CN105867115A - Method for controlling non-stationary random vibration test - Google Patents

Abstract

The invention discloses a method for controlling a non-stationary random vibration test. The method comprises the steps of setting test parameters; setting non-stationary random reference signals; performing automatic system check; identifying a transfer function; performing vibration control over a uniaxial periodic cycle: within a single frame time period, respectively executing the steps of calculating a normalization control power spectrum; performing correction calculation to obtain a normalization driving spectrum; performing phase randomization and inverse fast Fourier transform (IFFT) on the normalization driving spectrum, and performing frame lap joint to obtain true random signals; performing calculation update to obtain current driving root mean square gain; and multiplying the driving root mean square gain by driving frame data after the lap joint to calculate and obtain driving output signals of the current frame. Compared with conventional random vibration test control, through the method, not only can the satisfactory spectrum control of the non-stationary random vibration be obtained, but also the non-stationary random signal root mean square can be stably controlled to vary with a time curve, and it is effective and feasible to adopt the control method to implement non-stationary random vibration test control.

Description

A kind of non-stationary random vibration experimental control method

Technical field

The present invention relates to a kind of with electronic or hydraulic vibration gen (band product), excitation system and vibrator as control object Non-stationary random vibration experimental control method, is particularly suited for simulating delivery vehicle and (includes ground, aerial and marine delivery Instrument) for the purpose of non-stationary random vibration environment under unstable period, and the product functionality carried out and Reliability Check Vibration test control aspect, belongs to automatic control technology field.

Background technology

Random vibration test is inspection product reliability, and exposed structure material and defective workmanship analyze mechanical vibration performance Necessary measure, is the important means solving various mechanically and structurally vibration problems；Random vibration test is also military One of essential project of facility environment test, the examination to product has played huge effect；It addition, random vibration test is also Simulating the necessary means of all kinds of vibration operating mode scene, most of vibration environments are all random vibrations, such as vehicle, locomotive and high Railway transportation Defeated, many occasions such as vibration environment airborne, carrier-borne and weapon vibration environmental.

Random vibration is divided into again steadily and non-stationary random vibration two class by its statistical property.Traditional random vibration test It is all to use there is gauss of distribution function, the stationary random vibration of ergodic characteristic.For non-stationary random vibration environment Simulation, traditional way is to take statistics and spectrum envelope to add tight method by non-stationary random vibration Approximate Equivalent for the most random Test is carried out in vibration, is a kind of approximate simulation to vibration environment, for some test type (such as environment adaptation of test specimen Property test or long duration test), the method is feasible.But for functional or Reliability Check Test, this will cause very Differing greatly and affecting test effect of real environment simulation, brings difficulty to the functional and reliability of assessment product.Along with number Word controls technology development and random vibration test understanding is goed deep into by people, the most of carrier environments stood for weapon All there is non-stationary random vibration, such as fighter plane fistfight, automobile is exercised on uneven road surface, and naval vessels navigate by water in deep-sea condition, The flight course etc. of weapon.For its destruction being caused product, non-stationary random vibration is the most sub-on the impact of product The product bug pattern inspired in stationary random vibration, stationary random vibration and non-stationary random vibration also differs farther out, Non-stationary is replaced to be irrational at random by stationary random vibration.Simultaneously as in functional and Reliability Check Test Simulation to true environment has higher requirement, needs urgently to carry out non-stationary random vibration test technology investigation, in order to carry Go out a kind of non-stationary random experiment control method, solve the environmental simulation problem of non-stationary random vibration.

At present, vibrating controller product includes recalling perseverance, rising the controller that shakes of China, and international SD, DP and LMS are controlled The random vibration control function that device processed etc. are comprised, all be use stationary random vibration control mode, the most do not have non-stationary with Machine vibration controls function.Domestic in terms of Nonstationary Random Vibration Signals analysis and process, do certain research, such as document: " the Instantaneous Power Spectrum Analysis of Nonstationary Random Vibration Signals ", Jin Xianlong, vibration engineering journal, 1990, Vol.3No.3:26- 31；" the instantaneous power spectrum of Vehicular system non-stationary random vibration ", Liu Gang etc., 1997, vol.10No.3:121-123, it is proposed that Instantaneous power spectrum mode over time describes nonstationary random signal；Additionally document: " local stationary random vibration Process and application ", He Baiyang, system engineering and electronic technology, 1993, No.3:59-63, it is proposed that approximate with product model Describing nonstationary random process, this model is easy to process in time-domain and frequency domain.And try about non-stationary random vibration Test control document the fewest.

Through search patent and non-patent literature data, not disclosed about non-stationary random vibration experimental control method Pertinent literature, more has no that Related product uses in the application.

Summary of the invention

The purpose of the present invention is that provides a kind of non-stationary random vibration controlling test side to solve the problems referred to above Method.

The present invention is achieved through the following technical solutions above-mentioned purpose:

A kind of non-stationary random vibration experimental control method, is realized, including following by the software being located in control system Step:

(1) test parameters is arranged: basic parameter includes but not limited to that test name, date, operating path select and remarks, Self-inspection parameter includes but not limited to self-inspection starting voltage, self-inspection magnitude and self-inspection maximum voltage, controls parameter and includes but not limited to Spectral line number, sample frequency, frequency range, channel parameters but be not limited to include channel selecting, control mode, sensitivity coefficient, be System identified parameters includes but not limited to System Discrimination magnitude, identification signal frame number and average time, and security parameter includes but do not limits In extreme displacement, limit velocity, limit acceleration and maximum voltage；

(2) non-stationary random reference signal is arranged: under frequency coordinate, arranges random signal with reference to merit according to list mode Rate is composed, and the frequency scale of frequency coordinate is the integral multiple of frequency resolution；It is normalized to 1g, g representation unit gravity by root mean square Acceleration, root mean square is according to 0s moment i.e. initial time to ns moment i.e. finish time, under time coordinate, according to list mode Arranging the root mean square change of random signal, the time scale of time coordinate is the integral multiple of temporal resolution；

(3) System self-test: according to self-inspection starting voltage, sends pseudo-random signal that reference power spectrum is converted to progressively Raise, reach self-inspection magnitude, by the output of detecting system, carry out decision-making system the most working properly, the most then enter next step Suddenly, otherwise terminate；

(4) transmission Function identification: according to the identification magnitude arranged, utilize the pseudorandom letter that reference power spectrum is converted to Number, and carry out frame overlap joint according to the signal frame number arranged, directly carry out ssystem transfer function identification；

(5) vibration control of single shaft loop cycle: the transmission function that obtains according to identification is also inverted, calculates initial driving Frame data, then, carry out real-time circulation process by frame data, are respectively completed following steps: A, meter within the single frame time cycle Calculate normalization and control power spectrum；B, utilize reference power spectrum with normalization control power spectrum compare, corrected Calculation is normalized Drive spectrum；C, normalization drives spectrum carry out phase randomization and IFFT convert, and carry out frame overlap joint and obtain true random signal；D、 Proportional integral model-following control will be carried out with reference to root mean square time changing curve with controlling root mean square, calculate the most newly obtained current driving Root-mean-square gain；E, will drive root-mean-square gain with overlap joint after drive frame data be multiplied be calculated present frame driving export Signal.

As preferably, in described step (2), the computing formula of described frequency resolution is as follows:

$\mathrm{\Δ}f=\frac{f_{max}}{L}$

Wherein, Δ f represents frequency resolution, f_maxHighest frequency, L represents spectral line number；

Described temporal resolution Δ t is the inverse of frequency resolution Δ f.

In described step (4), by following linear averaging formula manipulation transmission function to obtain more preferable smooth effect:

$H_m=H_{m-1}+\frac{1}{m}({\tilde{H}}_m-H_{m-1})$

Wherein,Represent m sampleIn sample, H_mRepresent that m sample does m sublinear and put down All transmission functions of gained, m is the average time arranged；

System coherent function represents the frequency dependence degree of system input and output, uses coherent function penalty method to make up The adverse effect that the poor transmission characteristic of system is brought to vibration control, the final transmission function formula calculated is as follows:

H (ω)=H_m(ω)*γ^-1(ω)

In above formula, H_m(ω) represent m time averagely after transmission function, γ (ω) is the system coherent function calculated.

The method of step A of described step (5) is as follows:

Carry out single frames measurement to controlling response signal, and signal is carried out windowing, FFT process, be re-transformed into power spectrum, will The single frames obtained controls power spectrum and is normalized, and the method calculating normalization control power spectrum is as follows:

${\stackrel{\‾}{C}}_{new}=\left(\frac{1}{C_{rms}}\right)\·C_{new}$

In above formula,Represent that the normalization of present frame controls power spectrum, C_newFor the control power spectrum of present frame, C_rmsFor The frequency domain root mean square of present frame power spectrum；

Exponential average is carried out to controlling spectrum as follows again by following exponential average formula:

${\stackrel{\‾}{C}}_n=\mathrm{\θ}\·{\stackrel{\‾}{C}}_{new}+(1-\mathrm{\θ}){\stackrel{\‾}{C}}_{n-1}$

In above formula,Represent that the normalization of previous frame controls spectrum,Normalization for present frame controls spectrum,For Normalization after renewal controls spectrum, and θ is for updating weight coefficient, and its size is relevant to number of degrees of freedom, θ=16/ (DOF+8), and DOF is 8 ～200；

Finally the normalization after updating is controlled spectrum and carries out the arithmetic average process of 4 frames.

In step B of described step (5), corrected Calculation obtains normalization and drives the computing formula composed as follows:

$log\left({\stackrel{\‾}{D}}_{n+1}\right)=log\left({\stackrel{\‾}{D}}_n\right)+K_{DOF}\·{\stackrel{\‾}{E}}_n$

${\stackrel{\‾}{E}}_n=log(\stackrel{\‾}{R}/{\stackrel{\‾}{C}}_n)$

Wherein,Represent that present frame normalization drives spectrum,Represent that next frame normalization drives spectrum,Represent current Frame normalization drives the error of spectrum,Compose for reference power, K_DOFFor Ratio for error modification, relevant to number of degrees of freedom,.

In step D of described step (5), the computing formula driving root-mean-square gain u (n) is as follows:

$u\left(n\right)=u(n-1)+k_p\[\frac{T}{T_i}e\left(n\right)+e\left(n\right)-e(n-1)\]$

E (n)=R_rms(n)-C_rms(n)

Wherein, k_pFor proportionality coefficient,For integral coefficient, e (n) is root-mean-square error, R_rmsN () is current time With reference to root mean square, C_rmsN () is the frequency domain root mean square of present frame power spectrum.

The beneficial effects of the present invention is:

The present invention is on tradition random vibration test control method basis, and the one of proposition is for non-stationary random continuous Vibration control method under mode of vibration, first, arranges normalized non-stationary random reference spectrum, arranges non-stationary random simultaneously Change over curve with reference to root mean square, then, carry out the transmission Function identification of control object, it is thus achieved that ssystem transfer function is also Invert, finally, control the normalization driving spectrum after the real-time correction of spectrum is updated according to normalizing reference spectrum with normalizing, drive Spectrum phase randomization converts and carries out frame overlap joint with IFFT, simultaneously according to reference to root mean square versus time curve, introduces ratio Example integration model-following control i.e. PI model-following control, obtains the root-mean-square gain driving signal in real time, utilizes normalization to drive frame overlap joint Data and driving signal root-mean-square gain product, update output drive signal.

Compare traditional random vibration test to control, can not only obtain, by the present invention, the spectrum that non-stationary random vibration is satisfied Shape controls, and nonstationary random signal root mean square also can be stablized controllably according to time graph change, and this control method is by non- Stationary random vibration process is approximately a root mean square and the product model of power spectrum, and wherein root mean square changes over reaction Non-stationary random vibration process energy variation process in time；Power spectrum shape is constant, shows the frequency in whole vibration processes Rate characteristic remains stable substantially, and this delivers vibration environment operating mode closely with real, when root mean square does not changes over time, This non-stationary random vibration is stationary random vibration.Therefore, this control method is used to realize non-stationary random vibration test Control is effective and feasible, it is possible to simulating vehicle, locomotive and high ferro transport more truly, and vibration environment airborne, carrier-borne And the non-stationary random vibration environment such as weapon vibration environmental.

Accompanying drawing explanation

Fig. 1 is the structured flowchart of the control system that non-stationary random vibration experimental control method of the present invention uses.

Detailed description of the invention

The invention will be further described below in conjunction with the accompanying drawings:

As it is shown in figure 1, the control system that non-stationary random vibration experimental control method of the present invention uses includes commonly Computer, PXI 8880 controller, the multi-functional input-output card of PXI 6221 (containing 2 tunnel simulation outputs, 8 tunnel simulation inputs).General Having non-stationary random vibration test control software in logical computer, show for user operation and data, PXI 6221 is multi-functional Input-output card is used for gathering multi-channel control point acceleration transducer, and transfers signals to participate in PXI8880 controller calculating Method calculates, calculation of transfer function, calculates and updates driving signal and overlap, then being multiplied by driving root-mean-square gain, by working as of obtaining Front wheel driving frame signal, is exported to controlled device by PXI 6221 analog output channel, realizes according to picture frame period circulation work Non-stationary random continuous vibrates.PXI 6221 and PXI 8880 controller is arranged on the bus slot of PXI 1071 controller box In.

Non-stationary random vibration experimental control method of the present invention, is realized by the software being located in control system, bag Include following steps:

(1) test parameters is arranged: basic parameter includes but not limited to that test name, date, operating path select and remarks, Self-inspection parameter includes but not limited to self-inspection starting voltage, self-inspection magnitude and self-inspection maximum voltage, controls parameter and includes but not limited to Spectral line number, sample frequency, frequency range, channel parameters but be not limited to include channel selecting, control mode, sensitivity coefficient, be System identified parameters includes but not limited to System Discrimination magnitude, identification signal frame number and average time, and security parameter includes but do not limits In extreme displacement, limit velocity, limit acceleration and maximum voltage；

(2) non-stationary random reference signal is arranged: under frequency coordinate, arranges random signal with reference to merit according to list mode Rate is composed, and the frequency scale of frequency coordinate is the integral multiple of frequency resolution；It is normalized to 1g, g representation unit gravity by root mean square Acceleration, root mean square is according to 0s moment i.e. initial time to ns moment i.e. finish time, under time coordinate, according to list mode Arranging the root mean square change of random signal, the time scale of time coordinate is the integral multiple of temporal resolution；Described frequency discrimination The computing formula of rate is as follows:

$\mathrm{\Δ}f=\frac{f_{max}}{L}$

Wherein, Δ f represents frequency resolution, f_maxHighest frequency, L represents spectral line number；

Described temporal resolution Δ t is the inverse of frequency resolution Δ f；

(3) System self-test: according to self-inspection starting voltage, sends pseudo-random signal that reference power spectrum is converted to progressively Raise, reach self-inspection magnitude, by the output of detecting system, carry out decision-making system the most working properly, the most then enter next step Suddenly, otherwise terminate；

(4) transmission Function identification: according to the identification magnitude arranged, utilize the pseudorandom letter that reference power spectrum is converted to Number, and carry out frame overlap joint according to the signal frame number arranged, directly carry out ssystem transfer function identification；In this step, by following Linear averaging formula manipulation transmission function is to obtain more preferable smooth effect:

$H_m=H_{m-1}+\frac{1}{m}({\tilde{H}}_m-H_{m-1})$

Wherein,Represent m sampleIn sample, H_mRepresent that m sample does m sublinear and put down All transmission functions of gained, m is the average time arranged；

System coherent function represents the frequency dependence degree of system input and output, uses coherent function penalty method to make up The adverse effect that the poor transmission characteristic of system is brought to vibration control, the final transmission function formula calculated is as follows:

H (ω)=H_m(ω)*γ^-1(ω)

In above formula, H_m(ω) represent m time averagely after transmission function, γ (ω) is the system coherent function calculated；

(5) vibration control of single shaft loop cycle: the transmission function that obtains according to identification is also inverted, calculates initial driving Frame data, then, carry out real-time circulation process by frame data, are respectively completed following steps: A, meter within the single frame time cycle Calculate normalization and control power spectrum；B, utilize reference power spectrum with normalization control power spectrum compare, corrected Calculation is normalized Drive spectrum；C, normalization drives spectrum carry out phase randomization and IFFT convert, IFFT conversion i.e. inverse discrete Fourier transform, and Carry out frame overlap joint and obtain true random signal；D, will with reference to root mean square time changing curve with control root mean square carry out proportional integral with With control, calculate the most newly obtained current driving root-mean-square gain；E, will drive root-mean-square gain with overlap joint after driving frame data It is multiplied and is calculated the drive output signal of present frame；

In above-mentioned steps (5), the method for step A is as follows:

Carrying out single frames measurement to controlling response signal, and signal carries out windowing, FFT process, the i.e. fast Fourier of FFT becomes Change, be re-transformed into power spectrum, the single frames obtained control power spectrum is normalized, calculates normalization and control power spectrum Method is as follows:

${\stackrel{\‾}{C}}_{new}=\left(\frac{1}{C_{rms}}\right)\·C_{new}$

In above formula,Represent that the normalization of present frame controls power spectrum, C_newFor the control power spectrum of present frame, C_rmsFor The frequency domain root mean square of present frame power spectrum；

Exponential average is carried out to controlling spectrum as follows again by following exponential average formula:

${\stackrel{\‾}{C}}_n=\mathrm{\θ}\·{\stackrel{\‾}{C}}_{new}+(1-\mathrm{\θ}){\stackrel{\‾}{C}}_{n-1}$

In above formula,Represent that the normalization of previous frame controls spectrum,Normalization for present frame controls spectrum,For more Normalization after Xin controls spectrum, and θ is for updating weight coefficient, and its size is relevant to number of degrees of freedom, θ=16/ (DOF+8), DOF 8～ 200；

Finally the normalization after updating is controlled spectrum and carries out the arithmetic average process of 4 frames；

In step B of described step (5), corrected Calculation obtains normalization and drives the computing formula composed as follows:

$log\left({\stackrel{\‾}{D}}_{n+1}\right)=log\left({\stackrel{\‾}{D}}_n\right)+K_{DOF}\·{\stackrel{\‾}{E}}_n$

${\stackrel{\‾}{E}}_n=log(\stackrel{\‾}{R}/{\stackrel{\‾}{C}}_n)$

Wherein,Represent that present frame normalization drives spectrum,Represent that next frame normalization drives spectrum,Represent current Frame normalization drives the error of spectrum,Compose for reference power, K_DOFFor Ratio for error modification, relevant to number of degrees of freedom,；

In step D of described step (5), the computing formula driving root-mean-square gain u (n) is as follows:

$u\left(n\right)=u(n-1)+k_p\[\frac{T}{T_i}e\left(n\right)+e\left(n\right)-e(n-1)\]$

E (n)=R_rms(n)-C_rms(n)

Wherein, k_pFor proportionality coefficient,For integral coefficient, e (n) is root-mean-square error, R_rmsN () is current time With reference to root mean square, C_rmsN () is the frequency domain root mean square of present frame power spectrum.

Above-described embodiment is presently preferred embodiments of the present invention, is not the restriction to technical solution of the present invention, as long as The technical scheme that can realize on the basis of above-described embodiment without creative work, is regarded as falling into patent of the present invention Rights protection in the range of.

Claims (6)

1. a non-stationary random vibration experimental control method, is realized by the software being located in control system, it is characterised in that: Comprise the following steps:

(1) test parameters is arranged: basic parameter includes but not limited to that test name, date, operating path select and remarks, self-inspection Parameter includes but not limited to self-inspection starting voltage, self-inspection magnitude and self-inspection maximum voltage, controls parameter and includes but not limited to spectral line Number, sample frequency, frequency range, channel parameters but be not limited to include channel selecting, control mode, sensitivity coefficient, system is distinguished Knowing parameter and include but not limited to System Discrimination magnitude, identification signal frame number and average time, security parameter includes but not limited to pole Spacing shifting, limit velocity, limit acceleration and maximum voltage；

(2) non-stationary random reference signal is arranged: under frequency coordinate, arrange random signal reference power according to list mode Spectrum, the frequency scale of frequency coordinate is the integral multiple of frequency resolution；It is normalized to 1g, g representation unit gravity by root mean square add Speed, root mean square, according to 0s moment i.e. initial time to ns moment i.e. finish time, under time coordinate, sets according to list mode Putting the root mean square change of random signal, the time scale of time coordinate is the integral multiple of temporal resolution；

(3) System self-test: according to self-inspection starting voltage, sends reference power and composes the pseudo-random signal being converted to and progressively rise Height, reaches self-inspection magnitude, by the output of detecting system, carrys out decision-making system the most working properly, the most then enter next step, Otherwise terminate；

(4) transmission Function identification: according to the identification magnitude arranged, utilizes reference power to compose the pseudo-random signal being converted to, and Carry out frame overlap joint according to the signal frame number arranged, directly carry out ssystem transfer function identification；

(5) vibration control of single shaft loop cycle: the transmission function that obtains according to identification is also inverted, and calculates and initially drives frame number According to, then, carry out real-time circulation process by frame data, within the single frame time cycle, be respectively completed following steps: A, calculate and return One changes control power spectrum；B, utilize reference power spectrum with normalization control power spectrum compare, corrected Calculation obtain normalization driving Spectrum；C, normalization drives spectrum carry out phase randomization and IFFT convert, and carry out frame overlap joint and obtain true random signal；D, general's ginseng Examine root mean square time changing curve and carry out proportional integral model-following control with controlling root mean square, calculate the most newly obtained current driving the most square Root gain；E, will drive root-mean-square gain with overlap joint after drive frame data be multiplied be calculated present frame driving output believe Number.

Non-stationary random vibration experimental control method the most according to claim 1, it is characterised in that: in described step (2), The computing formula of described frequency resolution is as follows:

$\mathrm{\Δ}f=\frac{f_{max}}{L}$

Wherein, Δ f represents frequency resolution, f_maxHighest frequency, L represents spectral line number；

Described temporal resolution Δ t is the inverse of frequency resolution Δ f.

Non-stationary random vibration experimental control method the most according to claim 1, it is characterised in that: in described step (4), By following linear averaging formula manipulation transmission function to obtain more preferable smooth effect:

$H_m=H_{m-1}+\frac{1}{m}({\tilde{H}}_m-H_{m-1})$

Wherein,Represent m sampleIn sample, H_mRepresent that m sample does the average institute of m sublinear The transmission function obtained, m is the average time arranged；

System coherent function represents the frequency dependence degree of system input and output, uses coherent function penalty method to make up system The adverse effect that poor transmission characteristic is brought to vibration control, the final transmission function formula calculated is as follows:

H (ω)=H_m(ω)*γ^-1(ω)

In above formula, H_m(ω) represent m time averagely after transmission function, γ (ω) is the system coherent function calculated.

Non-stationary random vibration experimental control method the most according to claim 1, it is characterised in that: described step (5) The method of step A is as follows:

Carry out single frames measurement to controlling response signal, and signal is carried out windowing, FFT process, be re-transformed into power spectrum, will obtain Single frames control power spectrum be normalized, calculate normalization control power spectrum method as follows:

${\stackrel{\‾}{C}}_{new}=\left(\frac{1}{C_{rms}}\right)\·C_{new}$

In above formula,Represent that the normalization of present frame controls power spectrum, C_newFor the control power spectrum of present frame, C_rmsFor currently The frequency domain root mean square of frame power spectrum；

Exponential average is carried out to controlling spectrum as follows again by following exponential average formula:

${\stackrel{\‾}{C}}_n=\mathrm{\θ}\·{\stackrel{\‾}{C}}_{new}+(1-\mathrm{\θ}){\stackrel{\‾}{C}}_{n-1}$

In above formula,Represent that the normalization of previous frame controls spectrum,Normalization for present frame controls spectrum,After updating Normalization control spectrum, θ for update weight coefficient, its size is relevant to number of degrees of freedom, θ=16/ (DOF+8), and DOF is 8～200；

Finally the normalization after updating is controlled spectrum and carries out the arithmetic average process of 4 frames.

Non-stationary random vibration experimental control method the most according to claim 1, it is characterised in that: described step (5) In step B, corrected Calculation obtains normalization and drives the computing formula composed as follows:

$log\left({\stackrel{\‾}{D}}_{n+1}\right)=log\left({\stackrel{\‾}{D}}_n\right)+K_{DOF}\·{\stackrel{\‾}{E}}_n$

${\stackrel{\‾}{E}}_n=log(\stackrel{\‾}{R}/{\stackrel{\‾}{C}}_n)$

Wherein,Represent that present frame normalization drives spectrum,Represent that next frame normalization drives spectrum,Represent that present frame is returned One changes the error driving spectrum,Compose for reference power, K_DOFFor Ratio for error modification, relevant to number of degrees of freedom,.

Non-stationary random vibration experimental control method the most according to claim 1, it is characterised in that: described step (5) In step D, the computing formula driving root-mean-square gain u (n) is as follows:

$u\left(n\right)=u(n-1)+k_p\[\frac{T}{T_i}e\left(n\right)+e\left(n\right)-e(n-1)\]$

E (n)=R_rms(n)-C_rms(n)

Wherein, k_pFor proportionality coefficient,For integral coefficient, e (n) is root-mean-square error, R_rmsN () is the reference of current time Root mean square, C_rmsN () is the frequency domain root mean square of present frame power spectrum.