It enables the reshaping of the drive current waveform before disturbances occur, actively suppresses load disturbances, improves the control accuracy and stability of the drive system, and avoids the limitations of traditional communication bottlenecks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123122382A_ABST
    Figure CN123122382A_ABST
Patent Text Reader

Abstract

This invention discloses a vision-control fusion electrical drive method and system executed by an industrial intelligent gateway, belonging to the field of industrial automation control technology. It includes acquiring a visual data stream of the working scene and a real-time drive current signal stream; extracting features from the visual data stream of the working scene to generate a workpiece geometric description tensor; converting the real-time drive current signal stream into a drive current time-frequency spectrum and performing noise energy analysis to generate a current noise energy fluctuation spectrum; inputting this spectrum into a cross-physics field interference mapping model; executing the interference mapping to generate a current waveform modulation strategy; reshaping the waveform of the basic current command generated in the current control cycle to obtain a fused drive current signal, which is then sent to the electrical driver. This invention employs a cross-physics field mapping and forward-looking compensation technology that fuses visual data of the working scene and drive current signals, enabling the reshaping of the drive current waveform before disturbances occur, thereby actively suppressing load disturbances and improving the control performance of the drive system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of industrial automation control technology, and in particular to a visual control fusion electrical drive method and system executed by an industrial intelligent gateway. Background Technology

[0002] Currently, electric drive systems are the core execution units in modern industrial manufacturing equipment. Their performance, especially the smoothness and precision of motion control, directly determines the final product's processing quality and production efficiency. In high-end equipment such as CNC machine tools and industrial robots, these drive systems typically operate under pre-programmed instructions, achieving complex motion trajectories and technological processes through precise control of motor torque and speed. This falls under the typical category of program control systems.

[0003] In related technologies, Chinese invention patent CN120768185B discloses an intelligent control method and system for a motor driver, including: constructing a fault feature library by collecting harmonic current and sound pressure fluctuation data during motor operation and building an acoustic-electric correlation matrix; identifying usable fault modes to determine energy harvesting points and performing phase shift processing on the harvesting points to generate auxiliary drive signals; performing spectral analysis on the sound pressure fluctuation data to obtain acoustic feature spectra, extracting efficiency indicator tones through feature decomposition and mapping them to torque optimization parameters, and establishing a noise reduction control table; performing spectral analysis on the enhanced control domain to identify high-energy frequency bands, performing frequency shifting operations to generate safe frequency band signals, and combining the noise reduction control table to form a multi-frequency band management strategy; decomposing the management strategy into parallel control flows and arranging them in a time-interleaved manner to generate an energy-saving control matrix; generating PWM control signals based on the matrix and performing edge reconstruction, obtaining softened edge waveforms through anti-resonance injection technology, and completing the intelligent control of the motor driver.

[0004] However, the aforementioned technologies lack a forward-looking perception of the physical nature of the load, and their processing is based on electrical and acoustic signals after the disturbance occurs, which is a form of delayed compensation. In scenarios such as high-speed machining, transient load disturbances caused by sudden changes in workpiece geometry or transmission characteristics cannot be predicted and actively suppressed in advance by this solution, resulting in dynamic errors such as trajectory overshoot and resonance that are difficult to eliminate, thus limiting control accuracy and stability. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a visual-controlled fusion electrical drive method and system executed by an industrial intelligent gateway. It employs a cross-physical field mapping and forward-looking compensation technology that integrates visual data from the work scene with drive current signals. This technology can reshape the drive current waveform before disturbances occur, thereby proactively suppressing load disturbances and improving the control performance of the drive system.

[0006] The above objectives can be achieved through the following approach: A vision-control fusion electrical drive method and system executed by an industrial intelligent gateway includes: acquiring a visual data stream of a working scene collected by an industrial vision sensing device and a real-time drive current signal stream associated with an electrical driver; extracting spatial structural features from the visual data stream of the working scene to generate a workpiece geometric description tensor; converting the real-time drive current signal stream into a drive current time-frequency spectrum and performing noise energy analysis on the drive current time-frequency spectrum to generate a current noise energy fluctuation spectrum; inputting the workpiece geometric description tensor and the current noise energy fluctuation spectrum into a preset cross-physics field interference mapping model, performing interference mapping, and generating a current waveform modulation strategy; using the current waveform modulation strategy to reshape the waveform of the basic current command generated in the current control cycle to obtain a fused drive current signal, and sending the fused drive current signal to the electrical driver.

[0007] Optionally, generating the workpiece geometric description tensor includes: identifying the workpiece surface morphology change area and its edge contour features associated with the current motion axis of the electric drive from the visual data stream of the work scene, and obtaining the workpiece contour change matrix; encoding the shape and position change gradient of the workpiece contour change matrix along the motion direction of the drive to generate a shape and position change gradient vector; obtaining the texture inertia mapping coefficient that characterizes the influence of workpiece material texture on the load, and fusing the shape and position change gradient vector with the texture inertia mapping coefficient to generate the workpiece geometric description tensor.

[0008] Optionally, obtaining the texture inertia mapping coefficient that characterizes the influence of workpiece material texture on load includes: obtaining historical driving current signals corresponding to the surface morphology change area of ​​the workpiece, constructing a texture inertia association dataset; performing regression learning on the texture inertia association dataset to obtain the texture inertia mapping coefficient, wherein the texture inertia mapping coefficient characterizes the mapping relationship between the surface texture features of the workpiece and the equivalent load inertia fluctuation of the driver.

[0009] Optionally, generating the current noise energy fluctuation spectrum includes: performing a sliding window time-frequency transformation on the real-time drive current signal stream to obtain a drive current time-frequency spectrum containing fundamental frequency energy and disturbance frequency band energy; separating the disturbance frequency band energy unrelated to the ideal fundamental frequency operation of the motor from the drive current time-frequency spectrum to obtain a disturbance noise spectrum; and extracting the envelope of the frequency band energy associated with the main frequency of mechanical fluctuations at the transmission interface from the disturbance noise spectrum as a function of time to generate the current noise energy fluctuation spectrum.

[0010] Optionally, the generated current waveform modulation strategy includes: projecting the workpiece geometric description tensor into the flow direction to generate a load form and position fluctuation prediction curve along the driver's motion direction; performing phase alignment and coupling strength calculation on the load form and position fluctuation prediction curve and the current noise energy fluctuation spectrum to generate an energy geometric coupling coefficient; and driving a modulation parameter generator according to the energy geometric coupling coefficient to output a current waveform modulation strategy including modulation phase, modulation depth, and modulation envelope shape parameters.

[0011] Optionally, the step of performing phase alignment and coupling strength calculation on the load form and position fluctuation prediction curve and the current noise energy fluctuation spectrum includes: inputting the load form and position fluctuation prediction curve and the current noise energy fluctuation spectrum into the complex wavelet domain for complex cross-spectral analysis to calculate the interference coherence; and fusing the interference coherence with the amplitude change rate of the load form and position fluctuation prediction curve to obtain the energy geometric coupling coefficient.

[0012] Optionally, obtaining the fused drive current signal includes: acquiring a basic AC current waveform; generating a modulation disturbance waveform with the same frequency as the basic AC current waveform according to the modulation phase and modulation depth in the current waveform modulation strategy; and combining the modulation disturbance waveform with the basic AC current waveform point by point to generate a fused drive current signal with active harmonic suppression.

[0013] Optionally, the method further includes: acquiring continuous inter-frame difference information in the visual data stream of the work scene to generate a workpiece motion trend vector; projecting the workpiece motion trend vector onto the motion axis of the driver, and combining the energy fluctuation characteristics in the current noise energy fluctuation spectrum to generate an external disturbance and internal state cross-verification signal; using the cross-verification signal as a gating parameter to dynamically weight the current noise energy fluctuation spectrum to obtain a new current noise energy fluctuation spectrum with load correlation enhancement.

[0014] Optionally, the method further includes: obtaining the actual operating current signal fed back by the electrical driver after execution based on the fused driving current signal; comparing the actual disturbance fluctuation spectrum contained in the actual operating current signal with the expected disturbance fluctuation spectrum generated after delay processing of the workpiece geometric description tensor to generate a mapping residual signal; and using the mapping residual signal to update the parameters of the cross-physical field interference mapping model on which the current waveform modulation strategy is based.

[0015] Based on the same inventive concept, this invention also provides a vision-control fusion electrical drive system executed by an industrial intelligent gateway. The system includes: a vision data acquisition module for acquiring a visual data stream of the working scene collected by an industrial vision sensing device and a real-time drive current signal stream associated with the electrical drive; a geometric feature extraction module for extracting spatial structural features from the visual data stream of the working scene to generate a workpiece geometric description tensor; a noise spectrum analysis module for converting the real-time drive current signal stream into a drive current time-frequency spectrum and performing noise energy analysis on the drive current time-frequency spectrum to generate a current noise energy fluctuation spectrum; an interference mapping modulation module for inputting the workpiece geometric description tensor and the current noise energy fluctuation spectrum into a preset cross-physical field interference mapping model, performing interference mapping, and generating a current waveform modulation strategy; and a current waveform reshaping module for reshaping the waveform of the basic current command generated in the current control cycle using the current waveform modulation strategy to obtain a fused drive current signal, and sending the fused drive current signal to the electrical drive.

[0016] Compared with the prior art, the present invention has the following advantages: This invention utilizes an industrial intelligent gateway to synchronously process visual data streams and current signal streams at the device end, fundamentally avoiding the protocol conversion delays and clock deviations caused by uploading heterogeneous data to a centralized controller. This ensures the microsecond-level synchronization accuracy required for cross-physical field interference mapping. This real-time fusion processing at the edge eliminates the limitations of traditional communication bottlenecks in current waveform reshaping, providing a reliable time reference for forward-looking disturbance suppression.

[0017] This invention treats the workpiece geometric shape change trend implied by visual perception and the transmission friction fluctuation characteristics implicit in the drive current noise as two dynamic representations of the same physical process. It directly performs physical interference at the waveform level through cross-physical field interference mapping, rather than numerical superposition. This synergistic effect evolves visual information from macroscopic command superposition to microscopic current waveform-level intervention, enabling the drive current to anticipate and resist upcoming load changes, producing a smooth driving effect far exceeding that of rigid parameter control and independent feedforward compensation.

[0018] This invention continuously updates the parameters of the interferometric mapping model based on the deviation between the actual current feedback after execution and the visual predicted trend, enabling the system to automatically adapt to mechanical wear, changes in lubrication conditions, or long-term drift of tooling fixtures. This closed-loop adaptive mechanism ensures that the fluctuation suppression effect does not decay during long-term operation, overcoming the inherent defects of single-domain techniques such as fixed notch filters or order tracking compensation that fail as operating conditions deteriorate.

[0019] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims, and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating a visual-control fusion electrical drive method executed by an industrial intelligent gateway according to an embodiment of the present invention.

[0022] Figure 2 This is a diagram showing the changes in the surface morphology of a workpiece and the recognition of its edge contour features, according to an embodiment of the present invention.

[0023] Figure 3 This is a workpiece contour change matrix diagram according to an embodiment of the present invention.

[0024] Figure 4 This is a current noise energy fluctuation spectrum of an embodiment of the present invention.

[0025] Figure 5 This is a schematic diagram of the mapping residual signal and model parameter update in an embodiment of the present invention.

[0026] Figure 6 This is a schematic diagram of the structure of a vision-control fusion electrical drive system executed by an industrial intelligent gateway according to an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0028] Reference Figure 1One embodiment of the present invention proposes a vision-control fusion electrical drive method executed by an industrial intelligent gateway. It adopts a cross-physical field mapping and forward-looking compensation technology that integrates visual data of the working scene and drive current signal. This technology can reshape the drive current waveform before disturbance occurs, thereby actively suppressing load disturbance and improving the control performance of the drive system.

[0029] The method described in this embodiment specifically includes: Acquire visual data streams of the work scene collected by industrial vision sensing devices and real-time drive current signal streams associated with electrical actuators; Spatial structure features are extracted from the visual data stream of the work scene to generate a workpiece geometric description tensor; Optionally, the generated workpiece geometry description tensor includes: Identify the workpiece surface morphology change area and its edge contour features associated with the current motion axis of the electric drive from the visual data stream of the work scene, and obtain the workpiece contour change matrix. The shape and position change gradient of the workpiece contour change matrix along the direction of driver movement is encoded to generate a shape and position change gradient vector. Acquire historical driving current signals corresponding to the surface morphology change areas of the workpiece, and construct a texture inertia association dataset; Regression learning is performed on the texture inertia association dataset to obtain the texture inertia mapping coefficients, wherein the texture inertia mapping coefficients characterize the mapping relationship between the workpiece surface texture features and the equivalent load inertia fluctuation of the driver. The shape and position change gradient vector is fused with the texture inertia mapping coefficient to generate a workpiece geometric description tensor.

[0030] Specifically, the implementation process begins with acquiring basic data. Industrial vision sensors, such as high-frame-rate industrial cameras, deployed in the work area continuously capture images of the operation process, forming a visual data stream of the work scene. This data stream is a temporally continuous sequence of images, recording all visually visible changes in the workpiece during processing or handling. Simultaneously, real-time drive current signals are acquired from the servo controllers or frequency converters controlling the electrical drives. Electrical drives are components that perform physical motion, such as the feed axis motors of machine tools; their real-time drive current signals are direct electrical signals reflecting their load torque and operating status. After acquiring the data stream, the next stage is the extraction of spatial structural features. From the visual data stream of the work scene, image processing algorithms identify areas of surface morphological change in the workpiece. These areas refer to local regions where the workpiece's geometry or surface texture changes due to interaction with the tool or fixture during drive movement. The identification method can be differential calculation between consecutive image frames, highlighting dynamically changing parts. After identifying the region of change, edge detection techniques, such as the Canny operator or the Sobel operator, are used to extract the edge contour features of that region. Figure 2 As shown, the circular shaded area represents the macroscopic contour of the currently processed workpiece, while the rectangular area with a central depression represents the real-time captured area of ​​workpiece surface morphology change. The sharp black lines superimposed on the edges of the change area precisely reflect the extracted workpiece edge contour features. Edge contour features are a series of pixel coordinates describing the geometric boundaries of the workpiece. This set of contour points, which changes over time, is organized into a workpiece contour change matrix. (See diagram...) Figure 3 As shown, the horizontal axis represents the index of spatial contour points extracted along the workpiece edge, and the vertical axis represents the continuously acquired time frame sequence. The grayscale intensity of each point in the spectrum quantitatively maps the deformation position or geometric feature value of that contour point in space. The wavy texture evolution pattern on the spectrum intuitively reflects the macroscopic fluctuations and local geometric abrupt changes of the workpiece contour along the mechanical axis movement direction over time. The workpiece contour change matrix is ​​encoded to quantify the dynamic evolution of its geometric shape. Specifically, the shape and position change gradient of this matrix along the drive movement direction is calculated. The shape and position change gradient describes the rate and trend of change of the workpiece contour in a specific direction. The shape and position change gradient vector for calculating this process is... ,have: ; in, This is the time interval between acquiring two frames of images; Represents the time interval The change in position of a specific point on the workpiece contour in the direction of the driver's movement; This represents the current speed of the electric actuator. By dividing the displacement change by the theoretical displacement within that time period, the shape and position changes are normalized, eliminating the influence of speed changes on gradient calculation. To correlate the surface micro-properties of the workpiece with the driving load, a texture inertia correlation dataset needs to be constructed. Historical driving current signals corresponding to previously identified workpiece surface morphology change areas in time are obtained. Texture features are extracted from the workpiece surface morphology change areas in the image, and the equivalent load inertia fluctuations of the corresponding historical driving current signal segments are calculated. A texture inertia correlation dataset is constructed by pairing a series of texture feature values ​​with the corresponding equivalent load inertia fluctuation values. This dataset is processed to establish a mapping relationship between the two. Texture inertia mapping coefficients are obtained by performing regression learning on the texture inertia correlation dataset, such as using least squares linear regression. The texture inertia mapping coefficient is a key parameter that quantifies the intrinsic relationship between the complexity or roughness of the workpiece surface texture and the load inertia fluctuations sensed by the actuator. This relationship can be expressed as: ; in, The equivalent load inertia fluctuation is calculated by analyzing the fluctuation amplitude of the drive current signal, because the current fluctuation is proportional to the torque fluctuation, and the torque fluctuation is directly related to the change in load inertia. The texture inertia mapping coefficient is designed to characterize the conversion ratio from texture to inertia. It is set based on the statistical results obtained by regression analysis of, for example, 200 sets of industrial sensor measured data collected under different working conditions. This refers to texture feature metrics extracted from images of areas showing surface morphological changes on a workpiece, such as contrast or energy values ​​obtained by calculating the gray-level co-occurrence matrix of a local image region. The previously calculated shape and position change gradient vector is fused with the newly obtained texture inertia mapping coefficients to generate the final output, the workpiece geometric description tensor. The fusion process combines macroscopic geometric change information with the loading effects caused by microscopic surface characteristics. A direct fusion method is to combine the two into a multidimensional vector or tensor. That is, for this workpiece geometric description tensor... ,have: ; The fusion operation here involves juxtaposing the shape and position change gradient vector with the result of multiplying it by the texture inertia mapping coefficient, forming a two-dimensional tensor. The first component of this tensor represents the macroscopic geometric change rate of the workpiece, while the second component predicts the magnitude of the equivalent load inertia fluctuation caused by the surface texture characteristics accompanying this geometric change. In this way, the workpiece geometry description tensor simultaneously encapsulates the potential impact of the workpiece's geometric shape changes and its physical surface properties on the drive system.

[0031] For example, in a metal cutting scenario, an electric actuator controls a cutting tool to turn a workpiece with a rough surface along the X-axis. An industrial vision sensor acquires an image frame at time 1.0s. The location of a key point on the workpiece contour was identified as (50.0, 20.0), in millimeters (mm). An image frame was acquired at time 1.1s. At this point, the tool advances, and the newly generated surface position corresponding to the same key point becomes (51.5, 20.0). The movement speed of the electrical actuator is set to... Calculate the gradient vector of geometrical changes. Time interval. The key point is the change in position along the direction of the driver's movement (X-axis). Calculate the shape and position change gradient vector according to the formula. The shape and position change gradient vector here is a scalar of 1.0, indicating that the actual shape and position change rate is consistent with the expected motion rate. Simultaneously, based on historical data, a texture inertia association dataset has been constructed, and texture inertia mapping coefficients have been obtained through regression learning. Assuming based on 200 sets of industrial sensor measurement data, regression analysis yields the texture inertia mapping coefficients for the current processing materials and processes. The physical meaning of this coefficient is that each unit of dimensionless geometrical gradient will trigger... The equivalent load inertia fluctuation is calculated. The shape and position change gradient vector is fused with the texture inertia mapping coefficient to generate a workpiece geometric description tensor. The data of this tensor has been effectively validated; its first element reflects the rate of change of the workpiece's macroscopic geometry, while the second element quantitatively predicts the equivalent load inertia fluctuation value that will accompany this geometric change. This tensor comprehensively describes the geometric and physical composite characteristics of the current work step, providing sufficient and publicly available evidence for subsequent control strategy generation.

[0032] The real-time driving current signal stream is converted into a driving current time-frequency spectrum, and noise energy is analyzed on the driving current time-frequency spectrum to generate a current noise energy fluctuation spectrum. Optionally, the generated current noise energy fluctuation spectrum includes: Perform a sliding window time-frequency transformation on the real-time drive current signal stream to obtain a drive current time-frequency spectrum that includes fundamental frequency energy and disturbance frequency band energy; The disturbance frequency band energy that is unrelated to the ideal fundamental frequency operation of the motor is separated from the time-frequency spectrum of the driving current to obtain the disturbance noise spectrum; Extract the envelope of the frequency band energy that is associated with the main frequency of mechanical fluctuations at the transmission interface from the disturbance noise spectrum and generate the current noise energy fluctuation spectrum.

[0033] Specifically, the implementation process begins with time-frequency analysis of the real-time drive current signal flow. A sliding window time-frequency transform technique is used to convert the one-dimensional time-series current signal into a two-dimensional drive current time-frequency spectrum. The sliding window time-frequency transform is an analytical method that can simultaneously display the frequency components of a signal and their evolution over time, such as the short-time Fourier transform. By sliding a window function of a specific length along the time axis, a Fourier transform is performed on the signal segment within the window, thus obtaining the frequency distribution at each time point. The generated drive current time-frequency spectrum is an information graph with time as one axis, frequency as another, and energy or amplitude as the third dimension. This spectrum contains two key pieces of information: fundamental frequency energy and disturbance frequency band energy. Fundamental frequency energy corresponds to the main operating frequency components of the electrical drive under ideal operating conditions; the energy is concentrated and stable. Disturbance frequency band energy refers to energy distributed at frequencies other than the fundamental frequency, usually caused by non-ideal factors such as load changes, mechanical vibration, and transmission clearance. The generation of the drive current time-frequency spectrum can be described by the following formula: ; in, The time-frequency spectrum of the driving current is represented by a complex function, the square of which is... This represents the energy density at time point t and frequency f; It is the input real-time drive current signal flow, which changes with time. The changing current value; It is a window function, such as the Hanning window, which truncates the signal on the time axis with t as the center. Its length determines the time resolution and frequency resolution of the analysis; j is the imaginary unit. This is the core part of the Fourier transform, used to decompose a time-domain signal into the frequency domain. It's necessary to separate disturbance information unrelated to the ideal operating state of the motor from the time-frequency spectrum of the drive current. By filtering out the fundamental frequency energy, the disturbance noise spectrum can be obtained. Specifically, the ideal fundamental frequency of the electrical drive is determined, and an isolation band is set around this frequency. The energy within this band is considered fundamental frequency energy and is then eliminated. The ideal fundamental frequency is usually determined by the drive's power supply frequency or control commands and is a known parameter. This is crucial for calculating the disturbance noise spectrum. ,have: ; in, It is the ideal fundamental frequency of the motor; This is half the frequency bandwidth set to filter out the fundamental frequency energy. Its value is determined based on the stability of the driver frequency and the measurement accuracy; for example, based on the analysis of 200 sets of measured data from industrial sensors, it is set to 5% of the fundamental frequency. The final step is to focus on a specific disturbance frequency band and extract its energy variation trend over time. Gear meshing and bearing rolling in a transmission system will generate mechanical fluctuations at specific frequencies, which are called the dominant frequency of mechanical fluctuations at the transmission interface. This dominant frequency can be calculated based on mechanical design parameters such as the number of gear teeth and rotational speed. By integrating the energy within a narrow band around this dominant frequency in the disturbance noise spectrum, the envelope of the energy of this specific mechanical disturbance source over time can be obtained. This envelope is the current noise energy fluctuation spectrum, a one-dimensional signal that varies with time t, reflecting the fluctuation of the specific mechanical disturbance energy. The calculation of this current noise energy fluctuation spectrum... ,have: ; in, It is the dominant frequency of mechanical fluctuations at the transmission interface, for example, determined by the meshing frequency of the gearbox; It is half of the integral bandwidth around the main frequency, and its setting is based on the need to fully capture the peak value of the main frequency and its slight frequency shift caused by load changes. It is usually set to cover 95% of the peak energy of the main frequency. Figure 4 As shown, the horizontal axis represents the continuous operation time, and the vertical axis represents the intensity of microscopic disturbance energy extracted by integration. This time-series curve purely filters out the ideal operating fundamental frequency of 50Hz, smoothly and sensitively outlining the main frequency of mechanical fluctuations at the transmission interface, such as the dynamic envelope of energy evolution over time at 180Hz. Each peak on the curve accurately maps the energy accumulation process of electrical noise caused by sudden changes in mechanical impact or friction within the transmission.

[0034] For example, an electrical drive rotates the spindle, and its real-time drive current signal is acquired for analysis. Assuming the drive is powered by 50 Hz AC, its ideal fundamental frequency is... Meanwhile, the dominant frequency of the mechanical fluctuations generated by gear meshing in a reduction gearbox within the transmission system at the current speed is calculated to be... A sliding window time-frequency transformation is performed on a segment of the acquired real-time drive current signal to obtain the drive current time-frequency spectrum. On the spectrum, a very bright and continuous energy band at 50Hz can be observed, representing the fundamental frequency energy. Near f=180Hz, there exists a darker energy band whose intensity fluctuates over time; this belongs to the disturbance frequency band. To separate the disturbance information, a fundamental frequency filtering bandwidth is set. Assuming based on driver performance, the bandwidth is set... By applying the separation formula, the energy values ​​of the driving current time-frequency spectrum X(t,f) within the frequency range [47.5Hz, 52.5Hz] are all set to zero, thus obtaining the disturbance noise spectrum. At this point, the strong energy band at 50Hz is removed, leaving only the disturbance energy information around 180Hz. The integration bandwidth is then set. To fully capture the energy of the 180Hz main frequency, energy integration is performed on the disturbance noise spectrum, calculating the total energy from 175Hz to 185Hz at each time point t. At time point t_1 = 2.0s, the calculated integrated energy value is... As the cutting load increases, at time point 2.5s, the gear meshing impact intensifies, and the calculated integral energy value increases accordingly. Connecting the values ​​calculated at all time points forms a continuously changing curve, namely the current noise energy fluctuation spectrum. The data from this curve effectively verifies the technical characteristics set, quantifying the dynamic process of the disturbance energy caused by the specific mechanical factor of gear meshing over time, and providing key internal state observation signals for subsequent cross-physics field analysis.

[0035] The workpiece geometric description tensor and the current noise energy fluctuation spectrum are input into a preset cross-physics field interference mapping model, and interference mapping is performed to generate a current waveform modulation strategy. Optionally, the generated current waveform modulation strategy includes: The workpiece geometric description tensor is projected in the flow direction to generate a load form and position fluctuation prediction curve along the direction of driver movement. The load shape and position fluctuation prediction curve and the current noise energy fluctuation spectrum are input into the complex wavelet domain for complex cross spectrum analysis to calculate the interference coherence. The energy-geometric coupling coefficient is obtained by fusing the interference coherence degree with the amplitude change rate of the load form and position fluctuation prediction curve. The modulation parameter generator is driven by the energy geometric coupling coefficient to output a current waveform modulation strategy that includes modulation phase, modulation depth, and modulation envelope shape parameters.

[0036] Specifically, the objective is to establish a cross-physics field interferometric mapping model that correlates future load changes predicted from visual information with the current internal state of the system perceived from electrical signals, thereby generating a forward-looking current waveform modulation strategy. The cross-physics field interferometric mapping model consists of four functional levels: flow direction projection, complex wavelet cross-spectral analysis, coupling coefficient fusion, and modulation parameter generation, connected by deterministic mathematical formulas. Initial parameters in the model include texture inertia mapping coefficients, calibration coefficients, and load energy conversion coefficients, obtained through offline calibration experiments: before deployment, various standard specimens with known geometric features are run, visual data and current signals are collected simultaneously, accumulating no fewer than 200 sets of calibration data pairs, and the initial values ​​of each parameter are determined using least squares regression and ratios. After offline calibration, the model is put into online operation, and the parameters are continuously updated online adaptively through a mapping residual signal feedback mechanism. This process begins with processing the generated workpiece geometric description tensor. Flow direction projection is performed on the workpiece geometric description tensor to generate a load form and position fluctuation prediction curve along the actuator's motion direction. The workpiece geometry description tensor contains the gradient of form and position changes and the associated inertia fluctuation prediction. The flow projection operation essentially extracts and expands the time-dependent prediction components of this tensor, forming a time-series curve. This curve, the load form and position fluctuation prediction curve, depicts the prediction of the equivalent load inertia fluctuation generated by the actuator due to changes in workpiece geometry and surface texture over a short future period. For this load form and position fluctuation prediction curve... ,have: ; in, This is the gradient vector of form and position changes over time. This load form and position fluctuation prediction curve, representing "external prediction," is deeply correlated with the current noise energy fluctuation spectrum, generated step-by-step, representing "internal perception." The load form and position fluctuation prediction curve and the current noise energy fluctuation spectrum are input into the complex wavelet domain for complex cross-spectral analysis. Complex wavelet analysis is a powerful signal processing tool that can not only analyze the energy of a signal at different frequencies but also provide phase information, which is crucial for understanding the lead or lag relationship between two signals. By calculating the interferometric coherence, the extent to which two signals from different physical domains are synchronized or have a fixed phase difference can be quantified. The interferometric coherence is a complex number, where the magnitude represents the strength of the correlation and the argument represents the phase difference. For the interferometric coherence calculated at frequency f... ,have: ; in, The cross power spectral density of the load form and position fluctuation prediction curve and the current noise energy fluctuation spectrum; and These are the self-power spectral density of the load form-position fluctuation prediction curve and the current noise energy fluctuation spectrum, respectively. Calculations typically focus on the vicinity of the dominant frequency of mechanical fluctuations at the transmission interface, as this is the most sensitive reflection of the internal state. After obtaining the interference coherence, it needs to be combined with the urgency of the load change. The energy geometric coupling coefficient is obtained by fusing the interference coherence with the amplitude change rate of the load form-position fluctuation prediction curve. The amplitude change rate of the load form-position fluctuation prediction curve represents how drastic the predicted load change is. Fusing these two yields a comprehensive index that reflects both the reliability of the prediction (high coherence) and the severity of the event (high change rate). The calculation of this energy geometric coupling coefficient... ,have: ; in, The magnitude of the interferometric coherence at the dominant frequency represents the coupling strength between prediction and perception. The derivative of the load form-position fluctuation prediction curve with respect to time is the rate of change of amplitude. A modulation parameter generator is driven by the energy geometric coupling coefficient, outputting a current waveform modulation strategy containing modulation phase, modulation depth, and modulation envelope shape parameters. The modulation parameter generator is a set of preset mapping rules or functions that converts the dynamically calculated energy geometric coupling coefficient into specific control parameters. The modulation phase determines whether the compensation effect is advanced or delayed; it is mainly determined by the phase angle of the interference coherence. The modulation depth determines the strength of the compensation effect; it is positively correlated with the magnitude of the energy geometric coupling coefficient. The modulation envelope shape parameter determines the specific form of the compensation effect over time, such as whether it is abrupt or gradual, which is usually related to the local shape of the load form-position fluctuation prediction curve. For this process, the modulation phase... Modulation depth and modulation envelope shape parameters ,have: ; in, The calibration coefficient is used to convert the units of the energy geometric coupling coefficient into a dimensionless depth value. It is set based on experimental data of the system response to ensure that the modulation depth is within a reasonable range, such as 0 to 1. and They are respectively The first and second derivatives; This serves as a normalized reference for the load change rate, which is a pre-defined reference for normal operation under typical conditions. The typical maximum value or statistical upper limit; This serves as a normalized reference for the acceleration due to load changes, which is the system's performance under typical operating conditions. The typical maximum value or statistical upper limit; The weighting coefficient for the rate of change is recommended to be between 0.2 and 0.8. For the weighting coefficient of changing acceleration, a value range of 0.05 to 0.3 is recommended. The recommended value for the base envelope width is 0.8 to 2.0, with a typical default value of 1.0.

[0037] For example, the workpiece geometric description tensor at a certain moment is [1.0, 0.0005]. The resulting current noise energy fluctuation spectrum shows fluctuations within the corresponding time period. By projecting the workpiece geometric description tensor directionally, a point of the load form and position fluctuation prediction curve at that moment is obtained. Suppose that through continuous visual acquisition and processing, a prediction curve is obtained, and its amplitude change rate at the current moment is... The load form and position fluctuation prediction curve is compared with the current noise energy fluctuation spectrum extracted from the current signal using complex cross-spectral analysis. The analysis assumes that the dominant frequency of mechanical fluctuations at the transmission interface... At this point, the calculated interference coherence between the two is as follows: This result indicates a strong correlation (0.85) between the predicted load fluctuations and the actual internal vibrations of the system, and that the response of the internal vibrations, i.e., current noise, lags behind the visually predicted geometric events. Phase of radians. The energy-geometric coupling coefficient is calculated by fusing interference coherence with the load change rate. This value is a quantified coupling strength indicator, and the data effectively verifies the technical characteristics set. A current waveform modulation strategy is generated based on this energy geometric coupling coefficient. Assume the calibration coefficient is experimentally set to... Calculate the modulation phase: ; Calculate the modulation depth: The modulation envelope shape parameter is 2.0. The final generated current waveform modulation strategy is: modulation phase. The modulation depth is 0.1275, and the modulation envelope shape parameter is 2.0. This set of parameters will guide the next step of current waveform reshaping to actively counteract upcoming disturbances caused by changes in workpiece geometry.

[0038] The waveform of the basic current command generated in the current control cycle is reshaped using the current waveform modulation strategy to obtain a fused drive current signal, which is then sent to the electrical driver.

[0039] Optionally, obtaining the fused drive current signal includes: Obtain the basic AC current waveform, and generate a modulation disturbance waveform with the same frequency as the basic AC current waveform according to the modulation phase and modulation depth in the current waveform modulation strategy. The modulated disturbance waveform is combined point by point with the basic AC current waveform to generate a fused drive current signal with active harmonic countermeasure.

[0040] Specifically, the implementation process begins with acquiring the baseline AC current waveform generated by the drive system within the current control cycle. The baseline AC current waveform is a standard sinusoidal current signal calculated by the drive controller based on the target speed or position command. It represents the ideal current required by the drive motor without considering any external disturbances or system nonlinearities, and its mathematical expression is a standard sine function. For the baseline AC current waveform varying with time t... ,have: ; in, The peak amplitude of the current is determined by the required output torque of the current motor. Based on the current waveform modulation strategy, a modulated disturbance waveform for superposition is generated. This modulated disturbance waveform is the core of active compensation; its frequency is consistent with the basic AC current waveform (i.e., same frequency), but its amplitude and phase are precisely modulated. Modulation phase and modulation depth are key inputs. The modulation phase determines whether the disturbance waveform leads or lags the basic waveform, precisely aligning the timing of the compensation torque. The modulation depth determines the amplitude of the disturbance waveform, i.e., the intensity of the compensation. Simultaneously, the modulation envelope shape parameter controls the intensity change profile of this disturbance waveform over time, allowing for smooth intervention and withdrawal, avoiding the introduction of new shocks. For the calculated modulated disturbance waveform... ,have: ; in, The dimensionless envelope function, determined by the shape parameters of the modulation envelope and varying over time (between 0 and 1), is used to control the onset, termination, and strength of modulation. Specifically, it can have... , The predicted disturbance occurrence time is derived from the peak position detection of the load form and position fluctuation prediction curve, i.e., the moment when the visually predicted equivalent load inertia fluctuation reaches its maximum value. The generated modulated disturbance waveform is then composited point-by-point with the original basic AC current waveform to generate the final fused drive current signal. Point-by-point composite is a direct signal superposition operation, adding the instantaneous values ​​of the two waveforms at each time sampling point. The resulting fused drive current signal retains the basic energy required for normal operation of the drive motor while adding precisely designed compensation components for actively combating harmonics and load disturbances. The final generated fused drive current signal... ,have: ; This signal is then sent to the power amplifier module of the electrical driver to drive the motor.

[0041] For example, a current waveform modulation strategy has been generated, with the following parameters: modulation phase. The modulation depth is 0.1275, and the modulation envelope shape parameter is 2.0. In the current control cycle, the peak value of the basic AC current waveform corresponding to the torque required for the electrical driver to output is... The ideal fundamental frequency is Obtain the basic AC current waveform. According to the formula, the waveform is: The modulation disturbance waveform is generated based on the current waveform modulation strategy. The modulation envelope function is a Gaussian function centered at the predicted disturbance occurrence time of 3.0s. Therefore, the modulation disturbance waveform is: The two signals are then combined point by point to obtain the fused drive current signal. At the moment the disturbance is about to occur, for example, t=3.0s, the value of the fused drive current signal at that moment is calculated. At t=3.0s, the value of the basic AC current waveform is: The value of the modulated perturbation waveform is 1, at which point the envelope function value is 1. Therefore, the value of the fusion drive current signal at this moment is: The data from this calculation process effectively verifies the setting of the technical features. Without the method of this invention, the controller would output a current command of 0A at t=3.0s. However, through the waveform reshaping of this invention, the actual output fused drive current signal is 0.9015A. This additional injected current component with specific phase and amplitude will generate a precise electromagnetic torque in the motor to actively counteract the load disturbance caused by workpiece geometric changes, which reaches its peak at t=3.0s, thereby achieving proactive maintenance of the system's operational stability.

[0042] Optionally, the method further includes: Obtain continuous inter-frame difference information from the visual data stream of the work scene to generate a workpiece motion trend vector; The workpiece motion trend vector is projected onto the motion axis of the driver, and combined with the energy fluctuation characteristics in the current noise energy fluctuation spectrum, an external disturbance and internal state cross-verification signal is generated. Using the cross-validation signal as a gating parameter, the current noise energy fluctuation spectrum is dynamically weighted to obtain a new current noise energy fluctuation spectrum with load-related enhancement.

[0043] Specifically, independent visual information is used to confirm that the noise energy extracted from the current signal is indeed related to the physical motion of the workpiece, thereby filtering out or suppressing artifacts caused by other irrelevant factors such as electrical noise and unrelated mechanical vibrations, resulting in a signal more strongly correlated with load changes. The implementation process begins with further analysis of the visual data stream of the work scene. A workpiece motion trend vector is generated by acquiring continuous inter-frame difference information from the visual data stream of the work scene. Continuous inter-frame difference information is obtained by calculating the difference in brightness or color values ​​of corresponding pixels between two temporally adjacent image frames. Regions with significant differences indicate motion or change. The workpiece motion trend vector is a quantitative description of the overall motion direction and velocity of these changed regions, for example, by calculating the centroid displacement of the highlighted areas in the difference image. The workpiece motion trend vector calculated at time point t is... ,have: ; in, and These are the centroid coordinate vectors of the motion regions identified by inter-frame difference in the current and previous frames, respectively. After obtaining the workpiece motion trend vector representing the visually observed motion, it needs to be correlated with the motion of the electrical actuator. By projecting the workpiece motion trend vector onto the actuator's motion axis, the motion component along the actuator's direction of action, as revealed by the visual information, can be obtained. This projection operation requires prior knowledge of the calibration parameters of the industrial vision sensor and the direction vector of the actuator's motion axis in the camera coordinate system. This projected component is combined with the energy fluctuation characteristics in the obtained current noise energy fluctuation spectrum to generate an external disturbance and internal state cross-validation signal. Here, the external disturbance refers to the visually captured workpiece motion, while the internal state is reflected by the current noise energy fluctuation spectrum. The combination aims to confirm whether the two occur synchronously. A high cross-validation signal means that when the vision system observes the workpiece motion, a corresponding energy fluctuation is also detected in the current signal, which enhances the confidence that the energy fluctuation is caused by the workpiece load. For the generated external disturbance and internal state cross-validation signal... ,have: ; in, The workpiece motion trend vector is the unit direction vector along the drive's motion axis. The projection on the screen represents the visual motion speed along the drive axis; This is a normalization constant with the same unit as the numerator. It is set based on statistical analysis of historical data, taking the typical maximum value of the numerator product under normal operating conditions, for example, based on the analysis of 200 sets of measured data from industrial sensors. Finally, this cross-validation signal is used as a gating parameter to dynamically weight the original current noise energy fluctuation spectrum. The gating parameter acts like an adjustable valve; its value determines the proportion of the original signal that can pass through. Through this operation, a new load-correlated enhanced current noise energy fluctuation spectrum can be obtained. The final calculated new load-correlated enhanced current noise energy fluctuation spectrum is then used. ,have: ; when When the value is close to 1, it indicates that visual motion and current noise are highly synchronized, and the original energy spectrum value is basically preserved; when When the value is close to 0, it indicates that the two are out of sync, and the original energy spectrum value is suppressed because it is likely not caused by the workpiece load.

[0044] For example, in a metal cutting scenario, assuming at time t=4.5s, it is necessary to verify the current noise energy fluctuation spectrum. By comparing two frames at t=4.5s and t=4.4s, it is found that the centroid of the workpiece cutting area has moved from pixel coordinates (250, 180) to (254, 181). Therefore, the workpiece motion trend vector can be calculated: Project this vector onto the driver's motion axis. The unit direction vector of the driver's motion axis (X-axis) in the camera coordinate system is (1,0). Then... The 0.8 cross-validation signal value is a dimensionless gating parameter, indicating a strong correlation between the current noise and visually observed motion. Using this gating parameter to dynamically weight the original energy spectrum yields a new current noise energy fluctuation spectrum with enhanced load correlation: The calculation results show that the original energy spectrum value of 0.12 was reduced to 0.096. This new value, after verification and weighting, more accurately reflects the true disturbance energy caused by the workpiece load. This data effectively verifies the setting of the technical characteristics. This new load-correlated enhanced current noise energy fluctuation spectrum will replace the original spectrum and be fed into the above steps in subsequent control cycles to generate a more accurate current waveform modulation strategy.

[0045] Optionally, the method further includes: Obtain the actual operating current signal fed back by the electrical driver after it executes according to the fused drive current signal; The actual disturbance fluctuation spectrum contained in the actual operating current signal is compared with the expected disturbance fluctuation spectrum generated after the workpiece geometric description tensor is delayed to generate a mapping residual signal. The parameters of the cross-physical field interference mapping model on which the current waveform modulation strategy is based are updated using the mapped residual signal.

[0046] Specifically, the first step is to obtain the system's true response. After the electrical driver executes one or more control cycles based on the generated fused drive current signal, the actual operating current signal is obtained from the driver's current sensor. This signal represents the true electrical behavior of the system after interacting with the actual load under active compensation. After obtaining the actual operating current signal, information reflecting the actual disturbance needs to be extracted. Using the same signal processing procedure as described above, this actual operating current signal is analyzed, converted, and parsed to obtain the actual disturbance fluctuation spectrum. This actual disturbance fluctuation spectrum quantifies the residual energy fluctuations near the dominant frequency of the mechanical fluctuations at the transmission interface, even after compensation is applied. Simultaneously, a benchmark for comparison, the expected disturbance fluctuation spectrum, needs to be generated. This spectrum is constructed based on the generated workpiece geometry description tensor. Specifically, the expected disturbance fluctuation spectrum is generated by delaying the workpiece geometry description tensor. Delaying is crucial because the workpiece geometry description tensor is based on visual advance prediction, while the actual operating current signal is a feedback after the event occurs. The two must be aligned in time. The expected disturbance wave spectrum represents the theoretical magnitude of the disturbance energy caused solely by changes in workpiece geometry without any compensation. Its generation can be based on a calibrated physical model: ; in, The load energy conversion coefficient characterizes how much energy a unit load inertia fluctuation will generate in the current spectrum. It is set based on offline system identification or statistical analysis of a large amount of historical data, such as calibration based on 200 sets of industrial sensor measured data. This represents the total system delay time from visual prediction to current feedback. For generating, corresponding to a certain moment in the past The load form and position fluctuation prediction curve value. With the actual and expected values, a comparison can be made to generate a mapped residual signal. The mapped residual signal is the difference between the actual disturbance fluctuation spectrum and the expected disturbance fluctuation spectrum; it directly reflects the error of the entire prediction compensation chain. For calculating this mapped residual signal... ,have: ; in, This represents the actual perturbation fluctuation spectrum at time point t. A positive residual indicates that the actual perturbation is larger than expected, suggesting insufficient compensation or an underestimation; a negative residual indicates overcompensation or an overestimation. The final step is to use this error signal to correct the model. The parameters of the cross-physics interference mapping model upon which the current waveform modulation strategy is based are updated using the mapped residual signal. One of the core parameters of this model, the obtained texture inertia mapping coefficient, is a crucial bridge connecting visual features and physical loads. Therefore, an effective update strategy is to adjust the value of the texture inertia mapping coefficient based on the residual. The updated texture inertia mapping coefficient is calculated... ,have: ; in, The value before the update; The learning rate is a dimensionless positive number used to control the update step size. Its setting is based on empirical tuning, requiring a balance between convergence speed and stability. The logic of this update formula is that if the residual is positive, meaning the actual perturbation is large, the texture inertia mapping coefficient is increased, causing a larger load to be predicted for the same visual features in the future, thus enhancing compensation; conversely, the opposite is also true. Figure 5 As shown, sub-figure (a) is the mapping residual signal generation diagram, where the solid line represents the actual disturbance fluctuation spectrum fed back by the current sensor after execution, and the dashed line represents the theoretical expected disturbance fluctuation spectrum generated after the workpiece geometric description tensor is processed by the delay time; the two curves are aligned and compared on the time axis, and their difference is the mapping residual signal shown by the red dot dashed line; sub-figure (b) shows the parameter online update trajectory. When a positive residual is generated above, that is, when the actual disturbance is greater than expected, it indicates that the current model's prediction of load fluctuation is too low and the compensation is insufficient. Online update is initiated, driving the correction of the key model parameter - texture inertia mapping coefficient, and finally converges and remains stable.

[0047] An exemplary, continuing metal cutting scenario. Assume an adaptive update calculation is completed at time t=6.0s. Obtain the actual operating current signal fed back by the electrical driver at t=6.0s. Through analysis, the actual disturbance fluctuation spectrum value is extracted from it. Generate the expected perturbation fluctuation spectrum for comparison. Total system delay. We need to find the prediction made at t=5.8s for t=6.0s. The load form and position fluctuation prediction curve value corresponding to the workpiece geometric description tensor generated at that time was... The load energy conversion factor has been calibrated to be... ,but Calculate the mapped residual signal: The positive residual indicates that the actual residual perturbation is larger than the theoretical value based on visual prediction, suggesting inadequacies in previous predictions and compensation. This residual is used to update the parameters of the cross-physics interference mapping model. The texture inertia mapping coefficients are also updated. Assume the coefficients before the update are... The update result from the previous step, with the learning rate set to [value missing]. .but With this update, the texture inertia mapping coefficient increased from 0.000536 to 0.00054195. This data effectively validates the technical feature settings, indicating that the model self-adjusted based on the current underestimation, improving its sensitivity to similar visual features. In future operations, when encountering similar workpiece geometric changes again, the system will predict higher load disturbances, thereby generating stronger compensation signals to suppress actual disturbances to a lower level, achieving online adaptive optimization of the model.

[0048] Based on the same inventive concept, such as Figure 6 As shown, the present invention also provides a vision-control fusion electrical drive system executed by an industrial intelligent gateway, the system comprising: The visual data acquisition module is used to acquire the visual data stream of the working scene collected by the industrial vision sensing device and the real-time drive current signal stream associated with the electrical driver; The geometric feature extraction module is used to extract spatial structural features from the visual data stream of the work scene and generate a workpiece geometric description tensor. The noise spectrum analysis module is used to convert the real-time drive current signal stream into a drive current time-frequency spectrum, and to perform noise energy analysis on the drive current time-frequency spectrum to generate a current noise energy fluctuation spectrum. The interference mapping modulation module is used to input the workpiece geometric description tensor and the current noise energy fluctuation spectrum into a preset cross-physics field interference mapping model, perform interference mapping, and generate a current waveform modulation strategy. The current waveform reshaping module is used to reshape the waveform of the basic current command generated in the current control cycle using the current waveform modulation strategy to obtain a fused drive current signal, and then send the fused drive current signal to the electrical driver.

[0049] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0050] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A visual-control fusion electrical drive method executed by an industrial intelligent gateway, characterized in that, The method includes: Acquire visual data streams of the work scene collected by industrial vision sensing devices and real-time drive current signal streams associated with electrical actuators; Spatial structure features are extracted from the visual data stream of the work scene to generate a workpiece geometric description tensor; The real-time driving current signal stream is converted into a driving current time-frequency spectrum, and noise energy is analyzed on the driving current time-frequency spectrum to generate a current noise energy fluctuation spectrum. The workpiece geometric description tensor and the current noise energy fluctuation spectrum are input into a preset cross-physics field interference mapping model, and interference mapping is performed to generate a current waveform modulation strategy. The waveform of the basic current command generated in the current control cycle is reshaped using the current waveform modulation strategy to obtain a fused drive current signal, which is then sent to the electrical driver.

2. The visual-control fusion electrical drive method executed by an industrial intelligent gateway according to claim 1, characterized in that, The generated workpiece geometric description tensor includes: Identify the workpiece surface morphology change area and its edge contour features associated with the current motion axis of the electric drive from the visual data stream of the work scene, and obtain the workpiece contour change matrix. The shape and position change gradient of the workpiece contour change matrix along the direction of driver movement is encoded to generate a shape and position change gradient vector. Obtain the texture inertia mapping coefficients that characterize the influence of workpiece material texture on load, and fuse the shape and position change gradient vector with the texture inertia mapping coefficients to generate a workpiece geometric description tensor.

3. The visual-control fusion electrical drive method executed by an industrial intelligent gateway according to claim 2, characterized in that, The process of obtaining the texture inertia mapping coefficients that characterize the influence of workpiece material texture on load includes: Acquire historical driving current signals corresponding to the surface morphology change areas of the workpiece, and construct a texture inertia association dataset; Regression learning is performed on the texture inertia association dataset to obtain the texture inertia mapping coefficients, wherein the texture inertia mapping coefficients characterize the mapping relationship between the workpiece surface texture features and the equivalent load inertia fluctuation of the driver.

4. The visual-control fusion electrical drive method executed by an industrial intelligent gateway according to claim 1, characterized in that, The generated current noise energy fluctuation spectrum includes: Perform a sliding window time-frequency transformation on the real-time drive current signal stream to obtain a drive current time-frequency spectrum that includes fundamental frequency energy and disturbance frequency band energy; The disturbance frequency band energy that is unrelated to the ideal fundamental frequency operation of the motor is separated from the time-frequency spectrum of the driving current to obtain the disturbance noise spectrum; Extract the envelope of the frequency band energy that is associated with the main frequency of mechanical fluctuations at the transmission interface from the disturbance noise spectrum and generate the current noise energy fluctuation spectrum.

5. The visual-control fusion electrical drive method executed by an industrial intelligent gateway according to claim 1, characterized in that, The generated current waveform modulation strategy includes: The workpiece geometric description tensor is projected in the flow direction to generate a load form and position fluctuation prediction curve along the direction of driver movement. Phase alignment and coupling strength calculations are performed on the load shape and position fluctuation prediction curve and the current noise energy fluctuation spectrum to generate the energy geometric coupling coefficient; The modulation parameter generator is driven by the energy geometric coupling coefficient to output a current waveform modulation strategy that includes modulation phase, modulation depth, and modulation envelope shape parameters.

6. The visual-control fusion electrical drive method executed by an industrial intelligent gateway according to claim 5, characterized in that, The step of performing phase alignment and coupling strength calculation on the load shape and position fluctuation prediction curve and the current noise energy fluctuation spectrum includes: The load shape and position fluctuation prediction curve and the current noise energy fluctuation spectrum are input into the complex wavelet domain for complex cross spectrum analysis to calculate the interference coherence. The energy geometric coupling coefficient is obtained by fusing the interference coherence degree with the amplitude change rate of the load shape and position fluctuation prediction curve.

7. The visual-control fusion electrical drive method executed by an industrial intelligent gateway according to claim 1, characterized in that, The obtained fusion drive current signal includes: Obtain the basic AC current waveform, and generate a modulation disturbance waveform with the same frequency as the basic AC current waveform according to the modulation phase and modulation depth in the current waveform modulation strategy. The modulated disturbance waveform is combined point by point with the basic AC current waveform to generate a fused drive current signal with active harmonic countermeasure.

8. The visual-control fusion electrical drive method executed by an industrial intelligent gateway according to claim 4, characterized in that, The method further includes: Obtain continuous inter-frame difference information from the visual data stream of the work scene to generate a workpiece motion trend vector; The workpiece motion trend vector is projected onto the motion axis of the driver, and combined with the energy fluctuation characteristics in the current noise energy fluctuation spectrum, an external disturbance and internal state cross-verification signal is generated. Using the cross-validation signal as a gating parameter, the current noise energy fluctuation spectrum is dynamically weighted to obtain a new current noise energy fluctuation spectrum with load-related enhancement.

9. The visual-control fusion electrical drive method executed by an industrial intelligent gateway according to claim 1, characterized in that, The method further includes: Obtain the actual operating current signal fed back by the electrical driver after it executes according to the fused drive current signal; The actual disturbance fluctuation spectrum contained in the actual operating current signal is compared with the expected disturbance fluctuation spectrum generated after the workpiece geometric description tensor is delayed to generate a mapping residual signal. The parameters of the cross-physical field interference mapping model on which the current waveform modulation strategy is based are updated using the mapped residual signal.

10. A vision-control fusion electrical drive system executed by an industrial intelligent gateway, applied to a vision-control fusion electrical drive method executed by an industrial intelligent gateway as described in any one of claims 1-9, characterized in that, The system includes: The visual data acquisition module is used to acquire the visual data stream of the working scene collected by the industrial vision sensing device and the real-time drive current signal stream associated with the electrical driver; The geometric feature extraction module is used to extract spatial structural features from the visual data stream of the work scene and generate a workpiece geometric description tensor. The noise spectrum analysis module is used to convert the real-time drive current signal stream into a drive current time-frequency spectrum, and to perform noise energy analysis on the drive current time-frequency spectrum to generate a current noise energy fluctuation spectrum. The interference mapping modulation module is used to input the workpiece geometric description tensor and the current noise energy fluctuation spectrum into a preset cross-physics field interference mapping model, perform interference mapping, and generate a current waveform modulation strategy. The current waveform reshaping module is used to reshape the waveform of the basic current command generated in the current control cycle using the current waveform modulation strategy to obtain a fused drive current signal, and then send the fused drive current signal to the electrical driver.

Citation Information

Patent Citations

  • Intelligent control method and system of motor driver

    CN120768185B