AC Motor Fault Diagnosis Method and Device for Stripping the Influence of Control Parameters

Abstract

The present invention provides a method and device for diagnosing faults of an AC motor by peeling off the influence of control parameters. The method includes: obtaining a preset constant value of d-axis voltage and a preset constant value of q-axis voltage; processing a preset excitation signal, the constant value of d-axis voltage, and the constant value of q-axis voltage by using a diagnostic vector control model corresponding to the AC motor to be diagnosed to obtain the response current of each phase of the motor; the excitation signal is a variable-frequency and variable-amplitude excitation signal whose frequency varies within a first range and whose amplitude varies within a second range; according to the response current of each phase of the motor and a pre-trained convolutional neural network, obtaining the fault probability distribution of the AC motor to be diagnosed. The present application peels off the influence of control parameters on fault feature data, and can diagnose motor faults more accurately, effectively, and with high precision. By injecting excitation signals with different frequencies and amplitudes, the motor fault type and fault degree can be more truly and effectively reflected, and the accuracy of motor fault diagnosis is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of AC motor fault diagnosis, and particularly to an AC motor fault diagnosis method and device that strip the influence of control parameters. Background Art

[0002] AC motors are mainly divided into asynchronous motors and synchronous motors. Various types of AC motors are widely used in industrial manufacturing and daily production. The possible faults of AC motors mainly include winding turn-to-turn short circuit or open circuit faults, phase-to-phase short circuit or open circuit faults, grounding faults, bearing eccentricity faults, etc. Permanent magnet motors may also have demagnetization, loss of magnetism and other faults. After a motor fails, if the motor system cannot accurately detect, diagnose the fault and take corresponding measures, it may induce more serious faults in the generator, damage the equipment, and even endanger the life safety of personnel.

[0003] The existing AC motor fault diagnosis solutions and related difficulties or defects are mainly as follows:

[0004] 1) The solution of installing magnetic sensors inside the motor. Since the motor is a high-speed rotating device, and large motors also have characteristics such as high voltage and large current, the solution of changing the motor structure to install sensors is not only difficult and uneconomical, but also may affect the design performance of the motor.

[0005] 2) The solution of establishing a corresponding mathematical model for motor faults and solving relevant information through the mathematical model to judge motor faults. Since there are many types of possible fault points of the motor and the fault occurrence positions are relatively random, it is difficult to establish an effective and unified motor fault mathematical model in practical applications.

[0006] 3) The solution of collecting information such as the current, voltage, and sound of the motor, and then using methods such as spectrum analysis and neural networks for processing to perform motor fault diagnosis. This solution mainly has two problems. One is that traditional spectrum analysis or ordinary neural networks are difficult to distinguish effective fault features, resulting in a low fault diagnosis rate and a high false judgment rate. The other is that the collected signals such as voltage and current are affected by system control parameters, and the quality of control parameter adjustment completely affects the fault diagnosis result. Summary of the Invention

[0007] To solve the problems existing in the prior art, in a first aspect, the present application provides an AC motor fault diagnosis method that strips the influence of control parameters, including:

[0008] Obtain a preset d-axis voltage constant value and a q-axis voltage constant value;

[0009] Processing a preset excitation signal, a d-axis voltage constant value, and a q-axis voltage constant value by using a diagnostic vector control model corresponding to an AC motor to be diagnosed to obtain response currents of each phase of the motor; the excitation signal is a variable-frequency and variable-amplitude excitation signal with a frequency varying within a first range and an amplitude varying within a second range;

[0010] Obtaining a fault probability distribution of the AC motor to be diagnosed according to the response currents of each phase of the motor and a pre-trained convolutional neural network.

[0011] In one embodiment, the processing the preset excitation signal, the d-axis voltage constant value, and the q-axis voltage constant value by using the diagnostic vector control model corresponding to the AC motor to be diagnosed to obtain response currents of each phase of the motor includes:

[0012] Superposing the excitation signal and the d-axis voltage constant value to obtain a superposed voltage;

[0013] Performing an inverse Park transformation on the superposed voltage and the q-axis voltage constant value to obtain a first voltage and a second voltage;

[0014] Performing pulse width modulation on the first voltage and the second voltage to obtain a pulse modulation signal;

[0015] Generating a corresponding power voltage according to the pulse modulation signal and applying the power voltage to the motor to be diagnosed to obtain the response currents of each phase of the motor corresponding to the motor to be diagnosed.

[0016] In one embodiment, the convolutional neural network is a 2-D convolutional neural network;

[0017] The obtaining the fault probability distribution of the AC motor to be diagnosed according to the response currents of each phase of the motor and the pre-trained convolutional neural network includes:

[0018] Performing data preprocessing on the response currents of each phase of the motor to obtain a combined current; the data preprocessing includes combination, normalization, and decorrelation processing;

[0019] Performing two-dimensional conversion on the combined current to obtain a corresponding time-frequency diagram;

[0020] Inputting the time-frequency diagram into the 2-D convolutional neural network to obtain the fault probability distribution of the AC motor to be diagnosed.

[0021] In one embodiment, the steps of training the convolutional neural network include:

[0022] Obtaining a batch of AC motor data, where the AC motor data includes normal data and fault data; wherein, both the normal data and the fault data include an excitation signal, a d-axis voltage constant value, a q-axis voltage constant value, and a working condition type;

[0023] Establish a training data set based on the excitation signal, the d-axis voltage constant value, the q-axis voltage constant value, and the working condition type;

[0024] Use the training data set for model training to obtain the convolutional neural network, so as to obtain the fault probability distribution of the AC motor to be diagnosed according to the combined current of the motor to be diagnosed.

[0025] In one embodiment, the establishing a training data set according to the excitation signal, the d-axis voltage constant value, the q-axis voltage constant value, and the working condition type includes:

[0026] Input the d-axis voltage constant value, the q-axis voltage constant value, and the excitation signal into the diagnostic vector control model to obtain the response current of each phase of the motor;

[0027] Perform data preprocessing on the response current of each phase of the motor to obtain the corresponding combined current; the data preprocessing includes combination, normalization, and decorrelation processing;

[0028] Perform two-dimensional conversion on the combined current to obtain the corresponding time-frequency diagram;

[0029] Generate the training data set according to the time-frequency diagram and the working condition type.

[0030] In one embodiment, the AC motor fault diagnosis method for stripping the influence of control parameters further includes:

[0031] Establish a conventional vector control model of the AC motor system according to the structure of the AC motor to be diagnosed;

[0032] Delete the current loop control structure in the conventional vector control model of the AC motor system to obtain the diagnostic vector control model;

[0033] Wherein, the diagnostic vector control model includes a Park inverse transformation module and a space vector pulse width modulation module.

[0034] In a second aspect, the present application further provides an AC motor fault diagnosis device for stripping the influence of control parameters, including:

[0035] An input data acquisition module, configured to acquire a preset d-axis voltage constant value and a q-axis voltage constant value;

[0036] A response current acquisition module, configured to process a preset excitation signal, the d-axis voltage constant value, and the q-axis voltage constant value by using a diagnostic vector control model corresponding to the AC motor to be diagnosed to obtain the response current of each phase of the motor; the excitation signal is a variable-frequency and variable-amplitude excitation signal whose frequency varies within a first range and whose amplitude varies within a second range;

[0037] A fault diagnosis module, configured to obtain a fault probability distribution of the AC motor to be diagnosed according to the response currents of each phase of the motor and a pre-trained convolutional neural network.

[0038] In one embodiment, the response current acquisition module includes:

[0039] A voltage superposition unit, configured to superpose the excitation signal and the d-axis voltage constant value to obtain a superposed voltage;

[0040] A Park inverse transformation unit, configured to perform Park inverse transformation on the superposed voltage and the q-axis voltage constant value to obtain a first voltage and a second voltage;

[0041] A modulation unit, configured to perform pulse width modulation on the first voltage and the second voltage to obtain a pulse modulation signal;

[0042] A converter unit, configured to generate a corresponding power voltage according to the pulse modulation signal and apply the power voltage to the motor to be diagnosed;

[0043] A response current acquisition unit, configured to acquire the corresponding response currents of each phase of the motor to be diagnosed to which the power voltage has been applied.

[0044] In one embodiment, the convolutional neural network is a 2-D convolutional neural network;

[0045] The fault diagnosis module includes:

[0046] A combined current acquisition unit, configured to perform data preprocessing on the response currents of each phase of the motor to obtain a combined current; the data preprocessing includes combination, normalization, and decorrelation processing;

[0047] A current conversion unit, configured to perform two-dimensional conversion on the combined current to obtain a corresponding time-frequency diagram;

[0048] A fault diagnosis unit, configured to input the time-frequency diagram into the 2-D convolutional neural network to obtain a fault probability distribution of the AC motor to be diagnosed.

[0049] In one embodiment, the AC motor fault diagnosis device that eliminates the influence of control parameters further includes:

[0050] A training dataset generation module, configured to obtain a batch of AC motor data, where the AC motor data includes normal data and fault data, and both the normal data and the fault data include an excitation signal, a d-axis voltage constant value, a q-axis voltage constant value, and a working condition type; and establish a training dataset according to the excitation signal, the d-axis voltage constant value, the q-axis voltage constant value, and the working condition type;

[0051] A convolutional neural network training module, which is used to train a model using the training dataset to obtain the convolutional neural network, so as to obtain the fault probability distribution of the AC motor to be diagnosed based on the combined current of the motor to be diagnosed.

[0052] In one embodiment, the training dataset generation module is specifically used for:

[0053] Input the d-axis voltage constant value, q-axis voltage constant value, and the excitation signal into the diagnostic vector control model to obtain the response current of each phase of the motor;

[0054] Perform data preprocessing on the response current of each phase of the motor to obtain the corresponding combined current; the data preprocessing includes combination, normalization, and decorrelation processing;

[0055] Perform two-dimensional conversion on the combined current to obtain the corresponding time-frequency diagram;

[0056] Generate the training dataset according to the time-frequency diagram and the working condition type.

[0057] In one embodiment, the AC motor fault diagnosis device that strips the influence of control parameters further includes an AC motor modeling module, which is used for:

[0058] Establish a conventional vector control model of the AC motor system according to the structure of the AC motor to be diagnosed;

[0059] Delete the current loop control structure in the conventional vector control model of the AC motor system to obtain the diagnostic vector control model; wherein, the diagnostic vector control model includes a Park inverse transformation module and a space vector pulse width modulation module.

[0060] In a third aspect, the present application further provides an electronic device, including:

[0061] A central processing unit, a memory, and a communication module. A computer program is stored in the memory. The central processing unit can call the computer program. When the central processing unit executes the computer program, it implements any AC motor fault diagnosis method provided by the present application that strips the influence of control parameters.

[0062] In a fourth aspect, the present application further provides a computer-readable storage medium for storing a computer program, and when the computer program is executed by a processor, it implements any AC motor fault diagnosis method provided by the present application that strips the influence of control parameters.

[0063] The AC motor fault diagnosis method and device affected by the peeling control parameters of the present application do not require adding any hardware devices, changing the mechanical structure and electrical circuit of the existing motor system, or adding any fault feature data acquisition devices. It only needs to use the phase current data sampled by the existing current sensor in the motor system. The present application peels off the influence of the control parameters on the fault feature data, and can diagnose motor faults more accurately, effectively, and with high precision. In addition, by injecting excitation signals with different frequencies and amplitudes, the present application can more truly and effectively reflect the type and degree of motor faults, improving the accuracy of motor fault diagnosis. Description of the Drawings

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0065] Figure 1 It is a schematic diagram of the AC motor fault diagnosis method that peels off the influence of the control parameters provided by the present application.

[0066] Figure 2 It is a framework schematic diagram of the conventional vector control model of the AC motor system provided by the present application.

[0067] Figure 3 It is a framework schematic diagram of the diagnostic vector control model corresponding to the AC motor to be diagnosed provided by the present application.

[0068] Figure 4 It is a schematic diagram of a variable-frequency and variable-amplitude excitation signal provided by the present application.

[0069] Figure 5 It is another schematic diagram of the AC motor fault diagnosis method that peels off the influence of the control parameters provided by the present application.

[0070] Figure 6 It is a waveform diagram of the response current of each phase of the motor under a certain excitation condition provided by the present application.

[0071] Figure 7 It is another schematic diagram of the AC motor fault diagnosis method that peels off the influence of the control parameters provided by the present application.

[0072] Figure 8 For Figure 7 The corresponding data processing flow chart.

[0073] Figure 9 It is a waveform diagram of the combined current obtained by combining the response currents of each phase of the motor provided by the present application.

[0074] Figure 10 A time-frequency diagram of combined current provided for this application.

[0075] Figure 11 Schematic diagram of the structure of a 2-D convolutional neural network provided for this application.

[0076] Figure 12 Schematic diagram of an AC motor fault diagnosis device affected by peeling control parameters provided for this application.

[0077] Figure 13 Another schematic diagram of an AC motor fault diagnosis device affected by peeling control parameters provided for this application.

[0078] Figure 14 Another schematic diagram of an AC motor fault diagnosis device affected by peeling control parameters provided for this application.

[0079] Figure 15 Another schematic diagram of an AC motor fault diagnosis device affected by peeling control parameters provided for this application.

[0080] Figure 16 Another schematic diagram of an AC motor fault diagnosis device affected by peeling control parameters provided for this application.

[0081] Figure 17 Another schematic diagram of an AC motor fault diagnosis device affected by peeling control parameters provided for this application.

[0082] Figure 18 Schematic diagram of an electronic device provided for this application. Detailed implementation manners

[0083] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0084] This application provides an AC motor fault diagnosis method affected by peeling control parameters. As Figure 1 shown, the method includes the following steps S101 to step S103:

[0085] Step S101, obtain the preset d-axis voltage constant value and q-axis voltage constant value.

[0086] Specifically, the AC motors used in this application include asynchronous motors and synchronous motors. The d-axis voltage constant value and the q-axis voltage constant value in this step are both preset constants, for example, they can be 0. This application removes the current closed-loop in the conventional vector control model and uses constant values to replace the d-axis voltage and q-axis voltage output by the current closed-loop PI regulator, stripping the influence of control parameters, which is beneficial to improving the accuracy of AC motor fault diagnosis. The "d-axis voltage constant value" and "q-axis voltage constant value" in this step are only preset constants and have nothing to do with the "d-axis voltage" and "q-axis voltage". The reason for calling them "d-axis voltage constant value" and "q-axis voltage constant value" is only to illustrate that they replace the actual "d-axis voltage" and "q-axis voltage" as input information.

[0087] Step S102: Process the preset excitation signal, the d-axis voltage constant value, and the q-axis voltage constant value by using the diagnostic vector control model corresponding to the AC motor to be diagnosed to obtain the response current of each phase of the motor; the excitation signal is a variable-frequency and variable-amplitude excitation signal whose frequency varies within the first range and whose amplitude varies within the second range.

[0088] Specifically, the diagnostic vector control model corresponding to the motor is introduced here first. The generation of the diagnostic vector control model can be based on the following steps:

[0089] Step 1: Establish a conventional vector control model of the AC motor system according to the structure of the AC motor to be diagnosed.

[0090] Figure 2 Shown is a schematic diagram of the framework of a common conventional vector control model of an AC motor system, and the specific parameter settings can be adjusted according to the actual parameters of the AC motor to be diagnosed. As Figure 2 shown, the AC motor system model includes a control module 1 (current loop control structure), a Park inverse transformation module 2, a space vector pulse width modulation (SVPWM) module 3, an inverter 4, a motor 5, and a current sampling module 6.

[0091] Step 2: Delete the current loop control structure in the conventional vector control model of the AC motor system to obtain the diagnostic vector control model.

[0092] In order to strip the influence of the control parameters of the AC motor on the fault characteristic data, this application deletes the control module 1 in the AC motor system model to obtain the diagnostic vector control model corresponding to the AC motor to be diagnosed. As Figure 3As shown, the diagnostic vector control model includes a Park inverse transformation module, a space vector pulse width modulation module, a converter, a motor, and a current sampling module. In step S102, the excitation signal, the preset constant values of the d-axis voltage and the q-axis voltage are used as the input signals of the diagnostic vector control model, and the corresponding phase response currents of the motor can be collected from the output terminal of the converter. The phase response currents of the motor in this application can be three-phase response currents, four-phase response currents, five-phase response currents, etc., which are not limited here. Figure 2 and Figure 3 only the three-phase response current is taken as an example for illustration here.

[0093] In the excitation signal in step S102, it is a variable-frequency and variable-amplitude excitation signal with a frequency varying within a first range and an amplitude varying within a second range. For example, the frequency variation range of the excitation signal can be expressed as f min ≤f k ≤f max , and the amplitude variation range can be expressed as u min ≤u j (f k )≤u max , where f min , f max , u min , u max are all real numbers. The frequency and amplitude of the excitation signal u j (f k ) vary according to a fixed rule within the corresponding variation ranges above, that is, the variable-frequency and variable-amplitude excitation signal of this application is formed. The fixed rule here can be set arbitrarily as long as its frequency and amplitude do not exceed the corresponding variation ranges. The variable-frequency and variable-amplitude excitation signal here includes but is not limited to square wave signals and sine wave signals. Figure 4 That is, it is a schematic diagram of the variable-frequency and variable-amplitude excitation signal formed by the sine wave signal changing according to a fixed rule.

[0094] Step S103, according to the phase response currents of the motor and the pre-trained convolutional neural network, obtain the fault probability distribution of the AC motor to be diagnosed, where the convolutional neural network can be, for example, a 2-D convolutional neural network.

[0095] Specifically, please refer to Figure 3 simultaneously. After the phase response currents of the motor output by the current sampling module are processed by the preprocessing module, they are input into the convolutional neural network model. The input data of the convolutional neural network model in this step is the time-frequency diagram obtained by processing the phase response currents of the motor, and the output data is the fault type corresponding to the time-frequency diagram and the probability of each fault type occurring, that is, the fault probability distribution of the AC motor to be diagnosed. The training process of the convolutional neural network will be described in detail in the subsequent embodiments.

[0096] The fault types of the AC motor involved in this application include, but are not limited to, inter-turn short circuit or open circuit fault of the motor winding, phase-to-phase short circuit or open circuit fault, grounding fault, bearing eccentricity fault, permanent magnet demagnetization or loss of magnetism fault, etc. The fault probability distribution of the AC motor to be diagnosed output by the convolutional neural network can be shown, for example, in the form of Table 1 below:

[0097] Table 1: Fault Probability Distribution Table Output by Convolutional Neural Network

[0098] Serial number Fault type Fault probability 1 Ground fault a％ 2 Bearing eccentricity fault b％ 3 Inter-turn short circuit of motor winding c％ …… …… ……

[0099] The sum of the fault probability values corresponding to each fault type in Table 1 is less than or equal to 1. The reason why the sum of the fault probability values may be less than 1 is that there is a normal operating condition in addition to the fault conditions. Here, the normal operating condition refers to the condition without any faults, and the fault condition refers to the condition with at least one fault.

[0100] In practical applications, in addition to the table, the fault probability distribution can also be shown in the form of a statistical chart (such as a pie chart), or can be shown in combination with multiple forms. This application does not limit this.

[0101] In one embodiment, as Figure 5 shown, in step S102, the preset excitation signal and the d-axis voltage constant value and q-axis voltage constant value are processed by using the diagnostic vector control model corresponding to the AC motor to be diagnosed to obtain the response current of each phase of the motor, including the following steps:

[0102] Step S1021, superimpose the excitation signal and the d-axis voltage constant value to obtain a superimposed voltage.

[0103] Specifically, in combination with Figure 3 , the excitation signal u j (f k ) is superimposed with the d-axis voltage constant value u d to obtain the superimposed voltage u dj (f k ).

[0104] Step S1022, perform an inverse Park transformation on the superimposed voltage and the q-axis voltage constant value to obtain a first voltage and a second voltage.

[0105] Specifically, referring to Figure 3 , the superimposed voltage u dj (f k ), the q-axis voltage constant value u q and the motor rotor magnetic field angle θ are input into the inverse Park transformation module for inverse Park transformation (i.e., dq-αβ transformation) to obtain two output voltages, namely the first voltage u α (f k)(Generally referred to as the alpha voltage) and the second voltage u β (f k )(Generally referred to as the beta voltage). The superimposed voltage u dj (f k ), the q-axis voltage constant value u q and the motor rotor magnetic field angle θ are subjected to cross-coupling calculations in the Park inverse transformation module.

[0106] Step S1023, perform pulse width modulation on the first voltage and the second voltage to obtain a pulse modulation signal.

[0107] Specifically, refer to Figure 3 , input the first voltage u α (f k ) and the second voltage u β (f k ) into the space vector pulse width modulation (SVPWM) module for PWM modulation to obtain a PWM wave.

[0108] Step S1024, generate a corresponding power voltage according to the pulse modulation signal, and apply the power voltage to the motor to be diagnosed to obtain the phase response currents of the motor corresponding to the motor to be diagnosed.

[0109] The PWM wave output by the space vector pulse width modulation module usually cannot be directly used to drive the motor. Therefore, the PWM wave output by the space vector pulse width modulation module will be converted into a corresponding power voltage through an inverter, and the power voltage will be applied to the motor to be diagnosed to realize motor drive control. The function of the inverter is to convert the PWM wave into a power voltage that can drive the motor.

[0110] After applying the power voltage corresponding to the PWM wave output by the space vector pulse width modulation module to the motor, the phase response currents of the motor can be collected. In this application, the three-phase response currents i a (f k ), i b (f k ), i c (f k ) are taken as examples, and their waveforms are respectively as shown in the A-phase current, B-phase current, and C-phase current in Figure 6 .

[0111] In an embodiment, please refer to Figure 7 and Figure 8 at the same time. When the convolutional neural network is a 2-D convolutional neural network, step S103, according to the phase response currents of the motor and the pre-trained convolutional neural network, obtain the fault probability distribution of the AC motor to be diagnosed, which specifically includes the following steps:

[0112] Step S1031: Perform data preprocessing on the phase response currents of the motor to obtain a combined current. Here, the data preprocessing includes, but is not limited to, combination, normalization processing, and decorrelation processing.

[0113] Specifically, combine the phase response currents i a (f k )、i b (f k )、i c (f k ) of the motor to obtain the combined current i abc . From the waveform perspective, the combined current i abc obtained by combination is the waveform formed by connecting the phase response currents i a (f k )、i b (f k )、i c (f k ) in sequence. See Figure 9 . Then perform conventional data processing such as normalization processing and decorrelation processing on the combined current i abc . Figure 9 The horizontal axis in the figure is the data sequence number, representing the sequence numbers of the sample data of the phase response currents i a (f k )、i b (f k )、i c (f k ) of the motor. The examples in the figure respectively collect 1000 sample data for the three signals of i a (f k )、i b (f k )、i c (f k ). Therefore, the combined current obtained has a total of 3000 sample data.

[0114] Step S1032: Perform two-dimensional transformation on the combined current to obtain the corresponding time-frequency diagram.

[0115] Here, convert the one-dimensional data i abc into a two-dimensional time-frequency diagram as the input data of the 2-D convolutional neural network. Figure 10 This is an example of the time-frequency diagram of a combined current provided by this application.

[0116] Step S1033: Input the time-frequency diagram into the 2-D convolutional neural network to obtain the fault probability distribution of the AC motor to be diagnosed.

[0117] Specifically, the 2-D convolutional neural network analyzes and processes the input time-frequency diagram to obtain the corresponding fault probability distribution Ψ. Ψ contains the data in Table 1 mentioned above, representing the probability distributions of various types of faults. The fault types can be referred to the previous description.

[0118] Further, after obtaining the fault probability distribution Ψ of the AC motor to be diagnosed, multiply the fault probability distribution Ψ by a preset weight coefficient λ(i a ,i b ,i c ,f k ), and take the result of the product as the final fault diagnosis result. Among them, λ(i a ,i b ,i c ,f k ) is the weight function of i a ,i b ,i c ,f k , which is a preset fixed value.

[0119] Figure 11 is the structural schematic diagram of the 2-D convolutional neural network provided by this application. As Figure 11 shown, the 2-D convolutional neural network structure includes the first convolutional layer, the first pooling layer, the second convolutional layer, the second pooling layer, and the fully connected neural network. This structure is only an example of this application. In practical applications, the structure of the 2-D convolutional neural network can be changed as needed, and this application does not make any limitations in this regard. Input the time-frequency diagram of the combined current into the 2-D convolutional neural network. After data processing by the first convolutional layer, the first pooling layer, the second convolutional layer, and the second pooling layer, take the processing result as the input of the fully connected neural network; the fully connected neural network finally outputs the probability distribution Ψ of various types of faults of the AC motor to be diagnosed.

[0120] In one embodiment, as Figure 12 shown, the steps of training the convolutional neural network include:

[0121] Step S104, obtain a batch of AC motor data, where the AC motor data includes normal data and fault data; among them, both the normal data and the fault data include the excitation signal, the d-axis voltage constant value, the q-axis voltage constant value, and the working condition type. Here, the "excitation signal", "d-axis voltage constant value", and "q-axis voltage constant value" are all the same as those in the previous embodiments.

[0122] It is understandable that the purpose of motor fault diagnosis is to identify whether the motor is in a normal operating condition or a fault condition, and what the specific fault type is when it is in a fault condition. Therefore, when training the convolutional neural network, it is necessary to obtain normal data corresponding to the normal operating condition and fault data corresponding to the fault condition. Among them, the operating condition type in the normal data is no fault, and the operating condition type in the fault data is the fault type of the AC motor. For specific examples, refer to the fault types of the AC motor given in the previous step S103 and Table 1.

[0123] Step S105: Establish a training dataset according to the excitation signal, the d-axis voltage constant value, the q-axis voltage constant value, and the operating condition type.

[0124] Specifically, as Figure 13 shown, step S105 includes the following steps:

[0125] Step S1051: Input the d-axis voltage constant value, the q-axis voltage constant value, and the excitation signal into the diagnostic vector control model to obtain the response current of each phase of the motor. Among them, the type and variation law of the excitation signal in this step are the same as those of the variable-frequency and variable-amplitude excitation signal in step S102 of the foregoing embodiment, and will not be elaborated here. In this step, the excitation signal is also superimposed with the corresponding d-axis voltage constant value to obtain a superimposed voltage, and the superimposed voltage and the corresponding q-axis voltage constant value are used as input signals and input into the diagnostic vector control model. For the diagnostic vector control model in this step, refer to Figure 3 and the relevant parts of the specification.

[0126] Step S1052: Perform data preprocessing on the response current of each phase of the motor to obtain the corresponding combined current. Among them, the data preprocessing includes but is not limited to combination, normalization, and decorrelation processing. In this step, for the combination, normalization processing, and decorrelation processing of the three-phase current, refer to the description in step S1031 of the foregoing embodiment, and will not be elaborated here.

[0127] Step S1053: Perform two-dimensional conversion on the combined current to obtain the corresponding time-frequency diagram. For specific reference, refer to the description in step S1032 of the foregoing embodiment.

[0128] Step S1054: Generate the training dataset according to the time-frequency diagram and the operating condition type. The training dataset contains a large number of data groups, and each group of data corresponds to a type of fault and its corresponding time-frequency diagram.

[0129] Step S106: Use the training dataset to perform model training to obtain the convolutional neural network, so as to obtain the fault probability distribution of the AC motor to be diagnosed according to the combined current of the motor to be diagnosed.

[0130] The convolutional neural network obtained by training the model using the training data set can analyze and process the time-frequency diagram of the input combined current, and output the probability distribution of various faults of the AC motor to be diagnosed.

[0131] In summary, the AC motor fault diagnosis method that strips the influence of control parameters of the present application does not require adding any hardware devices, does not require changing the mechanical structure and electrical circuit of the existing motor system, and does not require adding any fault feature data acquisition devices. It only needs to use the phase current data sampled by the existing current sensor of the motor system. The present application strips the influence of control parameters on the fault feature data, and can diagnose motor faults more accurately, effectively and with high precision. In addition, by injecting excitation signals with different frequencies and amplitudes, the present application can more truly and effectively reflect the types and degrees of motor faults, and improve the accuracy of motor fault diagnosis.

[0132] Based on the same inventive concept, the embodiment of the present application also provides an AC motor fault diagnosis device that strips the influence of control parameters, which can be used to implement the method described in the above embodiment, as described in the following embodiment. Since the principle of the AC motor fault diagnosis device that strips the influence of control parameters to solve problems is similar to the AC motor fault diagnosis method that strips the influence of control parameters, the implementation of the AC motor fault diagnosis device that strips the influence of control parameters can refer to the implementation of the AC motor fault diagnosis method that strips the influence of control parameters, and the repeated parts will not be elaborated. Hereinafter, the term "unit" or "module" may be a combination of software and / or hardware that can implement a predetermined function. Although the system described in the following embodiments is preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0133] As Figure 14 shown, the AC motor fault diagnosis device that strips the influence of control parameters provided by the present application includes:

[0134] An input data acquisition module 201, configured to acquire a preset d-axis voltage constant value and a q-axis voltage constant value;

[0135] A response current acquisition module 202, configured to process a preset excitation signal and the d-axis voltage constant value and the q-axis voltage constant value by using a diagnostic vector control model corresponding to the AC motor to be diagnosed, to obtain the response current of each phase of the motor; the excitation signal is a variable-frequency and variable-amplitude excitation signal whose frequency varies within a first range and whose amplitude varies within a second range;

[0136] A fault diagnosis module 203, configured to obtain the fault probability distribution of the AC motor to be diagnosed according to the response current of each phase of the motor and a pre-trained convolutional neural network.

[0137] In an embodiment, as Figure 15As shown, the response current acquisition module 202 includes:

[0138] A voltage superposition unit 2021, configured to superpose the excitation signal and the d-axis voltage constant value to obtain a superposed voltage;

[0139] A Park inverse transformation unit 2022, configured to perform Park inverse transformation on the superposed voltage and the q-axis voltage constant value to obtain a first voltage and a second voltage;

[0140] A modulation unit 2023, configured to perform pulse width modulation on the first voltage and the second voltage to obtain a pulse modulation signal;

[0141] A converter unit 2024, configured to generate a corresponding power voltage according to the pulse modulation signal and apply the power voltage to the motor to be diagnosed;

[0142] A response current acquisition unit 2025, configured to acquire the corresponding phase response currents of the motor to be diagnosed to which the power voltage has been applied.

[0143] In one embodiment, as Figure 16 shown, the convolutional neural network is a 2-D convolutional neural network;

[0144] The fault diagnosis module 203 includes:

[0145] A combined current acquisition unit 2031, configured to perform data preprocessing on the phase response currents of the motor, where the data preprocessing includes combination, normalization, and decorrelation processing, to obtain a combined current;

[0146] A current conversion unit 2032, configured to perform two-dimensional conversion on the combined current to obtain a corresponding time-frequency diagram;

[0147] A fault diagnosis unit 2033, configured to input the time-frequency diagram into the 2-D convolutional neural network to obtain the fault probability distribution of the AC motor to be diagnosed.

[0148] In one embodiment, as Figure 17 shown, the AC motor fault diagnosis device that eliminates the influence of control parameters further includes:

[0149] A training data set generation module 204, configured to acquire a batch of AC motor data, where the AC motor data includes normal data and fault data, and both the normal data and the fault data include an excitation signal, a d-axis voltage constant value, a q-axis voltage constant value, and a working condition type; and establish a training data set according to the excitation signal, the d-axis voltage constant value, the q-axis voltage constant value, and the working condition type;

[0150] The convolutional neural network training module 205 is used to perform model training using the training dataset to obtain the convolutional neural network, so as to obtain the fault probability distribution of the AC motor to be diagnosed according to the combined current of the motor to be diagnosed.

[0151] In one embodiment, the training dataset generation module 204 is specifically configured to:

[0152] Input the d-axis voltage constant value, q-axis voltage constant value, and the excitation signal into the diagnostic vector control model to obtain the response current of each phase of the motor;

[0153] Perform data preprocessing on the response current of each phase of the motor. The data preprocessing includes combination, normalization, and decorrelation processing to obtain the corresponding combined current;

[0154] Perform two-dimensional conversion on the combined current to obtain the corresponding time-frequency diagram;

[0155] Generate the training dataset according to the time-frequency diagram and the working condition type.

[0156] In one embodiment, please continue to refer to Figure 17 , the AC motor fault diagnosis device that eliminates the influence of control parameters further includes an AC motor modeling module 206, which is used to:

[0157] Establish a conventional vector control model of the AC motor system according to the structure of the AC motor to be diagnosed;

[0158] Delete the current loop control structure in the conventional vector control model of the AC motor system to obtain the diagnostic vector control model; wherein, the diagnostic vector control model includes a Park inverse transformation module and a space vector pulse width modulation module.

[0159] The present invention also provides an electronic device. Refer to Figure 18 , the electronic device 100 specifically includes:

[0160] A central processor 110, a memory 120, a communication module 130, an input unit 140, an output unit 150, and a power supply 160.

[0161] Among them, the memory 120, the communication module 130, the input unit 140, the output unit 150, and the power supply 160 are respectively connected to the central processor 110. A computer program is stored in the memory 120, and the central processor 110 can call the computer program. When the central processor 110 executes the computer program, all steps in the AC motor fault diagnosis method affected by the peeling control parameters in the above embodiments are implemented.

[0162] An embodiment of the present application further provides a computer-readable storage medium for storing a computer program, and the computer program can be executed by a processor. When the computer program is executed by the processor, any AC motor fault diagnosis method affected by the peeling control parameters provided by the present invention is implemented.

[0163] Those skilled in the art should understand that the embodiments of this specification can be provided as a method, a system, or a computer program product. Therefore, the embodiments of this specification can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, 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. The embodiments in this specification are described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiment. In the description of this specification, the description of the above terms does not necessarily refer to the same embodiment or example. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples. The above is only the embodiments of the embodiments of this specification and is not used to limit the embodiments of this specification. For those skilled in the art, the embodiments of this specification can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of this specification shall be included in the scope of the claims of the embodiments of this specification.

[0164] In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. In addition, without contradiction, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples. The above is only the embodiments of the embodiments of this specification and is not used to limit the embodiments of this specification. For those skilled in the art, the embodiments of this specification can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of this specification shall be included in the scope of the claims of the embodiments of this specification.

Claims

1. A method for diagnosing AC motor faults by eliminating the influence of peeling control parameters, characterized in that Including: Obtain preset d-axis voltage constant value and q-axis voltage constant value; Process the preset excitation signal, the d-axis voltage constant value, and the q-axis voltage constant value by using a diagnostic vector control model corresponding to the AC motor to be diagnosed to obtain the response current of each phase of the motor; the excitation signal is a variable-frequency and variable-amplitude excitation signal with a frequency varying within a first range and an amplitude varying within a second range; According to the response current of each phase of the motor and a pre-trained convolutional neural network, obtain the fault probability distribution of the AC motor to be diagnosed; The step of processing the preset excitation signal, the d-axis voltage constant value, and the q-axis voltage constant value by using a diagnostic vector control model corresponding to the AC motor to be diagnosed to obtain the response current of each phase of the motor includes: Superimpose the excitation signal and the d-axis voltage constant value to obtain a superimposed voltage; Perform Park inverse transformation on the superimposed voltage and the q-axis voltage constant value to obtain a first voltage and a second voltage; Perform pulse width modulation on the first voltage and the second voltage to obtain a pulse modulation signal; Generate a corresponding power voltage according to the pulse modulation signal and apply the power voltage to the motor to be diagnosed to obtain the response current of each phase of the motor corresponding to the motor to be diagnosed.

2. The AC motor fault diagnosis method affected by peeling control parameters according to claim 1, characterized in that The convolutional neural network is a 2-D convolutional neural network; The step of obtaining the fault probability distribution of the AC motor to be diagnosed according to the response current of each phase of the motor and a pre-trained convolutional neural network includes: Perform data preprocessing on the response current of each phase of the motor to obtain a combined current; wherein, the data preprocessing includes combination, normalization, and decorrelation processing; Perform two-dimensional transformation on the combined current to obtain a corresponding time-frequency diagram; Input the time-frequency diagram into the 2-D convolutional neural network to obtain the fault probability distribution of the AC motor to be diagnosed.

3. The AC motor fault diagnosis method affected by the peeling control parameters according to any one of claims 1 to 2, characterized in that The steps of training the convolutional neural network include: Obtain a batch of AC motor data, where the AC motor data includes normal data and fault data; wherein, both the normal data and the fault data include an excitation signal, a d-axis voltage constant value, a q-axis voltage constant value, and a working condition type; Establish a training data set according to the excitation signal, the d-axis voltage constant value, the q-axis voltage constant value, and the working condition type; Use the training data set to perform model training to obtain the convolutional neural network, so as to obtain the fault probability distribution of the AC motor to be diagnosed according to the combined current of the motor to be diagnosed.

4. The AC motor fault diagnosis method affected by the peeling control parameters according to claim 3, characterized in that, The step of establishing a training data set according to the excitation signal, the d-axis voltage constant value, the q-axis voltage constant value, and the working condition type includes: Input the d-axis voltage constant value, the q-axis voltage constant value, and the excitation signal into the diagnostic vector control model to obtain the response current of each phase of the motor; Perform data preprocessing on the response current of each phase of the motor to obtain a corresponding combined current; the data preprocessing includes combination, normalization, and decorrelation processing; Perform two-dimensional transformation on the combined current to obtain a corresponding time-frequency diagram; Generate the training data set according to the time-frequency diagram and the working condition type.

5. The AC motor fault diagnosis method affected by peeling control parameters according to any one of claims 1 to 2, characterized in that, Also including: Establish a conventional vector control model of the AC motor system according to the structure of the AC motor to be diagnosed; Delete the current loop control structure in the conventional vector control model of the AC motor system to obtain the diagnostic vector control model; Among them, the diagnostic vector control model includes a Park inverse transformation module and a space vector pulse width modulation module.

6. An AC motor fault diagnosis device that eliminates the influence of peeling control parameters, characterized in that, It includes: An input data acquisition module for acquiring a preset d-axis voltage constant value and a q-axis voltage constant value; A response current acquisition module for processing a preset excitation signal, the d-axis voltage constant value, and the q-axis voltage constant value by using the diagnostic vector control model corresponding to the AC motor to be diagnosed to obtain the response current of each phase of the motor; the excitation signal is a variable frequency and variable amplitude excitation signal with a frequency varying within a first range and an amplitude varying within a second range; A fault diagnosis module for obtaining the fault probability distribution of the AC motor to be diagnosed according to the response current of each phase of the motor and a pre-trained convolutional neural network; The response current acquisition module includes: A voltage superposition unit for superposing the excitation signal and the d-axis voltage constant value to obtain a superposed voltage; A Park inverse transformation unit for performing Park inverse transformation on the superposed voltage and the q-axis voltage constant value to obtain a first voltage and a second voltage; A modulation unit for performing pulse width modulation on the first voltage and the second voltage to obtain a pulse modulation signal; An inverter unit for generating a corresponding power voltage according to the pulse modulation signal and applying the power voltage to the motor to be diagnosed; A response current acquisition unit for acquiring the response current of each phase of the motor corresponding to the motor to be diagnosed to which the power voltage has been applied.

7. The AC motor fault diagnosis device affected by the peeling control parameters according to claim 6, characterized in that, The convolutional neural network is a 2-D convolutional neural network; The fault diagnosis module includes: A combined current acquisition unit for performing data preprocessing on the response current of each phase of the motor to obtain a combined current; the data preprocessing includes combination, normalization, and decorrelation processing; A current conversion unit for performing two-dimensional conversion on the combined current to obtain a corresponding time-frequency diagram; A fault diagnosis unit for inputting the time-frequency diagram into the 2-D convolutional neural network to obtain the fault probability distribution of the AC motor to be diagnosed.

8. The AC motor fault diagnosis device affected by the peeling control parameter according to any one of claims 6 to 7, characterized in that It also includes: A training data set generation module for acquiring a batch of AC motor data, where the AC motor data includes normal data and fault data, and both the normal data and the fault data include an excitation signal, a d-axis voltage constant value, a q-axis voltage constant value, and a working condition type; and establishing a training data set according to the excitation signal, the d-axis voltage constant value, the q-axis voltage constant value, and the working condition type; A convolutional neural network training module for using the training data set to perform model training to obtain the convolutional neural network, so as to obtain the fault probability distribution of the AC motor to be diagnosed according to the combined current of the motor to be diagnosed.

9. The AC motor fault diagnosis device affected by the peeling control parameters according to claim 8, characterized in that The training data set generation module is specifically used for: Inputting the d-axis voltage constant value, the q-axis voltage constant value, and the excitation signal into the diagnostic vector control model to obtain the response current of each phase of the motor; Performing data preprocessing on the response current of each phase of the motor to obtain a corresponding combined current; the data preprocessing includes combination, normalization, and decorrelation processing; Perform two-dimensional conversion on the combined current to obtain a corresponding time-frequency diagram; Generate the training dataset according to the time-frequency diagram and the working condition type.

10. The AC motor fault diagnosis device affected by the peeling control parameter according to any one of claims 6 to 7, characterized in that It further includes an AC motor modeling module for: Establish a conventional vector control model of the AC motor system according to the structure of the AC motor to be diagnosed; Delete the current loop control structure in the conventional vector control model of the AC motor system to obtain the diagnostic vector control model; wherein, the diagnostic vector control model includes a Park inverse transformation module and a space vector pulse width modulation module.

11. An electronic device, characterized in that, It includes: A central processing unit, a memory, and a communication module. A computer program is stored in the memory. The central processing unit can call the computer program. When the central processing unit executes the computer program, it implements the AC motor fault diagnosis method for stripping the influence of control parameters as described in any one of claims 1 to 5.

12. A computer-readable storage medium for storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the AC motor fault diagnosis method for stripping the influence of control parameters as described in any one of claims 1 to 5.

Images