Autonomous intelligent calibration method for servo control system based on multi-source information fusion

Abstract

The invention relates to the technical field of multi-source servo intelligent calibration, and particularly discloses a servo control system autonomous intelligent calibration method based on multi-source information fusion, and the method comprises the steps: obtaining the equipment attribute information of all servo equipment and multi-source sensor data synchronized through a global time protocol in response to a system starting or calibration instruction; and according to the physical space distribution data and the transmission connection relation data in the equipment attribute information, constructing an equipment incidence matrix representing the coupling strength between the equipment. According to the method, various cross-device coupling conflict characteristics including transmission, electromagnetism and loads can be intelligently recognized, external interference noise is accurately stripped from sensor data through dynamic weight adjustment and a coupling projection reconstruction mechanism, it is ensured that a control system only carries out calibration for the real intrinsic error of the device, and the reliability of the system is improved. Therefore, false compensation and system instability caused by coupling interference are avoided.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of multi-source servo intelligent calibration, and in particular to a self-intelligent calibration method for a servo control system based on multi-source information fusion. BACKGROUND

[0002] With the development of high-end equipment manufacturing towards high speed and precision, multi-axis linkage servo control systems have been widely used in industrial robots, precision CNC machine tools and automated production lines. In the traditional servo control architecture, the calibration of each axis is usually based on the assumption of single-axis independent closed loop, that is, it is believed that the sensor feedback of each servo device only represents the running state of the device itself. The control system often uses standard PID regulation or simple synchronization control strategy, assuming that the physical environment is static decoupled. This traditional method can still meet the requirements in low speed or weak coupling scenarios, but in complex working conditions involving rigid mechanical connection, common bus electrical drive or high frequency load switching, there are complex physical coupling relationships between the servo axes.

[0003] However, the existing technology faces a serious conflict between sensor reliability and physical coupling interference when dealing with such multi-device linkage scenarios. Specifically, when there is strong mechanical transmission or electromagnetic interference between devices, the sensor readings of a servo device often superimpose vibration noise or signal crosstalk from associated devices. The existing control algorithm lacks the ability to distinguish between local intrinsic errors and cross-device coupling interference, and often misjudges external coupling interference as local device running deviation, and then forcibly executes incorrect calibration compensation.

[0004] Such mis-calibration not only cannot eliminate system errors, but also can cause system oscillation due to excessive regulation, and even in the asynchronous coordination process of load rapid switching, neglecting the physical inertia lag can cause destructive internal stress between devices, seriously restricting the overall precision and stability of the servo system. SUMMARY

[0005] The present application aims to at least solve one of the technical problems in the related art. To this end, the purpose of the present application is to propose a self-intelligent calibration method for a servo control system based on multi-source information fusion to solve the problems of mis-calibration and system oscillation caused by physical coupling in multi-device linkage.

[0006] To achieve the above-mentioned purpose, the first aspect of the present application proposes a self-intelligent calibration method for a servo control system based on multi-source information fusion, comprising the following steps: In response to a system start or calibration instruction, obtaining device attribute information of each servo device and multi-source sensor data synchronized through a global time protocol; According to the physical space distribution data and the transmission connection relationship data in the device attribute information, a device correlation matrix representing coupling strength between devices is constructed; The multi-source sensor data is preprocessed to extract cross-device coupling conflict features, the cross-device coupling conflict features are input into a preset conflict recognition model, and an output conflict type label is obtained; According to the conflict type label and the currently recognized device linkage scene type, a corresponding fusion strategy is called, and the data fusion weight of each servo device is dynamically and cooperatively adjusted based on the device correlation matrix, and cross-device fusion features are generated; Based on the cross-device fusion features and a preset cooperative calibration strategy, the calibration parameter compensation value of each servo device is calculated, and a calibration instruction is issued to each servo device to execute a multi-device synchronous calibration process. The process of constructing the device correlation matrix representing the coupling strength between devices includes: according to the physical distance value between devices, the quantization coefficient of the transmission connection mode and the load distribution proportion value, the correlation coefficient between each pair of devices is calculated, and the device correlation matrix is composed of all correlation coefficients.

[0007] To achieve the above purpose, the second aspect embodiment of the present application proposes a servo control system autonomous intelligent calibration system based on multi-source information fusion, comprising: A data acquisition and correlation modeling module is configured to acquire device attribute information of each servo device and multi-source sensor data synchronized through a global time protocol in response to a system start or calibration instruction, and construct a device correlation matrix representing coupling strength between devices according to physical space distribution data and transmission connection relationship data in the device attribute information; wherein the process of constructing the device correlation matrix representing the coupling strength between devices includes: according to the physical distance value between devices, the quantization coefficient of the transmission connection mode and the load distribution proportion value, the correlation coefficient between each pair of devices is calculated, and the device correlation matrix is composed of all correlation coefficients. A conflict feature recognition module is configured to preprocess the multi-source sensor data to extract cross-device coupling conflict features, input the cross-device coupling conflict features into a preset conflict recognition model, and obtain an output conflict type label. A dynamic weight cooperation module is configured to call a corresponding fusion strategy according to the conflict type label and the currently recognized device linkage scene type, and dynamically and cooperatively adjust the data fusion weight of each servo device based on the device correlation matrix, and generate cross-device fusion features. A cooperative calibration execution module is configured to calculate the calibration parameter compensation value of each servo device based on the cross-device fusion features and a preset cooperative calibration strategy, and issue a calibration instruction to each servo device to execute a multi-device synchronous calibration process.

[0008] To achieve the above object, the third aspect of the present application provides an electronic device, comprising a memory, a processor and a computer program stored in the memory, wherein the computer program is executed by the processor to realize the above-mentioned autonomous intelligent calibration method of the servo control system based on multi-source information fusion.

[0009] The autonomous intelligent calibration method of the servo control system based on multi-source information fusion provided by the embodiments of the present application can intelligently identify various cross-device coupling conflict characteristics including transmission, electromagnetism and load, and accurately strip external interference noise from sensor data through dynamic weight adjustment and coupling projection reconstruction mechanism, so as to ensure that the control system only calibrates the real intrinsic error of the device itself, thereby avoiding false compensation and system instability caused by coupling interference. Meanwhile, for the load fast switching in the asynchronous collaborative scene, the present application introduces a dynamic reference correction mechanism based on physical inertia, which eliminates the pseudo-synchronous error caused by the asynchronization of electrical signals and mechanical responses. In addition, the establishment of the global redundant resource pool enables the system to maintain high-precision collaborative control through the scheduling of associated sensors when some sensors fail or are seriously disturbed, thereby significantly improving the robustness and adaptive ability of the servo control system under complex working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0010] Figure 1 is a flowchart of the autonomous intelligent calibration method of the servo control system based on multi-source information fusion provided by the present application; Figure 2 is an implementation execution schematic diagram of the autonomous intelligent calibration system of the servo control system based on multi-source information fusion provided by the present application; Figure 3 is a structural schematic diagram of the electronic device provided by the present application. DETAILED DESCRIPTION

[0011] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the drawings, wherein the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0012] The autonomous intelligent calibration method, system and electronic device of the servo control system based on multi-source information fusion provided by the embodiments of the present application are described below with reference to the drawings.

[0013] Embodiment One

[0014] The embodiment details a kind of self-learning intelligent calibration method of servo control system based on multi-source information fusion.The method aims to solve the calibration deviation problem of multi-axis linkage servo system caused by mechanical transmission coupling, electromagnetic interference and dynamic load variation in complex industrial scenarios. The method of the embodiment mainly runs in the global edge controller of servo control system, and the controller has high-power processing unit and real-time industrial bus interface.

[0015] Referring to Figure 1 The method of the embodiment specifically includes the following steps as shown in the flowchart: Step S1: in response to system startup or calibration instruction, the device attribute information of each servo device and the multi-source sensor data synchronized through global time protocol are acquired.

[0016] For example, the global edge controller of servo control system monitors the running state of the system in real time. When the system completes power-on initialization, or receives the calibration instruction issued by the operation and maintenance personnel through the human-computer interaction interface, or the system detects that the running accuracy decreases and automatically triggers the calibration request, the calibration process is started immediately.

[0017] The process of acquiring device attribute information is a key link to establish the static physical model of the system. The global edge controller reads the electronic nameplate data stored in each servo driver through the bus, or downloads the configuration file from the cloud configuration database. The device attribute information covers the motor parameters such as rated power, rated torque, moment of inertia and pole pair number of servo motor, and also includes the transmission component parameters such as reduction ratio of speed reducer, lead of screw and stiffness coefficient of shaft coupling. What is particularly important is that the device attribute information clearly records the physical space distribution coordinates of each device in the industrial field and the mechanical connection topological structure between devices, for example, whether devices A and B share the same base, or whether they are mechanically coupled through belt or gear.

[0018] In order to solve the alignment problem of multi-source data in time domain, the embodiment strictly adopts global time protocol, specifically IEEE1588 Precision Time Protocol (PTP). In the data acquisition stage, the global edge controller acts as the PTP master clock, and each servo driver and independent sensor node acts as the slave clock. Through the delay measurement and clock deviation correction mechanism of PTP protocol, the system realizes sub-microsecond level time synchronization accuracy. Under the driving of this synchronous clock source, the system synchronously triggers the collection of multi-source sensor data. The multi-source sensor data not only includes the position data, speed data and torque current data feedback by the encoder of servo motor, but also includes the vibration accelerometer data of key points in mechanical structure, the laser displacement sensor data installed in end effector and the temperature sensor data monitoring environmental changes. These data are marked with uniform high-precision time stamp at the moment of acquisition, which lays a solid time sequence foundation for subsequent correlation analysis.

[0019] Step S2: Based on the physical space distribution data and transmission connection relationship data in the device attribute information, construct a device association matrix that characterizes the coupling strength between devices.

[0020] For example, traditional control strategies often assume that axes are independent of each other, while the core of this embodiment lies in quantifying the mutual influence between devices. Constructing a device correlation matrix is ​​the first step in achieving this goal. This matrix is ​​a... The square array, in which, This represents the number of servo devices in the system. Each element in the matrix quantifies the likelihood of two devices physically interfering with or interacting with each other.

[0021] The process of constructing the device association matrix, which characterizes the coupling strength between devices, specifically includes: calculating the association coefficient between each pair of devices based on the physical distance between devices, the quantification coefficient of the transmission connection method, and the load distribution ratio; and constructing the device association matrix from all association coefficients. During the calculation, the system iterates through each pair of devices, comprehensively considering spatial proximity, mechanical connection tightness, and load sharing relationship.

[0022] The calculation of the correlation coefficient between each pair of devices includes calculating the device correlation coefficient using the following formula. Correlation coefficient between equipment : ; in, Characterization equipment With equipment The physical distance between the devices is calculated using spatial coordinates from the device attribute information. The closer the physical distance, the greater the likelihood of vibration transmission or electromagnetic crosstalk between the devices via their bases. The formula above uses... This inverse relationship is reflected by the reciprocal of the distance; that is, the closer the distance, the larger the value of the correlation term. This coefficient characterizes the transmission link and is used to quantify the tightness of a mechanical connection. Depending on whether it's direct transmission, indirect transmission, or no transmission relationship, this coefficient takes a preset first value, a second value, or zero. For example, if the equipment... With equipment Through direct meshing of rigid couplings or gears, mechanical vibration and torque transmission are extremely direct. The value is the first numerical value, such as If the two are connected by a belt or a long drive shaft, there is a certain degree of flexibility and buffering. The value is the second number, such as If the two are completely physically isolated in terms of mechanical structure and have no transmission relationship, then... is zero; The device load proportion coefficient is characterized. In the scenario of multiple motors driving the same load cooperatively, for example, a double-drive gantry, two motors share the load torque. If the load sharing ratio of the device is close to that of the device , it indicates that the two are strongly coupled in dynamics, and the torque fluctuation of one will directly affect the running stability of the other. Therefore, is used to reflect the depth of such load-level coupling; , , are preset weight coefficients, respectively corresponding to the importance proportion of the spatial distance factor, the mechanical transmission factor and the load distribution factor in the overall correlation degree. According to different actual application scenarios, the three weight coefficients can be flexibly configured. For example, in a compact precision machining center, spatial electromagnetic interference is the main problem, so the value of is adjusted higher; in a heavy transmission line, mechanical transmission is the main problem, so the value of is adjusted higher.

[0023] In addition, in order to facilitate subsequent calculation, the value range of the correlation degree coefficient is normalized to the closed interval of zero to one. When the calculation result is greater than , it is forcibly truncated to . The finally generated device correlation matrix can be intuitively presented in the form of a heat map to guide the system to quickly locate the high-coupling risk area.

[0024] Step S3: pre-processing the multi-source sensor data to extract cross-device coupling conflict features, inputting the cross-device coupling conflict features into a pre-set conflict recognition model, and obtaining an output conflict type label.

[0025] For example, the collected original sensor data often contains noise and has different dimensions, and direct input into the model has poor effect. The pre-processing link first cleans the data, removes measurement white noise using Kalman filtering, and uses the Z-score standardization method to map different dimension data (such as ampere value of current and micron value of position) to the standard normal distribution interval. Subsequently, the system performs sliding window interception on the time series data to form a fixed-length time slice sequence.

[0026] The cross-device coupling conflict features at least include: transmission coupling conflict features representing mechanical transmission mutual influence, electromagnetic coupling conflict features representing electrical signal interference, and load coupling conflict features representing load distribution deviation. Instead of simply inputting the original data into the model, the system first extracts feature vectors with clear engineering significance based on physical mechanism.

[0027] The specific generation process of the cross-device coupling conflict feature includes the following sub-steps: First, computing devices With equipment The Pearson correlation coefficient of torque fluctuations between the two devices is used as the torque coupling coefficient in the transmission coupling conflict characteristics. The system extracts the torque-current sequences of the two devices within the same time window and calculates the ratio of the product of their covariance and standard deviation. If the Pearson correlation coefficient is close to 1 or -1, it indicates that the torque fluctuations of the two devices are highly positively or negatively correlated, indicating strong mechanical transmission coupling oscillations; if it is close to 0, it indicates that they do not affect each other. Simultaneously, the cross-spectral density of the vibration signals can also be calculated to further characterize the transmission coupling.

[0028] Secondly, the percentage of overlapping intervals of the same harmonic frequencies between devices is calculated as the current harmonic spectrum overlap degree in the electromagnetic coupling conflict characteristics. The system performs a Fast Fourier Transform (FFT) on the phase current data of each device to obtain the amplitude-frequency response curve in the frequency domain. Subsequently, the system analyzes the spectrum of different devices to identify characteristic harmonic frequencies with significant amplitudes. If the devices... With equipment If there is a high degree of overlap in a specific high-order harmonic frequency range, and this frequency is not the fundamental frequency, then it is highly likely that the electromagnetic coupling conflict is caused by common bus interference or electromagnetic radiation in space. The higher the proportion of the overlapping range, the greater the possibility of electromagnetic interference.

[0029] Finally, the difference between the actual load distribution ratio and the preset ratio is calculated as the load deviation rate in the load coupling conflict characteristic. The system calculates the actual load torque borne by each axis in real time and compares it with the preset theoretical load distribution value in the control algorithm. If the actual load distribution deviates significantly from the preset value at a certain moment, for example, if the dual motors that should have evenly distributed the load become a push-pull state, this constitutes a significant load coupling conflict characteristic.

[0030] After feature extraction is completed, the step of inputting the cross-device coupling conflict features into a pre-set conflict recognition model to obtain the output conflict type label includes: inputting the extracted cross-device coupling conflict features and the device association matrix as input vectors into the conflict recognition model containing a convolutional neural network layer and a self-attention mechanism layer. This model is a deep neural network, where the convolutional neural network layer (CNN) is used to extract the spatial correlation of frequency domain features and local temporal domain features, and the self-attention mechanism layer (Self-Attention) is used to capture long-distance temporal dependencies and global association weights between features of different devices.

[0031] Optionally, probability distribution values of each type of conflict output by the conflict identification model are obtained, and it is determined whether the highest probability value is greater than a preset conflict confidence threshold; if yes, the type corresponding to the highest probability value is determined as the conflict type label. The model output layer uses a Softmax function to output the probability that the current state belongs to a transmission coupling conflict, an electromagnetic coupling conflict, a load coupling conflict, or no conflict. If the highest probability is lower than the confidence threshold, the system can determine that it is a mixed complex conflict, and a more conservative calibration strategy needs to be started.

[0032] Step S4: According to the conflict type label and the currently identified device linkage scene type, a corresponding fusion strategy is called, and the data fusion weights of each servo device are dynamically coordinated and adjusted based on the device association matrix to generate cross-device fusion features.

[0033] For example, in the process of multi-source information fusion, not all sensor data has the same credibility. The core of this step is to dynamically suppress the data weight affected by the interference source according to the specific type of conflict and highlight the role of high-credibility data.

[0034] The dynamic coordinated adjustment of the data fusion weights of each servo device based on the device association matrix includes: retrieving a preset weight adjustment rule library according to the conflict type label to determine the local basic weight adjustment strategy corresponding to the affected device. For example, the rule library defines that when an electromagnetic conflict occurs, the weight of the current sensor should be preferentially reduced while the weight of the position encoder is maintained; when a transmission conflict occurs, the weight of the accelerometer should be reduced.

[0035] For example, based on the association degree coefficient in the device association matrix, the weight adjustment amplitude of the affected device is calculated, and the weight adjustment amplitude is transmitted to the associated device for coordinated adjustment. The adjustment of the weight is not isolated. If device A is determined to be untrustworthy, the data of device B, which has a very high association degree with device A, may also be contaminated. Therefore, the system uses the association degree coefficient as a transmission factor to proportionally transmit the weight reduction operation of device A to device B, achieving coordinated optimization of the global weight.

[0036] If the conflict type label is a transmission coupling conflict, a weight decay factor is calculated according to the association degree coefficient, and the weight decay factor is used to down-regulate the weight proportion of the affected sensor in the fusion calculation. Specifically, the weight decay factor can be defined as where k is an adjustment coefficient. The higher the association degree, the more severe the decay, thereby effectively suppressing the transmission of mechanical coupling noise in the data fusion stage and preventing false vibration signals from misleading the calibration algorithm.

[0037] For example, for such an extreme case of strong coupling, simply reducing the weight may cause the system to lose awareness of the real state of the device. Therefore, if the conflict type label is a transmission coupling conflict, before calculating the weight decay factor according to the correlation coefficient, an abnormal handling step in the strong coupling scenario is included: judging whether the correlation coefficient between devices in the device correlation matrix is greater than a preset strong coupling threshold. If the correlation coefficient is less than or equal to the threshold, the weight reduction is performed according to the normal process.

[0038] If it is greater than the preset strong coupling threshold, the step of using the weight decay factor to reduce the weight proportion of the affected sensor is suspended, and a coupling projection reconstruction process is started. Because in strong coupling, although the signal is mixed, it also contains strong correlation information, and directly discarding (reducing the weight) is a waste of information.

[0039] The coupling projection reconstruction process includes: obtaining real-time state data of associated devices, combining the correlation coefficient and the calculated mechanical transmission delay time, and calculating the coupling interference projection amount of the associated devices to the current device. This step actually establishes an interference observer to inversely deduce the size of the interference using the known correlation. Then, the original sensor data of the current device is subtracted from the coupling interference projection amount to extract the local intrinsic motion component representing the actual motion state of the current device itself. This component is a pure signal after removing external coupling noise, which truly reflects the error of the device itself. Finally, the signal-to-noise ratio of the local intrinsic motion component is calculated, and the signal-to-noise ratio is used as a forward correction factor to adjust the gain of the local basic weight of the current device. Through this strategy of subtracting interference and increasing weight, the system can still capture the small local fault features in a strong coupling environment.

[0040] Optionally, before performing the multi-device synchronization calibration process, in order to further improve the robustness of the system, the local redundant sensor information of each servo device is read to construct a global redundant resource pool, which records the type, accuracy parameter and current occupation state of the sensor. When the main sensor is reduced in weight or even disabled due to conflict, the standby sensor in the resource pool will appear as a substitute. The standby sensor is, for example, an auxiliary grating ruler and a redundant encoder.

[0041] For example, the long short-term memory network and the attention mechanism model are used to perform time series analysis on the cross-device coupling conflict features in the historical time window to predict the conflict occurrence probability in the future time window. The LSTM network is good at processing time series and can predict whether a more serious coupling conflict will occur at the next moment according to the current shock trend.

[0042] If the predicted conflict probability exceeds the preset intervention probability threshold, based on the device association matrix, the reading data fusion process of the idle cross-device redundant sensor with an association coefficient less than a preset safety threshold is scheduled from the global redundant resource pool. The scheduling strategy here is very intelligent: the system will intentionally select those sensors with low association with the current conflict source (i.e. The small ones, introduce these bystander external data sources, which can effectively break the local coupling dead loop and provide an objective calibration reference.

[0043] Step S5: Based on the cross-device fusion feature and the preset cooperative calibration strategy, the calibration parameter compensation value of each servo device is calculated, and a calibration instruction is issued to each servo device to execute a multi-device synchronous calibration process.

[0044] For example, after the above steps, the system obtains a cross-device fusion feature that integrates multi-source information and eliminates coupling interference. This feature can truly reflect the overall operation deviation of the system.

[0045] The calculation of the calibration parameter compensation value of each servo device based on the cross-device fusion feature and the preset cooperative calibration strategy includes identifying the current device linkage scene type. Different linkage modes require different calibration logic.

[0046] If the device linkage scene type is a synchronous transmission scene, such as electronic gear synchronization or gantry synchronization, the device with the largest association coefficient value in the device association matrix is selected as the reference device, and the calibration parameters of other devices are calculated based on the difference between the reference device state data and the local state data and the association coefficient weighting. This master-slave calibration strategy ensures that the entire system follows the strongest and most stable shaft movement, ensuring the consistency of the trajectory.

[0047] If the device linkage scene type is an asynchronous cooperative scene, such as cooperative handling or flexible processing, there is no absolute master shaft. Then, based on the load distribution ratio, a load weighted reference value is calculated, and the calibration parameters are calculated based on the difference between the load weighted reference value and the local state data and the load coupling coefficient. The system calculates a virtual center of gravity as a reference, and each axis moves towards the center of gravity to achieve force balance.

[0048] For example, in an asynchronous cooperative scene, load switching is often very fast and can easily lead to misjudgment. If the device linkage scene type is an asynchronous cooperative scene, the calculation process of the calibration parameter also includes an inertia lag compensation step for rapid load switching: before calculating the load weighted reference value, the rate of change of the load distribution ratio of each servo device over time is monitored in real time.

[0049] determining whether the change rate exceeds a preset transient switching threshold. If the change rate is low, it indicates a steady-state switching and no compensation is needed.

[0050] If the threshold is exceeded, it is determined that the current is in a load dynamic switching stage, triggering an inertia hysteresis compensation mechanism. The inertia hysteresis compensation mechanism includes: obtaining a preset device rotational inertia coefficient, and calculating a theoretical inertia hysteresis amount as the product of the change rate and the rotational inertia coefficient. According to Newton's second law, a change in torque produces acceleration, and changes in speed and position require time, and this time difference is inertia hysteresis. If this is not considered, the system will forcibly calibrate the normal physical hysteresis as a synchronous error, resulting in overshoot. Then, the theoretical inertia hysteresis amount is superimposed on the load-weighted reference value calculated based on static proportion to generate a dynamic inertia correction reference value. Finally, the dynamic inertia correction reference value is used to replace the load-weighted reference value as the basis for calculating the difference with the local state data. By artificially slowing down the reference value to match the inertia response of the physical system, the system successfully shields false errors during the transient process and achieves smooth transition.

[0051] Step S6: After performing the multi-device synchronization calibration process, further comprising: collecting the calibrated multi-device linkage data, and calculating a closed-loop verification index including a cross-device collaboration error and a coupling conflict resolution rate. After the calibration instruction is issued, the system continues to monitor subsequent operation data to evaluate the calibration effect. The cross-device collaboration error reflects the synchronization accuracy, and the coupling conflict resolution rate reflects the vibration suppression effect.

[0052] Optionally, it is determined whether the closed-loop verification index meets a preset calibration qualified standard. If the index meets the standard, it indicates that the current conflict identification model and weight adjustment strategy are effective.

[0053] If not, the single-device feature extraction layer parameters in the conflict identification model are kept unchanged, and the cross-device coupling feature extraction layer parameters are updated and iterated until the closed-loop verification index meets the calibration qualified standard. This online learning mechanism enables the system to have evolution ability. By freezing the bottom-layer single-device feature parameters, catastrophic forgetting of the model is prevented; only the high-layer coupling feature parameters are fine-tuned, so that the model can quickly adapt to new mechanical aging states or working condition changes, and continuously improve the recognition accuracy of complex coupling conflicts.

[0054] In summary, this embodiment constructs a complete autonomous intelligent calibration closed loop for a servo control system by introducing a device association matrix and a deep learning conflict identification model, combined with dynamic weight adjustment and inertial lag compensation techniques. This method not only theoretically solves the conflict resolution problem in multi-source information fusion, but also provides an effective means to cope with extreme conditions such as strong coupling and rapidly changing loads in engineering practice, significantly improving the control accuracy and intelligence level of multi-axis linkage servo systems.

[0055] Example 2

[0056] Building upon Example 1, this embodiment further elaborates on an enhanced calibration processing mechanism based on coupled projection reconstruction, specifically addressing the potentially severe, strongly coupled conditions in servo control systems. In Example 1, for conventional transmission coupling conflicts, the system primarily employs a strategy of reducing the weights of affected sensors to suppress noise. However, in certain high-precision or heavy-load industrial applications, such as dual-arm robots for semiconductor wafer transport or dual-drive feed axes in heavy-duty gantry machining centers, the mechanical connection stiffness between devices is extremely high. This causes the coupling effect to become more than just interference noise; it becomes a dominant system characteristic. In such strongly coupled scenarios, simply using a weight reduction strategy not only fails to effectively eliminate interference but also significantly reduces the control system's ability to perceive the true dynamics of the equipment due to the loss of crucial state information. Therefore, this embodiment introduces a coupled projection reconstruction process. By actively modeling and stripping coupled components, it transforms what was originally considered interference—strongly coupled signals—into a core basis for improving calibration accuracy.

[0057] During the multi-source information fusion process, when the conflict type label output by the conflict identification model is determined to be a transmission coupling conflict, the system does not immediately perform a weight decay operation. Instead, it first enters a crucial decision branch, namely, a quantitative assessment of the coupling strength. This assessment process is based on the device association matrix that has already been constructed in Implementation Example 1.

[0058] Optionally, the system first determines whether the correlation coefficient between devices in the device association matrix is ​​greater than a preset strong coupling threshold. This strong coupling threshold is a pre-set empirical value used to distinguish between ordinary coupling and strong coupling. In practical engineering applications, this threshold is usually set to a relatively high value, such as 0.7 or 0.8. The system reads the device... With equipment correlation coefficient between and compare it with a preset strong coupling threshold. Compare them. If the correlation coefficient... Less than or equal to the strong coupling threshold , the system will maintain the processing logic in Embodiment One, i.e. considering that the coupling at this time mainly manifests as noise interference, and then calculating a weight attenuation factor and using the weight attenuation factor to down-regulate the weight proportion of the affected sensor in the fusion calculation.

[0059] If the correlation coefficient is greater than a preset strong coupling threshold , it indicates that there is a very strong physical binding relationship between the device and the device . At this time, the vibration or torque fluctuation signal collected by the sensor contains a large amount of deterministic dynamic transmission components from the associated device. In this case, the system determines that simple weight reduction processing will cause serious loss of useful information, so the system will suspend the step of using the weight attenuation factor to down-regulate the weight proportion of the affected sensor, and immediately start the coupling projection reconstruction process. The core idea of this process is to decouple and separate the mixed signals to restore the true behavior of the device itself.

[0060] Optionally, the first step of the coupling projection reconstruction process is to obtain the real-time state data of the associated device, and calculate the coupling interference projection amount of the associated device on the current device in combination with the correlation coefficient and the calculated mechanical transmission delay time. In this process, the global edge controller reads the sensor data sequence of the associated device as the interference source in real time through the high-speed bus, denoted as . At the same time, the system needs to calculate the time required for mechanical vibration waves or torque waves to transmit from the device to the device , i.e. the mechanical transmission delay time.

[0061] The calculation of the mechanical transmission delay time is based on the principles of physical acoustics and the geometric parameters of the system. The system calls the physical distance between devices stored in the device attribute information, and the material attribute parameters of the connecting medium. Assuming that the mechanical wave propagation speed of the connecting medium, such as a cast iron base or a steel screw, is , the mechanical transmission delay time can be obtained by dividing the physical distance by the propagation speed . More accurately, the system can also use cross-correlation analysis algorithm to dynamically correct the theoretically calculated mechanical transmission delay time by calculating the time lag corresponding to the peak value of the cross-correlation function of the historical vibration data sequences of the device and the device , so as to ensure the accuracy of the time parameter.

[0062] After determining the delay time, the system further calculates the coupling interference projection of the associated device on the current device. This projection quantity characterizes the coupling interference projection of the associated device. The motion state is transmitted through the mechanical structure and then into the equipment. The theoretical response value generated. System construction projection function. This function uses the state data of the associated device before the delay time. As the independent variable, and using the correlation coefficient As the transfer gain. Specifically, the projection amount of coupling interference. It can be expressed as the correlation coefficient. Related device status data after delay The product of these terms, plus a correction term representing transmission losses. This projected quantity is physically equivalent to: if the device It is completely stationary, and only the device The equipment caused by the action Theoretical readings of the sensor.

[0063] The second step of the coupled projection reconstruction process is to subtract the coupled interference projection from the original sensor data of the current device to extract the local intrinsic motion components characterizing the actual motion state of the current device. This is the most critical signal separation step in the entire reconstruction process. Assume the device... The raw sensor data is This data is a superposition of local motion and external coupled disturbances. The system performs a subtraction operation to calculate the local intrinsic motion components. Its value is equal to the original sensor data. Subtract the calculated coupling interference projection amount .

[0064] Through the above subtraction operation, the system effectively separates strongly correlated devices. The resulting externally forced vibration or torque following component. The remaining local intrinsic motion components. Although its amplitude may be much smaller than the original signal, it reflects the device's signal extremely purely. Its own load characteristics, friction and wear conditions, and potential local fault characteristics. For example, in a dual-drive gantry crane, the two motors theoretically move synchronously, but if the motors... The guide rails show signs of corrosion; this minute resistance characteristic is often detected by the motor. The immense traction force transmitted through the crossbeam masks this characteristic. Through the extraction process described above, this minute drag feature, i.e., the local intrinsic motion component, becomes apparent, serving as the basis for subsequent precise calibration.

[0065] The third step of the coupling projection reconstruction process is to calculate the signal-to-noise ratio of the local intrinsic motion component, and use the signal-to-noise ratio as a forward correction factor to adjust the gain of the local base weight of the current device. After extracting the local intrinsic motion component, the system needs to evaluate the quality of this component to determine its position in the final multi-source information fusion. The system selects a time window of a preset length, calculates the effective value power of the local intrinsic motion component as the signal power, and calculates the high-frequency residual filtered out in the preprocessing stage as the noise power, and the ratio of the two is the signal-to-noise ratio. .

[0066] Optionally, unlike the weight reduction processing in Embodiment One for ordinary coupling, in the strong coupling scenario, once the local intrinsic motion component with high signal-to-noise ratio is successfully extracted, the system needs to increase the weight of this data. This is because the reconstructed data has excluded external interference, and the information density contained is extremely high, which plays an irreplaceable role in revealing the true health status of the device. The system constructs a weight gain function to map the signal-to-noise ratio to a forward correction factor . The correction factor is a number greater than one, and is positively correlated with the signal-to-noise ratio.

[0067] The system reads the local base weight of the device , which is the default value assigned by the system during initialization. Then, the system uses the calculated forward correction factor to perform multiplication on the local base weight to obtain the final dynamic fusion weight , that is equals times . Through this gain adjustment, the influence of the sensor data of the device in the multi-source information fusion algorithm is significantly amplified. This means that the system will trust the cleaned local data more and mainly rely on this data to generate calibration instructions.

[0068] ​This strategy based on coupled projection reconstruction not only solves the signal aliasing problem under strong coupling, but also constitutes a kind of virtual active noise reduction mechanism. It enables the servo control system to maintain similar perceptual clarity of single-axis independent control even in an extremely complex mechanical linkage environment. For example, in the moment of heavy load fast start-stop, strong coupling causes the sensor readings of all related axes to fluctuate dramatically, and the conventional algorithm may misjudge the system as a whole instability and trigger an emergency stop or false severe reverse compensation. However, after adopting the reconstruction method of the embodiment, the system can clearly distinguish which fluctuations are caused by normal following due to mechanical transmission and which fluctuations are caused by abnormal vibration due to the excessive backlash of a certain axis, and accordingly execute micron-level accurate position compensation for that specific axis, thereby realizing extreme precision calibration under the premise of ensuring system safety.

[0069] In summary, the embodiment introduces a strong coupling threshold determination mechanism and innovatively applies a coupled projection reconstruction process under strong coupling conditions, realizing a technical leap from passive interference suppression to active signal separation. Through the calculation of mechanical transmission delay, the construction of interference projection model, the extraction of intrinsic motion components, and the weight gain adjustment based on signal-to-noise ratio, the method of the embodiment effectively solves the error masking problem in strong coupling servo systems, ensuring that each servo device can obtain autonomous intelligent calibration based on its real operating state in complex linkage scenarios, greatly expanding the application range and engineering value of the method.

[0070] Embodiment Three

[0071] Based on Embodiment One and Embodiment Two, the embodiment discloses a calibration parameter calculation process containing an inertia lag compensation mechanism for the most complex and easily causing control oscillation asynchronous collaborative scenario in multi-device linkage. In Embodiment One, the system solves the coupled interference problem in steady state or quasi-steady state by constructing a device correlation matrix and weight adjustment. However, in actual industrial production, scenarios such as over-range switching of double-motor anti-backlash driving, load transfer in multi-robot collaborative handling, and process conversion on flexible production lines are typical asynchronous collaborative scenarios. In such scenarios, the load distribution ratio of each servo device is not fixed, but will change dramatically and rapidly with the needs of the process flow. This change is often completed within milliseconds or even microseconds at the electrical signal level, but the rotors of servo motors and connected mechanical loads have objective physical mass, so the change of their motion state is inevitably limited by the moment of inertia, resulting in a physical lag that cannot be eliminated. If the control system ignores this physical law and forcibly calibrates the mechanical body that has not responded according to the instantaneous change of the electrical signal, it will inevitably lead to serious overshoot and oscillation. Therefore, the embodiment discloses in detail how to introduce inertia lag compensation steps for load rapid switching in the process of calculating calibration parameters.

[0072] For example, the method in this embodiment also runs in the global edge controller of the servo control system and relies on multi-source sensor data synchronized by a global time protocol. The system first needs to identify the current device linkage scenario type. When the system determines that it is currently in an asynchronous collaborative scenario, it means that there is no absolute master-slave rigid constraint between the servo devices, but rather the system maintains balance and cooperation by dynamically allocating load torque. In this context, the core basis of the calibration algorithm is the load-weighted reference value. However, to avoid misjudgments caused by physical inertia, the system must perform a crucial pre-step before calculating the load-weighted reference value: real-time monitoring of the rate of change of the load allocation ratio of each servo device with respect to time.

[0073] The load distribution ratio is a torque distribution command coefficient that the control algorithm sends to each servo drive in real time according to process requirements. For example, in a scenario where two motors drive the same load, the sum of the load distribution ratios of motor A and motor B is usually 100%. At the moment of load switching, the ratio of motor A may drop rapidly from 80% to 20%, while that of motor B increases accordingly. The system uses a differential algorithm or differential filter to perform real-time differentiation on this ratio signal. Let the load distribution ratio at time t be... The system calculates its first derivative with respect to time to obtain the rate of change. The rate of change It intuitively reflects the drastic extent to which the control system attempts to change the physical load state.

[0074] The system then proceeds to a judgment process to determine whether the rate of change exceeds a preset transient switching threshold. The transient switching threshold... This is a pre-calibrated critical value based on the system's dynamic characteristics. This threshold defines a boundary between the system's steady-state operating region and its transient switching region. During steady-state or slowly changing load adjustments, the rate of change... Typically small, less than or equal to the transient switching threshold. At this point, the inertial lag effect of the physical system is not significant, and the inertial influence can be ignored, allowing the static proportional calculation of the baseline value to be used directly. However, when the process requires the load to be transferred significantly within a very short time, the rate of change... It will spike instantly and exceed the transient switching threshold. .

[0075] If the rate of change Exceeding the transient switching threshold , the system determines that it is currently in the load dynamic switching phase. The triggering of this determination means that the system is immediately aware that at this moment, the electrical signal command has undergone a mutation, but the physical mechanical structure is definitely still in the old state or in the process of slowly accelerating. At this time, the huge position deviation or speed deviation fed back by the sensor is not largely a loss of control accuracy, but a reasonable delay given to the system by the laws of physics. Therefore, in order to prevent the calibration algorithm from making a false correction to this reasonable delay, the system immediately triggers the inertia lag compensation mechanism.

[0076] The first step of the inertia lag compensation mechanism is to quantify this physical lag. The system obtains a preset device moment of inertia coefficient. This coefficient is a physical parameter obtained by the system in the debugging stage through frequency response analysis or adaptive identification algorithm, which represents the ability of the servo motor rotor and the transmission chain and load connected to it to resist changes in motion state. Denote the device moment of inertia coefficient as . Then, the system performs a calculation step to calculate the product of the change rate and the moment of inertia coefficient as the theoretical inertia lag amount. Let the theoretical inertia lag amount be , and its calculation formula can be expressed as: ; In the above formula, the change rate represents the speed of change of the torque command, and the moment of inertia represents the difficulty of the system response, and the product of the two quantifies the state lag amplitude that the physical system will theoretically produce under the current drastic command change. This value is a directional vector, and its positive or negative sign depends on the trend of the load change.

[0077] The second step of the inertia lag compensation mechanism is to dynamically correct the reference value. When the mechanism is not triggered, the system only calculates an idealized static reference value based on the current static load distribution ratio, denoted as . This static reference assumes that the physical system is a zero-mass, infinitely fast response perfect model, so it will jump instantaneously with the jump of the load ratio. After triggering the compensation, the system superimposes the theoretical inertia lag amount on the load-weighted reference value calculated based on the static ratio to generate a dynamic inertia correction reference value. Denote the dynamic inertia correction reference value as , and the correction formula is: ; wherein is a dimensionless coefficient for mapping the dimension of the lag amount to the same dimension as the reference value, such as position or speed dimension.

[0078] Through this superimposition operation, the system artificially constructs a slow half-beat reference value. When the electrical signal command When the target point has been reached, the dynamic inertial correction reference value Due to the addition of hysteresis Its value was temporarily pulled towards the old state. Over time, the rate of load change... Gradually declining, lagging quantity It also decreases accordingly, eventually It will converge smoothly to This trajectory of change closely matches the response trajectory of a real physical mechanical system under inertia.

[0079] The final step of the inertial hysteresis compensation mechanism is to replace the load-weighted reference value with the dynamic inertial correction reference value as the basis for calculating the difference between the reference value and the local state data. When performing calibration parameter calculations, the system no longer forces the current servo device to chase that unattainable static reference. Instead, it requires it to follow this dynamic benchmark that fully considers physical inertia. The system calculates the real-time status data fed back by local sensors. With dynamic inertia correction reference value The difference between them is denoted as the effective calibration deviation. .

[0080] For example, due to dynamic benchmarks The initial design intent was to simulate the real response of a physical system; therefore, if the servo device is operating normally, its feedback data... Should with They are very close, thus making the calculated effective calibration deviation... This is maintained within a very small range. This means that by introducing inertial hysteresis compensation, the system successfully eliminates large, reasonable deviations caused by inertia from the calibration deviations. The system only performs calibration compensation for abnormal errors that deviate from the normal inertial trajectory and are truly caused by mechanical clearance, sudden changes in friction, or mismatched control parameters.

[0081] To further illustrate the beneficial effects of this mechanism, we can compare it with the case where it is not used. In the traditional method without inertia compensation, the system directly calculates... Compared with static reference The difference. At the moment of sudden load change, due to Delay, A sudden change occurs, with a significant difference between the two values. The calibration algorithm might mistakenly interpret this as a large positional loss and issue an excessively large acceleration command to compensate. However, when the equipment actually accelerates to catch up due to inertia, this extra compensation command can cause the equipment to overshoot, resulting in severe overshoot. This can lead to intense torque conflicts between multiple devices and even damage to mechanical transmission components.

[0082] In the method of the embodiment, the calibration instruction becomes extremely restrained and accurate. In the transient process of load switching, although the physical position lags, as long as the lag conforms to the inertia law, the system considers the device to be healthy and does not intervene or only applies a slight synchronization correction. Only when the actual lag degree of the device exceeds the allowed range of the theoretical inertia lag amount, the system will intervene. This intelligent calibration strategy enables the servo control system to maintain a smooth running state even in the face of high-frequency load switching conditions such as several times per second, ensuring both the stability of the dynamic process and the high-precision synchronization after the steady state, perfectly solving the control problem in the asynchronous cooperation scenario.

[0083] In summary, the embodiment accurately identifies the load dynamic switching stage by monitoring the load change rate in real time and setting a transient switching threshold. On this basis, by introducing the device rotational inertia coefficient to calculate the theoretical inertia lag amount and constructing a dynamic inertia correction reference value, the embodiment creatively realizes mathematical compensation for physical inertia lag. This mechanism fundamentally eliminates the time mismatch between electrical signal instructions and mechanical physical responses, ensuring the correctness and stability of the calibration algorithm in extreme dynamic conditions, and providing a solid technical guarantee for high-speed high-precision cooperative control of high-end servo equipment.

[0084] Embodiment Four

[0085] As shown in Figure 2 Based on the logical basis of the aforementioned method embodiments one, two and three, the embodiment elaborates in detail a servo control system autonomous intelligent calibration system based on multi-source information fusion from the perspective of system architecture and hardware deployment. The system aims to solve the conflict between sensor reliability and physical coupling interference in the multi-device linkage scenario through modular functional design and close signal flow mechanism, thereby realizing high-precision autonomous cooperation of the servo control system.

[0086] The servo control system autonomous intelligent calibration system based on multi-source information fusion provided by the embodiment deploys its core logical entity in the global edge controller of the multi-device linkage servo control system. As the central brain of the system, the global edge controller has high-performance parallel computing capability and compatible communication interfaces for industrial real-time buses such as EtherCAT or Profinet. From the perspective of functional module division, the system mainly includes data acquisition and association modeling module, conflict feature identification module, dynamic weight cooperation module and cooperative calibration execution module. These four modules interact with each other through shared memory or message queues, together forming a closed-loop intelligent calibration system.

[0087] Firstly, the data acquisition and correlation modeling module is the foundation of the system to perceive the physical world. This module is configured to actively initiate the scanning and data collection task of the physical system in response to system startup or calibration instructions. It reads the electronic nameplate or configuration file of each servo device through the industrial bus to obtain device attribute information. More importantly, this module integrates the master clock function of the IEEE 1588 precise time protocol to ensure that it can obtain multi-source sensor data synchronized through the global time protocol. This means that whether it is the built-in encoder data of the servo motor or the externally installed vibration sensor data, they are strictly aligned on the time axis, thereby eliminating the interference of timing deviation on subsequent correlation analysis.

[0088] On this basis, the module performs the core modeling task, i.e. according to the physical space distribution data and transmission connection relationship data in the device attribute information, a device correlation matrix representing the coupling strength between devices is constructed. This matrix is not just a mathematical array, it is a digital mapping of physical coupling relationship. The process of constructing the device correlation matrix representing the coupling strength between devices is a multi-dimensional quantization process. This module will evaluate the potential risk of spatial electromagnetic interference according to the physical distance value between devices, evaluate the transmission efficiency of mechanical vibration according to the quantization coefficient of the transmission connection mode, and evaluate the mutual influence on the dynamics level according to the load distribution proportion value. The module finally calculates the correlation coefficient between each pair of devices, and all the correlation coefficients constitute the device correlation matrix, providing prior knowledge about the mutual influence between devices for subsequent modules.

[0089] Secondly, the conflict feature recognition module undertakes the responsibility of signal analysis and pattern recognition. This module receives raw data from the upstream and pre-processes the multi-source sensor data to extract cross-device coupling conflict features. Here, the pre-processing includes filtering and denoising and data normalization, while feature extraction focuses on mining the physical laws hidden behind the data, such as the correlation of torque fluctuations or the spectral overlap of current harmonics. Subsequently, the module inputs the cross-device coupling conflict features into the pre-set conflict recognition model. This model is a deep neural network trained with a large number of samples, with strong non-linear mapping ability. Through the inference operation of the model, the module obtains the output conflict type label, clearly indicating whether the current system is facing a transmission coupling conflict, an electromagnetic coupling conflict or a load coupling conflict, thereby providing accurate decision basis for solving the pain point mentioned in the background technology that there is a lack of ability to distinguish local intrinsic error and cross-device coupling interference.

[0090] Again, the dynamic weight coordination module is the decision-making core of the system, which re-evaluates the data reliability according to the identification result. The module is used to call the corresponding fusion strategy according to the conflict type label and the current identified device linkage scene type. Its work is not isolated, but based on the device association matrix, the data fusion weight of each servo device is dynamically coordinated and adjusted. In the conventional conflict, the module will reduce the weight of the disturbed sensor; and in the strong coupling extreme working condition as described in embodiment two, the module will activate the coupling projection reconstruction logic to separate the intrinsic component from the mixed signal and improve its weight. In addition, the module also has the ability to call the global redundant resource pool, and introduce third-party sensing data when necessary. After a series of complex operations, the module finally generates a cross-device fusion feature. This feature eliminates external coupling noise and truly reflects the running state of each device, solving the root problem of mis-calibration.

[0091] Finally, the cooperative calibration execution module is the execution mechanism that converts data into control actions. The module is used to calculate the calibration parameter compensation value of each servo device based on the cross-device fusion feature and the preset cooperative calibration strategy. During the calculation process, the module will intelligently distinguish between synchronous transmission scenes and asynchronous cooperative scenes according to the logic described in embodiment three. Especially in the load fast switching stage of the asynchronous cooperative scene, the module will introduce an inertia lag compensation mechanism to shield the pseudo error caused by physical inertia by building a dynamic reference. After the calculation is completed, the module issues calibration instructions to each servo device to execute the multi-device synchronous calibration process. These instructions reach each servo driver in real time through the industrial bus, correcting the control parameters of its position loop or speed loop, thereby eliminating the deviation at the physical level.

[0092] In summary, the system of the present embodiment realizes the technical solutions in the method embodiment through the close cooperation of the above four modules. The data acquisition and association modeling module solves the digitization problem of physical relationships, the conflict feature recognition module solves the qualitative problem of interference sources, the dynamic weight coordination module solves the problem of signal authenticity, and the cooperative calibration execution module solves the problem of accurate generation of control instructions. This system architecture ensures that the control system can intelligently strip external interference, calibrate only the real intrinsic error of the device itself, and effectively cope with the inertia effect caused by load mutation, thereby achieving the beneficial effects of improving the overall precision, stability and robustness of the servo system.

[0093] Embodiment Five

[0094] Corresponding to the above-mentioned embodiments, the present application also provides an electronic device.

[0095] As Figure 3As shown is a structural schematic diagram of an electronic device in the present application, the electronic device 100 comprises a processor 101 and a memory 103. Wherein, the processor 101 and the memory 103 are connected, such as connected through a bus 102. Optionally, the electronic device 100 can further comprise a transceiver 104. It needs to be explained that the transceiver 104 is not limited to one in the actual application, and the structure of the electronic device 100 does not constitute a limitation to the embodiments of the present application.

[0096] The processor 101 can be a CPU, a general processor, a DSP, an ASIC, an FPGA or other programmable logic device, a transistor logic device, a hardware component or any combination thereof. It can realize or execute various exemplary logic blocks, modules and circuits described in combination with the disclosure. The processor 101 can also be a combination of computing functions, such as one or more microprocessor combinations, DSP and microprocessor combinations, etc.

[0097] The bus 102 can include a channel for transmitting information between the above-mentioned components. The bus 102 can be a PCI bus or an EISA bus, etc. The bus 102 can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, Figure 3 In the present application, only one thick line is used to represent the bus, but it does not mean that there is only one bus or only one type of bus.

[0098] The memory 103 is used to store a computer program corresponding to the above-mentioned embodiment of the present application, that is, a self-learning intelligent calibration method of a servo control system based on multi-source information fusion. The computer program is controlled and executed by the processor 101. The processor 101 is used to execute the computer program stored in the memory 103 to realize the content shown in the above-mentioned method embodiments.

[0099] Wherein, the electronic device 100 includes but is not limited to mobile terminals such as notebook computers, PADs (tablet computers) and the like, and fixed terminals such as desktop computers and the like. Figure 3 The electronic device 100 shown is only an example, which should not bring any limitation to the function and use range of the embodiments of the present application.

[0100] Although the embodiments of the present application have been shown and described above, it can be understood that the above-mentioned embodiments are exemplary and cannot be understood as a limitation to the present application. Those skilled in the art can make changes, modifications, replacements and variations to the above-mentioned embodiments within the scope of the present application.

Claims

1. An autonomous intelligent calibration method for a servo control system based on multi-source information fusion, characterized in that, Includes the following steps: In response to system startup or calibration commands, acquire device attribute information of each servo device and multi-source sensor data synchronized via a global time protocol; Based on the physical spatial distribution data and transmission connection relationship data in the equipment attribute information, a device association matrix representing the coupling strength between devices is constructed. The multi-source sensor data is preprocessed to extract cross-device coupling conflict features, and the cross-device coupling conflict features are input into a preset conflict identification model to obtain the output conflict type label. Based on the conflict type label and the currently identified device linkage scenario type, the corresponding fusion strategy is invoked, and the data fusion weights of each servo device are dynamically and collaboratively adjusted based on the device association matrix to generate cross-device fusion features; Based on the cross-device fusion feature and the preset collaborative calibration strategy, the calibration parameter compensation value of each servo device is calculated, and calibration instructions are sent to each servo device to execute the multi-device synchronous calibration process. The process of constructing the device association matrix that characterizes the coupling strength between devices includes: calculating the association coefficient between each pair of devices based on the physical distance between devices, the quantification coefficient of the transmission connection method, and the load distribution ratio, and constructing the device association matrix from all the association coefficients.

2. The method according to claim 1, characterized in that, The calculation of the correlation coefficient between each pair of devices includes: calculating the device correlation coefficient using the following formula. With equipment correlation coefficient between : ； in, Characterization equipment With equipment The physical distance between them; The transmission link coefficient is characterized by taking a preset first value, second value, or zero value depending on whether it is a direct transmission, indirect transmission, or no transmission relationship. Characterizing the load ratio of equipment; , , The weighting coefficients are preset, and the correlation coefficients are... The range of values ​​is normalized to a closed interval between zero and one.

3. The method according to claim 1, characterized in that, The step of inputting the cross-device coupling conflict features into a preset conflict identification model to obtain the output conflict type label includes: The extracted cross-device coupling conflict features and the device association matrix are used as input vectors and input into the conflict recognition model, which includes a convolutional neural network layer and a self-attention mechanism layer. Obtain the probability distribution values ​​of various types of conflicts output by the conflict identification model, and determine whether the highest probability value is greater than the preset conflict confidence threshold. If so, the category corresponding to the highest probability value is determined as the conflict type label; The cross-device coupling conflict characteristics include at least: transmission coupling conflict characteristics characterizing the mutual influence of mechanical transmissions, electromagnetic coupling conflict characteristics characterizing electrical signal interference, and load coupling conflict characteristics characterizing load distribution deviation.

4. The method according to claim 3, characterized in that, The specific process for generating the cross-device coupling conflict feature includes: The Pearson correlation coefficient of torque fluctuation between device i and device j is calculated and used as the torque coupling coefficient in the transmission coupling conflict characteristics. The percentage of overlapping intervals of the same harmonic frequencies between the calculated devices is used as the current harmonic spectrum overlap degree in the electromagnetic coupling conflict characteristics. The difference between the actual load distribution ratio and the preset ratio is calculated as the load deviation rate in the load coupling conflict characteristics.

5. The method according to claim 1, characterized in that, The dynamic and coordinated adjustment of the data fusion weights of each servo device based on the device association matrix includes: Based on the conflict type label, a preset weight adjustment rule library is retrieved to determine the local basic weight adjustment strategy corresponding to the affected device; Based on the correlation coefficient in the device correlation matrix, the weight adjustment range of the affected devices is calculated, and the weight adjustment range is transmitted to the associated devices for coordinated adjustment. If the conflict type label is transmission coupling conflict, then the weight attenuation factor is calculated based on the correlation coefficient, and the weight ratio of the affected sensor in the fusion calculation is reduced using the weight attenuation factor.

6. The method according to claim 1, characterized in that, Before performing the multi-device synchronous calibration process, the following is also included: Read the local redundant sensor information of each servo device and construct a global redundant resource pool. The global redundant resource pool records the sensor type, accuracy parameters and current occupancy status. The cross-device coupling conflict characteristics within a historical time window are analyzed using a long short-term memory network and attention mechanism model to predict the probability of conflict occurrence in future time windows. If the predicted probability of conflict exceeds a preset intervention probability threshold, then based on the device association matrix, the readings of idle cross-device redundant sensors with an association coefficient less than a preset safety threshold are scheduled from the global redundant resource pool to intervene in the data fusion process.

7. The method according to claim 1, characterized in that, The calculation of calibration parameter compensation values ​​for each servo device based on the cross-device fusion features and the preset collaborative calibration strategy includes: Identify the current device interaction scenario type; If the device linkage scenario type is a synchronous transmission scenario, then the device with the largest correlation coefficient value in the device correlation matrix is ​​selected as the reference device. Based on the difference between the status data of the reference device and the local status data, the calibration parameters of other devices are calculated by weighting the correlation coefficient. If the device linkage scenario type is an asynchronous collaborative scenario, then the load weighting benchmark value is calculated based on the load distribution ratio, and the calibration parameters are calculated based on the difference between the load weighting benchmark value and the local status data, combined with the load coupling coefficient.

8. The method according to claim 1, characterized in that, After performing the multi-device synchronous calibration process, the following is also included: Collect calibrated multi-device linkage data and calculate closed-loop verification indicators, including cross-device coordination error and coupling conflict resolution rate. Determine whether the closed-loop verification index meets the preset calibration qualification standard; If the conditions are not met, the single-device feature extraction layer parameters in the conflict identification model remain unchanged, and the cross-device coupled feature extraction layer parameters are updated and iterated until the closed-loop verification index meets the calibration qualification standard.

9. The method according to claim 5, characterized in that, If the conflict type label is a transmission coupling conflict, before calculating the weight attenuation factor based on the correlation coefficient, an anomaly handling step for strongly coupled scenarios is also included: Determine whether the correlation coefficient between devices in the device association matrix is ​​greater than a preset strong coupling threshold; If so, then pause the step of adjusting the weight ratio of the affected sensor using the weight attenuation factor and start the coupling projection reconstruction process. The coupled projection reconstruction process includes: acquiring real-time status data of associated devices, combining the correlation coefficient and the calculated mechanical transmission delay time to calculate the coupling interference projection amount of the associated devices on the current device; subtracting the coupling interference projection amount from the original sensor data of the current device to extract the local intrinsic motion components characterizing the actual motion state of the current device; calculating the signal-to-noise ratio of the local intrinsic motion components, and using the signal-to-noise ratio as a positive correction factor to adjust the gain of the local base weights of the current device.

10. The method according to claim 7, characterized in that, If the device linkage scenario type is an asynchronous collaborative scenario, the calculation process of its calibration parameters also includes an inertial hysteresis compensation step for rapid load switching: Before calculating the load-weighted baseline value, monitor the rate of change of the load distribution ratio of each servo device over time in real time. Determine whether the rate of change exceeds a preset transient switching threshold; If so, it is determined that the current state is in the dynamic load switching phase, triggering the inertial lag compensation mechanism; The inertial hysteresis compensation mechanism includes: obtaining a preset equipment rotational inertia coefficient; calculating the theoretical inertial hysteresis by multiplying the rate of change by the rotational inertia coefficient; superimposing the theoretical inertial hysteresis onto the load-weighted reference value calculated based on a static ratio to generate a dynamic inertial correction reference value; and using the dynamic inertial correction reference value to replace the load-weighted reference value as the basis for calculating the difference between the data and the local state data.

Images