CN105841867A - Measuring method for tooth groove torque of permanent magnet motor - Google Patents

Abstract

The invention discloses a wavelet-signal-analysis-based measuring method for tooth groove torque of a permanent magnet motor. The method comprises: a torque sensor is used for detecting a measured permanent magnet motor; a signal pre-treatment module carries out conditioning on a sensing signal; according to a preset sampling frequency fs and a sampling point number N, a signal acquisition module extracts a torque signal output frequency f of the sensing signal after conditioning, wherein the f meets a formula: f=fs/N, and a torque signal of the measured permanent magnet motor is calculated; a feature extraction module carries out multi-resolution decomposition on the torque signal to obtain a wavelet coefficient of the torque signal and corrects the wavelet coefficient; and wavelet reconstruction is carried out by using the corrected wavelet coefficient, thereby obtaining a tooth groove torque of the measured permanent magnet motor. According to the invention, the method is easy to implement; the measurement difficulty is low; the measurement precision is high, and the cost is low.

Description

Measurement method of cogging torque of permanent magnet motor

technical field

The invention relates to the technical field of optimization design and control of permanent magnet motors, in particular to a method for measuring cogging torque of permanent magnet motors.

Background technique

With the continuous improvement of the performance of permanent magnet materials, permanent magnet motors are more and more widely used in high-performance speed and position electromechanical transmission control systems, such as numerical control machine tools, robots, electronic manufacturing, elevators and other fields. However, in a permanent magnet motor, the interaction between the permanent magnet and the slotted armature core inevitably produces cogging torque, which results in torque fluctuations, causing vibration and noise, and affecting the control accuracy of the system.

Cogging torque is a problem specific to permanent magnet motors and is a position-related quantity that will not change after the motor is manufactured. Cogging torque is a key issue that must be considered and solved in the design and manufacture of high-performance permanent magnet motors. In a high-performance electromechanical servo system, a permanent magnet motor with a large cogging torque cannot meet the requirements of high-performance control. For example, the basic requirement of permanent magnet motor design in high-speed high-precision CNC machine tools, industrial robots and other systems is low cogging torque, because excessive cogging torque will affect positioning accuracy. The cogging torque measurement of the permanent magnet motor, at present, usually adopts two methods in China: one is the lever measurement method. This method is simple, intuitive, and easy to implement, but the accuracy is difficult to guarantee, and it can only reflect the approximate information of the cogging torque, which cannot be used in occasions that require high measurement accuracy. The second is to use the static measurement method of the stepper motor, that is, the commonly used dragging method. The specific measurement method is: use a stepping motor with very small torque ripple and a reducer with a high reduction ratio, reduce the speed of the motor to be tested by connecting the reducer, and measure the torque ripple on the shaft with a speed torque meter , which is the cogging torque. This method requires high precision torque sensor, complex measurement method, high instrument cost, and single test function.

Contents of the invention

The technical problem to be solved by the present invention is to provide an easy-to-implement measurement method for the cogging torque of the permanent magnet motor, reduce the cost of the cogging torque measurement system of the permanent magnet motor, and improve its measurement accuracy.

In order to solve the above technical problems, an embodiment of the present invention provides a method for measuring cogging torque of a permanent magnet motor, including:

Step 101: The torque sensor detects the measured permanent magnet motor, and sends the sensing signal output by the torque sensor to the signal preprocessing module;

Step 102: The signal preprocessing module conditions the sensing signal;

Step 103: The signal acquisition module extracts the measured torque output frequency value f=f _s /N in the conditioned sensing signal according to the preset sampling frequency f _s and the number of sampling points N;

Step 104: The signal acquisition module calculates the torque signal M of the measured permanent magnet motor according to the conversion formula M=A×(ff ₀ )/(f _p -f ₀ ); wherein, A is the full torque Range, f is the measured torque output frequency value, f ₀ is the torque zero output frequency value, f _p is the positive full-scale output frequency value;

Step 105: The signal acquisition module transmits the acquired torque signal to the feature extraction module; the torque signal includes a cogging torque signal;

Step 106: The feature extraction module uses discrete wavelet transform to perform multi-resolution decomposition on the torque signal to obtain wavelet coefficients of the torque signal;

Step 107: the feature extraction module corrects the wavelet coefficients according to the pulse frequency of the cogging torque signal of the measured permanent magnet motor;

Step 108: The feature extraction module uses the corrected wavelet coefficients to perform wavelet reconstruction to obtain the cogging torque signal of the measured permanent magnet motor.

Preferably, the signal preprocessing module includes: a voltage follower, a signal amplification circuit, a filter, and a shaping circuit connected in sequence; the step 102 specifically includes:

Step 201: the voltage follower adjusts the voltage value of the sensing signal, so that the output voltage of the voltage follower matches the input voltage;

Step 202: the signal amplifying circuit amplifies the sensing signal output by the voltage follower;

Step 203: the filter filters the interference signal in the amplified sensing signal;

Step 204: the shaping circuit shapes the filtered sensing signal into a rectangular pulse signal, and provides it to the data acquisition card.

In a preferred implementation manner, the filter is a second-order active low-pass filter implemented by a hardware circuit.

In another preferred implementation manner, the filter uses software filtering to filter the sensing signal.

Preferably, the step 106 specifically includes:

Step 601: After the sampling rate of the torque signal satisfies the sampling theorem, the feature extraction module defines the total frequency band occupied by the collected signal as space V ₀ , and the total frequency band is 0-f _s /2; After one-level decomposition, V ₀ is divided into low-frequency subspace V ₁ and high-frequency subspace W ₁ ; the frequency band of V ₁ is 0～f _s /4, and the frequency band of W ₁ is f _s /4～f _s /2;

Step 602: After two-level decomposition, decompose the V ₁ into a low-frequency subspace V ₂ and a high-frequency subspace W ₂ , the frequency band of V ₂ is 0-f _s /8, and the frequency band of W ₂ is f _s /8～f _s /4; and so on, divide the frequency subspace into: V ₀ ＝V ₁ ⊕W ₁ , V ₁ ＝V ₂ ⊕W ₂ ,...,V _j-1 ＝V _j ⊕ W _j , so:

V ₀ ＝W ₁ ⊕W ₂ ⊕W ₃ ......W _j ⊕V _j

Among them, each W _j is the high-frequency subspace reflecting the details of V _j-1 spatial signal, V _j is the low-frequency subspace reflecting the details of V _j-1 spatial signal, and W _j is the orthogonal complement of V _j at V _j-1 space;

Step 603: From multi-resolution analysis and spatial orthogonal decomposition theory, Where J is any scale, and the signal x(t)∈L ² (R) is expanded on the space L ² (R), and the following expression is obtained:

Among them, the two-scale equation is:

Assume d _j+1,n ＝<f _j ,ψ _j+1,n >, and then decomposed to get the wavelet coefficient as:

${\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}}$

${\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}}$

Among them, k=0,1,2...n-1, represents the translation position; j corresponds to the frequency range of the signal; c _j+1,k , d _j+1,k are the wavelet coefficients; h(k) is the low Pass filter sampling response; g (k) is the high-pass filter sampling response; is the scaling function; ψ _j,k (t) is the wavelet function.

Preferably, said step 107 includes the following steps:

Step 701: Calculate the cogging torque ripple frequency f _c =v×Z/60 of the measured permanent magnet motor according to the rotational speed v and the cogging number Z of the measured permanent magnet motor;

Step 702: Select a wavelet according to the torque signal output frequency f and the cogging torque ripple frequency f _c ; and determine the number of layers n of wavelet decomposition, where n>0;

Step 703: Perform discrete wavelet transform of 1 to n layers on the torque signal, and decompose to obtain wavelet coefficients of each layer;

Step 704: Keep the wavelet coefficient components corresponding to the _cogging torque ripple frequency fc, and set the wavelet coefficient components corresponding to the remaining frequency bands to zero to obtain corrected wavelet coefficients.

Preferably, the wavelet is a Dobesian wavelet.

Preferably, said step 108 includes the following steps:

Step 801: Using the multi-resolution characteristic of wavelet, search for the frequency band signal corresponding to the cogging torque ripple frequency band;

Step 802: make the other frequency band signals except the corresponding frequency band signal be zero, and process to obtain new wavelet coefficients;

Step 803: Reconstruct the signal using the wavelet coefficients through the following equation:

${\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>$

Among them, c _j,k is the cogging torque signal, and the reconstructed signal is the cogging torque signal.

The technical solution provided by the embodiment of the present invention has the beneficial effect of processing the sensing signal of a commonly used torque sensor, using the signal acquisition module and the data acquisition card to extract the torque signal of the measured permanent magnet motor, and according to the wavelet The multi-resolution characteristics of the decomposition, using the feature extraction module and the data analysis software in the industrial control computer to perform wavelet decomposition on the torque signal, and extract the wavelet coefficients that match the cogging torque ripple frequency in the torque signal for wavelet reconstruction structure, and finally obtain the cogging torque signal of the permanent magnet motor under test, the measurement accuracy is high and easy to implement, and the torque signal output by the data acquisition card and the cogging torque signal output by the industrial control computer can be output to the display for further measurement. Display, the measurement is more intuitive and can reflect the parameters of the measurement signal in real time.

Description of drawings

Fig. 1 is a schematic flow chart of an embodiment of a permanent magnet motor cogging torque measurement method provided by the present invention;

FIG. 2 is a schematic flow chart of an implementable manner of step S102 shown in FIG. 1;

FIG. 3 is a schematic flow chart of an implementable manner of step S106 shown in FIG. 1;

FIG. 4 is a schematic flow chart of an implementable manner of step S107 shown in FIG. 1;

Fig. 5 is a schematic structural diagram of a system test platform provided by the present invention;

Fig. 6 is a kind of hardware circuit connection diagram of signal preprocessing module;

Fig. 7 is a waveform diagram of motor cogging torque.

detailed description

The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention.

Referring to FIG. 1 , it is a schematic flowchart of an embodiment of a method for measuring cogging torque of a permanent magnet motor provided by the present invention.

In this embodiment, by utilizing a measuring device composed of a torque sensor, a signal preprocessing module, a signal acquisition module, a feature extraction module, and an industrial control computer, the method for measuring cogging torque of a permanent magnet motor includes the following steps S101～step S108:

Step S101: The torque sensor detects the torque of the permanent magnet motor under test, and sends the sensing signal output by the torque sensor to the signal preprocessing module.

Step S102: the signal preprocessing module conditions the sensing signal. Specifically, its signal conditioning process includes signal following, signal amplification, signal filtering, and signal shaping. The purpose of its preprocessing or conditioning is to make the signal output by the signal preprocessing module meet the requirements of the input signal of the next-level processing module.

In a practicable manner, the signal preprocessing module is composed of hardware circuits.

Referring to FIG. 2 , it is a schematic flowchart of an implementable manner of step S102 shown in FIG. 1 .

The signal preprocessing module shown includes a voltage follower, a signal amplifying circuit, a filter and a shaping circuit connected in sequence.

Then the step S102 specifically includes steps S201 to S204:

Step S201: The voltage follower adjusts the voltage value of the sensing signal, so that the output voltage of the voltage follower matches the input voltage. The voltage follower is a circuit that makes the output voltage the same as the input voltage. It is characterized by high input impedance and low output impedance. Therefore, the voltage follower can be used as a buffer stage and isolation stage for the previous circuit and the subsequent circuit, thereby increasing the load of the circuit. ability.

Step S202: The signal amplifying circuit reasonably amplifies the sensing signal output by the voltage follower. In this embodiment, the purpose of amplifying the sensing signal is to facilitate the processing of the signal acquisition module of the subsequent circuit, so as to obtain as high precision as possible. During specific implementation, the sensing signal is amplified near the amplitude of the signal source, and the destructive effect of noise will be reduced. Preferably, the amplitude of the sensing signal amplification is based on the maximum input range of the signal acquisition module.

Step S203: the filter filters the interference signal in the amplified sensing signal. During the specific implementation, due to the high-frequency interference of electrical equipment such as motors at the measurement site, when the measured signal is very weak, it will be covered by interference signals (noise, etc.), resulting in errors in the collected data. According to the output signal characteristics of the torque sensor, a hardware circuit can be designed to implement a second-order active low-pass filter to filter out unwanted components or noise in the sensor signal.

Optionally, in this embodiment, software filtering can also be used, and the collected sensing signal can be obtained by using LabVIEW (Laboratory Virtual Instrument Engineering Workbench) software, that is, the laboratory virtual instrument engineering workbench, wherein the filtering module is used for The sensing signal is filtered.

Step S204: the shaping circuit shapes the filtered sensing signal into a rectangular pulse signal. During specific implementation, the sensing signal processed by the torque sensor and filter is often a sawtooth wave or a non-standard square wave. After the sensing signal of this waveform is sent to the signal acquisition module, the output frequency of the torque signal collected is And other data will form errors. Therefore, it is necessary to design a shaping circuit to shape the sensing signal before collecting the torque signal on the sensing signal.

Step S103: The signal acquisition module extracts the measured torque output frequency value f in the conditioned sensing signal according to the preset sampling frequency f _s and the number of sampling points N, where f=f _s /N. Specifically, the signal acquisition module is realized by a data acquisition card, preferably, the data acquisition card is implemented by selecting NI (National Instruments), namely the PCI-6251 multifunctional data acquisition card of National Instruments Corporation of America.

Specifically, when extracting the measured torque output frequency value f, according to the preset sampling frequency f _s and the number of sampling points N, the sampling time T=N/f _s for the torque signal can be obtained, that is, at the time interval T is collected to obtain a torque value, therefore, the output frequency of the torque signal is f=1/T.

In this embodiment, it is considered that if the sampling frequency cannot be divisible by the number of sampling points in the subsequent analysis of the torque signal, it will cause energy leakage and amplitude distortion. Therefore, the ratio between the number of sampling points and the sampling frequency is preferably preset as an integer. For example, in the measurement process, when the sampling frequency f _s of the data acquisition card is 1.25MHz (megahertz), the number of sampling points can be set to 1250 points, and the calculated sampling period is 1ms (milliseconds). That is, a torque value is transmitted every 1 ms, and the corresponding output frequency of the torque value is 1000 Hz (Hertz). The speed of the permanent magnet motor is adjusted to 10 rpm. Then the time required for the motor to make one revolution is 6 seconds. The number of torque values collected during this period is 6000. During specific implementation, the collected waveform data of the torque signal can be saved in the form of electronic form to a designated location of the hard disk of the host computer.

During specific implementation, the PCI-6251 multi-function data acquisition card can collect the torque signal by means of analog input or counter. When the analog input method is used, the pulse signal output by the torque sensor is collected to the computer through the analog channel; by setting the sampling number and sampling frequency, the sampling time includes several pulse periods, and the analog signal input channel of the data acquisition card will After the pulse signal is completely collected, the output frequency of the torque signal can be obtained by using the extract single frequency information subVI in LabVIEW. When using a counter for measurement, set the maximum and minimum values of the measurement frequency range, use the measurement engineering function to calculate the frequency, measure multiple frequency values each cycle, and use the average value of the multiple frequencies measured as The desired signal is acquired.

Step S104: The signal acquisition module calculates the torque signal M of the measured permanent magnet motor according to the conversion formula M=A×(ff ₀ )/(f _p −f ₀ ). Among them, A is the torque full scale, f ₀ is the torque zero output frequency value, f _p is the positive full scale output frequency value. In this embodiment, when the torque signal M>0, it is a forward torque output value; when the torque signal M<0, it is a reverse torque output value.

Step S105: the signal collection module transmits the collected torque signal to the feature extraction module; the torque signal includes a cogging torque signal.

Step S106: The feature extraction module uses discrete wavelet transform to perform multi-resolution decomposition on the torque signal to obtain wavelet coefficients of the torque signal. During specific implementation, the MATLAB (Matrix Laboratory) software working platform can be called in LabVIEW, and the algorithm of wavelet transform mixed characteristic signal can be realized through programming, and the torque signal can be decomposed by wavelet to obtain the wavelet coefficient of the torque signal.

MATLAB is a high-level technical computing language and interactive environment for algorithm development, data visualization, data analysis, and numerical computation.

Referring to FIG. 3 , it is a flow chart of the steps of an implementable way of performing multi-resolution analysis on torque signals by using discrete wavelet transform provided by an embodiment of the present invention.

In the specific implementation, when the discrete wavelet transform is used to perform multi-resolution analysis on the torque signal, the specific decomposition process is:

Step S601: After the sampling rate of the torque signal satisfies the sampling theorem, the feature extraction module defines the total frequency band occupied by the collected signal (for example, the frequency band is 0-f _s /2) as the space V ₀ After decomposition, V ₀ is divided into two subspaces: low frequency subspace V ₁ (with a frequency band of 0～f _s /4) and high frequency subspace W ₁ (with a frequency band of f _s /4～f _s /2).

Step S602: After two-level decomposition, V ₁ is decomposed into low-frequency V ₂ (frequency band is 0-f _s /8) and high-frequency subspace W ₂ (frequency band is f _s /8-f _s /4), and so on analogy. The subdivision of this frequency subspace is as follows:

V ₀ =V ₁ ⊕W ₁ , V ₁ =V ₂ ⊕W ₂ ,...,V _j-1 =V _j ⊕W _j , so:

V ₀ ＝W ₁ ⊕W ₂ ⊕W ₃ ......W _j ⊕V _j (1)

Step S603: From multi-resolution analysis and spatial orthogonal decomposition theory, Where J is of any scale, and the signal x(t)∈L ² (R) is expanded on the space L ² (R), and the following expression is obtained:

Among them, the two-scale equation is:

Assume d _j+1,n ＝<f _j ,ψ _j+1,n >Then the wavelet coefficient obtained by decomposing is:

${\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

Among them, k=0,1,2...n-1, represents the translation position, and only needs to take a value within a limited range; j corresponds to the frequency range of the signal; c _j+1,k , d _j+1,k is the Wavelet coefficients; h(k) is the sampling response of the low-pass filter; g(k) is the sampling response of the high-pass filter; is the scaling function; ψ _j,k (t) is the wavelet function.

The multi-resolution analysis of discrete wavelet transform decomposes mixed signals composed of various frequencies (such as torque signals) into signals of different frequency bands, and has the ability to process signals according to frequency bands.

Step S107: the feature extraction module corrects the wavelet coefficients according to the pulse frequency of the cogging torque signal of the measured permanent magnet motor. During concrete implementation, after calculating by equation (5) and equation (6), obtain the torque signal wavelet coefficient; Then according to the pulsation frequency of the cogging torque signal in the torque signal, extract the wavelet coefficient of the torque signal and The cogging torque signal pulse frequency matches the wavelet coefficient component, and the other wavelet coefficient components are set to zero to correct the wavelet coefficient.

Referring to FIG. 4 , it is a schematic flowchart of an implementable manner of step S107 shown in FIG. 1 .

Specifically, the step S107 includes the following steps S701 to S704:

Step S701 : Calculate the cogging torque ripple frequency f _c =v*Z/60 of the measured permanent magnet motor according to the rotational speed v and the cogging number Z of the measured permanent magnet motor.

Step S702: According to the torque signal output frequency f and the cogging torque pulsation frequency f _c , select a wavelet; and determine the number n of wavelet decomposition layers, where n>0.

In this embodiment, preferably, the wavelet is a Daubechies Wavelet. Dobesy wavelet, also known as DB wavelet, is a compactly supported orthogonal wavelet with high vanishing moment, which plays an important role in signal compression, denoising and singularity detection.

Specifically, after the cogging torque ripple frequency f _c is obtained, the number of layers n of wavelet decomposition can be determined after selecting wavelets. A signal with a signal length of m can be decomposed into n=log ₂ m layers at most, and each layer has a corresponding frequency band. During specific implementation, the range of each frequency band can be obtained through calculation, and then the frequency band where it is located can be found according to the cogging torque ripple frequency, so as to determine the number of layers n to be decomposed by the wavelet transform.

Step S703: Perform discrete wavelet transformation of 1 to n layers on the torque signal, and decompose to obtain wavelet coefficients of each layer.

Step S704: Keep the wavelet coefficient components corresponding to the _cogging torque ripple frequency fc, and set the wavelet coefficient components corresponding to the remaining frequency bands to zero to obtain corrected wavelet coefficients.

Further, the method for measuring the cogging torque of the permanent magnet motor further includes step S108: the feature extraction module uses the corrected wavelet coefficients to perform wavelet reconstruction to obtain the cogging torque signal of the measured permanent magnet motor .

In a practicable manner, the step S108 specifically includes the following steps S801-S803:

Step S801: Using the multi-resolution characteristic of wavelet, find the frequency band signal corresponding to the cogging torque ripple frequency band.

Step S802: Make the other frequency band signals except the corresponding frequency band signal be zero, process and obtain new wavelet coefficients, and the reconstructed signal is the cogging torque signal.

Step S803: Using the wavelet coefficients to reconstruct the signal through the following equation:

${\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}{\mathrm{<span\; class="notranslate">\; d</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; k</span>}}\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$

The wavelet reconstruction process is the inverse operation of the decomposition process, and the reconstructed signal is the cogging torque signal.

As shown in Figure 5, in order to further verify the effectiveness of the magnetic groove torque measurement method proposed in this embodiment, this embodiment has constructed a permanent magnet synchronous motor system test platform with LabVIEW as the upper computer software, wherein the main parts are: magnetic powder Brake, JN338 intelligent digital torque speed sensor, NI PCI-6251 data acquisition card, SCB-68 junction box, GSK SJT series permanent magnet motor of Guangzhou CNC Equipment Co., Ltd. and supporting DAH01 series servo driver, of which JN338 torque The parameters of the speed sensor are: the torque range is 30N.m, the number of teeth is 60 teeth, the accuracy is 0.5, and the linearity is ≤0.5% F S; the parameters of the permanent magnet motor of the measured object are: rated power 1.5kW, rated speed 2500r/ min, rated torque 6N.m, cogging number 60.

After the software and hardware part of the experimental platform is designed, the permanent magnet motor magnetic groove torque measurement experiment is carried out. The JN338 torque sensor is used to detect the permanent magnet motor under test, and the signal preprocessing module shown in Figure 6 is used to condition the measured sensor signal. According to the preset sampling frequency f _s and the number of sampling points N, the torque signal output frequency f=f _s /N of the conditioned sensing signal is extracted, and the torque signal of the measured permanent magnet motor is calculated. Carry out multi-resolution decomposition on the torque signal according to the formula, calculate the wavelet coefficient of the measured torque signal, and correct the wavelet coefficient, and then perform wavelet reconstruction on the corrected wavelet coefficient to obtain the measured permanent The cogging torque signal of the magneto.

Firstly, set the motor speed to be constant at 10r/min through the servo driver setting, set the sampling frequency of the PCI-6251 data acquisition card to 600KHz in the LabVIEW host computer software, and set the number of acquisition points to 600, according to the formula f=f _s /N, calculate The measured torque output frequency is 1000 Hz, and the torque signal of the measured permanent magnet motor is calculated through the frequency and torque conversion formula M=A×(ff ₀ )/(f _p −f ₀ ).

Then according to the measured torque signal according to the step S106, the discrete wavelet transform is used to carry out multi-resolution decomposition of the torque signal to obtain the wavelet coefficient of the torque signal, and the obtained wavelet coefficient is corrected by the step S107, and finally by the step S108 Reconstruct the corrected wavelet coefficients to obtain the cogging torque signal. The waveform is shown in Figure 7.

The method for measuring the cogging torque of a permanent magnet motor provided by the embodiment of the present invention can process the sensing signal of a torque sensor with ordinary precision, and does not need to be used with an expensive high-precision torque sensor. The module or data acquisition card extracts the weak torque signal in the sensing signal, and according to the multi-resolution characteristics of wavelet decomposition, performs wavelet decomposition on the torque signal through the MATLAB platform, and extracts the cogging torque ripple in the torque signal The frequency-matched wavelet coefficients are used for wavelet reconstruction, and finally the cogging torque signal of the measured permanent magnet motor is obtained, which has high measurement accuracy and is easy to implement.

The above description is a preferred embodiment of the present invention, and it should be pointed out that for those skilled in the art, without departing from the principle of the present invention, some improvements and modifications can also be made, and these improvements and modifications are also considered Be the protection scope of the present invention.

Claims (8)

1. the measuring method of a cogging torque of permanent magnet motor, it is characterised in that comprise the following steps:

Step 101: tested magneto is detected by torque sensor, exports described torque sensor Transducing signal sends to signal pre-processing module；

Step 102: described transducing signal is nursed one's health by described signal pre-processing module；

Step 103: signal acquisition module is according to sample frequency f preset_sWith sampling number N, extract conditioning After transducing signal in actual measurement torque output frequency value f=f_s/N；

Step 104: described signal acquisition module is according to conversion formula M=A × (f-f₀)/(f_p-f₀), meter Calculate the dtc signal M of described tested magneto；Wherein, A is torque full scale, and f is defeated for actual measurement torque Go out frequency values, f₀For torque offset output frequency values, f_pFor forward Full-span output frequency values；

Step 105: the described dtc signal collected is transmitted to feature extraction mould by described signal acquisition module Block；Described dtc signal includes cogging torque signal；

Step 106: described characteristic extracting module utilizes wavelet transform to differentiate described dtc signal more Rate is decomposed, it is thus achieved that the wavelet coefficient of described dtc signal；

Step 107: described characteristic extracting module is according to the pulsation of the cogging torque signal of described tested magneto Frequency, is modified described wavelet coefficient；

Step 108: described characteristic extracting module uses revised wavelet coefficient to carry out wavelet reconstruction, it is thus achieved that institute State the cogging torque signal of tested magneto.

2. the measuring method of cogging torque of permanent magnet motor as claimed in claim 1, it is characterised in that described Signal pre-processing module includes: voltage follower, signal amplification circuit, wave filter and the shaping being sequentially connected with Circuit；Described step 102 specifically includes:

Step 201: the magnitude of voltage of described transducing signal is adjusted by described voltage follower, so that described electricity The output voltage of pressure follower matches with input voltage；

Step 202: the transducing signal that described voltage follower is exported by described signal amplification circuit is amplified；

Step 203: the interference signal in the transducing signal after amplification process is filtered by described wave filter；

Step 204: filtered transducing signal is shaped as rectangular pulse signal by described shaping circuit, and provides To data collecting card.

3. the measuring method of cogging torque of permanent magnet motor as claimed in claim 2, it is characterised in that described Wave filter is the second order active low pass filter using hardware circuit to realize.

4. the measuring method of cogging torque of permanent magnet motor as claimed in claim 2, it is characterised in that described Wave filter uses the mode of software filtering to be filtered transducing signal.

5. the measuring method of cogging torque of permanent magnet motor as claimed in claim 1, it is characterised in that described Step 106 specifically includes:

Step 601: after described characteristic extracting module treats that the sample rate of described dtc signal meets sampling thheorem, The total frequency band occupied by the signal collected is defined as SPACE V₀, described total frequency band is 0～f_s/2；Through a fraction Xie Hou, by V₀It is divided into low frequency subspace V₁With high frequency subspace W₁；, described V₁Frequency band be 0～f_s/ 4, Described W₁Frequency band is f_s/ 4～f_s/2；

Step 602: after two grades are decomposed, by described V₁It is decomposed into low frequency subspace V₂With high frequency subspace W₂, described V₂Frequency band be 0～f_s/ 8, described W₂Frequency band be f_s/ 8～f_s/4；And so on, by frequency Uniformly subdivision is: V₀=V₁⊕W₁, V₁=V₂⊕W₂..., V_j-1=V_j⊕W_j, therefore:

V₀=W₁⊕W₂⊕W₃......W_j⊕V_j

Wherein, each W_jIt is reflection V_j-1The high frequency subspace of spacing wave details, V_jIt is reflection V_j-1Spacing wave The low frequency subspace of details, W_jFor V_jAt V_j-1The orthogonal complement space；

Step 603: by multiresolution analysis and orthogonal space resolution theory,Wherein J is Arbitrarily yardstick, by signal x (t) ∈ L²(R) at space L²(R) upper launch, obtain following expression:

Wherein, Double-scaling equation is:

Ifd_{J+1, n}=< f_j, ψ_{J+1, n}> and then decompose and obtain wavelet coefficient and be:

$c_{j+1,k}=\underset{n}{\&Sigma;}h(n-2k)c_{j,n}$

$d_{j+1,k}=\underset{n}{\&Sigma;}g(n-2k)d_{j,n}$

Wherein, k=0,1,2 ... n-1, represent translation position；The j frequency range to induction signal；c_j+1,k, d_j+1,kFor Described wavelet coefficient；H (k) is low pass filter sampling response；G (k) is high-pass filter sampling response； For scaling function；ψ_j,kT () is wavelet function.

6. the measuring method of cogging torque of permanent magnet motor as claimed in claim 1, it is characterised in that described Step 107 comprises the following steps:

Step 701: according to rotating speed v and teeth groove quantity Z of described tested magneto, calculate described tested Cogging torque ripple frequency f of magneto_c=v × Z/60；

Step 702: according to described dtc signal output frequency f and described cogging torque ripple frequency f_c, select Small echo；And determine number of plies n of wavelet decomposition, wherein n > 0；

Step 703: described dtc signal is carried out 1 wavelet transform arriving n-layer, decomposes and obtain each layer Wavelet coefficient；

Step 704: retain and described cogging torque ripple frequency f_cThe component of corresponding wavelet coefficient, and will The component zero setting of the wavelet coefficient that remaining frequency range is corresponding, it is thus achieved that revised wavelet coefficient.

7. the measuring method of cogging torque of permanent magnet motor as claimed in claim 6, it is characterised in that described Small echo is the western small echo of many shellfishes.

8. the measuring method of cogging torque of permanent magnet motor as claimed in claim 1, it is characterised in that described Step 108 comprises the following steps:

Step 801: utilize the multi-resolution characteristics of small echo, searches the frequency corresponding to cogging torque ripple frequency section Segment signal；

Step 802: order remaining frequency band signals in addition to the frequency band signals of described correspondence is zero, and process obtains new Wavelet coefficient；

Step 803: utilize described wavelet coefficient that signal is reconstructed by below equation:

$c_{j,k}=\underset{n}{\&Sigma;}{c}_{j+1,k}h(n-2k)+\underset{n}{\&Sigma;}{d}_{j+1,k}g(n-2k)$

Wherein, c_j,kFor cogging torque signal, the signal that reconstruct obtains is cogging torque signal.