CN103207600B - Multi-node network vibration abatement device and method - Google Patents

Abstract

The invention discloses a multi-node network vibration suppression device and a vibration suppression method. The vibration suppression device includes a display and control unit, a communication bus, and a plurality of node suppression units. Estimation of Transfer Properties of Structural Points And the control frequency w _i is sent to each node suppression unit to monitor the response of each structural point in real time, thus playing the role of global control. The node suppression unit takes the DSP processor as the core, by measuring the structural response Y(e ^jw ) of a certain point, using the improved feedback harmonic frequency suppression algorithm, and taking the minimum model of the structural response Y(e ^jw ) as the evaluation criterion to iteratively calculate the control in real time The signal U(e ^jw ) reduces the response of a certain structural point, thereby playing the role of local inhibition.

Description

A multi-node network vibration suppression device and vibration suppression method

technical field

The invention relates to a multi-node network vibration suppression device and vibration suppression method for suppressing large flexible structures.

Background technique

Large flexible structures are widely used in engineering, such as satellite antennas, suspension bridges, etc. In order to suppress the vibration of large flexible structures under disturbance excitation and maintain their geometry (especially satellite antennas), some control methods are needed. Existing methods usually adopt a main control point to suppress the vibration of the whole structure locally, but for complex whole structures, the control effect achieved by one control point often cannot meet the design requirements.

During the working process, the mechanical structure mainly experiences random vibration, but it is worth pointing out that the vibration response of the mechanical structure is a comprehensive reflection of its inherent characteristics and externally applied loads. When the frequency of the external load is equal to or close to the natural frequency of the structure, it will cause a relatively large resonance response of the structure. Or when the energy of a certain frequency contained in the external load is relatively large, as long as it does not act on the nodes of the structural vibration mode, it will also cause a large forced vibration of the structure. If several order frequency components with larger vibration response energy at a certain point of the structure can be directly suppressed, a good vibration damping effect at that point can be obtained.

Contents of the invention

The purpose of the present invention is to provide a global-local combined multi-node network vibration suppression device and vibration suppression method by arranging multiple sensors and controllers on the key points of the structure through the analysis of large flexible structures, so that the entire structure The vibration damping effect has been significantly improved.

To achieve the above object, the present invention is achieved by adopting the following technical solutions:

A multi-node network vibration suppression device is characterized in that it includes a display and control unit, and the display and control unit is connected to n node suppression units through a communication bus, wherein, n≥2, the control output of each node suppression unit The end is connected to a node of the vibration suppression object through a power amplifier and a vibration table, and the node is connected to the data acquisition end of the node suppression unit through a charge amplifier and a sensor; the display control unit includes a structural vibration display module and a principal component analysis module and the finite element analysis module, the vibration response data of each node of the suppression object from the suppression unit of each node is input to the structural vibration display module, the principal component analysis module and the finite element analysis module through the communication bus, wherein the finite element analysis module outputs each Estimation of Transfer Characteristics of Nodal Structures After analysis, the principal component analysis module outputs the third-order main frequency w _i of each node structure, i=1, 2, 3, and sends it to each node suppression unit through the communication bus, and the structural vibration display module performs the vibration response data of each node and point Analysis and display; the node suppression unit takes the DSP processor as the core, obtains the vibration response data of a node of the vibration suppression object through the data acquisition module, and locally controls the vibration of the node through the signal synthesis module with the assistance of the programmable logic device PLD vibration.

In the above device, the data acquisition module includes two channels, each of which is composed of a programmable amplifier PGA and an analog-to-digital converter AD, wherein one channel collects the time-domain response signal of the structural vibration of a node of the vibration suppression object, and transforms it through the DSP It is the frequency domain response signal Y (e ^jw ); the other way collects the frequency domain control signal U (e ^jw ) fed back from the signal synthesis module as a phase reference for calculating the phase of Y (e ^jw ); the signal synthesis module It is composed of three-way digital frequency synthesizer DDS and an adder. The three-way DDS respectively generate sinusoidal signals corresponding to the three main vibration components in Y(e ^jw ), and the adder synthesizes these three-way signals as the control signal U(e ^jw ) million output.

The communication bus includes a usb bus to CAN bus converter and a CAN bus connected to each node suppression unit.

A vibration suppression method, which is realized by the aforementioned multi-node network vibration suppression device, is characterized in that it includes the following steps:

(1) Global monitoring and control

First, the display and control unit analyzes the vibration response data from the corresponding nodes of each node suppression unit in the finite element analysis module and the principal component analysis module to obtain the transfer characteristic estimation of each node structure And the third-order main frequency w _i i = 1, 2, 3, and then sent to each node suppression unit through the communication bus; the structural vibration display module analyzes and displays the vibration response data of each node, and uses the overall response amplitude of the vibration suppression object As an evaluation criterion, adjust the control of the structural vibration of each node suppression unit corresponding to the node structure in real time;

(2) Nodal suppression unit corresponds to the control of structural vibration of each node

The suppression units of each node receive the vibration response data of their respective node structures through the data acquisition module, and DSP and w _i , convert the vibration response data of the node structure into a frequency domain response signal Y(e ^jw ), and then calculate Y(e ^jw ) at least three times according to the existing feedback harmonic suppression algorithm or the improved feedback harmonic suppression algorithm The amplitude and phase of the main sinusoidal component of the first order, the synthesized frequency domain control signal U(e ^jw ) is output to the corresponding node, and the U(e ^jw ) is fed back to the data acquisition module as a phase reference to calculate Y( e ^jw ) phase.

In the above vibration suppression method, the specific method of using the principal component analysis module to obtain the third-order main frequency w _i of the structural response is:

Step 1: Collect the overall structural response signal Y(t)=[Y ₁ (t), Y ₂ (t),...,Y _i (t)] ^T at time t, where i is the number of node suppression units ;Y _i (t) is the measured value of the structural response on the i-th node, i=1, 2,...,n;

Step 2: Calculate the autocorrelation matrix; C _XX =E[YY ^T ]∈R ^n×n ; Y is the abbreviated form of the overall response signal Y(t) of the structure, and C _XX is an n×n real number matrix;

Step 3: Calculate the eigenvalue λ of C _XX , and sort it from large to small, so that: λ ₁ ≥λ ₂ ≥...≥λ _i ≥...≥λ _n , i=1, 2,...n ;

Step 4: Take the largest first m eigenvalues of C _XX to form a diagonal square matrix The corresponding eigenvectors form the transformation matrix If it is truncated, C _XX can be approximately decomposed into an eigenvalue equation: The variance contribution rate of the first m principal components is Among them, λ _i is the i-th eigenvalue sorted from large to small, m<n; the variance truncation error caused by observation noise is When the number m of irrelevant latent variables is unknown, set the threshold value of η to truncate m, and η≥95%;

Step 5: Pass the first three-order principal component eigenvalue λ _i through the transformation matrix Convert to the corresponding frequency W _i , i=1, 2, 3.

The expression of the improved feedback harmonic suppression algorithm is:

${\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\frac{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; |</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; |</span>}$

${\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}$

in is the estimated transfer characteristic of a node structure in matrix form, is the vector form of the frequency domain response signal Y(e ^jw ) of a node structure, is the vector form of the control quantity; max|Y| is the maximum value of the structural response in the iterative process, the subscript k indicates the current moment, and k+1 indicates the next moment; the subscripts r and i represent the real part and the imaginary part respectively ; w _j subscript j=1, 2, 3, respectively represent the third-order frequency; the initial value of the iteration coefficient is

The advantage of the device of the present invention is that the display and control unit and several node suppression units are connected to form a network through the communication bus, and under the monitoring of the display and control unit, the multiple node suppression units suppress the vibration of several key points of the large flexible structure, It plays a role in suppressing the structure as a whole, and achieves the effect that a single suppressor cannot achieve. The node suppression unit with DSP and PLD as the core has a fast processing speed and can realize real-time control. The device can drive a variety of actuators and should have a wide range.

The advantage of the method of the present invention is that the division of labor is carried out according to the difference in hardware structure between the display and control unit and the node suppression unit. The display and control unit based on the PC platform has strong computing power, can monitor the response of each point in real time, and perform offline structural analysis to provide structural information for the node suppression unit. The node suppression unit is a suppressor with high-speed DSP and PLD as the core, which can quickly calculate the iterative algorithm to realize real-time vibration suppression on a certain structural point. On the basis of the overall characteristic information of the structure, the vibration suppression effect on large flexible structures can be improved. The improved feedback harmonic suppression algorithm adopted in this method has the advantages of insensitivity to system parameters, fast convergence speed, and stable convergence process.

Description of drawings

The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

FIG. 1 is a block diagram of the overall structure of the vibration suppressing device of the present invention.

Fig. 2 is a schematic diagram of a specific embodiment of the vibration suppression method of the present invention. Among them: (a) The figure is the functional structure block diagram of the display and control unit in Figure 1; (b) The figure is the flow chart of the principal component analysis algorithm in (a) Figure 1; (c) The figure is the node suppression unit and vibration suppression in Figure 1 Schematic diagram of the data flow of the object; (d) is the flow chart of the node vibration suppression algorithm.

Fig. 3 is a comparison diagram of the effect of single control and multi-node network control of the present invention on vibration suppression of a slender cylindrical shell. Among them, the abscissa is the number of test points distributed on the slender cylindrical shell, and the ordinate represents the grms value of each point.

Fig. 4 is a comparative diagram of local vibration control effects of the present invention. Among them, the figure (a) is the response convergence process using the feedback harmonic frequency suppression method; the figure (b) is the response convergence process using the method of the present invention.

Detailed ways

As shown in Figure 1, the multi-node network vibration suppression device of the present invention includes a display and control unit, a communication bus, n node suppression units and corresponding n vibration suppression objects (nodes a-1, a-2, ..., a-n), n≥2. The display control unit uses a PC as a platform, and uses a communication bus to connect with each node suppression unit (No. 1-N), which plays the role of global detection of vibration suppression objects and global vibration suppression. The communication bus includes a usb bus to CAN bus converter and a CAN bus connected to each node suppression unit.

Each node suppression unit takes the DSP processor as the core, and the programmable logic device PLD is connected to the DSP through a 16-bit data bus and a 19-bit address bus. The vibration response data of a node of the vibration suppression object is obtained through the data acquisition module. With the assistance of the PLD Next, the vibration of the node is locally controlled through the signal synthesis module. The node suppression unit is connected to the communication bus through the DSP-integrated CAN (Controller Area Network) bus transceiver. The CAN bus adopts a two-wire communication system with a simple structure. It adopts an ingeniously designed CAN communication protocol, which makes it highly reliable and error-free The detection ability can be applied to the industrial environment with harsh ambient temperature, strong electromagnetic radiation and large vibration.

The data acquisition module is divided into two routes, one route consists of a programmable amplifier PGA-0 and an analog-to-digital converter AD, the input of PGA-0 is connected to a node of the vibration suppression object through a charge amplifier and a sensor, and the point is collected The time-domain response signal Y(t) and time-domain control signal U(t) of the structure are processed by DSP to obtain the frequency-domain response signal Y(e ^jw ) and frequency-domain control signal U(e ^jw ) of the structure point; One route consists of a programmable amplifier PGA-1 and an analog-to-digital converter AD, which collects the frequency domain control signal U (e ^jw ) output from the signal synthesis module. The PGA can adjust the magnitude of the collected signal and control the transmission signal at - Between 10V and +10V. What AD adopts is a high-speed 16-bit AD converter, which sends data to DSP through the 16-bit data external data bus. The control signal U(e ^jw ) is used as a phase reference to calculate the phase of the response signal.

The signal synthesis module consists of three direct digital frequency synthesizers DDS and an adder. The three-way DDS synthesizes the corresponding sinusoidal signals of the three main vibration components in Y(e ^jw ) through an adder, and outputs them as a node control signal U(e ^jw ) to the node through a power amplifier and a vibration table. Programmable logic device PLD assists DSP to control and manage external ICs, such as PGA, DDS, etc. PLD and DSP work in coordination through 16-bit external data bus and 19-bit external address bus. Using the VHDL hardware description statement, PLD can write the control timing of the external IC, map the external device as a register address to the DSP, reduce its burden, and simplify the control of the external IC by the main control chip.

As shown in Figure 2, the vibration suppression method of the present invention includes two parts: global monitoring and node suppression:

(1) Overall monitoring and control.

As shown in Figure 2(a), first, the display and control unit performs principal component and finite element analysis in the finite element analysis module and principal component analysis module through the vibration response data of each node of the suppression object from each node suppression unit, Estimates of the transfer characteristics of the structure at each point are obtained separately And the third-order main frequency w _i in the structural response, i=1, 2, 3, and then sent to each node suppression unit through the communication bus; the structural vibration display module analyzes and displays the vibration response data of each node, and uses the vibration suppression object The overall response amplitude is the evaluation criterion, and the suppression units of each node are adjusted in real time.

Referring to Figure 2(b), the specific method of using the principal component analysis module to obtain the third-order main frequency w _i of the structural response, i=1, 2, 3 is:

Step 1: Collect the overall structural response signal Y(t)=[Y ₁ (t), Y ₂ (t),...,Y _i (t)] ^T at time t, where i is the number of node suppression units ;Y _i (t) is the measured value of the structural response on the i-th node, i=1, 2,...,n;

Step 2: Calculate the autocorrelation matrix Y is the abbreviated form of the overall response signal Y(t) of the structure, and C _XX is an n×n real number matrix;

Step 3: Calculate the eigenvalue λ of C _XX , and sort it from large to small, so that: λ ₁ ≥λ ₂ ≥...≥λ _i ≥...≥λ _n , i=1, 2,..., n;

Step 4: Take the largest first m eigenvalues of C _XX to form a diagonal square matrix The corresponding eigenvectors form the transformation matrix If it is truncated, C _XX can be approximately decomposed into an eigenvalue equation: The variance contribution rate of the first m principal components is Among them, λ _i is the i-th eigenvalue sorted from large to small, m<n; the variance truncation error caused by observation noise is When the number m of irrelevant latent variables is unknown, set the threshold value of η to truncate m, and η≥95%;

Step 5: Pass the first three-order principal component eigenvalue λ _i through the transformation matrix Convert to the corresponding frequency w _i , i=1, 2, 3.

(2) Control of node suppression unit.

As shown in Figure 2(c) and (d), firstly, the DSP receives the information sent by the display and control unit The frequency w _i corresponding to the principal component, i=1, 2, 3. Then, the data acquisition module measures the time-domain response signal Y(t) and the time-domain control signal U(t) of a certain structural point, the sampling frequency is selected as 2048hz, and 512 points are collected in each frame; ) and U(t) to do fast Fourier transform (FFT) to obtain the frequency domain response signal Y(e ^jw ) of the structure point; finally, DPS is based on the feedback harmonic suppression algorithm (Veres, S., Adaptiveharmonic control. International Journal of Control, 2001.74 (12): p.1219-1225.) or the improved feedback harmonic suppression algorithm of the present invention calculates the amplitude and phase of the third-order main sinusoidal component of Y (e ^jw ), and finally the third-order amplitude, phase and principal component The corresponding frequency w _i is sent to the three-way DDS, and the synthesized frequency domain control signal U(e ^jw ) is output, and the U(e ^jw ) is fed back to PGA-0 as a phase reference (used to calculate the frequency domain response signal Y(e ^jw ) phase).

Wherein, the expression of the improved feedback harmonic suppression algorithm of the present invention is:

${\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\frac{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; |</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; |</span>}$

${\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}$

in is an estimate of the transfer properties of the structure at a point in matrix form, is the vector form of a point structure response Y(e ^jw ), is the vector form of the control quantity. μ is the iteration coefficient, which changes as the iteration progresses. The initial value of the iteration coefficient is max|Y| is the maximum value of the structural response that occurs during the iterative process. The subscript k indicates the current moment, and k+1 indicates the next moment; the subscripts r and i represent the real part and the imaginary part, respectively; the subscripts j=1, 2, and 3 of w _j represent the third-order frequencies respectively.

The following is an application example of the present invention: a slender cylindrical shell made of aluminum alloy is supported to vibrate under the complex periodic excitation of a vibrating table. The acceleration sensor located at the control point transmits the signal to the node controller (node suppression unit), and the node controller uses the improved feedback harmonic frequency suppression algorithm to output the control signal to drive the shaking table to suppress the vibration response of the control point. Accelerometers are arranged on the slender cylindrical shell according to the 30 points arranged in the finite element analysis, and the response changes of the shell as a whole are detected during the experiment.

As shown in Fig. 3, the histogram marked with the slash texture reflects the vibration distribution of the single-node suppressor arranged at 14 points. The histogram marked with the real texture reflects the vibration distribution at 12 o'clock and 21 o'clock using the multi-node network vibration suppressor. It can be seen from the figure that the structural vibration of the multi-node network vibration suppressor has been greatly suppressed as a whole, and the suppression effect is better than that of a single node.

As shown in Figure 4, comparing the two figures (a) and (b), it can be seen that the response convergence process of the improved feedback harmonic frequency suppression method proposed by the present invention is faster, and there is no oscillation phenomenon in the figure (a) , the suppression effect is better.

Claims (6)

1. A multi-node network vibration suppression device is characterized in that it includes a display and control unit, which is connected to n node suppression units by a communication bus, wherein, n≥2, each node suppression unit The control output end is connected to a node of the vibration suppression object through a power amplifier and a vibrating table, and the node is connected to the data acquisition end of the node suppression unit through a charge amplifier and a sensor; the display and control unit includes a structural vibration display module, a main The component analysis module and the finite element analysis module, the vibration response data of each node of the suppression object from the suppression unit of each node are input to the structural vibration display module, the principal component analysis module and the finite element analysis module through the communication bus, among which, the finite element analysis module analyzes Then output the transfer characteristic estimation of each node structure The principal component analysis module outputs the third-order main frequency of each node structure after analysis Send it to each node suppression unit through the communication bus, and the structural vibration display module analyzes and displays the vibration response data of each node; the node suppression unit takes the DSP processor as the core, and obtains the vibration of a node of the vibration suppression object through the data acquisition module In response to the data, the vibration of the node is locally controlled through the signal synthesis module with the assistance of the programmable logic device PLD. 2. multi-node network vibration suppression device as claimed in claim 1, is characterized in that, described data acquisition module comprises two roads, and each road all is made of a programmable amplifier PGA and an analog-to-digital converter AD, wherein one road collects The structural vibration time-domain response signal of a node of the vibration suppression object is transformed into a frequency-domain response signal Y(e ^jw ) by DSP; the other channel collects the frequency-domain control signal U(e ^jw ) fed back from the signal synthesis module as a phase reference, It is used to calculate the phase of Y(e ^jw ); the signal synthesis module is composed of three-way digital frequency synthesizer DDS and an adder, and the three-way DDS respectively generates sinusoidal signals corresponding to the three main vibration components in Y(e ^jw ) , the adder synthesizes these three signals and outputs it as a control signal U(e ^jw ). 3. multi-node network vibration suppression device as claimed in claim 1, is characterized in that, described communication bus comprises usb bus to CAN bus converter and the CAN bus that connects each node suppression unit. 4. A vibration suppression method, which is realized by the multi-node network vibration suppression device according to claim 1, is characterized in that, comprising the steps of: (1) Global monitoring and control First, the display and control unit analyzes the vibration response data from the corresponding nodes of each node suppression unit in the finite element analysis module and the principal component analysis module to obtain the transfer characteristic estimation of each node structure and the third-order main frequency w _i , i=1, 2, 3, and then send them to the suppression units of each node through the communication bus; the structural vibration display module analyzes and displays the vibration response data of each node, and uses the overall response amplitude of the vibration suppression object The value is the evaluation criterion, and the control of the structural vibration of the corresponding node of each node suppression unit is adjusted in real time; (2) Nodal suppression unit corresponds to the control of structural vibration of each node The suppression units of each node receive the vibration response data of their respective node structures through the data acquisition module, and DSP and w _i , convert the vibration response data of the node structure into a frequency domain response signal Y(e ^jw ), and then calculate Y(e ^jw ) at least three times according to the existing feedback harmonic suppression algorithm or the improved feedback harmonic suppression algorithm The amplitude and phase of the main sinusoidal component of the first order, the synthesized frequency domain control signal U(e ^jw ) is output to the corresponding node, and the U(e ^jw ) is fed back to the data acquisition module as a phase reference to calculate Y(e ^jw ) phase. 5. vibration suppressing method as claimed in claim 4 is characterized in that, the concrete method that uses principal component analysis module to obtain the third-order main frequency _w of structural response is: Step 1: Collect the overall structural response signal Y(t)=[Y ₁ (t), Y ₂ (t),...,Y _i (t)] ^T at time t, where i is the number of node suppression units ;Y _i (t) is the measured value of the structural response on the i-th node, i=1, 2,..., n; Step 2: Calculate the autocorrelation matrix C _XX =E[YY ^T ]∈R ^n×n ; Y is the abbreviated form of the overall response signal Y(t) of the structure, and C _XX is an n×n real number matrix; Step 3: Calculate the eigenvalue λ of C _XX and sort it from large to small, so that: λ ₁ ≥ λ ₂ ≥... ≥ λ _i ≥... ≥ λ _n , i=1, 2, ..., n; Step 4: Take the largest first m eigenvalues of C _XX to form a diagonal square matrix The corresponding eigenvectors form the transformation matrix If it is truncated, C _XX can be approximately decomposed into an eigenvalue equation: The variance contribution rate of the first m principal components is Among them, λ _i is the i-th eigenvalue sorted from large to small, m<n; the variance truncation error caused by observation noise is When the number m of irrelevant latent variables is unknown, set the threshold of η to truncate m, η≥95%; Step 5: Pass the first three-order principal component eigenvalue λ _i through the transformation matrix Convert to corresponding frequency w _i , i=1, 2, 3. 6. vibration suppression method as claimed in claim 4, is characterized in that, the expression of described improved feedback harmonic frequency suppression algorithm is: ${\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}\frac{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; |</span>}{\mathrm{<span\; class="notranslate">\; max</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; |</span>}$ ${\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}{\mathrm{<span\; class="notranslate">\; Y</span>}}_{\mathrm{<span\; class="notranslate">\; k</span>}}$ in $P^T=\left[\begin{array}{cc}{\hat{P}}_r\left(w_j\right)& {\hat{P}}_i\left(w_j\right)\\ -{\hat{P}}_i\left(w_j\right)& {\hat{P}}_r\left(w_j\right)\end{array}\right]$ is the estimated transfer characteristic of a node structure in matrix form, $Y=\left[\begin{array}{c}{Y}_r\left(w_j\right)\\ Y_i\left(w_j\right)\end{array}\right]$ is the vector form of the frequency domain response signal Y(e ^jw ) of a node structure, $u=\left[\begin{array}{c}{u}_r\left(w_j\right)\\ u_i\left(w_j\right)\end{array}\right]$ is the vector form of the control quantity; max|Y| is the maximum value of the structural response in the iterative process, the subscript k indicates the current moment, and k+1 indicates the next moment; the subscripts r and i represent the real part and the imaginary part respectively ; w _j subscript j=1, 2, 3, respectively represent the third-order frequency; the initial value of the iteration coefficient is