Resonance control method and system for stepping drive motor in high-precision scene

Abstract

The invention relates to the technical field of stepping drive motor control, and discloses a stepping drive motor resonance control method and system for a high-precision scene, and the method comprises the steps: collecting motor operation data through a multi-dimensional detector array, and inputting a dynamic response model to obtain dynamic response features; extracting a current harmonic frequency spectrum key point as a resonance event mark point, aligning and matching a rotor angular displacement and winding temperature data fluctuation interval, and generating a mechanical response association pair; performing resonance event association and space-time positioning by using the association pair, and determining abnormal response distribution; and generating a self-adaptive suppression strategy based on the association pair and the abnormal distribution and triggering an execution device. The system comprises a multi-dimensional detector array, a dynamic response processing module, an event association analysis unit, an abnormal distribution modeling unit and a strategy execution unit. According to the invention, multi-physics field data fusion analysis and resonance precise control are realized, the method is suitable for high-precision scenes, and the motor operation stability and reliability are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of stepper drive motor control, and specifically to a method and system for controlling resonance of a stepper drive motor in high-precision scenarios. Background Art

[0002] Stepper motors are widely used in high-precision applications such as industrial automation, precision manufacturing, and medical equipment due to their precise position and speed control. However, during operation, stepper motors are susceptible to electromagnetic forces, mechanical structure resonance, and other factors, leading to reduced accuracy, increased vibration, and even equipment failure, severely limiting their effectiveness in high-precision applications.

[0003] Traditional stepper motor resonance control methods are primarily based on single-dimensional detection and control, such as simple closed-loop control using only current feedback or vibration sensors. These methods have significant drawbacks: Firstly, single-dimensional data cannot fully reflect the complex operating state of the motor, and key resonance excitation factors are easily missed, resulting in ineffective and ineffective control strategies. Secondly, traditional control strategies typically employ fixed-parameter control modes, which are difficult to adapt to the dynamic characteristics of the motor under varying loads and operating cycles. This can easily lead to control lag or over-adjustment, further exacerbating resonance.

[0004] As the demand for high-precision stepper motor control continues to increase, resonance issues caused by multi-physics coupling are becoming increasingly prominent. During motor operation, complex interactions exist between the current ripple generated by the electromagnetic system, the rotor angular displacement changes in the mechanical system, and the winding temperature rise in the thermal system. Traditional methods lack the ability to jointly analyze and collaboratively process multi-dimensional data, making it difficult to accurately capture the correlation characteristics of resonance events across multiple domains, such as electromagnetic, mechanical, and thermal fields. This makes it difficult to accurately locate and effectively suppress resonance events.

[0005] Furthermore, existing resonance control systems lack real-time performance and adaptability. High-precision scenarios often require control systems that can monitor the motor's operating status in real time and respond quickly. However, traditional systems struggle to meet these requirements due to the complexity of data processing algorithms and limited computing resources. Furthermore, when the motor's operating environment or load parameters change, traditional systems are unable to dynamically adjust control strategies based on real-time data, resulting in poor system robustness. Summary of the Invention

[0006] The object of the present invention is to provide a method and system for controlling resonance of a stepper drive motor in a high-precision scenario, so as to solve the problems raised in the above-mentioned background technology.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for controlling resonance of a stepper drive motor in a high-precision scenario, the method comprising: Step 1: Collecting motor operating status data using a multi-dimensional detector array, including a current ripple sensor, a rotor angular displacement sensor, and a winding temperature sensor. The data collected synchronously by each detector is input into a dynamic response model for parallel processing to obtain dynamic response characteristics corresponding to each detector. The characteristics include current harmonic spectrum, rotor oscillation phase, and thermal accumulation gradient curve. Step 2: Extract key frequency response points from the current harmonic spectrum as resonance event markers. Align and match the data fluctuation intervals of the rotor angular displacement and winding temperature within the same time period to obtain multiple sets of mechanical response association pairs. Each association pair contains the timestamp and amplitude parameters of the same mechanical response event in different detectors. Step 3: Using the mechanical response correlation pairs obtained in step 2, perform resonance event correlation and spatiotemporal location to determine the abnormal response distribution under a preset dynamic benchmark model during the motor operation cycle; Step 4: Based on the mechanical response association pairs obtained in step 2 and the abnormal response distribution determined in step 3, an adaptive suppression strategy is generated and an execution device is triggered.

[0008] Preferably, the resonance event association and spatiotemporal positioning in step 3 specifically include: Step 31, resonance event association and location: Determine whether the resonance event marker is within a preset threshold range of rotor oscillation and thermal gradient fluctuation. If so, associate it with the same mechanical resonance event and calculate the relative position of the event on the motor operation timeline; the operation timeline is based on the motor startup timeline as the origin and is divided into milliseconds. Step 32, spatiotemporal fusion positioning: Extract high-frequency resonance events of the motor within M consecutive operating cycles. Obtain the distribution positions P1 to PM of the M events on the operating time axis at time T in the same manner as in step 31. Select the first event as the reference point, normalize the positions of the other M-1 events relative to the reference point, and generate a dynamic response spectrum with the first event as the origin, the time axis as the horizontal axis, and the event energy as the vertical axis. By back-tracing and correlating historical data, combined with motor load parameters and drive signals, the offset of new events at time T+1 is calculated, and the abnormal response distribution map is dynamically updated.

[0009] Preferably, the adaptive suppression strategy in step 4 includes phase compensation parameter adjustment, drive frequency correction scheme and emergency braking instruction; wherein: Phase compensation parameter adjustment specifically includes: matching the preset damping control curve based on the resonance event energy and rotor oscillation amplitude, generating a compensation parameter set for the closed-loop control topology, and loading it in real time through the drive controller; The drive frequency correction scheme specifically includes: analyzing the timing correlation between the resonance event and the drive pulse width and duty cycle, locating the frequency band of the abnormal response, and generating pulse modulation waveform reconstruction instructions; The emergency braking instruction specifically includes: when the energy of continuous resonance events exceeds the preset safety threshold, the hardware protection circuit is triggered, and fault snapshot data including motor status and environmental parameters is automatically stored.

[0010] Preferably, the deployment method of the multi-dimensional detector array in step 1 includes: D1, embedded deployment: The detector is integrated into the motor stator slot or end cover structure to collect the mechanical response signal in real time during operation; D2, external deployment: The detector is placed at the monitoring point of the motor mounting base to synchronously collect the vibration spectrum and electromagnetic radiation signal.

[0011] Preferably, the method for constructing the dynamic response model includes: Step 21, obtaining calibration parameters of the detector and motor individual dynamics baseline data, and defining the target response type; Step 22: Collect and clean the multi-dimensional time series data, and divide it into a single-dimensional independent data set and a multi-dimensional correlation data set. The single-dimensional data set is used to train the independent feature extraction model of each detector, and the multi-dimensional data set is used to train the cross-dimensional response correlation model. Step 23: construct a hierarchical feature processing network based on the training data, wherein the first layer is a branch for extracting the spectrum slice features of the electromagnetic signal, the second layer is a branch for time domain analysis of the mechanical response, and the third layer is a branch for cross-dimensional response fusion; Step 24: Prune the model parameters using the validation dataset to adapt to the computing resource constraints of the edge computing unit. Step 25: Use the test data set to verify the feature processing accuracy and delay performance of the model in the real-time control loop.

[0012] Preferably, the present invention further includes a resonance control system for a stepper drive motor in a high-precision scenario, the system comprising: A multi-dimensional detector array, deployed on the motor body or mounting base, is used to collect current ripple, rotor displacement, and winding temperature data; A dynamic response processing module, which is used to receive the data stream collected synchronously by the detector, perform parallel processing and output the mechanical response characteristics; An event correlation analysis unit, which is used to extract resonance event markers and align and match the cross-detector data fluctuation intervals to generate mechanical response correlation pairs; An abnormal distribution modeling unit, which locates resonance events in time and space based on mechanical response correlation and constructs a distribution map of abnormal motor responses; A strategy execution unit generates an adaptive suppression strategy according to the abnormal response distribution and controls the execution device.

[0013] Preferably, the abnormal distribution modeling unit specifically includes: The event correlation subunit is used to determine the correlation between the resonance event marker and the mechanical parameter fluctuation, associate them as the same resonance excitation event, and calculate their relative positions on the operating time axis; The spectrum construction subunit is used to extract high-frequency events in a continuous operation cycle, generate a dynamic response spectrum through normalization processing, and dynamically update the abnormal response spectrum based on historical data.

[0014] Preferably, the policy execution unit specifically includes: Phase compensation subunit, which is used to match the damping control curve and generate parameter adjustment scheme, which is loaded in real time by the drive controller; A frequency correction subunit, which is used to analyze the correlation between driving parameters and resonance events and generate waveform reconstruction instructions; The brake protection subunit is used to trigger the hardware protection circuit when the safety threshold is exceeded and store the fault snapshot data.

[0015] Preferably, the deployment structure of the multi-dimensional detector array includes: Embedded structure: The detector is integrated into the stator slot, using high-temperature packaging and anti-interference transmission circuit; External structure: The detector is arranged on the mounting base and equipped with a multi-channel data synchronous acquisition device.

[0016] Preferably, the method for constructing the dynamic response processing module includes: D1, obtain the calibration parameters of the detector and the motor individual dynamics baseline data, and define the target response type; D2, collect and clean multi-dimensional time series data, and divide it into single-dimensional independent data sets and multi-dimensional related data sets; D3, constructing a hierarchical feature processing network, including an electromagnetic signal spectrum slicing branch, a mechanical response time domain analysis branch, and a cross-dimensional response fusion branch; D4, prune the model parameters to adapt to the edge computing unit; D5, verify the performance indicators of the model in the real-time control loop.

[0017] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method and system for high-precision stepper motor resonance control. This system uses a multi-dimensional detector array to synchronously collect multiple data types, including current ripple, rotor angular displacement, and winding temperature. This data is then processed in parallel using a dynamic response model. This method comprehensively captures the dynamic response characteristics of multiple physical fields, including electromagnetic, mechanical, and thermal, during motor operation. This multi-dimensional data fusion and analysis approach overcomes the limitations of traditional single-dimensional detection and accurately extracts key features, such as current harmonic spectra, rotor oscillation phase, and thermal accumulation gradient curves, providing rich data support for the precise identification of resonance events.

[0018] In terms of resonance event correlation and spatiotemporal positioning, the method extracts key frequency response points from the current harmonic spectrum as resonance event markers and aligns and matches the data fluctuation intervals across detectors to generate multiple sets of mechanical response correlation pairs. Based on these correlation pairs, it is not only possible to determine the correlation between the resonance event markers and mechanical parameter fluctuations and associate them as the same resonance excitation event, but also to generate dynamic response spectra through normalization processing within continuous operation cycles. The abnormal response distribution map is dynamically updated by combining historical data with motor load parameters, drive signals, and other information. This spatiotemporal positioning method achieves the precise distribution of resonance events on the time axis and dynamic tracking of energy characteristics, providing a powerful means for in-depth analysis of the mechanisms and laws of resonance generation.

[0019] The generation and execution of the adaptive suppression strategy is another significant advantage of the present invention. The strategy includes a variety of measures such as phase compensation parameter adjustment, drive frequency correction scheme and emergency braking instructions. The phase compensation parameter adjustment can match the preset damping control curve according to the energy of the resonance event and the rotor oscillation amplitude, generate a compensation parameter set of the closed-loop control topology and load it in real time, effectively adjust the dynamic characteristics of the motor, enhance the damping characteristics of the system, and suppress rotor oscillation; the drive frequency correction scheme locates the frequency band of abnormal response by analyzing the timing correlation between the resonance event and the drive pulse width and duty cycle, generates a pulse modulation waveform reconstruction instruction, suppresses the resonance from the source of the drive signal, and improves the stability of the motor operation; the emergency braking instruction triggers the hardware protection circuit in time when the energy of the continuous resonance event exceeds the preset safety threshold, and automatically stores the fault snapshot data containing the motor status and environmental parameters, which not only avoids serious damage to the equipment due to resonance, but also provides valuable data for subsequent fault analysis and system optimization.

[0020] The flexible deployment options of the multi-dimensional detector array (embedded and external) further enhance the system's applicability and reliability. Embedded deployment integrates detectors into the motor's stator slots or end caps, enabling real-time acquisition of mechanical response signals during operation and capturing critical internal motor data. External deployment places detectors at monitoring points on the motor's mounting base, simultaneously collecting vibration spectra and electromagnetic radiation signals, supplementing monitoring information from the external environment. These two deployment options complement each other to achieve comprehensive, three-dimensional monitoring of the motor's operating status.

[0021] The dynamic response model is constructed using a hierarchical feature processing network. By partitioning and training multi-dimensional time series data, this network achieves hierarchical processing for electromagnetic signal spectrum slice feature extraction, mechanical response time-domain analysis, and cross-dimensional response fusion. Parameter pruning is also employed to adapt to the computational resource constraints of edge computing units, and the model's feature processing accuracy and latency performance in real-time control loops are verified. This ensures the model's efficiency and reliability in practical applications, meeting the stringent requirements of high-precision scenarios for real-time data processing and control. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a working principle diagram of the resonance control method for a high-precision stepper drive motor according to the present invention; Figure 2 A design diagram for the spatiotemporal location of resonance events; Figure 3 Structural diagram for multi-dimensional detector array deployment; Figure 4 Design drawings constructed for dynamic response models; Figure 5 Design diagram constructed for the dynamic response processing module. DETAILED DESCRIPTION

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0024] See also Figure 1-Figure 5 The present invention relates to a resonance control method for a stepper drive motor in a high-precision scenario, and the specific implementation steps are as follows: Step 1: Collect motor operating status data using a multi-dimensional detector array. The detector array includes current ripple sensors, rotor angular displacement sensors, and winding temperature sensors. Each detector collects data in a synchronized sequence and feeds the data stream into a dynamic response model for parallel processing. The dynamic response model extracts the current harmonic spectrum from the current ripple data, analyzes the rotor oscillation phase from the rotor angular displacement data, and generates a thermal accumulation gradient curve from the winding temperature data. Ultimately, it outputs the dynamic response characteristics corresponding to each detector.

[0025] Step 2: Identify key frequency response points from the current harmonic spectrum as resonance event markers. Simultaneously, time-align and amplitude-match the fluctuation intervals of the rotor angular displacement and winding temperature data within the same time period to form multiple sets of mechanical response association pairs. Each association pair contains the timestamp and amplitude parameters of the same mechanical response event from different detectors, such as the rotor oscillation phase offset and winding temperature change rate corresponding to the current harmonic peak at a specific moment.

[0026] Step 3: Using the mechanical response correlation pairs generated in Step 2, perform resonance event correlation and spatiotemporal location. By analyzing the correlation between the data from each detector, the distribution of abnormal responses during the motor's operating cycle under the preset dynamic benchmark model is determined. For example, the concentration area of high-frequency resonance events on the time axis and their corresponding spatial location are located.

[0027] Step 4: Based on the mechanical response association pairs and abnormal response distribution, an adaptive suppression strategy is generated and the actuator is triggered. The suppression strategy includes different levels of control instructions such as phase compensation, drive frequency correction, or emergency braking to achieve dynamic suppression of resonance.

[0028] The present invention will be further described below in conjunction with Examples 1 to 5: Example 1: In step 3 of the resonance control method, the specific implementation method of the resonance event association and spatiotemporal positioning is as follows: first, the resonance event association and positioning operation is performed to determine whether the frequency response key point in the current harmonic spectrum extracted in step 2, that is, the resonance event marker point, is within the preset threshold range of the rotor oscillation and thermal gradient fluctuation. The rotor oscillation parameters here can be calculated from the data collected by the rotor angular displacement sensor, such as the fluctuation amplitude and fluctuation frequency of the rotor angular displacement, and the thermal gradient fluctuation parameters are calculated from the temperature data collected by the winding temperature sensor, such as the temperature change rate, temperature gradient distribution, etc. The preset threshold range is pre-set based on the normal operating parameters of the motor, historical resonance data and dynamic benchmark model, and motors of different types and specifications can be adjusted according to actual conditions. If the resonance event marker point meets the condition of being within the preset threshold range of the rotor oscillation and thermal gradient fluctuation, it will be associated with the same mechanical resonance event.

[0029] After confirming that the event is the same mechanical resonance, the relative position of the event on the motor's operating timeline needs to be calculated. The motor's operating timeline is based on the motor's start-up time and is divided into milliseconds. This provides a precise time coordinate for each resonance event. For example, if the motor is started at time t0 and a resonance event occurs at t milliseconds after the motor starts, the relative position of the event on the operating timeline is t milliseconds. In this way, the occurrence time of each resonance event in the time series can be accurately recorded, providing basic time information for subsequent analysis and processing.

[0030] Next, a spatiotemporal fusion positioning operation is performed to extract high-frequency resonance events of the motor within M consecutive operating cycles. An operating cycle here refers to the time it takes for the motor to complete a full working cycle. High-frequency resonance events are those that occur frequently within a certain time range. This can be achieved by setting a frequency threshold. When the frequency of a resonance event exceeds a preset value, it is identified as a high-frequency resonance event. Then, using the aforementioned resonance event association and positioning method, the distribution positions P1 to PM of the M events at time T on the operating timeline are obtained.

[0031] The first event at time T is selected as the reference point, and the positions of the other M-1 events are normalized relative to the reference point. The purpose of normalization is to eliminate the absolute position differences of different events on the time axis, so as to facilitate the analysis of relative time relationships. Specifically, the time position of the reference point is used as the origin, and the time positions of other events are converted into time offsets relative to the reference point. For example, if an event is after the reference point, milliseconds, then its normalized time position is This normalization process generates a dynamic response spectrum with the first event as the origin, time as the horizontal axis, and event energy as the vertical axis. Event energy is a parameter that comprehensively considers the data from each detector. It is calculated by weighting parameters such as the amplitude of the current harmonic spectrum, the amplitude of the rotor oscillation, and the rate of change of the heat accumulation gradient curve. The weights of different parameters can be set based on their impact on the resonance event.

[0032] The dynamic response spectrum can intuitively display the distribution of high-frequency resonance events along the time axis and the energy of each event over M consecutive operating cycles. By observing the distribution density of events and the location of energy peaks in the spectrum, we can understand the temporal patterns and energy distribution characteristics of resonance events.

[0033] Then, by retrospectively correlating historical data and combining it with motor load parameters and drive signals, the offset of the newly added event at time T+1 is calculated. Retrospective correlation of historical data involves analyzing the correlations and changing trends between resonance event data, motor load parameters, and drive signals from multiple previous operating cycles. Motor load parameters include load torque and load inertia, while drive signals include drive pulse width, duty cycle, and frequency. By establishing a mathematical model or using a data mining algorithm, the relationship between these parameters and the time and energy of the resonance event is analyzed, thereby predicting the time offset of the newly added event at time T+1 relative to the historical event.

[0034] The abnormal response distribution map is dynamically updated based on the calculated offset. This map comprehensively describes the location and extent of abnormal responses during the motor's operating cycle. It not only includes the distribution of resonance events along the time axis but also combines information about the motor's spatial structure, such as the different motor parts and winding phases, to show the spatial distribution of abnormal responses. By dynamically updating this map, changes in abnormal responses during motor operation can be reflected in real time, providing a basis for timely adjustments to suppression strategies.

[0035] Throughout the entire process of resonant event correlation and spatiotemporal localization, it is crucial to ensure the synchronization and accuracy of data from each detector. The current ripple sensors, rotor angular displacement sensors, and winding temperature sensors within the multidimensional detector array must acquire data with strict synchronization to ensure accurate correspondence between the timestamps and amplitude parameters of the same mechanical response event across different detectors. Furthermore, anti-interference measures, such as shielded cables and differential signal transmission, must be implemented during data transmission to minimize the impact of external interference on the data and ensure the accuracy of subsequent analysis and processing.

[0036] Furthermore, for different motor operating scenarios and working conditions, parameters such as the preset threshold range, M value, and weight coefficient need to be flexibly adjusted to meet the needs of different high-precision scenarios. For example, in scenarios with extremely high precision requirements, such as medical equipment and precision machining machine tools, stricter preset threshold ranges and more frequent event sampling frequencies may be required to ensure timely detection and processing of small resonance events, thereby preventing them from affecting the accuracy and performance of the equipment.

[0037] Example 2: The adaptive suppression strategy in step 4 includes three specific execution modes: phase compensation parameter adjustment, drive frequency correction scheme, and emergency braking command. The following is a detailed implementation of each mode: The core of phase compensation parameter adjustment is to suppress mechanical oscillations by dynamically adjusting motor control parameters. The specific implementation process is as follows: First, based on the mechanical response association pairs generated in step 2, the system extracts the key characteristic parameters of the resonance event, including the amplitude in the current harmonic spectrum, the peak and duration of the rotor oscillation amplitude, and the winding temperature change rate. The resonance event energy value is then calculated through weighted calculation. The weight of the current harmonic amplitude can be set based on the electromagnetic characteristics of the motor. For example, within the typical operating frequency band of a stepper motor, the weight of the influence of the high-order harmonic amplitude on resonance can be set to 40%. The weight of the rotor oscillation amplitude is determined by combining the mechanical structure stiffness. If the motor shaft stiffness is low, the weight of this parameter can be increased to 50%. The weight of the winding temperature change rate reflects the indirect influence of heat accumulation on resonance and is usually set to 10%.

[0038] Secondly, based on the resonance event energy value and the rotor oscillation amplitude, the system invokes a preset damping control curve matching mechanism. The damping control curve is a pre-built parameter mapping table based on the motor dynamics model. For example, the rotor oscillation amplitude is divided into several intervals (such as ±1°, ±3°, ±5°, etc.), each corresponding to a different damping ratio target value. The damping ratio target value is associated with the phase compensation parameters (such as the proportional coefficient and the integral time constant) through the transfer function relationship of the closed-loop control topology. For example, when the rotor oscillation amplitude is detected to be ±4° and the resonance energy is at a medium level, the system matches the corresponding damping control curve and determines that the damping ratio of the closed-loop system needs to be increased from the current 0.2 to 0.5. It then calculates that the proportional coefficient needs to be adjusted from 10 to 18 and the integral time constant needs to be adjusted from 0.05s to 0.03s, forming a set of compensation parameters.

[0039] Finally, the compensation parameter set is loaded into the motor control loop in real time via the drive controller. The drive controller can be a digital signal processor (DSP) or field-programmable gate array (FPGA), which features high-speed data processing capabilities and a real-time control interface. During the loading process, the system transmits the parameters to the controller via a serial communication protocol (such as the CAN bus or SPI bus). The controller then reconfigures the closed-loop control topology based on the new parameters. For example, it updates the proportional-integral-derivative parameters in the PID control algorithm, adjusts the current phase relationship between the motor's phase windings, and generates an electromagnetic torque that is opposite in phase to the mechanical oscillation, thereby suppressing abnormal rotor oscillations. The time delay of this process must be controlled within the motor control cycle (typically in the microsecond to millisecond range) to ensure real-time phase compensation.

[0040] The drive frequency correction scheme is implemented based on a temporal correlation analysis between resonance events and drive pulse parameters. The specific steps are as follows: First, the system extracts temporal features from the mechanical response association pairs generated in step 2, analyzing the correspondence between the resonance event occurrence time and drive pulse width, duty cycle, frequency, and other parameters. For example, by counting the timestamps of 100 consecutive resonance events, it was found that 80% of them occurred within the drive pulse duty cycle range of 55%-65% and the frequency range of 200-300Hz, thereby locating the frequency range of the abnormal response.

[0041] Secondly, the system generates a pulse modulation waveform reconstruction instruction, which contains an adjustment strategy for the drive pulse parameters. There are two adjustment methods: one is frequency offset, which is to adjust the drive pulse frequency as a whole to avoid the resonant frequency band. For example, if the original drive frequency is 250Hz (within the resonant frequency band of 200-300Hz), the frequency can be adjusted to 180Hz or 350Hz. The specific offset is determined according to the motor load characteristics and process requirements. It is necessary to ensure that the motor can still meet the speed and torque requirements after the frequency adjustment; the second is pulse width and duty cycle optimization. By changing the duty cycle of the pulse waveform, the proportion of the power-on time of each phase winding of the motor is adjusted, thereby changing the distribution and action time of the electromagnetic force. For example, adjusting the duty cycle from 60% to 40% can reduce the energy accumulation during the winding power-on period and reduce the possibility of resonance.

[0042] Before generating a reconstruction command, the system verifies that the adjusted drive parameters meet the motor's operating boundary conditions, such as the maximum allowable frequency and minimum pulse width. For example, if the motor's maximum safe operating frequency is 500Hz and its minimum pulse width is 2μs, the adjusted drive frequency must not exceed 500Hz, and the pulse width must not be less than 2μs. After verification, the command is executed by the drive controller's waveform generation module, which modifies the register parameters of the pulse generation circuit in real time to achieve dynamic reconstruction of the drive waveform. To avoid motor shock caused by sudden frequency changes, the system can use a ramped frequency adjustment method, gradually adjusting the drive frequency from the original value to the target value over a certain period of time (e.g., 50ms) to ensure smooth motor operation.

[0043] The emergency braking command is a protective measure implemented when the risk of resonance exceeds a safety threshold. The specific implementation is as follows: First, the system continuously monitors the accumulated energy of consecutive resonance events. When this value exceeds a preset safety threshold (set based on parameters such as the motor's rated power and mechanical strength, for example, 1.5 times the rated energy), the hardware protection circuit is triggered. This hardware protection circuit typically includes components such as fast-acting fuses and solid-state relays (SSRs), which can cut off power to the motor in microseconds to prevent mechanical damage or electrical failure caused by resonance.

[0044] At the same time, the system automatically stores fault snapshot data containing motor status and environmental parameters. Motor status parameters include the current ripple waveform, rotor angular displacement curve, winding temperature, and drive pulse parameters at the time of triggering. Environmental parameters include the supply voltage, ambient temperature, and real-time load torque. This data is stored in its original format in non-volatile memory (such as Flash memory or hard disk) via a high-speed data acquisition module (such as an ADC-based multi-channel synchronous acquisition card), with a storage frequency of no less than 10kHz to ensure data integrity and traceability. Fault snapshot data can be used for subsequent fault analysis. For example, by replaying the current waveform and rotor displacement curve, it can be determined whether the source of the resonance excitation is electromagnetic interference, mechanical resonance, or load mutation.

[0045] After triggering the emergency brake, the system enters a safe shutdown process, including shutting down all drive signals, activating the brake device (if any), and illuminating the fault indicator. Simultaneously, an alert is issued to the operator via the human-machine interface (HMI) or communication interface, indicating the fault type and time of occurrence. To prevent false triggering, the system can incorporate multiple verification mechanisms. For example, braking is triggered only after detecting three or more consecutive resonance events exceeding the safety threshold, thus avoiding malfunctions caused by a single abnormal fluctuation. Furthermore, the emergency brake command takes precedence over other control commands, ensuring a rapid system response in extreme situations and safeguarding equipment and personnel safety.

[0046] The three suppression strategies do not operate independently, but rather work synergistically based on the severity and stage of the resonance event. For example, when mild resonance is detected (resonance energy below 50% of the safety threshold), the phase compensation parameter adjustment strategy is prioritized, suppressing vibration by fine-tuning control parameters. If the resonance energy continues to rise to the 50%-100% safety threshold, the drive frequency correction strategy is activated, adjusting the drive parameters to avoid the resonant frequency band. If the energy exceeds the safety threshold, an emergency braking command is immediately triggered to ensure safety.

[0047] Furthermore, the system features adaptive parameter adaptation, dynamically adjusting key strategy parameters based on the motor's real-time operating status. For example, the weight coefficients in the phase compensation parameters can be automatically optimized using an online learning algorithm for motors under varying load conditions. The offset for drive frequency correction can be calculated in real time based on load torque, and the safety threshold can be dynamically adjusted based on winding temperature. This adaptive mechanism enables the system to adapt to the diverse and variability of motor operating conditions in high-precision scenarios, ensuring the effectiveness and robustness of the resonance control strategy.

[0048] Example 3: The multi-dimensional detector array can be deployed in two modes: embedded deployment and external deployment. The following are the specific implementation methods: Embedded deployment integrates sensors into the motor's structure to directly monitor the motor's operating status. Specifically, current ripple sensors, rotor angular displacement sensors, and winding temperature sensors are integrated into the motor's stator slots, end caps, or shaft through a specific process.

[0049] Current ripple sensors utilize high-precision Hall elements or Rogowski coils, embedded near the stator winding ends or the output terminals. The Hall element's placement must adhere to the principle of electromagnetic induction, with its sensing surface perpendicular to the magnetic field generated by the winding current to accurately capture real-time fluctuations in the current signal. For example, a separate Hall current sensor is installed at the output terminal of each phase winding of a three-phase stepper motor to ensure independent monitoring of the current ripple in each phase. The sensor is connected to the signal conditioning circuit via high-temperature conductors (such as silicone-insulated conductors with a temperature rating of 200°C or higher). The conductor path must avoid high-temperature areas of the motor, and a shielding layer must be used to suppress electromagnetic interference.

[0050] Rotor angular displacement sensors typically use magnetoelectric or photoelectric encoders, integrated into the motor shaft or rotor. The stator portion of a magnetoelectric encoder (containing the induction coil) is fixed to the inside of the motor end cap, while the rotor portion (containing permanent magnets or reluctance elements) is rigidly connected to the motor shaft and rotates synchronously with the rotor. As the rotor rotates, the stator induction coil generates a sinusoidal signal from the changing magnetic flux. This signal processing circuitry then converts it into rotor position and speed information. For high-precision applications, an incremental encoder with a resolution of 10,000 lines or higher can be used to ensure an angular displacement monitoring accuracy of ±0.036° (corresponding to 360° / 10,000 lines).

[0051] Winding temperature sensors utilize thermistors or thermocouples, embedded between stator winding turns or in the gaps between end windings. Thermistors must be high-precision NTC (negative temperature coefficient) elements, with resistance-to-temperature relationships conforming to a specific scale (for example, an NTC element with a B value of 3950K has a resistance of 10kΩ at 25°C and 1.2kΩ at 100°C). The sensors are connected to the external circuit via insulated enameled wire, which must be insulated from the winding conductors to avoid the risk of short circuits. To achieve distributed monitoring of the winding temperature field, multiple temperature sensors can be deployed at different locations on the same phase winding (such as the upper and lower turns, or at the center of the end winding) to form a temperature monitoring array.

[0052] The mechanical structure of embedded deployments must meet the motor's sealing and vibration resistance requirements. The sensor is connected to the motor using high-temperature curing adhesive (such as epoxy) or mechanical clips to prevent displacement under high-speed motor rotation and vibration. The sensor housing is made of stainless steel or ceramic to withstand the high temperatures (maximum operating temperature up to 180°C) and oil corrosion inside the motor. Signal transmission uses a differential signaling mode. For example, the current signal is transmitted using two wires, "+I" and "-I." A subtractor circuit eliminates common-mode interference and improves the signal-to-noise ratio.

[0053] External deployment involves placing detectors at monitoring points on the motor's mounting base or surrounding structures, collecting motor operating status data through non-contact or indirect contact. This mode is suitable for scenarios where space on the motor itself is limited or where modifications to the motor structure are undesirable.

[0054] Vibration spectrum monitoring is achieved using vibration accelerometers mounted on the base. These sensors, which utilize piezoelectric or piezoresistive principles, are located at the four corners or center of the base to monitor the vibration amplitude and frequency during motor operation. For example, a single-axis accelerometer is installed in each of the three orthogonal directions (X, Y, and Z) to form a three-axis vibration monitoring unit capable of collecting vibration signals in the frequency range of 0.1 Hz to 10 kHz. The sensors are secured to the base surface by magnets or bolts, and should be installed as close to the motor's feet as possible to minimize the vibration transmission path.

[0055] Electromagnetic radiation signals are collected using external current probes or magnetic field sensors. The current probes are clipped onto the motor power cables or winding leads, sensing the current harmonics through electromagnetic coupling. Magnetic field sensors (such as fluxgate sensors) are placed around the motor, 5-10 cm from the motor housing, to monitor the distribution and changes in the leakage magnetic field. For three-phase motors, electromagnetic radiation signals from all three phases can be monitored simultaneously. Fourier transform analysis of their spectral characteristics can be used to identify abnormal harmonic components (such as the fifth and seventh harmonics).

[0056] External monitoring of winding temperature can be achieved through infrared temperature sensors or indirect inference of base temperature. The infrared temperature sensor is aimed at the motor surface (such as the stator housing or end cap) and measures the surface temperature non-contact, with a measurement accuracy of ±2°C and a response time of ≤50ms. The base temperature sensor uses a patch thermocouple, attached to the contact area between the base and the motor. It monitors the heat transfer of the motor through heat conduction and infers the winding temperature by combining it with a motor thermal model (such as a thermal resistance network model). The thermal resistance network model can be expressed as:

[0057] in, is the winding temperature (unit: °C), is the base temperature measurement value (unit: °C), is the motor power loss (unit: W), is the thermal resistance between the motor body and the base (unit: °C / W). The motor power loss can be calculated by the effective value of current and voltage ( ,in is the line voltage, is the line current, is the power factor), and the thermal resistance parameters are obtained from the motor design drawings or measured data.

[0058] External deployments require a multi-channel data synchronization acquisition device to ensure that the signals from each detector are synchronized with microsecond-level timing accuracy. This synchronization acquisition device, based on a field-programmable gate array (FPGA) or a dedicated synchronization controller, achieves time alignment of the data across channels through hardware trigger signals (such as pulse signals) or software timestamp calibration mechanisms. For example, a 10MHz synchronization clock signal is generated internally by the FPGA and transmitted via differential clock lines to the analog-to-digital converter (ADC) of each detector, ensuring that the sampling time error for each channel is ≤1μs.

[0059] Embedded and external deployments can be combined based on actual needs to form a complementary monitoring system. For example, in high-precision machine tool applications, both embedded current ripple sensors and external vibration accelerometers are used: the former directly monitors subtle fluctuations in winding current, while the latter monitors the overall vibration response of the machine tool. This cross-dimensional data fusion improves the accuracy of resonance detection.

[0060] For different stepper motor types (e.g., two-phase, three-phase, and five-phase stepper motors), the placement and number of sensors must be adjusted accordingly. For example, a five-phase stepper motor requires five sets of current ripple sensors, one for each phase winding. The resolution of the rotor angular displacement sensor must be adapted to the motor's step angle (e.g., 0.72° / step) to ensure angular displacement monitoring accuracy is better than 1 / 100 of the step angle.

[0061] Embedded deployment requires reserved space for sensor installation during the motor design phase, such as sensor slots within the stator slots and encoder mounting holes on the end caps. External deployment requires planning sensor installation locations during base design, noting installation dimensions and orientation requirements. Signal transmission lines for both deployment modes must be kept away from the motor power lines to avoid electromagnetic coupling interference. If necessary, separate cable trays or shielded ducting should be used for routing.

[0062] Example 4: The dynamic response model construction method is used to process the current ripple, rotor angular displacement, and winding temperature data collected by the multi-dimensional detector array to achieve real-time extraction and correlation analysis of the motor resonance characteristics. The specific implementation method is as follows: The initial stage of model building requires basic data preparation and target definition. First, obtain the detector calibration parameters, such as the current ripple sensor sensitivity (measured in V / A, representing the voltage output per ampere of current change), the rotor angular displacement sensor resolution (e.g., pulses per revolution (PPR), used to convert pulse signals into angle values), and the winding temperature sensor accuracy (e.g., ±0.5°C). Simultaneously, collect baseline motor dynamics data, including the no-load current waveform (e.g., the RMS current per phase of a three-phase stepper motor at no-load), the rotor oscillation amplitude at rated speed (the angular displacement fluctuation range calculated from encoder data), and the normal temperature rise curve (the time-varying trend of the winding temperature during continuous motor operation). Based on this data, define the target response type. For example, a total harmonic distortion (THD) of less than 5% and a rotor oscillation amplitude of less than ±2° is defined as "normal operation," a THD ≥ 10% and a rotor oscillation amplitude ≥ ±5° is defined as "severe resonance," and anything in between is classified as "mild resonance."

[0063] Data acquisition and cleaning are key steps in model construction. Time series data is synchronously collected using a multi-dimensional detector array. The sampling frequency must satisfy the Nyquist sampling theorem. For example, for a current ripple signal with a maximum frequency of 10kHz, the sampling frequency is set to 25kHz. The collected data contains noise and outliers, requiring cleaning. First, a sliding average filter is used to remove high-frequency noise (such as power supply ripple interference). Then, the Laida criterion (3σ principle) is applied to remove outliers. Specifically, for each detector's data stream, the mean μ and standard deviation σ are calculated. Data points outside the range of μ ± 3σ are considered outliers and corrected through interpolation. The cleaned data is divided into single-dimensional independent datasets and multi-dimensional correlated datasets. Single-dimensional datasets contain only a single type of data from a single detector (for example, only the time domain waveform of current ripple) and are used to train independent feature extraction models for each detector. Multi-dimensional datasets contain current, angular displacement, and temperature data at the same timestamp and are used to train cross-dimensional response correlation models. For example, single-dimensional datasets can be used to train spectral analysis models for current ripple, while multi-dimensional datasets can be used to analyze the phase correlation between current harmonics and rotor oscillations.

[0064] The hierarchical feature processing network is constructed using a modular architecture and is divided into three layers. The first layer is the branch for extracting the spectrum slice features of the electromagnetic signal. For current ripple data, the time domain signal is first converted into a frequency domain spectrum through a fast Fourier transform (FFT), generating multiple frequency slices (such as the amplitude and phase of key frequency points such as 100Hz, 200Hz, and 500Hz). A two-dimensional convolutional neural network (CNN) is then used to extract features from the spectrum slices. The filter size of the convolution layer can be set to 3×3 with a step size of 1. Multi-layer convolution operations capture harmonic features in different frequency bands. For example, for the fifth harmonic (frequency 5× the fundamental frequency) commonly found in stepper motors, the CNN can automatically learn the correlation between its amplitude changes and resonant events.

[0065] The second layer is the time-domain analysis branch for mechanical response, processing time series data of rotor angular displacement and winding temperature. For rotor angular displacement data, a long short-term memory (LSTM) network is used to extract time series features. The LSTM unit has two hidden layers, each containing 128 neurons. A gating mechanism is used to capture long-term dependencies in rotor oscillations (such as the oscillation decay trend over multiple consecutive cycles). For winding temperature data, a one-dimensional convolutional neural network (1D-CNN) is used to analyze heat accumulation trends. The convolution kernel size is set to 5, extracting temperature change rate and gradient distribution features. For example, if the winding temperature continues to rise for 10 minutes and the rate of change exceeds 0.1°C / s, the 1D-CNN can identify it as a sign of abnormal heat accumulation.

[0066] The third layer is the cross-dimensional response fusion branch, which fuses the electromagnetic and mechanical features extracted by the first two layers through a fully connected layer. The number of neurons in the input layer is the sum of the lengths of the feature vectors output by the first two layers. For example, if the first layer outputs 50-dimensional electromagnetic features and the second layer outputs 80-dimensional mechanical features, the input layer has a total of 130 neurons. The hidden layer uses dropout technology (with a deactivation rate set to 0.2) to prevent overfitting. The output layer uses a softmax function to output the probability values of three target response categories (normal operation, mild resonance, and severe resonance), achieving a comprehensive assessment of the motor's operating status.

[0067] After model training, parameter pruning is required to adapt to the computational resource constraints of edge computing units. Edge computing units are typically embedded processors (such as the ARM Cortex-A72) or low-power FPGAs, which have limited CPU power and memory capacity. Parameter pruning is achieved through the following steps: First, the parameters of each layer of the hierarchical feature processing network are ranked by importance, and L1 regularization is used to identify redundant connections (connections with weights approaching zero). Connections with weights less than a set threshold (such as 0.01) are then removed, and the number of filters in the convolutional layers is pruned (for example, reducing the number of filters in the first CNN layer from 64 to 32). Finally, knowledge distillation techniques are used to transfer the knowledge of the complex model to a lightweight model. For example, a fully trained model is used as a teacher model to guide a student model (with 50% fewer parameters) to learn its output probability distribution, ensuring that the accuracy loss of the pruned model is within an acceptable range (for example, a decrease in target response classification accuracy of no more than 3%).

[0068] During the model validation phase, a test dataset was used to evaluate the performance metrics of the real-time control loop. The test dataset contained motor operating data under various operating conditions, including light load (20% rated load), full load (100% rated load), and overload (120% rated load), as well as various drive frequency combinations (100Hz, 500Hz, and 1000Hz). Verification focused on feature processing accuracy and latency performance. Feature processing accuracy was assessed by comparing model output features with manually annotated benchmark values. For example, the amplitude error of the current harmonic spectrum must be ≤5%, and the rotor oscillation phase error must be ≤±2°. Latency performance was assessed by measuring the time interval from data acquisition to feature output, requiring an overall processing delay of ≤10ms to meet real-time control requirements. For example, under full-load conditions with a drive frequency of 500Hz, the model was required to complete current ripple spectrum analysis, rotor angular displacement oscillation feature extraction, and cross-dimensional fusion within 10ms, outputting a resonance risk assessment result to provide timely support for the development of mitigation strategies.

[0069] In practical applications, the dynamic response model needs to be adaptively adjusted based on the motor type and operating conditions. For example, the electromagnetic characteristics of a two-phase hybrid stepper motor and a five-phase reactive stepper motor differ significantly, requiring independent model parameter training. For the frequent start-stop conditions of high-precision machine tool processing, the model can add a feature extraction module for the transient response during the startup phase. By capturing the current surge peak and rotor angular acceleration mutation characteristics at the startup moment, it can provide an early warning of the risk of startup resonance. For scenarios with long-term continuous operation of automated production lines, the model can optimize the weight of thermal accumulation features, strengthen the analysis of winding temperature gradients and thermal time constants, and avoid mechanical resonance caused by thermal expansion.

[0070] Example 5: The hardware architecture of the resonance control system of the present invention consists of a multi-dimensional detector array, a dynamic response processing module, an event correlation analysis unit, an abnormal distribution modeling unit, and a policy execution unit. Each unit works in coordination via a high-speed data link. The following describes the implementation method in detail in conjunction with specific application scenarios: Taking the application of three-phase stepper motors in precision assembly equipment as an example, the detector array adopts a hybrid deployment of embedded and external configurations. In the embedded configuration, the current ripple sensor uses three high-precision Hall elements (sensitivity 50mV / A) embedded in the three-phase stator winding output terminals and connected to the signal conditioning board via high-temperature-resistant shielded cable. The rotor angular displacement sensor uses a 20,000-line magnetoelectric encoder integrated into the motor shaft, outputting A / B / Z quadrature pulse signals (resolution 0.018° / pulse). The winding temperature sensor uses three NTC thermistors (B value 3950K) embedded between the turns of the U, V, and W phase windings and connected to the temperature acquisition module via enameled wire. In the external configuration, a triaxial vibration accelerometer (measurement range ±10g, frequency response 0.1-10kHz) is mounted at each corner of the base, secured by a magnetic mount. The signal is connected to a multi-channel synchronous data acquisition card via shielded cable. Electromagnetic radiation monitoring uses a clip-on current probe (bandwidth 100MHz) attached to the motor power line to sense high-frequency current harmonics in real time.

[0071] This module is implemented based on an ARM+FPGA architecture, using the Xilinx Zynq series SoC as an example. The ARM Cortex-A9 dual-core processor serves as the master control unit, responsible for system scheduling, model parameter updates, and communication interface management. The FPGA programmable logic unit implements parallel processing of multi-dimensional data. It includes three independent data channels: a current ripple channel, which uses an FFT IP to calculate the harmonic spectrum (updated every 1ms); a rotor angular displacement channel, which uses a pulse counting module to calculate position and velocity (outputting angle values every 500μs); and a temperature channel, which uses an ADC sampling module (12-bit precision, 100Hz sampling rate) to acquire thermistor resistance and convert it into temperature. After preprocessing, the data from each channel is stored in shared memory. The ARM processor then calls a dynamic response model (deployed in DDR3 memory) for feature extraction, outputting the current harmonic spectrum feature vectors, rotor oscillation phase parameters, and thermal accumulation gradient values.

[0072] This unit is built on the NVIDIA Jetson Nano edge computing platform, leveraging its CUDA acceleration capabilities for parallel computing. The specific process involves first acquiring real-time feature data from the dynamic response processing module via the PCIe interface. Resonance event markers are then extracted on the CPU (for example, when the amplitude of the 200Hz component in the current harmonics exceeds 1.2 times the baseline value, it is marked as a potential resonance point). Time series data across detectors (such as current, angular displacement, and temperature data within 50 ms before and after the marker) are then transferred to the GPU memory. A data alignment algorithm is then executed in parallel using CUDA cores. Using the timestamps of the current markers as a reference, the rotor angular displacement and temperature data are synchronized to the same time grid (with 1μs accuracy) through linear interpolation. The fluctuation range of each parameter is calculated (for example, ±3° for angular displacement and 0.3°C / ms for temperature). If these values fall within preset thresholds, a mechanical response correlation pair (including timestamps, parameter amplitudes, and energy values) is generated. The unit can process over 1,000 correlation pairs per second, meeting real-time analysis requirements.

[0073] This unit consists of an event correlation subunit and a spectrogram construction subunit, implemented on industrial-grade PC hardware (Intel i7 processor with 16GB of RAM). The event correlation subunit stores historical resonance event data in a database (such as SQLite) and receives correlation pairs output by the event correlation analysis unit in real time. It determines the temporal correlation between new events and historical events (for example, if a resonance event occurs at the same time axis position within three consecutive cycles, it is considered a high-frequency region) and calculates the relative position of the event on the running time axis (with the motor start-up time as the origin, and an accuracy of 1ms). The spectrogram construction subunit dynamically generates a two-dimensional spectrogram using the OpenGL graphics library: the abscissa is the normalized time axis (with the first event as the origin, ranging from -50ms to +50ms), and the ordinate is the event energy value (calculated as the current harmonic amplitude × 0.6 + the angular displacement fluctuation amplitude × 0.3 + the temperature change rate × 0.1). Each event is represented as a scattered point, and the color depth reflects the energy level. When a new event is detected at time T, the system automatically retraces the historical data of the previous 100 cycles, calculates the average time offset of the event relative to historical events (such as an average lag of 2.5ms), and updates the abnormal response distribution map to mark potential resonance risk areas.

[0074] The unit consists of three subunits, implemented with discrete electronic components and programmable logic devices: Phase compensation subunit: Based on TI's TMS320F28379DDSP chip, it receives the compensation parameters (such as proportional coefficient Kp=20 and integration time Ki=0.1) output by the abnormal distribution modeling unit, generates a phase modulation signal through the ePWM module, and loads it to the motor driver through a power amplifier (such as the IR2110 driver chip). It adjusts the conduction phase difference of each phase winding to generate a damping torque to suppress oscillation.

[0075] Frequency correction subunit: Using Altera Cyclone IV FPGA, it monitors the drive pulse parameters (frequency f = 300 Hz, duty cycle D = 60%) in real time. When it detects that the resonance event is concentrated in the f = 280-320 Hz range, it generates a new frequency control word (such as f = 400 Hz) through the PLL phase-locked loop module, adjusts the pulse width to D = 50%, and transmits it to the motor controller through the SPI interface to achieve drive waveform reconstruction.

[0076] The brake protection subunit consists of a hardware logic circuit, including a comparator (LM339) and a solid-state relay (G3MB-202P). When the energy of three consecutive resonance events exceeds the safety threshold, the comparator outputs a high level, triggering the SSR to disconnect the motor power supply. Simultaneously, it controls the EEPROM memory (24LC64) to store motor status data at the time of the fault (e.g., RMS current 5A, speed 1200rpm, temperature 75°C). The storage time window is 100ms before and after the event trigger to ensure data integrity.

[0077] Each hardware unit is interconnected via a CAN bus (1Mbps transmission rate). The signal conditioning board of the multi-dimensional detector array integrates a CAN transceiver (TJA1145), which packages raw data into CAN frames and sends them to the dynamic response processing module. The dynamic response processing module transmits feature data to the event correlation analysis unit and the anomaly distribution modeling unit via the CANopen protocol. The strategy execution unit receives suppression strategy instructions via the CAN bus and provides feedback on execution status (e.g., phase compensation parameters loaded, frequency corrections effective). The system uses an isolated DC-DC module (24V input, 5V / 3.3V output) to ensure independent power supply for each unit, reducing the impact of power supply interference on signal accuracy.

[0078] In a precision machine tool machining scenario, when a motor drives a ball screw for high-speed positioning, if the dynamic response processing module detects a sudden increase in the 400Hz component of the current harmonics (THD increases from 3% to 8%) and the rotor angular displacement fluctuates by ±4°, the system generates a mechanical response correlation pair through the event correlation analysis unit. The anomaly distribution modeling unit identifies this event as forming a continuous distribution on the time axis with the resonance events at the same location in the previous two cycles, identifying it as a high-frequency resonance region. The strategy execution unit first triggers the phase compensation subunit to load parameters Kp=25 and Ki=0.08. If the resonance energy does not decrease within 500ms, the frequency correction subunit is activated to adjust the drive frequency from 400Hz to 300Hz to avoid the resonance frequency range. If the energy exceeds the safety threshold continuously due to a sudden load change, the brake protection subunit immediately disconnects the power supply and stores fault data to prevent vibration-induced machining accuracy loss.

[0079] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0080] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for controlling resonance of a stepper motor for high-precision scenarios, characterized in that: include: Step 1: Collecting motor operating status data using a multi-dimensional detector array, including a current ripple sensor, a rotor angular displacement sensor, and a winding temperature sensor. The data collected synchronously by each detector is input into a dynamic response model for parallel processing to obtain dynamic response characteristics corresponding to each detector. The characteristics include current harmonic spectrum, rotor oscillation phase, and thermal accumulation gradient curve. Step 2: Extract key frequency response points from the current harmonic spectrum as resonance event markers. Align and match the data fluctuation intervals of the rotor angular displacement and winding temperature within the same time period to obtain multiple sets of mechanical response association pairs. Each association pair contains the timestamp and amplitude parameters of the same mechanical response event in different detectors. Step 3: Using the mechanical response correlation pairs obtained in step 2, perform resonance event correlation and spatiotemporal location to determine the abnormal response distribution under a preset dynamic benchmark model during the motor operation cycle; Step 4: Based on the mechanical response association pairs obtained in step 2 and the abnormal response distribution determined in step 3, an adaptive suppression strategy is generated and an execution device is triggered.

2. The resonance control method for a high-precision stepper motor according to claim 1, wherein: The resonance event association and spatiotemporal positioning in step 3 specifically include: Step 31, resonance event association and location: Determine whether the resonance event marker is within a preset threshold range of rotor oscillation and thermal gradient fluctuation. If so, associate it with the same mechanical resonance event and calculate the relative position of the event on the motor operation timeline; the operation timeline is based on the motor startup timeline as the origin and is divided into milliseconds. Step 32, spatiotemporal fusion positioning: Extract high-frequency resonance events of the motor within M consecutive operating cycles. Obtain the distribution positions P1 to PM of the M events on the operating time axis at time T in the same manner as in step 31. Select the first event as the reference point, normalize the positions of the other M-1 events relative to the reference point, and generate a dynamic response spectrum with the first event as the origin, the time axis as the horizontal axis, and the event energy as the vertical axis. By back-tracing and correlating historical data, combined with motor load parameters and drive signals, the offset of new events at time T+1 is calculated, and the abnormal response distribution map is dynamically updated.

3. The resonance control method for a high-precision stepper motor according to claim 1, wherein: The adaptive suppression strategy in step 4 includes phase compensation parameter adjustment, drive frequency correction scheme and emergency braking command; wherein: Phase compensation parameter adjustment specifically includes: matching the preset damping control curve based on the resonance event energy and rotor oscillation amplitude, generating a compensation parameter set for the closed-loop control topology, and loading it in real time through the drive controller; The drive frequency correction scheme specifically includes: analyzing the timing correlation between the resonance event and the drive pulse width and duty cycle, locating the frequency band of the abnormal response, and generating pulse modulation waveform reconstruction instructions; The emergency braking instruction specifically includes: when the energy of continuous resonance events exceeds the preset safety threshold, the hardware protection circuit is triggered, and fault snapshot data including motor status and environmental parameters is automatically stored.

4. The resonance control method for a high-precision stepper motor according to any one of claims 1 to 3, characterized in that: The deployment of the multi-dimensional detector array in step 1 includes: D1, embedded deployment: The detector is integrated into the motor stator slot or end cover structure to collect the mechanical response signal in real time during operation; D2, external deployment: The detector is placed at the monitoring point of the motor mounting base to synchronously collect the vibration spectrum and electromagnetic radiation signal.

5. The resonance control method for a high-precision stepper motor according to any one of claims 1 to 3, characterized in that: The construction methods of dynamic response models include: Step 21, obtaining calibration parameters of the detector and motor individual dynamics baseline data, and defining the target response type; Step 22: Collect and clean the multi-dimensional time series data, and divide it into a single-dimensional independent data set and a multi-dimensional correlation data set. The single-dimensional data set is used to train the independent feature extraction model of each detector, and the multi-dimensional data set is used to train the cross-dimensional response correlation model. Step 23: construct a hierarchical feature processing network based on the training data, wherein the first layer is a branch for extracting the spectrum slice features of the electromagnetic signal, the second layer is a branch for time domain analysis of the mechanical response, and the third layer is a branch for cross-dimensional response fusion; Step 24: Prune the model parameters using the validation dataset to adapt to the computing resource constraints of the edge computing unit. Step 25: Use the test data set to verify the feature processing accuracy and delay performance of the model in the real-time control loop.

6. A resonance control system for a stepper motor in a high-precision scenario, characterized in that: include: A multi-dimensional detector array, deployed on the motor body or mounting base, is used to collect current ripple, rotor displacement, and winding temperature data; A dynamic response processing module, which is used to receive the data stream collected synchronously by the detector, perform parallel processing and output the mechanical response characteristics; An event correlation analysis unit, which is used to extract resonance event markers and align and match the cross-detector data fluctuation intervals to generate mechanical response correlation pairs; An abnormal distribution modeling unit, which locates resonance events in time and space based on mechanical response correlation and constructs a distribution map of abnormal motor responses; A strategy execution unit generates an adaptive suppression strategy according to the abnormal response distribution and controls the execution device.

7. The resonance control system for a high-precision stepper motor according to claim 6, wherein: The abnormal distribution modeling unit specifically includes: The event correlation subunit is used to determine the correlation between the resonance event marker and the mechanical parameter fluctuation, associate them as the same resonance excitation event, and calculate their relative positions on the operating time axis; The spectrum construction subunit is used to extract high-frequency events in a continuous operation cycle, generate a dynamic response spectrum through normalization processing, and dynamically update the abnormal response spectrum based on historical data.

8. The resonance control system for a high-precision stepper motor according to claim 6, wherein: The policy execution unit specifically includes: Phase compensation subunit, which is used to match the damping control curve and generate parameter adjustment scheme, which is loaded in real time by the drive controller; A frequency correction subunit, which is used to analyze the correlation between driving parameters and resonance events and generate waveform reconstruction instructions; The brake protection subunit is used to trigger the hardware protection circuit when the safety threshold is exceeded and store the fault snapshot data.

9. The resonance control system for a high-precision stepper motor according to any one of claims 6 to 8, characterized in that: The deployment structure of the multi-dimensional detector array includes: Embedded structure: The detector is integrated into the stator slot, using high-temperature packaging and anti-interference transmission circuit; External structure: The detector is arranged on the mounting base and equipped with a multi-channel data synchronous acquisition device.

10. The resonance control system for a high-precision stepper motor according to any one of claims 6 to 8, characterized in that: The construction method of the dynamic response processing module includes: D1, obtain the calibration parameters of the detector and the motor individual dynamics baseline data, and define the target response type; D2, collect and clean multi-dimensional time series data, and divide it into single-dimensional independent data sets and multi-dimensional related data sets; D3, constructing a hierarchical feature processing network, including an electromagnetic signal spectrum slicing branch, a mechanical response time domain analysis branch, and a cross-dimensional response fusion branch; D4, prune the model parameters to adapt to the edge computing unit; D5, verify the performance indicators of the model in the real-time control loop.

Images