Demagnetization fault detection method and system based on high-precision torque distribution

Abstract

The application discloses a kind of demagnetization fault detection method and system based on high-precision torque distribution, which comprises the following steps: according to the electrical parameters of permanent magnet synchronous motor at multiple speeds, calculate flux linkage training data, and construct flux linkage distribution model;Based on the differential data corresponding to the flux linkage training data, a data-driven iron loss compensation model is established, the flux linkage distribution model is compensated to obtain flux linkage mapping data;Based on independent component analysis algorithm, the torque matrix distribution data composed of the flux linkage mapping data is processed to obtain demagnetization fault parameters;Determine whether the demagnetization fault parameters are greater than the preset signal threshold, if yes, determine that the permanent magnet synchronous motor has demagnetization fault. It can be seen that the application can realize high-precision torque distribution data based on flux linkage distribution model and iron loss compensation, independent component analysis fault parameter extraction and accurate demagnetization fault diagnosis, and improve the reliability and maintenance efficiency of motor operation.

Description

Technical Field

[0001] This invention relates to the field of demagnetization fault detection technology, and in particular to a demagnetization fault detection method and system based on high-precision torque distribution. Background Technology

[0002] Permanent magnet synchronous motors (PMSMs) are characterized by high efficiency, low torque ripple, and strong fault tolerance, and are widely used in new energy vehicles, aircraft, and robotics. However, PMSMs are susceptible to permanent demagnetization of the permanent magnets in the rotor due to factors such as abnormal temperature, armature reaction, mechanical stress, and environmental factors. This demagnetization significantly reduces the motor's output torque and efficiency, causing torque pulsation and speed fluctuations, leading to system instability and increased maintenance costs. Furthermore, in critical scenarios (such as industrial servo systems or rail transportation), it can potentially cause safety accidents. Detecting demagnetization can provide insights into the actual operating condition of the PMSM. Currently, methods for detecting demagnetization faults in PMSMs can be divided into two categories: active detection methods based on signal injection and passive detection methods based on parameter analysis.

[0003] Active detection methods based on signal injection directly detect demagnetization by injecting specific excitation signals (such as high-frequency square wave voltage or demagnetizing pulses) into the motor and analyzing the response characteristics. These methods offer high detection accuracy and fast response, but require additional hardware (such as signal generators and demagnetizing coils) and anti-interference design, which may increase system complexity and cost. Furthermore, high-frequency signals may interfere with the normal operation of the motor, and some techniques are only applicable to offline scenarios.

[0004] Passive detection methods based on parameter analysis indirectly determine demagnetization by monitoring changes in naturally generated electromagnetic parameters (such as back EMF, current, flux linkage, and torque) during motor operation. These methods do not require external excitation and rely on algorithms to process signal differences, but they are susceptible to interference from load fluctuations, sensor accuracy, and temperature drift. Furthermore, some methods depend on specific operating conditions (such as no-load or shutdown), limiting their real-time performance and versatility.

[0005] Existing solutions lack comprehensive analysis of multi-speed electrical parameters and modeling of magnetic flux linkage and iron loss compensation, making it difficult to accurately extract demagnetization fault parameters and optimize torque distribution. This results in insufficient accuracy in fault diagnosis, making it prone to equipment damage due to undetected demagnetization faults, thus limiting the reliability of motor operation and maintenance efficiency. Clearly, existing technologies have shortcomings that urgently need to be addressed. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a demagnetization fault detection method and system based on high-precision torque distribution, which can realize high-precision torque distribution data based on flux linkage distribution model and iron loss compensation, as well as fault parameter extraction and accurate demagnetization fault diagnosis based on independent component analysis, thereby improving the reliability of motor operation and maintenance efficiency, and reducing the risk of equipment damage caused by failure to detect demagnetization faults in a timely manner.

[0007] To address the aforementioned technical problems, the first aspect of this invention discloses a demagnetization fault detection method based on high-precision torque distribution, the method comprising:

[0008] Based on multiple electrical parameters of the permanent magnet synchronous motor at multiple speeds, flux linkage training data are calculated, and a flux linkage distribution model is constructed.

[0009] Based on the differential data corresponding to the flux linkage training data, a data-driven iron loss compensation model is established to compensate the flux linkage distribution model and obtain flux linkage mapping data.

[0010] Based on the independent component analysis algorithm, the torque matrix distribution data composed of the flux linkage mapping data is processed to obtain demagnetization fault parameters;

[0011] Determine whether the demagnetization fault parameter is greater than a preset signal threshold. If so, determine that the permanent magnet synchronous motor has a demagnetization fault.

[0012] A second aspect of this invention discloses a demagnetization fault detection system based on high-precision torque distribution, the system comprising:

[0013] The calculation module is used to calculate flux linkage training data and construct a flux linkage distribution model based on multiple electrical parameters of the permanent magnet synchronous motor at multiple speeds.

[0014] The mapping module is used to establish a data-driven iron loss compensation model based on the differential data corresponding to the magnetic flux training data, and to compensate the magnetic flux distribution model to obtain magnetic flux mapping data.

[0015] The analysis module is used to process the torque matrix distribution data composed of the flux linkage mapping data based on the independent component analysis algorithm to obtain demagnetization fault parameters.

[0016] The judgment module is used to determine whether the demagnetization fault parameter is greater than a preset signal threshold. If so, it is determined that the permanent magnet synchronous motor has a demagnetization fault.

[0017] A third aspect of this invention discloses another demagnetization fault detection system based on high-precision torque distribution, the system comprising:

[0018] Memory containing executable program code;

[0019] A processor coupled to the memory;

[0020] The processor calls the executable program code stored in the memory to execute some or all of the steps in the demagnetization fault detection method based on high-precision torque distribution disclosed in the first aspect of the present invention.

[0021] The fourth aspect of the present invention discloses a computer storage medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the demagnetization fault detection method based on high-precision torque distribution disclosed in the first aspect of the present invention.

[0022] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0023] This invention acquires the electrical parameters of a permanent magnet synchronous motor at multiple speeds, processes and generates flux linkage mapping data based on flux linkage and iron loss compensation models, and uses independent component analysis (ICA) to extract demagnetization fault parameters and determine whether they exceed a threshold to identify demagnetization faults. This enables high-precision torque distribution data based on flux linkage distribution models and iron loss compensation, as well as fault parameter extraction and accurate demagnetization fault diagnosis based on ICA, thereby improving motor operating reliability and maintenance efficiency, and reducing the risk of equipment damage due to untimely detection of demagnetization faults. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart illustrating a demagnetization fault detection method based on high-precision torque distribution disclosed in an embodiment of the present invention.

[0026] Figure 2 This is a schematic diagram of a demagnetization fault detection system based on high-precision torque distribution disclosed in an embodiment of the present invention.

[0027] Figure 3 This is a schematic diagram of another demagnetization fault detection system based on high-precision torque distribution disclosed in an embodiment of the present invention.

[0028] Figure 4 This is a schematic diagram illustrating the working principle of the permanent magnet synchronous motor disclosed in the embodiments of the present invention.

[0029] Figure 5 This is a schematic diagram of the original data of the demagnetization fault detection method disclosed in the embodiments of the present invention.

[0030] Figure 6 This is a schematic diagram of the output data of the magnetic flux linkage model of the demagnetization fault detection method disclosed in the embodiments of the present invention.

[0031] Figure 7 This is a schematic diagram of torque distribution data based on iron loss compensation in the demagnetization fault detection method disclosed in the embodiments of the present invention.

[0032] Figure 8 This is a schematic diagram of the background noise and demagnetization signal obtained after independent component analysis of the demagnetization fault detection method disclosed in the embodiments of the present invention.

[0033] Figure 9 This is a schematic diagram comparing the demagnetization trends of a specific implementation of the demagnetization fault detection method disclosed in this invention. Detailed Implementation

[0034] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0035] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0036] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0037] This invention discloses a demagnetization fault detection method and system based on high-precision torque distribution. By acquiring the electrical parameters of a permanent magnet synchronous motor at multiple speeds, and processing them based on a flux linkage model and an iron loss compensation model to generate flux linkage mapping data, the method uses independent component analysis (ICA) to extract demagnetization fault parameters and determines whether a threshold is exceeded to confirm a demagnetization fault. This enables high-precision torque distribution data based on the flux linkage distribution model and iron loss compensation, as well as fault parameter extraction and accurate demagnetization fault diagnosis using ICA, improving motor operational reliability and maintenance efficiency, and reducing the risk of equipment damage due to untimely detection of demagnetization faults. Detailed explanations follow.

[0038] Example 1

[0039] Please see Figure 1 , Figure 1 This is a flowchart illustrating a demagnetization fault detection method based on high-precision torque distribution disclosed in an embodiment of the present invention. Figure 1 The demagnetization fault detection method based on high-precision torque distribution described herein can be applied to data processing systems / data processing equipment / data processing servers (wherein, the server includes a local processing server or a cloud processing server). For example... Figure 1 As shown, the demagnetization fault detection method based on high-precision torque distribution can include the following operations:

[0040] 101. Based on multiple electrical parameters of the permanent magnet synchronous motor at multiple speeds, calculate the flux linkage training data and construct a flux linkage distribution model.

[0041] 102. Based on the differential data corresponding to the flux linkage training data, establish a data-driven iron loss compensation model to compensate the flux linkage distribution model and obtain flux linkage mapping data.

[0042] 103. Based on the independent component analysis algorithm, the torque matrix distribution data composed of flux linkage mapping data is processed to obtain demagnetization fault parameters.

[0043] 104. Determine whether the demagnetization fault parameters are greater than the preset signal threshold. If so, it is determined that the permanent magnet synchronous motor has a demagnetization fault.

[0044] As can be seen, the above-described embodiments of the invention acquire electrical parameters of the permanent magnet synchronous motor at multiple speeds, process and generate flux linkage mapping data based on flux linkage model and iron loss compensation model, and use independent component analysis algorithm to extract demagnetization fault parameters and determine whether the threshold is exceeded to identify demagnetization faults. This enables the acquisition of high-precision torque distribution data based on flux linkage distribution model and iron loss compensation, as well as fault parameter extraction and accurate demagnetization fault diagnosis based on independent component analysis, thereby improving the reliability of motor operation and maintenance efficiency, and reducing the risk of equipment damage caused by failure to detect demagnetization faults in a timely manner.

[0045] As an optional embodiment, the electrical parameters in the above steps include motor speed, motor torque, d-axis current, d-axis voltage, q-axis current, and q-axis voltage.

[0046] As an optional embodiment, specifically, Figure 4 This is a schematic diagram illustrating the working principle of a permanent magnet synchronous motor system. The motor is controlled by a programmable real-time controller and driven by an inverter.

[0047] The steps described above, including calculating flux linkage training data and constructing a flux linkage distribution model based on multiple electrical parameters of the permanent magnet synchronous motor at multiple speeds, include:

[0048] Determine the steady-state equations of the permanent magnet synchronous motor:

[0049]

[0050] Where: V d / q ,I d / q These represent the DC components of the voltage and current along the d / q axes of the permanent magnet synchronous motor, respectively; R is the winding resistance; ω is the motor speed; and V... dead For the distortion voltage term, F q F q These represent the flux linkages along the d-axis and q-axis, respectively. Here are the inverter nonlinear coefficients, where:

[0051]

[0052] The steady-state equations of the permanent magnet synchronous motor are cross-combined to obtain the following set of equations:

[0053]

[0054] in: For stator current, I d =-I s sinγ,I q =I s cosγ, c is a constant;

[0055] The flux linkage output value of the permanent magnet synchronous motor is obtained based on equation (2):

[0056]

[0057] Set the rotational speed to ω0 and measure the d-axis current I of the permanent magnet synchronous motor under different load conditions. d q-axis current I q With the corresponding d-axis flux linkage output value F d q-axis flux linkage output value F q Combined to form a set of training samples {Id,i ,I q,i ,F d,i ,F q,i}; By changing the d / q axis current, K sets of samples are obtained to construct a dataset {d1,d2,…,d} covering multiple operating conditions. K}, where d i ={I d,i ,I q,i ,F d,i ,F q,i};

[0058] Considering the nonlinear variation of d-axis and q-axis flux linkages with dq-axis current, the dq-axis flux linkage distribution model can be represented by a set of weighted kernel functions:

[0059]

[0060] Where f d / q () represents the d / q flux linkage distribution; H(·) represents the radial basis kernel function, I d,i I q,i Let represent the i-th element of the d-axis and q-axis current training datasets, respectively, and K represent the number of elements in the training dataset. d,i w q,i , respectively, are the i-th weights to be determined for the d-axis flux linkage distribution model and the q-axis flux linkage distribution model, and λ represents the permanent magnet flux linkage;

[0061] The sparse Bayesian learning method is employed, based on the training data {d1, d2, ..., dn}. k The model weights w are obtained by training with equation (4). d,i w q,i Thus, the flux linkage distribution model along the dq axis is established.

[0062] Specifically, Figure 5 This is a visualization of the raw data for the aforementioned electrical parameters. The reference current angle varies between 0-40 degrees. Real-time d / q-axis current and voltage are collected and recorded at motor speeds of 225 r / min, 300 r / min, 375 r / min, and 450 r / min. Specifically, the influence of rotor motion on flux linkage modeling is eliminated by taking short-time averages. Figure 5The left image shows the d-axis voltage distribution, and the right image shows the q-axis voltage distribution. Blue indicates the d / q-axis voltage distribution at a rotational speed of 225 r / min, with the stator current in the range of [0, 15] and the stator current angle in the range of [0, 40] degrees. Yellow indicates the d / q-axis voltage distribution at a rotational speed of 300 r / min, with the stator current in the range of [0, 15] and the stator current angle in the range of [0, 40] degrees. Red indicates the d / q-axis voltage distribution at a rotational speed of 375 r / min, with the stator current in the range of [0, 15] and the stator current angle in the range of [0, 40] degrees. Cyan indicates the d / q-axis voltage distribution at a rotational speed of 450 r / min, with the stator current in the range of [0, 15] and the stator current angle in the range of [0, 40] degrees.

[0063] As an optional embodiment, the above steps, including establishing a data-driven iron loss compensation model based on the differential data corresponding to the flux linkage training data, and compensating the flux linkage distribution model to obtain flux linkage mapping data, include:

[0064] Core losses can be quantified by the d / q axis flux differential data:

[0065]

[0066] Where: ΔF d / q,c The iron loss component along the d / q axis. These represent rotational speeds ω0 and ω, respectively. t At that time, based on the d-axis flux linkage output by equation (3), These represent rotational speeds ω0 and ω, respectively. t At that time, the q-axis flux linkage is based on the output of equation (3);

[0067] For a given reference rotational speed ω0, different stator currents I are collected. s The voltage value corresponding to the stator current angle γ is calculated. Changing different rotational speeds ω t Based on each rotational speed, multiple d / q-axis currents are collected, and calculations are performed based on the voltage values. ΔF is obtained using equation (5) d,c ΔF q,c And calculate the speed difference Δω=ω t -ω0, forming L sets of sample data {Y1,Y2,…Y} L}, where Y i ={I s,i ,γ i ,ΔF d / q,c,i ,Δω i};

[0068] Considering the nonlinearity of core losses, the core loss compensation model is represented by a set of weighted kernel functions:

[0069]

[0070] Among them, f d,c f q,c Represent the iron loss compensation distribution along the d-axis and q-axis, respectively. d,j V represents the j-th weight of the iron loss compensation model on the d-th axis. q,j I represents the j-th weight in the iron loss compensation model for the q-th axis. s,i I represents the i-th stator current. s,i γ i γ represents the i-th stator current angle. Represents the kernel function;

[0071] Using the least squares method, based on the L sets of sample data {Y1,Y2,…Y} L The model weights v are obtained by training with equation (6). d,j ,v q,j Thus, the iron loss compensation model was established;

[0072] The flux linkage mapping data is obtained by combining the dq-axis flux linkage distribution model and the iron loss compensation model:

[0073]

[0074] Among them, F D F Q These represent the flux linkage mapping data of the d-axis and q-axis after iron loss compensation, respectively.

[0075] It is worth noting that there are many kernel functions to choose from. This embodiment only uses the radial basis kernel function as an example for illustration and should not impose any restrictions on other kernel functions that achieve the purpose of this invention.

[0076] As an optional embodiment, in the above steps, a sparse Bayesian learning method is employed, based on the training data {d1, d2, ..., dn}. K The model weights w are obtained by training with equation (4). d,i w q,i Thus, the dq-axis flux linkage distribution model is established, including:

[0077] Rewrite the relationship between training data and weight coefficients in matrix form:

[0078]

[0079] Among them, F d =[F d,1 ,……F d,k ],F q =[F q,1 ,……F q,k ] represents the training data vector, w d / q=[w d / q,1 ,……w d / q,k ] represents the weight vector to be estimated, ξ d / q =[ξ d / q,1 ,……ξ d / q,k [ ] represents the model error along the d-axis or q-axis, Marix d / q It is a design matrix that depends on the kernel function, Marix d / q (i,j)=H([I d,i ,I q,i ] T , [I d,j ,I q,j ] T ), i,j∈{1,……,K};

[0080] A hierarchical noise model is introduced, including:

[0081] The first layer model is:

[0082] The second-layer model is:

[0083] Where, β=[β1,……β k ], This indicates that the mean is 0 and the variance is β. i Gaussian distribution, Gamma(β) i |τ,τ) represents a gamma distribution with both shape and scale parameters τ;

[0084] It should be noted that in the examples provided by this invention, the training dataset is known. And the functional relationship F d / q =G(I d / q ,λ(T m Under the condition of ), consider the problem of estimating the weighting coefficients. For the time-varying non-Gaussian distribution v d / q The above-mentioned hierarchical noise model is adopted.

[0085] In the first layer model, each v d / q It follows an independent Gaussian distribution, tuned by a variance parameter, enabling it to flexibly handle scenarios where the overall error is non-Gaussian. Within the Bayesian learning framework, this variance can be automatically learned from the training data, adapting to the actual data and addressing the problem of time-varying error distributions.

[0086] In the second-layer model, a unified gamma distribution is used to describe the parameters of the heteroscedastic Gaussian distribution. By limiting the overall distribution of variance through the gamma distribution, its value range is constrained, thus improving the efficiency of variance learning. The hierarchical structure can not only handle complex and variable errors but also ensure computational efficiency.

[0087] The sparsity promotion model is introduced as a zero-mean Gaussian distribution model, and the mathematical model of the zero-mean Gaussian distribution model is as follows:

[0088]

[0089] Among them, GD(w d / q,i |0,α i ) indicates that the mean is 0 and the variance is α. i Gaussian distribution;

[0090] Specifically, a zero-mean Gaussian distribution model is used to sparsify the weights of each kernel function: where α i During iterative learning, since the Automatic Relevance Determination (ARD) mechanism is likely to tend to 0, then ω i Approaching 0 allows for the elimination of relevant terms, accelerating the learning process. Only a small number of ω... i The values ​​of these weights are non-zero, and the output torque model consists of these weights and the corresponding kernel function. The model can efficiently obtain the output results when applied. According to the automatic correlation decision mechanism, most weights will tend to 0 during the learning process, which promotes the sparsity of the model, reduces the computational burden required for output torque modeling, and improves the efficiency of model application.

[0091] Based on the hierarchical noise model and the sparsity promotion model, the weights of each kernel function are iteratively estimated, including:

[0092] Initialize model parameters α i (i∈{1,…,N}),β i (i∈{1,…,K});

[0093] Based on the initialization model parameter α i ,β i Derive the weight w d,i ,w q,i The posterior distribution of w is obtained. d,i ,w q,i Maximum a posteriori estimate μ d / q,i With the covariance matrix ∑ d / q,i ;

[0094] According to μ d / q,i With ∑ d / q,i Derivation of model parameters α i ,β i The posterior probability distribution;

[0095] According to the model parameter α i ,β i Calculate α from the posterior probability distribution i ,β i The derivative of α, let αi ,β i The derivative is zero, thus α is obtained. i ,β i The update formula;

[0096] Iteratively calculate the maximum a posteriori estimate and model parameters until the weights w are reached. d,i ,w q,i The maximum a posteriori estimate no longer changes or the maximum number of iterations is reached, and the trained weight estimate is obtained.

[0097] Thus, the flux linkage distribution model along the dq axis is determined;

[0098] The posterior probability distribution of the weighting coefficients is as follows:

[0099]

[0100] in:

[0101]

[0102] Determine α i ,β i The learning principle is:

[0103]

[0104] in:

[0105] More specifically, Marix is ​​a matrix designed based on kernel functions, where:

[0106] For a flux linkage value F along the d-axis d,i In this regard, its linear model is represented as F d,i =Marix i ×w d

[0107]

[0108] For a set of flux linkage values ​​F d In this regard, its linear model is represented as F d =Marix×w d

[0109]

[0110] For F along the q-axis q Its design details are the same as above.

[0111] Figure 6This is a visualization of the flux linkage distribution model obtained based on the Bayesian learning method described above. It can be seen that the flux linkage value decreases as the rotational speed increases. Specifically, Figure 6 The left image shows the d-axis flux linkage distribution, and the right image shows the q-axis flux linkage distribution. Blue indicates the d / q-axis flux linkage distribution at a rotational speed of 225 r / min, a stator current within the range of [0, 15], and a stator current angle within the range of [0, 40] degrees. Yellow indicates the d / q-axis flux linkage distribution at a rotational speed of 300 r / min, a stator current within the range of [0, 15], and a stator current angle within the range of [0, 40] degrees. Red indicates the d / q-axis flux linkage distribution at a rotational speed of 375 r / min, a stator current within the range of [0, 15], and a stator current angle within the range of [0, 40] degrees. Cyan indicates the d / q-axis flux linkage distribution at a rotational speed of 450 r / min, a stator current within the range of [0, 15], and a stator current angle within the range of [0, 40] degrees.

[0112] As an optional embodiment, in the above steps, the least squares method is used, based on the L sets of sample data {Y1,Y2,…Y}. L The model weights v are obtained by training with equation (6). d,j ,v q,j Thus, the iron loss compensation model is established, including:

[0113] The weight updates for the iron loss compensation model include:

[0114] Rewrite the relationship between training data and model coefficients in matrix form:

[0115]

[0116] ε=[ε1,……,ε L [ ] represents the model error;

[0117] Where, Φ d / q Matrix designed based on kernel function v d / q =[v d / q,1 ,…,v d / q,L [ ] represents the weight vector to be estimated. For training data vectors;

[0118] v is obtained based on the least squares method. d / q :

[0119]

[0120] As an optional embodiment, in the above steps, processing the torque matrix distribution data composed of flux linkage mapping data based on the independent component analysis algorithm to obtain demagnetization fault parameters includes:

[0121] The torque distribution is calculated by the following formula:

[0122] T e =1.5P(F) D I q -F Q I d );

[0123] Wherein: T e P represents the torque, and P represents the number of pole pairs of the motor.

[0124] T e The matrix representation is as follows:

[0125]

[0126] Among them, T eN,M This is the Mth element in the Nth row of the torque distribution matrix;

[0127] Independent component analysis was selected for demagnetization signal detection. Independent component analysis is represented as follows:

[0128] T ec =AS;

[0129] Among them, T ec The torque distribution matrix T e The centered matrix is ​​A, where A is an unknown mixture matrix; S bg As background signal, S dm This is a demagnetization signal;

[0130] Principal component analysis (PCA) was used to analyze the torque distribution matrix T. e Dimensionality reduction processing includes:

[0131] Calculate the covariance matrix: in, For the set of real numbers, Represents an N x N matrix where all elements are real numbers;

[0132] Eigenvalue decomposition: C = EΛE T Λ=diag(λ1,……λ) N ),λ1>λ2>…>λ N ;

[0133] Take the eigenvector matrix corresponding to the two largest eigenvalues. And the diagonal matrix Λ2diag(λ1,λ2); Represents an N x 2 matrix with all real numbers; E2 and Λ2 are the first two columns of the eigenvector matrices corresponding to the first two eigenvalues ​​of E and Λ, respectively; Projection dimensionality reduction:

[0134]

[0135] Construct the whitening matrix:

[0136] Calculate the whitened data: X o =W white T ec ;

[0137] Find the orthogonal matrix in the whitening matrix using fast ICA separation. Make S = OX o The two rows are independent of each other, and for each row Iteration: where o i Let O be the i-th column, and let be the separation vector of the i-th independent component;

[0138] Iteration:

[0139] For all updated o i Perform orthogonalization and normalization: o i ←o i / ||o i ||;

[0140] Repeat the iteration until each o i Convergence: The separation matrix is ​​obtained.

[0141] in, The background signal S is respectively bg Demagnetization signal S dm The estimated value;

[0142] The likelihood estimate of A is given by the following equation:

[0143]

[0144] The demagnetization fault parameter E is calculated using the following formula. dm :

[0145] Where z is the sample point index, and M represents the number of columns in the torque distribution matrix. This represents the amplitude of the separated demagnetizing component at time z.

[0146] The visualization of the mapped torque distribution output after compensation by the above iron loss compensation model is shown below. Figure 7 Specifically, Figure 7In the diagram, cyan represents the torque distribution at a speed of 300 r / min, with the stator current in the range of [0, 15] and the stator current angle in the range of [0, 40] degrees; blue represents the torque distribution at a speed of 375 r / min, with the stator current in the range of [0, 15] and the stator current angle in the range of [0, 40] degrees; and green represents the torque distribution at a speed of 450 r / min, with the stator current in the range of [0, 15] and the stator current angle in the range of [0, 40] degrees.

[0147] Independent component analysis (ICA) was chosen for demagnetization signal detection in permanent magnet synchronous motors (PMSMs) because it can effectively separate statistically independent non-Gaussian source signals. Demagnetization fault signals exhibit significant non-Gaussian differences from background noise (such as electromagnetic interference and mechanical vibration) and satisfy the independence assumption. ICA, by maximizing the non-Gaussianity of the separated components, can accurately extract fault features from mixed observation data. Of course, there are many techniques based on background separation; this explanation uses ICA as an example only and does not constitute any limitation on other independent analysis methods that can achieve the objectives of this invention.

[0148] It is worth noting that in PMSM demagnetization detection, the energy distribution of the demagnetization fault signal and background noise is usually concentrated in a few principal components. To improve detection accuracy, the observation matrix T in this embodiment... ec Due to the high dimensionality, direct ICA separation would lead to a significant increase in computational cost. This embodiment employs PCA to reduce the dimensionality of the observation matrix. PCA selects the principal components with the largest variance through eigenvalue decomposition, discarding components with smaller variances. These minor components typically correspond to measurement noise or irrelevant signals; removing them improves the separation accuracy between demagnetization signals and background noise.

[0149] Specifically, for the given several mapped torque distributions, schematic diagrams of the background signal and demagnetization signal obtained from the above detection model can be found in [reference needed]. Figure 8 Specifically, Figure 8 The horizontal axis represents the sample point index, which can also be understood as the sampling point number of the time series. The vertical axis of the upper half of the graph represents the amplitude of the background noise component after ICA separation; the vertical axis of the lower half of the graph represents the amplitude of the demagnetizing signal separated by ICA, which is the independent component related to flux linkage attenuation.

[0150] Furthermore, the demagnetization signal is determined by the following formula:

[0151] E th <E dm ;

[0152] Among them: E th If the discriminant is true, a demagnetization fault is determined to exist, with a preset threshold.

[0153] Specifically, Figure 9 Based on the above demagnetization analysis scheme, the original curve and the corresponding smoothed curve are displayed side by side in some actual implementation scenarios. The demagnetization trend differences of the three channels of the observation matrix are compared. Specifically, channel 1 represents the demagnetization intensity curve at a rotation speed of 300 r / min, channel 2 represents the demagnetization intensity curve at a rotation speed of 375 r / min, and channel 3 represents the demagnetization intensity curve at a rotation speed of 450 r / min. Among them, "original" means the instantaneous demagnetization intensity obtained directly from the square of the demagnetization signal, and "smoothed" means the trend line after the moving average of the original curve.

[0154] Compared with existing technologies for detecting demagnetization faults in permanent magnet synchronous motors, a specific implementation scheme of this embodiment has the following advantages:

[0155] 1. It eliminates the influence of VSI nonlinear distortion, thus improving the accuracy of model estimation.

[0156] 2: By establishing the d / q axis flux linkage model with respect to the stator current and d / q axis voltage through the d / q axis current and voltage, the current distribution of the d / q axis can be obtained by giving the orthogonal axis current, and the d / q axis current can be directly measured.

[0157] 3: By using the magnetic flux linkage model and real-time measured stator current, the d / q axis flux linkage can be calculated quickly, avoiding the low efficiency problem caused by the signal injection method which requires signal injection and iterative calculation.

[0158] 4: By designing a Bayesian learning strategy that integrates hierarchical noise model and sparse weight model, a magnetic flux linkage model can be established with a small amount of training data, effectively reducing the time required for training data collection, enhancing engineering practicality, and avoiding the tedious operation of building tables using parametric methods.

[0159] 5: By using the iron loss compensation model, the torque distribution at different speeds is mapped to the same speed, effectively solving the problem that it is difficult to obtain the torque at the same speed in actual working conditions.

[0160] 6: Torque-based demagnetization judgment integrates input data from multiple dimensions, providing higher model accuracy compared to traditional demagnetization judgment based on a single dimension, such as current or voltage.

[0161] 7: Based on the ICA separation demagnetization background, only a small amount of data is needed to achieve fault detection, effectively reducing the time spent on training data collection and enhancing engineering practicality.

[0162] Example 2

[0163] Please see Figure 2 , Figure 2This is a schematic diagram of a demagnetization fault detection system based on high-precision torque distribution, as disclosed in an embodiment of the present invention. Figure 2 The demagnetization fault detection system based on high-precision torque distribution described herein can be applied to data processing systems / data processing equipment / data processing servers (wherein, the server includes a local processing server or a cloud processing server). For example... Figure 2 As shown, the demagnetization fault detection system based on high-precision torque distribution may include:

[0164] The calculation module 201 is used to calculate flux linkage training data and construct a flux linkage distribution model based on multiple electrical parameters of the permanent magnet synchronous motor at multiple speeds.

[0165] The mapping module 202 is used to establish a data-driven iron loss compensation model based on the differential data corresponding to the magnetic flux training data, to compensate the magnetic flux distribution model, and to obtain magnetic flux mapping data.

[0166] Analysis module 203 is used to process the torque matrix distribution data composed of flux linkage mapping data based on the independent component analysis algorithm to obtain demagnetization fault parameters.

[0167] The judgment module 204 is used to determine whether the demagnetization fault parameter is greater than the preset signal threshold. If so, it is determined that the permanent magnet synchronous motor has a demagnetization fault.

[0168] The details of the steps and techniques performed by the above modules can be found in the content disclosed in Embodiment 1, and will not be repeated here again.

[0169] As can be seen, the above-described embodiments of the invention acquire electrical parameters of the permanent magnet synchronous motor at multiple speeds, process and generate flux linkage mapping data based on flux linkage model and iron loss compensation model, and use independent component analysis algorithm to extract demagnetization fault parameters and determine whether the threshold is exceeded to identify demagnetization faults. This enables the acquisition of high-precision torque distribution data based on flux linkage distribution model and iron loss compensation, as well as fault parameter extraction and accurate demagnetization fault diagnosis based on independent component analysis, thereby improving the reliability of motor operation and maintenance efficiency, and reducing the risk of equipment damage caused by failure to detect demagnetization faults in a timely manner.

[0170] Example 3

[0171] Please see Figure 3 , Figure 3 This is another demagnetization fault detection system based on high-precision torque distribution disclosed in the embodiments of the present invention. Figure 3 The demagnetization fault detection system based on high-precision torque distribution described herein is applied in data processing systems / data processing equipment / data processing servers (wherein, the server includes a local processing server or a cloud processing server). For example... Figure 3As shown, the demagnetization fault detection system based on high-precision torque distribution may include:

[0172] Memory 301 storing executable program code;

[0173] Processor 302 coupled to memory 301;

[0174] The processor 302 calls the executable program code stored in the memory 301 to execute the steps of the demagnetization fault detection method based on high-precision torque distribution described in Embodiment 1.

[0175] Example 4

[0176] This invention discloses a computer read storage medium that stores a computer program for electronic data exchange, wherein the computer program causes a computer to execute the steps of the demagnetization fault detection method based on high-precision torque distribution described in Embodiment 1.

[0177] Example 5

[0178] This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps of the demagnetization fault detection method based on high-precision torque distribution described in Embodiment 1.

[0179] The foregoing has described specific embodiments of this specification; other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than those shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily have to follow the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0180] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0181] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware components.

[0182] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0183] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create a machine for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0184] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0185] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0186] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0187] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0188] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0189] Finally, it should be noted that the demagnetization fault detection method and system based on high-precision torque distribution disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A demagnetization fault detection method based on high-precision torque distribution, characterized in that, The method includes: Based on multiple electrical parameters of the permanent magnet synchronous motor at various speeds, flux linkage training data is calculated, and a flux linkage distribution model is constructed, including: Determine the steady-state equations of the permanent magnet synchronous motor: in: , These are the DC components of the voltage and current along the d / q axes of the permanent magnet synchronous motor, respectively. For the winding resistance; This refers to the motor speed. This is the distortion voltage term. , These represent the flux linkages along the d-axis and q-axis, respectively. , Here are the inverter nonlinear coefficients, where: ,in It is the current angle; The steady-state equations of the permanent magnet synchronous motor are cross-combined to obtain the following set of equations: in: For stator current, , , ; c It is a constant; The flux linkage output value of the permanent magnet synchronous motor is obtained based on equation (2): Set the speed to The d-axis current of the permanent magnet synchronous motor under different load conditions q-axis current Corresponding d-axis flux linkage output value q-axis flux linkage output value Combined to form a set of training samples Obtained by changing the d / q axis current K Group samples to construct a dataset covering multiple working conditions. ,in ; Considering the nonlinear variation of d-axis and q-axis flux linkages with dq-axis current, the dq-axis flux linkage distribution model can be represented by a set of weighted kernel functions: in f d / q () represents the d / q flux linkage distribution; Represents the radial basis kernel function. , These represent the d-axis and q-axis current training datasets, respectively. i One element, K This indicates the number of elements in the training dataset. , These are the first two magnetic flux distribution models, representing the d-axis and q-axis magnetic flux distribution models, respectively. i One weight to be determined Indicates permanent magnet flux linkage; Using the sparse Bayesian learning method, based on the training data The model weights are obtained by training with equation (4). , , Thus, the flux linkage distribution model along the dq axis was established; Based on the differential data corresponding to the flux linkage training data, a data-driven iron loss compensation model is established to compensate the flux linkage distribution model, resulting in flux linkage mapping data, including: Core losses can be quantified by the d / q axis flux differential data: in: The iron loss component along the d / q axis. , These represent rotational speeds of 1000 and 1000 respectively. , At that time, based on the d-axis flux linkage output by equation (3), , These represent the rotational speeds as follows: , At that time, the q-axis flux linkage is based on the output of equation (3); For a given reference speed Collect different stator currents Stator current angle The corresponding voltage value was calculated. ; change different speeds Based on each rotational speed, multiple d / q-axis currents are collected, and calculations are performed based on the voltage values. Using equation (5) to obtain , And calculate the speed difference. L sets of sample data are formed. ,in ; Considering the nonlinearity of core losses, the core loss compensation model is represented by a set of weighted kernel functions: in, , These represent the iron loss compensation distributions along the d-axis and q-axis, respectively. This represents the j-th weight in the iron loss compensation model for the d-th axis. This represents the j-th weight in the iron loss compensation model for the q-th axis. Represents the i-th stator current , Represents the i-th stator current angle , Represents the kernel function; Using the least squares method, based on the above The model weights are obtained by training with equation (6). , Thus, the iron loss compensation model was established; The flux linkage mapping data is obtained by combining the dq-axis flux linkage distribution model and the iron loss compensation model: in, , These represent flux linkage mapping data for the d-axis and q-axis after iron loss compensation, respectively. Based on the independent component analysis algorithm, the torque matrix distribution data composed of the flux linkage mapping data is processed to obtain demagnetization fault parameters; Determine whether the demagnetization fault parameter is greater than a preset signal threshold. If so, determine that the permanent magnet synchronous motor has a demagnetization fault.

2. The demagnetization fault detection method based on high-precision torque distribution according to claim 1, characterized in that, The electrical parameters include motor speed, motor torque, d-axis current, d-axis voltage, q-axis current, and q-axis voltage.

3. The demagnetization fault detection method based on high-precision torque distribution according to claim 1, characterized in that, The method employs sparse Bayesian learning based on the training data. The model weights are obtained by training with equation (4). , , Thus, the dq-axis flux linkage distribution model is established, including: Rewrite the relationship between training data and weight coefficients in matrix form: ； in, , For training data vectors, Let be the weight vector to be estimated. The model error is represented by the d-axis or q-axis. It is a design matrix that depends on the kernel function. , ; A hierarchical noise model is introduced, including: The first layer model is: ; The second-layer model is: ; in, This indicates that the mean is 0 and the variance is . Gaussian distribution, Both shape and scale parameters are represented. The gamma distribution; The sparsity promotion model is introduced as a zero-mean Gaussian distribution model, and the mathematical model of the zero-mean Gaussian distribution model is as follows: ； ； in, This indicates that the mean is 0 and the variance is . Gaussian distribution; Based on the hierarchical noise model and the sparsity promotion model, the weights of each kernel function are iteratively estimated, including: Initialize model parameters ( ), ( ); Based on the initial model parameters , Derive the weights , The posterior distribution is obtained. , Maximum a posteriori estimation With covariance matrix ; according to and Derivation of model parameters , The posterior probability distribution; According to the model parameters , Calculation of posterior probability distribution , The derivative of, let , The derivative is zero, thus obtaining , The update formula; Iteratively calculate the maximum a posteriori estimate and model parameters until the weights are calculated. , The maximum a posteriori estimate no longer changes or the maximum number of iterations is reached, and the trained weight estimate is obtained. Thus, the flux linkage distribution model along the dq axis is determined; The posterior probability distribution of the weighting coefficients is as follows: ； in: ； Sure , The learning principle is: , ; ； in: .

4. The demagnetization fault detection method based on high-precision torque distribution according to claim 1, characterized in that, The least squares method is used, according to the... The model weights are obtained by training with equation (6). , Thus, the iron loss compensation model is established, including: The weight updates for the iron loss compensation model include: Rewrite the relationship between training data and model coefficients in matrix form: ; The model error is the model error. in, Matrix designed based on kernel function ; Let be the weight vector to be estimated. For training data vectors; Obtained based on least squares method : 。 5. The demagnetization fault detection method based on high-precision torque distribution according to claim 1, characterized in that, The independent component analysis algorithm is used to process the torque matrix distribution data composed of the flux linkage mapping data to obtain demagnetization fault parameters, including: The torque distribution is calculated by the following formula: ； in: Indicates torque, Indicates the number of pole pairs of the motor; The matrix representation is as follows: ； in, This is the Mth element in the Nth row of the torque distribution matrix; Independent component analysis was selected for demagnetization signal detection. Independent component analysis is represented as follows: ； in, Torque distribution matrix The centered matrix is ​​A, where A is an unknown mixture matrix; ， For background signal, This is a demagnetization signal; Principal component analysis (PCA) was used to analyze the torque distribution matrix. Dimensionality reduction processing includes: Calculate the covariance matrix: ;in, For the set of real numbers, Represents an N x N matrix where all elements are real numbers; Eigenvalue decomposition: ; Take the eigenvector matrix corresponding to the two largest eigenvalues. and diagonal matrix ; Represent an N-row, 2-column matrix containing only real numbers; , E, respectively The first two columns of the eigenvector matrix corresponding to the two largest eigenvalues; Projection dimensionality reduction: ； Construct the whitening matrix: ; Calculate the data after whitening: ; Find the orthogonal matrix in the whitening matrix using fast ICA separation. , making The two rows are independent of each other, and for each row Iteration: where, for The i-th column is the... Separation vectors of independent components; Iteration: ; For all updated Perform orthogonalization and normalization: ; Repeat iterate to each Convergence: The separation matrix is ​​obtained. ; in, , Background signals Demagnetization signal The estimated value; The likelihood estimate of A is given by the following equation: ； Demagnetization fault parameters are calculated using the following formula. : ,in: z This refers to the sample point index, where M represents the number of columns in the torque distribution matrix. The separated demagnetizing component is in the first... z The amplitude at each moment.

6. The demagnetization fault detection method based on high-precision torque distribution according to claim 5, characterized in that, The step of determining whether the demagnetization fault parameter is greater than a preset signal threshold, and if so, determining that the permanent magnet synchronous motor has a demagnetization fault, includes: The threshold discrimination formula is determined as follows: ,in Custom threshold; If the threshold discrimination formula is true, then the permanent magnet synchronous motor is considered to have a demagnetization fault.

7. A demagnetization fault detection system based on high-precision torque distribution, characterized in that, The system executes the demagnetization fault detection method based on high-precision torque distribution as described in any one of claims 1-6, and the system comprises: The calculation module is used to calculate flux linkage training data and construct a flux linkage distribution model based on multiple electrical parameters of the permanent magnet synchronous motor at multiple speeds. The mapping module is used to establish a data-driven iron loss compensation model based on the differential data corresponding to the magnetic flux training data, and to compensate the magnetic flux distribution model to obtain magnetic flux mapping data. The analysis module is used to process the torque matrix distribution data composed of the flux linkage mapping data based on the independent component analysis algorithm to obtain demagnetization fault parameters. The judgment module is used to determine whether the demagnetization fault parameter is greater than a preset signal threshold. If so, it is determined that the permanent magnet synchronous motor has a demagnetization fault.

8. A demagnetization fault detection system based on high-precision torque distribution, characterized in that, The system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the demagnetization fault detection method based on high-precision torque distribution as described in any one of claims 1-6.

Images