Motor instantaneous torque measurement method and system based on high-bandwidth sensor

Abstract

The application relates to the technical field of motor detection, and discloses a motor instantaneous torque measurement method and system based on a high-bandwidth sensor, the method comprising the following steps: synchronously collecting torque signals and multi-source interference quantities through a high-bandwidth sensor, and forming an original torque sequence after time synchronization superposition; extracting disturbance frequency characteristics through multi-scale wavelet transformation and spectrum analysis; generating filter coefficients matched with working conditions by using a pre-trained adaptive filter model, and completing primary filtering; combining Kalman state estimation to suppress residual fluctuations, and obtaining an optimized torque estimation value; realizing parameter adaptive compensation through online calibration weight updating, and outputting a stable torque measurement sequence; realizing load mutation prediction and threshold dynamic updating based on time-frequency correlation analysis, and finally ensuring measurement rapid convergence after working condition mutation through a closed-loop reset mechanism. The method can solve the problem of low instantaneous torque measurement precision in the prior art.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of motor detection, and in particular to a motor instantaneous torque measurement method and system based on a high-bandwidth sensor. BACKGROUND

[0002] At present, motor instantaneous torque measurement is a basic link of electric drive system state monitoring and closed-loop control, and the output precision directly affects the speed stability, energy consumption evaluation and mechanical component life management. In new energy vehicle, industrial servo and robot scenes, the motor is often in acceleration, deceleration or load mutation working condition, and the torque signal needs to reflect the real change within milliseconds of time, so that the controller can make torque compensation or protection decision in time.

[0003] In one prior art, a strain type torque sensor is usually used to collect original signals with a fixed sampling frequency, and denoising is realized through static calibration and simple low-pass filtering. For example, the system samples at a constant 1kHz, compares the torque reading with the factory calibration table, and directly outputs the filtering result; when vibration or current mutation is detected, only the filtering order is increased or an alarm is triggered, without separating and dynamically compensating the disturbance component. In addition, the calibration parameters are not updated once set, and cannot be adaptively adjusted following the load change or temperature drift, resulting in obvious measurement deviation during high-speed commutation or load mutation period.

[0004] Due to the lack of online identification of disturbance frequency and dynamic adjustment of filtering coefficient in the existing method, high-frequency noise and mechanical vibration are partially retained in the measurement result; at the same time, the static calibration parameters are still used after the working condition changes, so that the system error between the torque estimated value and the true value gradually expands. Therefore, the prior art has the problem of low instantaneous torque measurement accuracy. SUMMARY

[0005] The present application provides a motor instantaneous torque measurement method and system based on a high-bandwidth sensor to solve the problem of low instantaneous torque measurement accuracy in the prior art.

[0006] In a first aspect, to solve the above technical problems, the present application provides a motor instantaneous torque measurement method based on a high-bandwidth sensor, comprising:

[0007] Obtaining a torque signal and an interference quantity, and performing time synchronization operation on the torque signal and the interference quantity to obtain an original torque sequence containing superimposed interference, and performing multi-scale wavelet transform and spectrum analysis on the original torque sequence containing superimposed interference to obtain disturbance frequency characteristics;

[0008] input the disturbance frequency feature into a pre-trained adaptive filter model to obtain a filter coefficient corresponding to the current working condition, and perform high-frequency disturbance filtering on the original torque sequence according to the filter coefficient corresponding to the current working condition to obtain a primary filtered torque signal;

[0009] If the time-domain fluctuation amplitude of the primary filtered torque signal exceeds a preset fluctuation threshold, Kalman state estimation is performed according to the primary filtered torque signal to obtain an optimized torque estimation value;

[0010] According to the optimized torque estimation value, the calibration weight is updated to obtain an adaptive calibration parameter, and the subsequent collected torque signal is compensated in real time according to the adaptive calibration parameter to obtain a stable torque measurement sequence;

[0011] According to the stable torque measurement sequence and the disturbance frequency feature, an association comparison is performed to obtain a load mutation precursor sign;

[0012] According to the load mutation precursor sign, the interference source is classified, and the preset mutation detection threshold is updated to obtain an optimized reset judgment basis;

[0013] According to the stable torque measurement sequence, load mutation is determined, and if load mutation occurs, the optimized reset judgment basis is used to trigger a parameter reset mechanism and re-filtering and compensation to output a final accurate torque.

[0014] Preferably, the torque signal and the interference quantity are obtained, and time synchronization operation is performed on the torque signal and the interference quantity to obtain an original torque sequence containing superimposed interference, and multi-scale wavelet transform and spectrum analysis are performed on the original torque sequence containing superimposed interference to obtain a disturbance frequency feature, including:

[0015] The torque signal and the interference quantity are obtained;

[0016] The network time protocol synchronization operation is performed on the torque signal and the interference quantity to obtain the time-synchronized torque signal and the disturbance quantity;

[0017] The time-synchronized torque signal and the disturbance quantity are superimposed to obtain an original torque sequence containing superimposed interference;

[0018] The multi-scale wavelet transform is performed on the original torque sequence containing superimposed interference to obtain a high-frequency disturbance component;

[0019] The fast Fourier transform is performed on the high-frequency disturbance component to extract the main disturbance frequency component to obtain the disturbance frequency feature.

[0020] Preferably, the disturbance frequency feature is input into a pre-trained adaptive filter model to obtain filter coefficients corresponding to the current working condition, and the original torque sequence is filtered of high-frequency disturbances according to the filter coefficients corresponding to the current working condition to obtain a primary filtered torque signal, including:

[0021] According to the disturbance frequency feature, an adaptive filter model loading operation is performed to obtain an adaptive filter model matched with the current working condition;

[0022] The disturbance frequency feature is input into the adaptive filter model to perform a filter coefficient generation operation to obtain filter coefficients corresponding to the current working condition;

[0023] A digital band-stop filter is constructed according to the filter coefficients corresponding to the current working condition, and a high-frequency disturbance filtering operation is performed on the original torque sequence according to the digital band-stop filter to obtain a primary filtered torque signal.

[0024] Preferably, if the time-domain fluctuation amplitude of the primary filtered torque signal exceeds a preset fluctuation threshold, Kalman state estimation is performed according to the primary filtered torque signal to obtain an optimized torque estimation value, including:

[0025] A time-domain fluctuation amplitude calculation operation is performed according to the primary filtered torque signal to obtain a current fluctuation amplitude;

[0026] If the current fluctuation amplitude is greater than the preset fluctuation threshold, a state equation and an observation equation are established with the primary filtered torque signal as an observation vector, and a Kalman filter recursive estimation operation is performed to obtain an optimized torque estimation value.

[0027] Preferably, calibration weight updating is performed according to the optimized torque estimation value to obtain adaptive calibration parameters, and real-time compensation is performed on a torque signal collected subsequently according to the adaptive calibration parameters to obtain a stable torque measurement sequence, including:

[0028] An initial calibration parameter is obtained;

[0029] A deviation quantization operation is performed according to the optimized torque estimation value and the initial calibration parameter to obtain an estimated deviation;

[0030] A gradient descent weight updating operation is performed according to the estimated deviation to obtain an updated calibration weight matrix;

[0031] An adaptive calibration parameter generation operation is performed according to the updated calibration weight matrix to obtain adaptive calibration parameters;

[0032] Real-time compensation is performed on a torque signal collected subsequently according to the adaptive calibration parameters to obtain a stable torque measurement sequence.

[0033] Preferably, according to the stable torque measurement sequence and the disturbance frequency feature, a correlation comparison is performed to obtain a load mutation premonition sign, including:

[0034] According to the stable torque measurement sequence, a time-domain trend extraction operation is performed to obtain a low-frequency torque trend;

[0035] According to the disturbance frequency feature, a frequency-domain energy calculation operation is performed to obtain a high-frequency energy proportion;

[0036] A correlation comparison operation is performed on the low-frequency torque trend and the high-frequency energy proportion, and if the high-frequency energy proportion rises and the low-frequency torque trend deviates, a load mutation premonition sign is generated.

[0037] Preferably, according to the load mutation premonition sign, an interference source classification is performed, and a preset mutation detection threshold is updated to obtain an optimized reset judgment basis, including:

[0038] According to the load mutation premonition sign, an interference source feature extraction operation is performed to obtain a mutation frequency component;

[0039] According to the mutation frequency component, a clustering classification operation is performed to obtain a service interference type result;

[0040] If the service interference type result indicates a load mutation, a dynamic update operation on the preset mutation detection threshold is performed to obtain an updated mutation detection threshold;

[0041] According to the updated mutation detection threshold, a reset judgment basis generation operation is performed to obtain an optimized reset judgment basis.

[0042] Preferably, according to the stable torque measurement sequence, a load mutation determination is performed, and if a load mutation occurs, according to the optimized reset judgment basis, a parameter reset mechanism is triggered and re-filtering and compensation are performed to output a final accurate torque, including:

[0043] According to the stable torque measurement sequence, a mutation determination operation is performed to compare a sequence fluctuation amplitude with the optimized reset judgment basis;

[0044] If the fluctuation amplitude exceeds the optimized reset judgment basis, a parameter reset mechanism is triggered, a filtering and compensation process is re-executed, and a torque measurement sequence obtained after re-filtering and compensation is output as a final accurate torque.

[0045] In a second aspect, the present application provides a motor instantaneous torque measurement system based on a high-bandwidth sensor, including:

[0046] The data acquisition preprocessing module is configured to acquire a torque signal and an interference quantity, perform time synchronization operation on the torque signal and the interference quantity, obtain an original torque sequence containing superimposed interference, perform multi-scale wavelet transform and spectrum analysis on the original torque sequence containing superimposed interference, and obtain a disturbance frequency feature.

[0047] The adaptive filtering processing module is configured to input the disturbance frequency feature into a pre-trained adaptive filter model, obtain a filter coefficient corresponding to a current working condition, and filter out high-frequency disturbance from the original torque sequence according to the filter coefficient corresponding to the current working condition, to obtain a primary filtered torque signal.

[0048] The state estimation optimization module is configured to, if a time-domain fluctuation amplitude of the primary filtered torque signal exceeds a preset fluctuation threshold, perform Kalman state estimation according to the primary filtered torque signal, to obtain an optimized torque estimation value.

[0049] The online calibration compensation module is configured to perform calibration weight updating according to the optimized torque estimation value, obtain adaptive calibration parameters, and perform real-time compensation on a torque signal collected subsequently according to the adaptive calibration parameters, to obtain a stable torque measurement sequence.

[0050] The mutation precursor detection module is configured to perform correlation comparison according to the stable torque measurement sequence and the disturbance frequency feature, to obtain a load mutation precursor flag.

[0051] The threshold dynamic updating module is configured to perform interference source classification according to the load mutation precursor flag, and update a preset mutation detection threshold, to obtain an optimized reset judgment basis.

[0052] The closed-loop reset output module is configured to perform load mutation determination according to the stable torque measurement sequence, and if a load mutation occurs, trigger a parameter reset mechanism and perform re-filtering and compensation according to the optimized reset judgment basis, to output a final accurate torque.

[0053] In a third aspect, the present application further provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the motor instantaneous torque measurement method based on a high-bandwidth sensor according to any one of the above aspects when executing the computer program.

[0054] In a fourth aspect, the present application further provides a computer readable storage medium, comprising a stored computer program, wherein the computer readable storage medium controls a device where the computer readable storage medium is located to execute the motor instantaneous torque measurement method based on a high-bandwidth sensor according to any one of the above aspects when the computer program runs.

[0055] Compared with the prior art, the present application has the following beneficial effects:

[0056] (1) The present application can extract disturbance frequency characteristics representing the current working condition by time synchronization and superposition of torque signals and interference quantities, and performing multi-scale wavelet transform and spectrum analysis. This method helps to separate effective interference information from complex background noise, providing a data basis for subsequent targeted filtering processing.

[0057] (2) The present application inputs the disturbance frequency characteristics into a pre-trained adaptive filter model to generate filter coefficients corresponding to the current working condition, and accordingly constructs a digital band-stop filter to filter out high-frequency disturbances. This method can match the filtering characteristics with real-time interference characteristics, thereby improving the quality of the primary filtered torque signal.

[0058] (3) The present application starts Kalman state estimation when the fluctuation of the primary filtered torque signal exceeds a preset threshold, obtaining an optimized torque estimation value. This step helps to suppress residual random fluctuations in the signal, improving the smoothness and stability of the torque estimation value.

[0059] (4) The present application updates the calibration weight based on the optimized torque estimation value, and generates adaptive calibration parameters for real-time compensation of subsequent torque signals. This mechanism can adapt to the slow drift of system parameters, helping to maintain the consistency of long-term measurement results.

[0060] (5) The present application generates a load mutation precursor sign by correlating and comparing the time domain trend of the stable torque measurement sequence with the frequency domain energy of the disturbance frequency characteristics. This design helps to identify potential working condition changes early, providing early warning for system adjustment.

[0061] (6) The present application performs interference source classification and threshold update operation according to the load mutation precursor sign, obtaining an optimized reset judgment basis. This process enables the system to distinguish between interference types and dynamically adjust the judgment criteria, improving the recognition ability of abnormal working conditions.

[0062] (7) The present application triggers the parameter reset mechanism and re-executes the filtering and compensation process after determining that the load mutation has occurred. This closed-loop design enables the system to quickly restore measurement accuracy after a significant change in working condition, enhancing the robustness and adaptability of the entire method. BRIEF DESCRIPTION OF DRAWINGS

[0063] Figure 1 is the flowchart of the motor instantaneous torque measurement method based on a high-bandwidth sensor provided by the first embodiment of the present application;

[0064] Figure 2 is the structural schematic diagram of the motor instantaneous torque measurement system based on a high-bandwidth sensor provided by the second embodiment of the present application. DETAILED DESCRIPTION

[0065] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0066] With reference to Figure 1 The first embodiment of the present application provides a motor instantaneous torque measurement method based on a high-bandwidth sensor, comprising the following steps:

[0067] S11, acquiring a torque signal and an interference quantity, and performing time synchronization operation on the torque signal and the interference quantity to obtain an original torque sequence containing superimposed interference, and performing multi-scale wavelet transform and spectrum analysis on the original torque sequence containing superimposed interference to obtain a disturbance frequency feature;

[0068] S12, inputting the disturbance frequency feature into a pre-trained adaptive filter model to obtain a filter coefficient corresponding to a current working condition, and filtering out high-frequency disturbance from the original torque sequence according to the filter coefficient corresponding to the current working condition to obtain a primary filtered torque signal;

[0069] S13, if the time-domain fluctuation amplitude of the primary filtered torque signal exceeds a preset fluctuation threshold, performing Kalman state estimation according to the primary filtered torque signal to obtain an optimized torque estimation value;

[0070] S14, performing calibration weight updating according to the optimized torque estimation value to obtain adaptive calibration parameters, and performing real-time compensation on a torque signal collected subsequently according to the adaptive calibration parameters to obtain a stable torque measurement sequence;

[0071] S15, performing correlation comparison according to the stable torque measurement sequence and the disturbance frequency feature to obtain a load mutation precursor sign;

[0072] S16, performing interference source classification according to the load mutation precursor sign, and updating a preset mutation detection threshold to obtain an optimized reset judgment basis;

[0073] S17, performing load mutation determination according to the stable torque measurement sequence, and if a load mutation occurs, triggering a parameter reset mechanism and re-filtering and compensation according to the optimized reset judgment basis to output a final accurate torque.

[0074] In step S11, the torque signal and the interference quantity are acquired, and time synchronization operation is performed on the torque signal and the interference quantity to obtain a raw torque sequence containing superimposed interference, and multi-scale wavelet transform and spectrum analysis are performed on the raw torque sequence containing superimposed interference to obtain disturbance frequency characteristics, including:

[0075] acquiring a torque signal and an interference quantity;

[0076] performing network time protocol synchronization operation on the torque signal and the interference quantity to obtain a time-synchronized torque signal and a disturbance quantity;

[0077] superimposing the time-synchronized torque signal and the disturbance quantity to obtain a raw torque sequence containing superimposed interference;

[0078] performing multi-scale wavelet transform on the raw torque sequence containing superimposed interference to obtain a high-frequency disturbance component;

[0079] performing fast Fourier transform on the high-frequency disturbance component to extract a main disturbance frequency component to obtain disturbance frequency characteristics.

[0080] This step provides a synchronous, complete and feature-specific raw torque-disturbance quantity data set for subsequent filtering through high-bandwidth acquisition and multi-scale decomposition, and is specifically implemented as follows:

[0081] First, a torque signal and an interference quantity are acquired, a strain high-bandwidth torque sensor is installed at a motor output shaft flange, a range covering 120% of a rated torque is covered, and a differential analog voltage is output; a three-axis micro-vibration accelerometer and a Hall current sensor are arranged at the same mechanical interface to record mechanical vibration and electromagnetic disturbance, respectively, and the three constitute an interference quantity acquisition channel, and all sensors share the same power bus to avoid ground potential drift.

[0082] Subsequently, network time protocol synchronization operation is performed on the torque signal and the interference quantity, sensor analog output is synchronously sampled by a 24-bit ADC, a sampling clock is distributed by a gateway PTP master clock, a timestamp resolution is 1 ms, and after the ADC buffer area is filled, PTP timestamps are immediately stamped on the torque channel and the interference quantity channel, respectively, to obtain a time-synchronized torque signal and a disturbance quantity.

[0083] The torque signal synchronized in time and the interference quantity are added in equal amounts according to the sampling sequence number, and the interference quantity is normalized to the torque dimension before addition to avoid dimension conflict, and an original torque sequence containing superimposed interference is generated. For example, the output of the vibration sensor is acceleration value, with a unit of m / s². The length of the force arm from the measuring point to the flange is obtained through field calibration, and the acceleration is multiplied by the mass and the force arm to convert it into an equivalent torque component, with a unit of N·m. The output of the current sensor is A, which is converted into an equivalent electromagnetic disturbance torque through the motor torque constant, with a unit of N·m. After the above conversion, the two equivalent torque components are added to the measured value of the torque sensor to form an original torque sequence containing superimposed interference with unified dimensions.

[0084] The original torque sequence containing superimposed interference is subjected to multi-scale wavelet transform, a Daubechies-4 wavelet basis is selected, and the third layer detail coefficient is obtained as a high-frequency disturbance component after three layers of decomposition. The approximation coefficient is retained but does not participate in subsequent spectrum analysis.

[0085] Finally, the high-frequency disturbance component is subjected to fast Fourier transform, the FFT length is an integer power of 2 and covers the latest 10s of data, the amplitude spectrum is obtained, the frequency points with an amplitude higher than 1.5 times the spectral mean value are determined as the main disturbance frequency components, the disturbance frequency characteristics are formed, and are used for subsequent adaptive filter coefficient generation. It is worth noting that the FFT length is the latest 10s of data, corresponding to a 0.1Hz frequency resolution, which can completely cover the common disturbance frequency band of the motor and mechanical transmission. The FFT length is an integer power of 2, which facilitates the implementation of the base two algorithm and reduces the calculation amount of zero padding. The selection basis for determining the frequency points with an amplitude higher than 1.5 times the spectral mean value as the main disturbance frequency components is that the amplitude higher than 1.5 times the spectral mean value is about 75% of the historical rain peak sample amplitude distribution, which can retain significant disturbance components without introducing too many false frequencies to form the disturbance frequency characteristics.

[0086] In step S12, the disturbance frequency characteristics are input into the pre-trained adaptive filter model to obtain the filter coefficients corresponding to the current working condition, and the original torque sequence is subjected to high-frequency disturbance filtering according to the filter coefficients corresponding to the current working condition to obtain a primary filtered torque signal, including:

[0087] According to the disturbance frequency characteristics, an adaptive filter model loading operation is performed to obtain an adaptive filter model matched with the current working condition;

[0088] The disturbance frequency characteristics are input into the adaptive filter model to perform a filter coefficient generation operation to obtain the filter coefficients corresponding to the current working condition;

[0089] According to the filter coefficient corresponding to the current working condition, a digital band-stop filter is constructed, and a high-frequency disturbance filtering operation is performed on the original torque sequence according to the digital band-stop filter to obtain a primary filtered torque signal.

[0090] This step converts the disturbance frequency characteristics into an executable digital filter, and completes the initial removal of high-frequency disturbances. The process is as follows:

[0091] First, the adaptive filter model loading operation is performed. The edge computing unit retrieves the corresponding entry in the locally stored filter coefficient table according to the main frequency value in the disturbance frequency characteristics. The table is obtained by offline training of historical rain peak period samples. The historical rain peak period refers to the period when the motor is running, and the multi-source disturbance such as mechanical vibration and electromagnetic noise reaches the peak level. For example, when a new energy vehicle accelerates rapidly, high-frequency vibration occurs in the motor, and high-frequency disturbance is significantly enhanced, showing concentrated and high-intensity characteristics similar to rain peaks. The table entries are indexed by the main frequency and store finite impulse response coefficients. After loading, the adaptive filter model matched with the current working condition is obtained. It is worth noting that the input layer of the adaptive filter model receives a single-channel disturbance frequency characteristic vector, and then two cascaded finite impulse response sub-filters are connected. Each sub-filter has a fixed length of 64 points, uses a rectangular window function and a linear phase structure, and is used to extract local frequency band characteristics. After the sub-filter, a summation fusion layer and two fully connected layers are connected. The number of nodes in the fully connected layer is set to 32, and the ReLU activation function is used. Finally, the filter coefficient vector corresponding to the current working condition is output. The training of this model is supervised learning. The training data comes from historical disturbance-filter samples of multiple known motor operating conditions. The historical disturbance frequency characteristics are used as input, and the filter coefficient vector representing the optimal filtering effect verified by expert evaluation in the same period is used as target output. The mean square error is used as the loss function, and the Adam optimizer is used for iterative training on a large amount of data until the model converges. The trained model can predict the filter coefficient corresponding to the current working condition from the disturbance frequency characteristics of a new and unknown working condition.

[0092] Subsequently, the filter coefficient generation operation is performed. The disturbance frequency characteristics are input into the adaptive filter model, and the model outputs the filter coefficient corresponding to the current working condition. The coefficient length is fixed at 64 points, and the dimension is 1, which is independent of the sampling frequency and can be directly used to construct a digital band-stop filter.

[0093] Next, a digital band-stop filter is constructed with the filter coefficient corresponding to the current working condition as the core. The center frequency is set to equal to the main frequency of the disturbance frequency characteristics, and the bandwidth covers the main frequency ± one frequency step. The step size is determined by the wavelet decomposition band width, forming a finite impulse response band-stop filter.

[0094] Finally, the high-frequency disturbance filtering operation is performed, and the original torque sequence is convolved with a digital band-stop filter. The convolution result is a primary filtered torque signal. The signal retains the low-frequency trend and suppresses the main frequency and the high-frequency disturbance in the adjacent frequency band, thereby providing a smooth input for subsequent Kalman state estimation.

[0095] In step S13, if the time-domain fluctuation amplitude of the primary filtered torque signal exceeds the preset fluctuation threshold, Kalman state estimation is performed according to the primary filtered torque signal to obtain an optimized torque estimation value, including:

[0096] According to the primary filtered torque signal, a time-domain fluctuation amplitude calculation operation is performed to obtain a current fluctuation amplitude.

[0097] If the current fluctuation amplitude is greater than the preset fluctuation threshold, the primary filtered torque signal is taken as an observation vector, state equations and observation equations are established, and a Kalman filter recursive estimation operation is performed to obtain an optimized torque estimation value.

[0098] This step performs Kalman state estimation after primary filtering, which is used to suppress residual fluctuations and output a smooth torque trajectory. The process is as follows:

[0099] First, a time-domain fluctuation amplitude calculation operation is performed. The primary filtered torque signal is taken as a sliding window of the latest N seconds, the sample standard deviation is calculated, and the current fluctuation amplitude is obtained. The dimension is consistent with the torque. N is determined by the autocorrelation time of the historical steady-state sample, and is set to 10 s by default, taking into account the calculation load and response sensitivity.

[0100] Subsequently, threshold comparison is performed. If the current fluctuation amplitude is greater than the preset fluctuation threshold, the Kalman filter recursive estimation operation is triggered. The preset fluctuation threshold is set by multiplying the 75% quantile of the standard deviation of the historical steady-state segment of the same motor by an empirical proportionality coefficient. The dimension is the same as the standard deviation, and is used to distinguish normal disturbance from abnormal fluctuation.

[0101] After triggering, the primary filtered torque signal is taken as an observation vector, and state equations and observation equations are established. The state equation is set as:

[0102]

[0103] The observation equation is set as:

[0104]

[0105] wherein is the current true torque state, is the true torque state at the previous moment, is the observation value, and respectively, both covariance are qualitatively set according to historical steady Allan variance, dimension is torque square, ensure the filter both follow the trend and suppress the burst peak.

[0106] The Kalman filter recursive estimation operation is executed in a prediction-update two-step cycle, and an optimized torque estimation value is output, which is used as a reference for subsequent calibration and compensation, and the dimension is consistent with the input torque.

[0107] In step S14, according to the optimized torque estimation value, the calibration weight update is performed to obtain the adaptive calibration parameter, and the subsequent collected torque signal is compensated in real time according to the adaptive calibration parameter to obtain a stable torque measurement sequence, including:

[0108] Obtain the initial calibration parameter;

[0109] According to the optimized torque estimation value and the initial calibration parameter, a bias quantization operation is performed to obtain an estimated bias;

[0110] According to the estimated bias, a gradient descent weight update operation is performed to obtain an updated calibration weight matrix;

[0111] According to the updated calibration weight matrix, an adaptive calibration parameter generation operation is performed to obtain an adaptive calibration parameter;

[0112] According to the adaptive calibration parameter, a real-time compensation operation is performed on the subsequent collected torque signal to obtain a stable torque measurement sequence.

[0113] This step uses the optimized torque estimation value to update the calibration weight online, so that the subsequent collected signal remains consistent in dimension and trend in real-time compensation. The specific steps are as follows:

[0114] First, obtain the initial calibration parameter, which is provided by the motor factory calibration data and stored in the local non-volatile memory as the weight update reference vector.

[0115] Then, the bias quantization operation is performed to obtain the estimated bias by subtracting the corresponding elements of the optimized torque estimation value and the initial calibration parameter, and accumulating in the form of Euclidean distance, which is consistent with the dimension of torque and is used to measure the degree of deviation of the current working condition from the calibration working condition.

[0116] Then, a gradient descent weight update operation is performed to estimate the bias as a loss function, perform a gradient descent on the calibration weight matrix, the step size is determined by the median value of the bias descent rate of historical samples, the dimension is 1, and the iteration is stopped until the change ratio of adjacent two biases is less than a set ratio (such as 1%), to obtain an updated calibration weight matrix, and avoid infinite loop. The calibration weight matrix refers to a linear transformation matrix used to map the original measurement value of the sensor to the standard torque dimension. The dimension of the matrix matches the number of sensor channels and the number of calibration parameters, and the initial value is determined by the multi-condition calibration experiment when the motor is shipped. When performing the gradient descent weight update operation, the system first reads the current value of the matrix from the non-volatile memory as the optimization starting point. The physical meaning of the calibration weight matrix is to establish a linear relationship model between the sensor output and the real torque.

[0117] Then, an adaptive calibration parameter generation operation is performed, principal component decomposition is performed on the updated calibration weight matrix, the eigenvector corresponding to the maximum eigenvalue is taken, the maximum value is normalized to form the adaptive calibration parameter, and the dimension is kept as 1, so that the parameter vector can be directly used for multiplicative compensation.

[0118] Finally, a real-time compensation operation is performed: the adaptive calibration parameter is multiplied by the corresponding elements of the subsequently collected torque signal, and the product is the stable torque measurement sequence, which has the same dimension as the input torque.

[0119] In step S15, according to the stable torque measurement sequence and the disturbance frequency feature, a correlation comparison is performed to obtain a load mutation precursor flag, including:

[0120] According to the stable torque measurement sequence, a time domain trend extraction operation is performed to obtain a low-frequency torque trend;

[0121] According to the disturbance frequency feature, a frequency domain energy calculation operation is performed to obtain a high-frequency energy proportion;

[0122] The correlation comparison operation is performed on the low-frequency torque trend and the high-frequency energy proportion, and if the high-frequency energy proportion rises and the low-frequency torque trend deviates, a load mutation precursor flag is generated.

[0123] This step synchronously monitors the time domain trend and the frequency domain energy to form the load mutation precursor flag, which provides a trigger basis for subsequent threshold updating and parameter resetting, and the process is as follows:

[0124] First, a time domain trend extraction operation is performed, a sliding window of the last N seconds is taken for the stable torque measurement sequence, N is determined by the autocorrelation half-life period of the historical steady state segment, and 10s is taken by default; linear fitting is performed in the window, the slope sign represents the trend direction, and the slope absolute value represents the trend strength, which together constitute the low-frequency torque trend, with dimensions of 1 and N·m / s, respectively, for capturing slow drift.

[0125] Subsequently, a frequency domain energy calculation operation is performed on the amplitude spectrum in the disturbance frequency feature, and an energy sum of a frequency band above 20 Hz is divided by a total frequency band energy sum to obtain a high frequency energy proportion, which is used to quantify high frequency disturbance intensity changes.

[0126] Next, a correlation comparison operation is performed. If the high frequency energy proportion monotonically increases for three consecutive windows, and the absolute value of the low frequency torque trend slope exceeds the upper limit of the historical steady state slope interval, a load mutation precursor flag is generated. The historical steady state slope interval is determined by the 90% quantile of the slope distribution of the same motor in the past year, with a dimension of N·m / s, to ensure that only significant deviations trigger the flag.

[0127] In step S16, according to the load mutation precursor flag, a disturbance source classification is performed, and a preset mutation detection threshold is updated to obtain an optimized reset judgment basis, including:

[0128] According to the load mutation precursor flag, a disturbance source feature extraction operation is performed to obtain a mutation frequency component;

[0129] According to the mutation frequency component, a clustering classification operation is performed to obtain a service disturbance type result;

[0130] If the service disturbance type result indicates a load mutation, a dynamic update operation is performed on the preset mutation detection threshold to obtain an updated mutation detection threshold;

[0131] According to the updated mutation detection threshold, a reset judgment basis generation operation is performed to obtain an optimized reset judgment basis.

[0132] This step, after the generation of the load mutation precursor flag, forms an optimized reset judgment basis through feature extraction, clustering classification, and threshold dynamic update, and the process is as follows:

[0133] First, a disturbance source feature extraction operation is performed to extract the peak frequency and its adjacent half-width in the amplitude spectrum of the disturbance frequency feature above 20 Hz, using the load mutation precursor flag as an index, to form a mutation frequency component for representing the center frequency and bandwidth of the mutation disturbance.

[0134] Subsequently, a clustering classification operation is performed, K-means clustering is performed on the mutation frequency components, the cluster number k is 3, covering three categories of normal, mutation, and noise; the clustering distance uses the Euclidean distance, and the center initialization uses the 25%, 50%, and 75% quantile values of the historical steady-state frequency components, to obtain a service disturbance type result for distinguishing the mutation from other disturbances. For example, a mutation frequency component has a feature of a peak frequency of 35 Hz and a half-height width of 5 Hz, and the Euclidean distances of the component from three initial center points are calculated. Assuming that one center point represents a historical normal working condition and has a feature of a peak frequency of 25 Hz and a half-height width of 3 Hz, another center point represents a historical mutation working condition and has a feature of a peak frequency of 30 Hz and a half-height width of 6 Hz, and the third center point represents a historical noise working condition and has a feature of a peak frequency of 40 Hz and a half-height width of 10 Hz. After calculation, the distance between the component and the center point representing the historical mutation working condition is the smallest, and therefore the component is classified into the load mutation type, so that the service disturbance type result is load mutation.

[0135] If the service disturbance type result indicates load mutation, a dynamic updating operation on the preset mutation detection threshold is performed: taking the energy of the mutation frequency component of the previous steady-state window as a reference, the updated mutation detection threshold is calculated in a manner of “reference energy plus adjustment ratio”, and the adjustment ratio is determined by the median value of the energy increment of the historical mutation sample. It should be noted that the first dynamic updating of the mutation detection threshold will be updated according to the preset initial value, and for example, the preset initial value of the mutation detection threshold can be set based on the 95% quantile of the energy of the mutation frequency component in the historical steady-state operation of the motor, such as 0.2.

[0136] Finally, a reset judgment basis generation operation is performed, the updated mutation detection threshold is combined with the upper limit of the low-frequency torque trend slope to form an optimized reset judgment basis in the form of a vector.

[0137] In step S17, according to the stable torque measurement sequence, a load mutation determination is performed, and if a load mutation occurs, a parameter reset mechanism is triggered according to the optimized reset judgment basis, and filtering and compensation are re-performed, and a final accurate torque is output, including:

[0138] According to the stable torque measurement sequence, a mutation determination operation is performed, and the fluctuation amplitude of the sequence is compared with the optimized reset judgment basis;

[0139] If the fluctuation amplitude exceeds the optimized reset judgment basis, the parameter reset mechanism is triggered, the filtering and compensation process is re-performed, and the torque measurement sequence obtained after re-filtering and compensation is taken as the final accurate torque output.

[0140] This step implements online determination based on the stable torque measurement sequence, and once a load mutation is identified, the parameter reset is triggered and the filtering-compensation chain is re-performed, and the final accurate torque matching the current working condition is output, and the process is as follows:

[0141] Firstly, a mutation determination operation is performed to calculate the standard deviation of the latest N-second stable torque measurement sequence, N being 10s; the standard deviation is compared with the threshold component in the optimized reset determination basis, and if the standard deviation is greater than the threshold, it is determined that a load mutation occurs.

[0142] After the determination is established, a parameter reset mechanism is triggered, and the edge computing unit restores the adaptive calibration parameters, Kalman covariance matrix and wavelet decomposition layer number to the initial values, and immediately re-executes the filtering and compensation process of S12 to S16, that is, reloads the filtering coefficient, re-runs the Kalman recursion, and re-updates the calibration weight to obtain a new stable torque measurement sequence.

[0143] The sequence is verified again by the same determination logic of S16, and if the standard deviation is lower than the updated threshold, the cycle is terminated, and the new stable torque measurement sequence is output as the final accurate torque; if it is still over the limit, the cycle continues until the threshold requirement is met, the S17 stage processing is completed, and the rolling adaptive measurement is realized.

[0144] In summary, the motor instantaneous torque measurement method based on a high-bandwidth sensor is disclosed, the torque signal and the multi-source interference quantity are synchronously collected by the high-bandwidth sensor, and the original torque sequence is formed after time synchronization superposition; the disturbance frequency characteristics are extracted through multi-scale wavelet transform and spectrum analysis; the working condition matched filtering coefficient is generated by using the pre-trained adaptive filter model to complete the primary filtering; the residual fluctuation is suppressed by combining Kalman state estimation to obtain the optimized torque estimation value; the parameter adaptive compensation is realized through online calibration weight updating to output the stable torque measurement sequence; the load mutation prediction and threshold dynamic updating are realized based on time-frequency correlation analysis, and finally the measurement is quickly converged after the working condition mutation through the closed-loop reset mechanism, thereby solving the problem of low instantaneous torque measurement accuracy in the prior art.

[0145] Referring to Figure 2 , the second embodiment of the present application provides a motor instantaneous torque measurement system based on a high-bandwidth sensor, comprising:

[0146] A data acquisition preprocessing module is configured to acquire a torque signal and an interference quantity, perform time synchronization operation on the torque signal and the interference quantity, obtain an original torque sequence containing superimposed interference, and perform multi-scale wavelet transform and spectrum analysis on the original torque sequence containing superimposed interference to obtain disturbance frequency characteristics.

[0147] An adaptive filtering processing module is configured to input the disturbance frequency characteristics into a pre-trained adaptive filter model to obtain filtering coefficients corresponding to a current working condition, and filter out high-frequency disturbances from the original torque sequence according to the filtering coefficients corresponding to the current working condition to obtain a primary filtered torque signal.

[0148] a state estimation optimization module, configured to, if a time-domain fluctuation amplitude of the primary filtered torque signal exceeds a preset fluctuation threshold, perform Kalman state estimation according to the primary filtered torque signal to obtain an optimized torque estimation value;

[0149] an online calibration compensation module, configured to perform calibration weight updating according to the optimized torque estimation value to obtain adaptive calibration parameters, and perform real-time compensation on a subsequently collected torque signal according to the adaptive calibration parameters to obtain a stable torque measurement sequence;

[0150] a mutation precursor detection module, configured to perform correlation comparison according to the stable torque measurement sequence and the disturbance frequency feature to obtain a load mutation precursor flag;

[0151] a threshold dynamic updating module, configured to perform interference source classification according to the load mutation precursor flag, and update a preset mutation detection threshold to obtain an optimized reset judgment basis;

[0152] a closed-loop reset output module, configured to perform load mutation determination according to the stable torque measurement sequence, and if a load mutation occurs, trigger a parameter reset mechanism and perform re-filtering and compensation according to the optimized reset judgment basis to output a final accurate torque;

[0153] It should be noted that the motor instantaneous torque measurement device based on a high-bandwidth sensor provided by the embodiment of the present application is used to execute all process steps of the motor instantaneous torque measurement method based on a high-bandwidth sensor of the above-mentioned embodiment, and the working principles and beneficial effects of the two are one-to-one correspondence, thus it is not repeated here.

[0154] The embodiment of the present application also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a motor instantaneous torque measurement program based on a high-bandwidth sensor. The processor implements the steps in each of the above motor instantaneous torque measurement methods based on a high-bandwidth sensor when executing the computer program, such as Figure 1 the step S11 shown. Alternatively, the processor implements the functions of each module in each of the above device embodiments when executing the computer program, such as the data acquisition module.

[0155] For example, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present application. The one or more modules can be a series of computer program instruction segments that can complete a specific function, which are used to describe the execution process of the computer program in the electronic device.

[0156] The electronic device can be a desktop computer, a notebook computer, a palm computer, a smart tablet, and the like. The electronic device can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the above components are only examples of the electronic device, and do not constitute a limitation on the electronic device, and can include more or less components than the above, or combine certain components, or different components, for example, the electronic device can also include an input / output device, a network access device, a bus, and the like.

[0157] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, and the like. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The processor is the control center of the electronic device, and connects various parts of the electronic device through various interfaces and lines.

[0158] The memory can be used to store the computer programs or modules, and the processor realizes various functions of the electronic device by running or executing the computer programs or modules stored in the memory, and calling data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required by a function (such as a sound playing function, an image playing function, and the like), and the like; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, and the like), and the like. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, for example, a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state memory devices.

[0159] The modules integrated in the electronic device can be stored in a computer readable storage medium if they are implemented in the form of software function units and sold or used as independent products. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware. The computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms, etc. The computer readable medium can include any entity or device, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. that can carry the computer program code. It should be noted that the content included in the computer readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0160] It should be noted that the above-described device embodiments are only schematic, and the units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, i.e. they can be located in one place or distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment. In addition, the connection relationship between the modules in the device embodiment provided by the present application indicates that there is a communication connection between them, which can be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement it without creative labor.

[0161] The above-described specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above-described specific embodiments are only examples of the present application and do not limit the protection scope of the present application. It is particularly pointed out that any modification, equivalent replacement, improvement, etc. made by those skilled in the art within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for measuring instantaneous torque of a motor based on a high-bandwidth sensor, characterized in that, include: The torque signal and interference quantity are acquired, and the torque signal and interference quantity are time-synchronized to obtain the original torque sequence containing superimposed interference. Multi-scale wavelet transform and spectrum analysis are performed on the original torque sequence containing superimposed interference to obtain the disturbance frequency characteristics. The disturbance frequency features are input into a pre-trained adaptive filter model to obtain the filter coefficients corresponding to the current operating condition. The original torque sequence is then filtered to remove high-frequency disturbances based on the filter coefficients corresponding to the current operating condition to obtain the primary filtered torque signal. If the time-domain fluctuation amplitude of the primary filtered torque signal exceeds a preset fluctuation threshold, then Kalman state estimation is performed based on the primary filtered torque signal to obtain an optimized torque estimate. Based on the optimized torque estimate, the calibration weight is updated to obtain adaptive calibration parameters. Based on the adaptive calibration parameters, the subsequently acquired torque signals are compensated in real time to obtain a stable torque measurement sequence. By correlating and comparing the stable torque measurement sequence with the disturbance frequency characteristics, a load change precursor indicator can be obtained; Based on the load mutation warning flags, interference sources are classified, and the preset mutation detection threshold is updated to obtain the optimized reset judgment criteria. Based on the stable torque measurement sequence, load change determination is performed. If a load change occurs, the parameter reset mechanism is triggered according to the optimized reset judgment criteria, and the filter and compensation are re-filtered and re-compensated to output the final accurate torque.

2. The method for measuring instantaneous torque of a motor based on a high-bandwidth sensor according to claim 1, characterized in that, The process involves acquiring the torque signal and interference quantity, performing time synchronization on the torque signal and interference quantity to obtain an original torque sequence containing superimposed interference, and then performing multi-scale wavelet transform and spectral analysis on the original torque sequence containing superimposed interference to obtain the disturbance frequency characteristics, including: Acquire torque signal and interference; Perform a network time protocol synchronization operation on the torque signal and the disturbance amount to obtain a time-synchronized torque signal and disturbance amount; The time-synchronized torque signal is superimposed with the disturbance to obtain the original torque sequence containing the superimposed disturbance. A multi-scale wavelet transform is performed on the original torque sequence containing superimposed interference to decompose it into high-frequency disturbance components; Perform a Fast Fourier Transform on the high-frequency disturbance component to extract the main disturbance frequency components and obtain the disturbance frequency characteristics.

3. The method for measuring instantaneous torque of a motor based on a high-bandwidth sensor according to claim 1, characterized in that, The step involves inputting the disturbance frequency features into a pre-trained adaptive filter model to obtain the filter coefficients corresponding to the current operating condition, and then filtering out high-frequency disturbances from the original torque sequence based on the filter coefficients corresponding to the current operating condition to obtain a primary filtered torque signal, including: Based on the disturbance frequency characteristics, an adaptive filter model loading operation is performed to obtain an adaptive filter model that matches the current operating conditions. The disturbance frequency characteristics are input into the adaptive filter model, and the filter coefficient generation operation is performed to obtain the filter coefficients corresponding to the current operating condition. A digital band-stop filter is constructed based on the filtering coefficients corresponding to the current operating condition. A high-frequency disturbance filtering operation is then performed on the original torque sequence based on the digital band-stop filter to obtain a primary filtered torque signal.

4. The method for measuring instantaneous torque of a motor based on a high-bandwidth sensor according to claim 1, characterized in that, If the time-domain fluctuation amplitude of the primary filtered torque signal exceeds a preset fluctuation threshold, then Kalman state estimation is performed based on the primary filtered torque signal to obtain an optimized torque estimate, including: Based on the primary filtered torque signal, a time-domain fluctuation amplitude calculation operation is performed to obtain the current fluctuation amplitude; If the current fluctuation amplitude is greater than the preset fluctuation threshold, then the primary filtered torque signal is used as the observation vector to establish the state equation and the observation equation, and the Kalman filter recursive estimation operation is performed to obtain the optimized torque estimate.

5. The method for measuring instantaneous torque of a motor based on a high-bandwidth sensor according to claim 1, characterized in that, The process of updating calibration weights based on the optimized torque estimate to obtain adaptive calibration parameters, and then using these parameters to compensate for subsequently acquired torque signals in real time to obtain a stable torque measurement sequence, includes: Obtain the initial calibration parameters; Based on the optimized torque estimate and the initial calibration parameters, a deviation quantization operation is performed to obtain the estimated deviation. Based on the estimated deviation, a gradient descent weight update operation is performed to obtain the updated calibration weight matrix; Based on the updated calibration weight matrix, an adaptive calibration parameter generation operation is performed to obtain the adaptive calibration parameters; Based on the adaptive calibration parameters, a real-time compensation operation is performed on the subsequently acquired torque signals to obtain a stable torque measurement sequence.

6. The method for measuring instantaneous torque of a motor based on a high-bandwidth sensor according to claim 1, characterized in that, The step of correlating and comparing the stable torque measurement sequence with the disturbance frequency characteristics to obtain load change warning signs includes: Based on the stable torque measurement sequence, a time-domain trend extraction operation is performed to obtain the low-frequency torque trend; Based on the perturbation frequency characteristics, a frequency domain energy calculation operation is performed to obtain the high-frequency energy ratio; A correlation comparison operation is performed between the low-frequency torque trend and the high-frequency energy ratio. If the high-frequency energy ratio increases and the low-frequency torque trend deviates, a load change warning sign is generated.

7. The method for measuring instantaneous torque of a motor based on a high-bandwidth sensor according to claim 1, characterized in that, The step of classifying interference sources based on the load mutation precursor flags and updating the preset mutation detection threshold to obtain the optimized reset judgment criteria includes: Based on the load mutation precursor flag, perform interference source feature extraction to obtain mutation frequency components; Based on the mutation frequency components, a clustering classification operation is performed to obtain the business interference type results; If the result of the service interference type indicates a load mutation, then a dynamic update operation is performed on the preset mutation detection threshold to obtain the updated mutation detection threshold. Based on the updated mutation detection threshold, a reset judgment criteria generation operation is performed to obtain the optimized reset judgment criteria.

8. The method for measuring instantaneous torque of a motor based on a high-bandwidth sensor according to claim 1, characterized in that, The process of determining load abrupt changes based on the stable torque measurement sequence, and triggering a parameter reset mechanism and re-filtering and compensating according to the optimized reset judgment criteria to output the final accurate torque, includes: Based on the stable torque measurement sequence, a sudden change determination operation is performed, and the sequence fluctuation amplitude is compared with the optimized reset determination criteria. If the fluctuation exceeds the optimized reset judgment criteria, the parameter reset mechanism is triggered, the filtering and compensation process is re-executed, and the torque measurement sequence obtained after re-filtering and compensation is used as the final accurate torque output.

9. A motor instantaneous torque measurement system based on a high-bandwidth sensor, characterized in that, include: The data acquisition and preprocessing module is used to acquire torque signals and interference quantities, and to perform time synchronization operations on the torque signals and interference quantities to obtain the original torque sequence containing superimposed interference. The module also performs multi-scale wavelet transform and spectral analysis on the original torque sequence containing superimposed interference to obtain the disturbance frequency characteristics. An adaptive filtering module is used to input the disturbance frequency features into a pre-trained adaptive filter model to obtain the filter coefficients corresponding to the current operating condition, and to filter out high-frequency disturbances in the original torque sequence according to the filter coefficients corresponding to the current operating condition to obtain a primary filtered torque signal. The state estimation optimization module is used to perform Kalman state estimation based on the primary filtered torque signal if the time-domain fluctuation amplitude of the primary filtered torque signal exceeds a preset fluctuation threshold, so as to obtain an optimized torque estimate value. The online calibration compensation module is used to update the calibration weights based on the optimized torque estimate to obtain adaptive calibration parameters, and to perform real-time compensation on the subsequently acquired torque signals based on the adaptive calibration parameters to obtain a stable torque measurement sequence. The sudden change precursor detection module is used to correlate and compare the stable torque measurement sequence with the disturbance frequency characteristics to obtain a load sudden change precursor indicator; The threshold dynamic update module is used to classify interference sources according to the load mutation warning flag and update the preset mutation detection threshold to obtain the optimized reset judgment basis. The closed-loop reset output module is used to determine the load change based on the stable torque measurement sequence. If a load change occurs, the parameter reset mechanism is triggered according to the optimized reset judgment criteria, and the filter and compensation are re-filtered to output the final accurate torque.

Images