JP2012233814A - Vibration test control device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a vibration test control device capable of performing a random vibration test setting the kurtosis of a random signal as a parameter.SOLUTION: In a random vibration test control apparatus (300) for performing a vibration test of a test piece by applying a random vibration signal to a vibration exciter device (308), probability density distribution function generation means (301) for obtaining a probability density distribution function Dist(x) of non-normal distribution satisfying the conditions of kurtosis, variance, and PSD spectrum given as a test target and an objective random sequence signal generation means (302) for generating a target random sequence signal Tg[t] of non-normal distribution from the probability density distribution function Dist(x) are provided, a control is carried out so that the target random sequence signal Tg[t] and a response sequence signal Rt coincide with each other in amplitude and phase, and a vibration test by the random signal of non-normal distribution is realized.

Description

  The present invention relates to a vibration test control apparatus. More specifically, the earthquake resistance and durability of an industrial product are evaluated by applying the same random vibration to the industrial product in the laboratory.

In order to verify the proper operation of an industrial product, testing various vibrations applied to the product in a state close to the natural environment is one of the important processes in product design.
In the vibration test, it is required to perform the vibration test by giving the test object (product) the complicated vibration recorded in the actual use / installation environment and the situation used / transported.
In the random vibration test, a normally distributed random signal (kurtosis = 3) having a power spectrum density (PSD) defined by a frequency distribution has been used.
In recent years, however, the need for non-normally distributed random signals (kurtosis> 3) has been screamed.
In the conventional vibration test control device, the random signal used for the test is a normal distribution random signal generated from the probability density distribution function of the normal distribution, and is controlled by comparing the target PSD of the random test with the average PSD of the response. As a result, random vibration tests using the kurtosis of random signals as a parameter could not be performed.

The configuration and operation of a conventional vibration test control apparatus 100 will be described with reference to FIG.
A response signal from the vibration exciter device (including the test body and sensor) 112 is converted into a digital signal by the AD converter 111.
The digital signal is processed by the FFT converter 110 to calculate a Fourier spectrum for each frame, and a response PSD 108 is generated by averaging for a certain period of time.
The target PSD 101 obtained from the PSD pattern to be tested and the response PSD 108 obtained from the response signal are compared, and a modified PSD 102 in which the target PSD 101 is corrected by the PSD operation amount 109 minutes that is the difference between the target PSD 101 and the response PSD 108 is generated.
A reverse transfer characteristic is calculated from the transfer characteristic of the vibration exciter apparatus 112, and a transfer characteristic correction process is performed on the modified PSD 102 in the reverse transfer characteristic multiplication 103 to generate a drive PSD.
Normal distribution random signal generated by normal distribution random number generator 105 is normalized by modified PSD102, converted to normal distribution random time series signal by IFFT converter (Inverse Fast Fourier Transform) 104, and driving waveform The synthesis 106 performs superposition synthesis processing of time series signals, and the DA converter 107 generates a drive signal to the vibration exciter device 112.

Special table 2007-531896 gazette

When the random vibration test is performed in the conventional vibration test apparatus 100, only the normal distribution random signal generated by the normal distribution random number generator 105 can be generated.
Therefore, because the random signal kurtosis (Kurtosis) including the normal distribution random signal (kurtosis = 3) and the non-normal distribution random signal (kurtosis> 3) cannot be specified at the time of random signal generation, Random vibration tests with different kurtosis were difficult.

 The vibration test control apparatus according to claim 1 of the present invention for solving the above-described problems is a random vibration test control apparatus (300) for performing a vibration test of a specimen by giving a random vibration signal to the vibration exciter apparatus (308). A response time series generating means (307) for generating a response time series signal Rt by converting a response signal obtained from a sensor for detecting the acceleration of the test body into a digital time series signal by AD conversion, and giving as a test target A probability density distribution function generating means (301) for obtaining a non-normal distribution probability density distribution function Dist (x) that satisfies the conditions of kurtosis, variance and PSD spectrum, and the probability density distribution function Dist (x) A target random time-series signal generating means (302) for generating a target random time-series signal Tg [t] having a non-normal distribution, and a frequency sequence to be a control target in a frequency space from the target random time-series signal Tg [t] A target that is a signal Target spectrum generating means (303) for converting to Kutor Tf, a transfer function and an inverse transfer function are calculated from the response time series signal Rt and the drive time series signal Dt, and the amplitude and phase of the target spectrum Tf in the frequency space And a frequency series correction control means (304) for generating a driving Fourier spectrum Dr, and a time series signal for generating a smooth driving time series signal Dt from the driving Fourier spectrum Dr by an inverse fast Fourier transform and an overlap synthesis method And a decoding means (305), wherein the control is performed such that the target random time series signal Tg [t] and the response time series signal Rt are matched, and a vibration test using a non-normally distributed random signal is realized. To do.

The vibration test control apparatus according to claim 2 of the present invention for solving the above problem is the vibration test control apparatus according to claim 1, wherein the probability density distribution function generating means (301) calculates an rms value from a given PSD pattern. A step of calculating and calculating an average PSD from the rms value (R01), a step of generating a normal distribution time series signal F (t) satisfying a flat spectrum of the average PSD (R02), and the normal distribution time series signal F The step of determining the variance σ ² and the kurtosis Kw by the step (R03) of determining the variance σ ² in (t) (R04), and the step of determining the probability density distribution function Dist (x) from the variance σ ² and the kurtosis Kw ( R05).

The vibration test control apparatus according to a third aspect of the present invention for solving the above-described problems is the vibration test control apparatus according to the first or second aspect, wherein the probability density distribution function generating means (301) sets the specified standard deviation σ. A step (D01) for setting an initial normal distribution function g ₀ (x) having, a step (D02) for newly defining a standard deviation σw for further expanding the initial standard deviation σ and a standard deviation σs for narrowing, and a standard deviation σw Generate a normal distribution function (Gaussian distribution) with standard deviation σs, add three normal distribution functions (Gaussian distribution) with standard deviation σ, standard deviation σw, and standard deviation σs, calculate the average, and gi ( x) (D03), a step (D04) of calculating the kurtosis Kg of gi (x) while simultaneously changing the standard deviation σw and the standard deviation σs, and the kurtosis Kg of gi (x) is predetermined. A step (D06) of obtaining a non-normal distribution function gi (x) = g (x) of the kurtosis Kw by a step (D02 to D05) of repeating the calculation until the kurtosis Kw reaches a value (D06), g (x) A step (D07) for calculating the standard deviation σg, σ / σg only a distribution function and the step of creating a g (x) which is transverse deformation (D08), g (x) with the maximum value Max _g of g (x) And a step (D09) of obtaining a probability density distribution function Dist (x) of a non-normal distribution having a maximum value of 1.0 and a kurtosis of Kw.

The vibration test control apparatus according to a fourth aspect of the present invention for solving the above-mentioned problems is the vibration test control apparatus according to the first, second or third aspect, wherein the target random time series signal generating means (302) uses a Rand function. As shown in equation (7) from X _Min to X _Max , a white random number RandX is generated, and the probability density Probab in RandX is obtained from the probability density distribution function Dist (x), and at the same time, white from 0.0 to 1.0 Generate random number RandC,
(Formula 7)
X _Min = −βσ
X _Max = + βσ
However, β is a value determined from the maximum excitation force of the vibration exciter device (308).

If RandC value <Probab, RandX is adopted as the output value, otherwise discarding is repeated, and the adopted RandX is stored in the first-in first-out (FIFO) memory to generate a probability of non-normal distribution A process (R06) in which a random time series signal Rn (t) having a density distribution function Dist (x) is generated, and the generated random time series signal Rn (t) is subjected to an FFT transform process to obtain a Fourier spectrum fft (i ) And power spectral density PSD (i) (R07), PSD (i) is a flat distribution pattern with minute fluctuations around a constant value, so PSD (i) The process of calculating the Ave PSD (i) by performing exponential averaging (R08), and Factor (i) = Target PSD (i) / Ave PSD (i) is calculated to match the PSD pattern with the target PSD pattern, and Factor (i) By multiplying Fourier spectrum fft (i) by Fourier transform fft (i) and inverse Fourier transform of Fourier spectrum fft (i) multiplied by Factor (i). And a step (R10) of generating a target random time series signal Tg [t] having a period Kw, σ, and PSD pattern.

  Analyze the measured data recorded at the time of actual use and transportation of the industrial product, obtain the dispersion, kurtosis and PSD spectrum specific to the measurement data from the frequency distribution, and use these kurtosis and dispersion as the specified parameters. A probability density distribution function satisfying the PSD spectrum, a random time series signal corresponding to the probability density distribution function is generated, and the random time series signal is set as a target time series signal waveform to match the response time series signal waveform. By performing this, it is possible to perform a random vibration test of a non-normal distribution according to the distribution in which the PSD of the Fourier spectrum is specified and the kurtosis and variance of the time series signal have specified values.

It is a block diagram of the vibration test control apparatus of the non-normal distribution random which concerns on Example 1 of this invention. It is a flowchart which shows the flow of the process which produces | generates a probability density distribution function. It is a flowchart which shows the flow of the process which produces | generates the probability density distribution function designated with standard deviation (sigma) and kurtosis Kw. FIG. 4A is a graph showing the probability density distribution function, and FIG. 4B is a graph showing a part of FIG. It is a flowchart which shows the flow of the process which produces | generates the random time series signal according to the distribution from a probability density distribution function. It is a frequency distribution chart which graphed measured value Xi of time series signal x [t], and the appearance frequency Ni. It is a block diagram of the non-normal distribution random 3 axis | shaft simultaneous vibration test control apparatus which concerns on Example 2 of this invention. It is a graph which shows the probability density distribution of the random vibration test control result of X, Y, Z each axis from which the kurtosis differs in the same PSD spectrum in the non-normal distribution random three-axis simultaneous vibration test control apparatus by this invention. It is a graph which shows the spectrum of the random vibration test control result of X, Y, Z each axis | shaft from which the kurtosis differs in the same PSD spectrum in the non-normal distribution random three-axis simultaneous vibration test control apparatus by this invention. It is a block diagram which shows the hardware constitutions of the non-normal distribution random three-axis simultaneous vibration test control apparatus by this invention. It is a block diagram which shows the conventional vibration test control apparatus. It is a graph which shows an example of a probability density distribution function. It is a graph which shows an example of the probability density distribution function which added RandX and RandC. 5 is a flowchart for generating a random time series signal Rn (t). It is a graph which shows the relationship between target PSD and average PSD. It is a flowchart which calculates | requires a normal distribution time series signal from average PSD. It is a graph which shows the relationship between the Fourier spectrum component and the deflection angle θ of the complex plane. It is explanatory drawing of probability density distribution and the random time series signal according to this.

Hereinafter, embodiments of the present invention will be described with reference to examples shown in the drawings. Prior to describing the present invention, terms used in the present invention will be defined. An asterisk (*) is an operator representing a product.
<Definition 1 PSD>
Between the time series signal f (t) and its Fourier transform F (ω)

There is a relationship between the real component a and the imaginary component b that are the outputs of the discrete Fourier transform FFT.

The PSD [n] for ω [n] is calculated using a [n] and b [n] for ω [n] in the FFT transform of the N-point time-series signal f (t).
PSD [n] = 2.0 * (a [n] ² + b [n] ² ) / (N * N * df)
It is defined as
<Definition 2 RMS>
Using PSD [n] calculated from N-point time series signal f (t), RMS of Fourier spectrum from ω [0] to ω [k] is

It is defined as
<Definition 3 Probability density distribution>
FIG. 6 is a graph showing the measured value Xi of the time series signal x [t] and the number of appearances Ni of the time series signal x [t] with Xi and Ni, and is called frequency distribution.
The probability density distribution is defined by recalculating the frequency distribution (FIG. 6) with the maximum number of appearances as 1.0.
<Definition 4 Number of valid cases M>

<Definition 5 Mean>

<Definition 6 variance σ ² >

<Definition 7 Kurtosis Kw>

A non-normal distribution random vibration test control apparatus according to Embodiment 1 of the present invention is shown in FIGS.
First, with reference to FIG. 1, the control process of the non-normal distribution random vibration test control apparatus will be described.
As shown in FIG. 1, the vibration test control apparatus 300 with a non-normal distribution random includes a probability density distribution function generation unit 301, a target random time series signal generation unit 302, a target spectrum generation unit 303, a frequency series correction control unit 304, a time series. A signal decoding unit 305, a DA conversion unit 306, and an AD conversion unit 307 are provided, and a vibration exciter device (including a power amplifier and a vibration response detector) 308 is connected to the vibration test control device 300.

<Probability density distribution function generator 301>
Details of the probability density distribution function generation unit 301 will be described with reference to FIG. 2 in the case where a PSD pattern and a kurtosis Kw as test targets are individually given.
First, an RMS value is calculated from a given PSD pattern, and an average PSD is obtained from the RMS value (R01).

That is, PSD [n] for ω [n] using a [n] and b [n] for ω [n] in FFT conversion of the N-point time-series signal f (t) is defined as <Definition 1 PSD>. As shown.
In addition, the RMS of the Fourier spectrum from ω [0] to ω [k] using PSD [n] calculated from the N-point time series signal f (t) is as shown in <Definition 2 RMS>. It is. In other words, RMS is equal to the square root of the shaded area of the target PSD (the sum of PSD (n) from n = 0 to k multiplied by df), as shown in FIG.

Since the average PSD (PSDAve) is a flat PSD having the same RMS value as the target PSD, the following equation holds.
PSDAve ＝ (RMS ² / df) k
That is, the average PSD pattern can be said to be a PSD pattern in which the height of all frequency bands is PSDAve as shown in FIG.
Next, a normal distribution time series signal F (t) satisfying the flat spectrum of the average PSD is generated (R02). This will be described with reference to FIG.

First, a Fourier spectrum having a random phase is generated from the average PSD pattern (step 501).
Assuming that the fundamental frequency f ₀ = 1 / T and the fundamental angular frequency ω ₀ = 2πf ₀ , the time series signal x (t) and its Fourier transform X _n

There is the following relationship between the real component a and the imaginary component b of the output of the FFT, which is a discrete Fourier transform.

Between the real component a _n and the imaginary component b _n and PSD (n),

However,
N: Number of sampling points of time series data at the time of Fourier transform N = 2 ⁱ i = 1, 2, 3, 4, 5.
n: 1, 2, 3, ..., (N / 2-1)
Δf: frequency resolution at the time of Fourier transform Δf = 1 / (dt · N)
dt: There is a relationship of sampling time intervals of time series data, and Absolute (n) of the Fourier spectrum uses PSD (n),

It is expressed. An asterisk (*) indicates a product operator. Hereinafter, the same applies to the drawings. The Fourier spectrum component indicating PSD (n), where Absolute is the base and the declination of the Fourier spectrum is θ,

Can be written.
The relationship between the Fourier spectrum component and the deviation angle θ of the complex plane is as shown in FIG.

The PSD is determined by the value of Absolute (PSD value) 601 shown in FIG. 17. However, since there is a degree of freedom in the phase value θ, the PSD distribution can be obtained by randomly varying the phase value θ. A random phase Fourier spectrum can be generated in which does not change.
If a Fourier spectrum having a random phase is subjected to IFFT processing (inverse FFT transform), a normal distribution time series signal F (t) having a spectrum distribution of an average PSD pattern in which sine waves having different phases are overlapped is generated (step). 502).

However, IFFT processing used for digital signal processing is a discrete process, so it does not include all frequency components without omission, and there is an effect that missing components are made up by mixing the phases, but there are still some missing components. Since this occurs, this normally distributed time series signal is called pseudo-random. In order to generate true random from pseudo-random, the following processing is performed.
A process of multiplying a time series signal for each pseudo-random frame by a Hanning Window is performed (step 503).

The normal distribution time series signal multiplied by the Hanning window is subjected to the overlap synthesis process every 1/2 frame (step 504).
The time series are combined by the process of multiplying the Hanning window and the process of overlap synthesis every 1/2 frame, and the true normal distribution time series signal F (t) maintaining the average PSD pattern is generated (step 505).
By this normal distribution time-series signal F (t) in determining the variance σ ² (R03), determines variance sigma ² and kurtosis Kw (R04), variance sigma ² probability density from Kurtosis Kw distribution function Dist (x ) (R05).

Also, if there is a sufficient amount of measured time series data, the average PSD, variance σ ² and kurtosis Kw are calculated from the measured time series data to obtain the probability density distribution function Dist (x) (R05 ) Is also possible.
The generation of the probability density distribution function Dist (x) having the specified PSD pattern and the kurtosis Kw can also be performed by offline processing as preprocessing of the random vibration test of non-normal distribution.
Details of the method of obtaining the probability density distribution function Dist (x) satisfying the variance σ ² and the kurtosis Kw of the test target (R05) will be described with reference to FIG. 3 (hereinafter, the square root σ of the variance σ ² is the standard deviation). Called).

The normal distribution (Gaussian distribution) function has the standard deviation σ (variance σ ² ) defined in (Equation 1) and the mean value μ as parameters, but the normal distribution (Gaussian distribution when the mean value μ is zero) The (distribution) function is defined by (Equation 2).
In the case of a vibration test, since a random signal is treated as a signal that vibrates randomly around zero, (Equation 2) is used.

Since the definite integral from -∞ to + ∞ of the cumulative distribution function in the normal distribution (Gaussian distribution) function is 1, the definite integral from -∞ to + ∞ in (Equation 2) is the standard deviation σ (variance σ ² ) It becomes 1 regardless of (Equation 3).

First, an initial normal distribution function having a designated standard deviation σ (variance σ ² ) is set as g ₀ (x) (Equation 4) (D01).

Furthermore, the standard deviation σw that expands the initial standard deviation σ and the standard deviation σs that narrows it are newly defined (D02), and a normal distribution function (Gaussian distribution) with the standard deviation σw and standard deviation σs is generated. Three normal distribution functions (Gaussian distribution) having σw and standard deviation σs are added, and the average is calculated (formula 5) to be g _i (x) (D03).
As initial conditions, σw = σs = σ, and changes in σw and σs are set to Δσw = Δσs = 0.01σ (D01).
The coefficient 0.01 affects the convergence speed, and an appropriate value is selected.

The kurtosis Kg of gi (x) is calculated while simultaneously changing the standard deviation σw and the standard deviation σs (D04). By repeating the operations of Definitions 3 to 7 until the kurtosis Kg of gi (x) reaches a predetermined kurtosis Kw (D02 to D05), the non-normal distribution function gi (x) of the kurtosis Kw is obtained (D06 ).
Since the addition average gi (x) of the normal distribution function (Gaussian distribution) has the property as shown in (Expression 3), the definite integral from −∞ to + ∞ in (Expression 5) is 1 (Expression 6).

The standard deviation σg (D07) of the non-normal distribution function g (x) obtained here is slightly different from the standard deviation σ (R03) in the normal distribution time series signal F (t), so the standard deviation of the distribution function g (x) By laterally deforming σg by σ / σg (D08), PSD (i) with a kurtosis Kw and standard deviation σ (variance σ ² ) is obtained as a non-normal distribution function g (x) (D09) The rms value is equal to the rms value of the average PSD (R01) of the given PSD pattern.
When g (x) is divided by the maximum value Max _g of g (x), the probability density distribution function Dist (x) of the non-normal distribution having the maximum value of 1.0 and the kurtosis of Kw is obtained (FIG. 4).

<Target random time series signal generator 302>
For details of the target random time series signal generation unit 302, refer to FIG. 4 and FIG. 5 when generating a target random time series signal Tg [t] having a non-normal distribution from a probability density distribution function Dist (x) having a non-normal distribution. To explain.
An example of the probability density distribution function Dist (x) of the non-normal distribution distributed from the signal range X _Min to X _Max is shown in FIG. In this probability density distribution function, Xi is a value of a signal calculated for each resolution interval ΔX of X, and Pr [i] is a probability density at which Xi appears. When this probability density distribution function is expanded into the probability density distribution function table Dist A, Table 1 is obtained.

As shown in FIG. 14, a white (uniform) random number RandX = Xi that appears evenly every resolution interval ΔX from the signal range X _Min to X _Max shown in (Expression 7) is generated based on the Rand function (s1). ), The probability density Probab in RandX is obtained from the probability density distribution function Dist (x) (s2), and at the same time, another white random number RandC uniformly distributed from 0.0 to 1.0 is generated (s3). RandX corresponds to the horizontal axis of the probability density distribution function shown in FIG. 12, and the probability density Probab in RandX corresponds to the vertical axis. Therefore, as shown in FIG. 13, the probability density Pr [i] = Probab at which Xi appears. Further, obtaining the probability density Probab from the probability density distribution function Dist (x) is equivalent to subtracting Pr [i] corresponding to Xi in the probability density distribution function table Dist A of Table 1.
(Formula 7)
X _Min = −βσ
X _Max = + βσ

However, β is a value determined from the maximum excitation force of the vibratory vibration exciter device 308. Here, RandC is used as a random signal generation value or discarded because Xi appears at Probab's probability density. used.
Further, as shown in FIG. 14, when the value of RandC <Probab, RandX is adopted as the output value (s4), and the others are discarded. For example, in the case of Probab = 0.2, on average, RandX is adopted as an output value twice out of 10 times, and 8 times is discarded. Therefore, RandX will be adopted with Probab's probability density. The adopted RandX is sequentially stored in the memory (a small computer described later) as a random value. In other words, n random values are generated at a time at the PC speed. Here, the PC is a small computer described later.

  The generation processing of the white random number RandX and the white random number RandC is repeated, and a random time series signal Rn (t) having a non-normal distribution probability density distribution function Dist (x) as shown in FIG. 18 is accumulated in the FIFO memory. (R06).

The generated random time series signal Rn (t) is subjected to an FFT transform to obtain a Fourier spectrum fft (i), and a power spectrum density (PSD) PSD (i) is obtained from the Fourier spectrum fft (i) ( R07).
Since PSD (i) is a flat distribution pattern having minute fluctuations around a certain value, Ave PSD (i) is obtained by exponential averaging of PSD (i) to obtain a stable flat distribution (R08).

Factor (i) = target PSD (i) / Ave PSD (i) is calculated to match the PSD pattern with the target PSD pattern, and Factor (i) is multiplied by the Fourier spectrum fft (i) (R09).
When a Fourier spectrum fft (i) multiplied by Factor (i) is subjected to inverse FFT transform, a target random time series signal Tg [t] having an intended Kw, σ and PSD pattern is generated (R10).
Recalculate the kurtosis Kw and variance σ ² (standard deviation σ) from the frequency distribution of the target random time series signal Tg [t], and constantly monitor and correct the probability density distribution function Dist (x) (R11) (R12) . The reason for performing the monitoring correction in this way is that the factor (i) is multiplied to cause a difference from the original Fourier spectrum fft (i).

By repeating R06 to R12, the processing system continues to output a random time series signal of non-normal distribution that matches the kurtosis Kw, the standard deviation σ (variance σ ² ), and the PSD pattern.
The process of repeating R06 to R12 is a real-time process performed within one frame time of random vibration control.

<Target spectrum generator 303>
The target spectrum generation unit 303 will be described.
The target random time series signal Tg [t]) of non-normal distribution generated by the target random time series signal generation unit 302 of non-normal distribution is used as the target waveform, and the response waveform from the vibration exciter device 308 is the target waveform. By controlling faithfully to match, a non-normally distributed random vibration test with the desired Kw, σ, and PSD patterns is realized.
Therefore, since it is necessary to control including the amplitude and phase for each frequency component, the target random time series signal Tg [t]) of the non-normal distribution as the target waveform is divided in the frequency band under test by FFT processing. Further, the target spectrum Tf in the frequency space is generated by obtaining the amplitude and phase for each frequency (Δf).

<Frequency series correction control unit 304>
The frequency series correction control unit 304 will be described.
The response time series signal Rt converted from the response analog signal of the vibration shaker device 308 to a digital signal by the AD conversion unit 307 and the DA conversion unit 306 that converts it to the drive analog signal of the vibration shaker device 308 are branched in the previous stage. The transfer characteristic (transfer function including amplitude and phase) in the frequency space is calculated from the measured drive time series signal Dt, and the transfer characteristic correction processing (reverse transfer) of the vibration exciter device 308 including the test sample in the target spectrum Tf The process is multiplied by a function to generate a driving Fourier spectrum Dr.
As a result, non-normal distribution random vibration control in which the target random time-series signal Tg [t] and the response time-series signal Rt coincide with each other becomes possible.

<Time-series signal decoding unit 305>
The time series signal decoding unit 305 will be described.
The function of the time series signal decoding unit 305 is an operation of decoding the drive Fourier spectrum Dr from the frequency series signal to the time series signal by IFFT transform (inverse fast Fourier transform).
Since signal operations in the frequency space are always performed collectively for time series signals of appropriate length (frames) determined by the FFT operation logic, time series signals obtained by IFFT transform (inverse fast Fourier transform) There is a discontinuity between frames.
Control operations are performed on the premise that the time series signal is piecewise continuous, so overlap processing is applied to remove the discontinuity of the time series signal obtained by IFFT conversion, and continuous true random driving A series signal Dt is generated.

<DA converter 306>
The DA conversion unit 306 will be described.
The function of the DA converter 306 is an operation of converting the drive time series signal Dt into an analog signal for driving the vibration exciter device 308.
Since the vibration exciter device 308 is driven by an analog electric signal, it is necessary to convert the driving time series signal Dt, which is a digital signal, into an analog signal.

<AD converter 307>
The AD conversion unit 307 will be described.
The function of the AD conversion unit 307 is to convert the response analog signal of the vibration exciter device 308 into a response time series signal Rt that is a digital signal.

The vibration applied to the industrial product in the actual use environment is three-dimensional.
Analyzing the three-dimensional (called X-axis, Y-axis, and Z-axis) measurement data recorded during actual use and transportation, the inherent dispersion, kurtosis and chromaticity of each of the X-axis, Y-axis, and Z-axis A PSD spectrum is obtained.
Generates a random time series signal that is replaced with a random signal that satisfies the statistical distribution and average energy of each of the X, Y, and Z axes, and uses this random time series signal as the target waveform for each of the X, Y, and Z axes. By simultaneously controlling the three axes so that the response waveform matches the target waveform, the vibration stress actually received by the specimen can be reproduced three-dimensionally.

A control process of the non-normally distributed random three-axis simultaneous vibration test control device 600 according to the second embodiment of the present invention will be described with reference to FIG.
As shown in FIG. 7, the random distribution random three-axis simultaneous vibration test control apparatus 600 includes the X-axis, Y-axis, and Z-axis probability density distribution function generators 601, 604, 607, X-axis, Y-axis, and Z-axis. Respective target random time series signal generation units 602, 605, 608, X-axis, Y-axis, Z-axis target spectrum generation units 603, 606, 609, 3-axis simultaneous frequency series correction control unit 610, X-axis time series signal Decoder 611, Y-axis time series signal decoder 612, Z-axis time series signal decoder 613, X-axis AD / DA converter 614, Y-axis AD / DA converter 615, Z-axis AD / DA converter 616, synchronization A clock 617 is provided, and a three-axis vibration exciter device (including a power amplifier and a vibration response detector) 618 is connected to the three-axis simultaneous vibration test control device 600.

<Probability density distribution function generators 601, 604, 607>
Non-normal distribution random probability density distribution function for each of the X, Y, and Z axes is the kurtosis given to each of the X, Y, and Z axes by the probability density distribution function generators 601, 604, and 607. It is generated from Kw, standard deviation σ (variance σ ² ) and PSD pattern by the same procedure as the probability density distribution function generation unit 301.
The probability density distribution function for each of the X-axis, Y-axis, and Z-axis can be generated in advance before starting the vibration test, so it is possible to perform a random vibration test with a probability density distribution function that suits the test purpose. It becomes.

<Target random time series signal generators 602, 605, 608>
The target random time-series signal generators 602, 605, and 608 use the same procedure as the target random time-series signal generator 302 from the probability density distribution functions of the X-axis, Y-axis, and Z-axis, It continues to generate and output a target random time series signal of non-normal distribution that matches the kurtosis Kw, standard deviation σ (variance σ ² ), and PSD pattern individually given for each Z axis.

<Target spectrum generator 603, 606, 609>
The target spectrum generators 603, 606, and 609) calculate the X-axis, Y-axis, and Z-axis in the frequency space from the target random time series signals of the X-axis, Y-axis, and Z-axis in the same procedure as the target spectrum generator 303. Non-normal distribution random target spectra Tfx, Tfy, and Tfz for each axis are generated.

  The three-axis simultaneous frequency sequence correction control unit 610 incorporates a transfer function identification calculation mechanism having a three-dimensional matrix calculation function necessary for correction control in the frequency sequence space, and a three-axis vibration exciter apparatus including a test body. 612 transfer characteristics (including interference components) are obtained, correction processing is performed on the target spectra Tfx, Tfy, and Tfz for the X, Y, and Z axes, and the drive Fourier spectra Drx for the X, Y, and Z axes are obtained. , Dry, Drz (formula 8) are generated.

  X-axis, Y-axis, and Z-axis drive Fourier spectra Drx, Dry, Drz are IFFT transformed by the X-axis time series signal decoding unit 611, the Y-axis time series signal decoding unit 612, and the Z-axis time series signal decoding unit 613, respectively. Generates continuous true non-normally distributed drive time series signals Dtx, Dty, and Dtz with the desired Kw, σ, and PSD patterns for each of the X, Y, and Z axes through overlap processing To do.

X-axis, Y-axis, and Z-axis drive time series signals Dtx, Dty, and Dtz are passed through the X-axis AD / DA converter 614, Y-axis AD / DA converter 615, and Z-axis AD / DA converter 616 for 3 axes. It is converted into a drive analog signal of the vibration exciter device 618.
As a result, a three-axis simultaneous vibration test using a non-normal distribution random with the desired Kw, σ, and PSD patterns is realized.

FIG. 8 shows the probability density distribution of the control results of the random vibration test for each of the X, Y, and Z axes with the same PSD spectrum and different kurtosis by the non-normally distributed random three-axis simultaneous vibration test controller 600 of the present invention. Indicates.
FIG. 9 shows a spectrum of control results of random vibration tests of the X-axis, Y-axis, and Z-axis with the same PSD spectrum and different kurtosis by the non-normally distributed random three-axis simultaneous vibration test control apparatus 600 of the present invention. Indicates.

The hardware configuration and operation of the non-normally distributed random three-axis simultaneous vibration test control apparatus 600 according to the third embodiment of the present invention will be described with reference to FIG.
The basic hardware configuration is a small computer 801 equipped with general commercial operating software, and a control operation that operates completely in synchronization with the 51.2KHz sampling clock, independent of the operation of the small computer 801. And an X-axis sub-function device (master) 803, an Y-axis sub-function device (slave 1) 804, and a Z-axis sub-function device (slave 2) 805 that perform analog signal input / output conversion. The small computer 801 includes a calculation control unit and a multi-channel FIFO, and the sub-function devices 803, 804, and 805 include a loop buffer unit, a calculation control unit, and an AD / DA conversion unit.

Since the commercial operating software of the small computer 801 does not have a real-time control function, the small computer 801 cannot perform the real-time operation required for the control of the random vibration test, but the operation speed is synchronized with the clock. This is much faster than the restricted sub-function devices 803, 804, and 805.
In each of the sub-function devices 803, 804, and 805, all function blocks including the AD / DA conversion unit oscillate from the inside of the sub-function device (master) 803 independently of the operation of the small computer 801. It operates in complete synchronization with the KHz sampling clock.

The small computer 801 takes advantage of the high-speed computation, generates a non-normally distributed random three-axis simultaneous vibration control target signal and part of the drive signal, regardless of the control progress of each sub-function device 803, 804, 805 , Write the signal data for up to 60 seconds in advance to the multi-channel FIFO.
Each sub-function device 803, 804, 805 has a loop buffer unit having the same capacity as that mounted on the small computer 801.
Each sub-function device 803, 804, 805 takes out one frame signal from the loop buffer unit in synchronization with the sampling clock and executes the control, and every time the control is completed, the loop buffer unit is vacant. Notify computer 801.
Since the practical speed for processing one frame of time-series data is much faster for the small computer 801, each sub-function device 803, 804, 805 always stores sufficient drive signals and is stable control. Is preserved.
The signal data previously stored in the multi-channel FIFO of the small computer 801 is automatically sent to an empty loop buffer section area in each sub-function device 803, 804, 805 through the DMA circuit 802.

  A small computer 801 that operates at its own timing speed and each of the sub-function devices 803, 804, and 805 that execute AD / DA processing in complete synchronization with the sampling clock operate asynchronously with each other. As the entire simultaneous vibration test control apparatus, real-time control in complete synchronization with the sampling clock is achieved.

  The present invention can be widely used industrially as a vibration test control device for evaluating the earthquake resistance and durability of industrial products.

301 Probability Density Distribution Function Generation Unit 302 Target Random Time Series Signal Generation Unit 303 Target Spectrum Generation Unit 304 Frequency Series Correction Control Unit 305 Time Series Signal Decoding Unit 306 DA Conversion Unit 307 AD Conversion Unit 308 Vibration Exciter Device

Claims (4)

In the random vibration test control device (300) that performs a vibration test of the test body by giving a random vibration signal to the vibration shaker device (308),
Response time series generating means (307) for generating a response time series signal Rt by converting a response signal obtained from a sensor for detecting the acceleration of the test body into a digital time series signal by AD conversion;
Probability density distribution function generating means (301) for obtaining a non-normal distribution probability density distribution function Dist (x) that satisfies the conditions of kurtosis (kurtosis), dispersion and PSD spectrum given as a test target;
Target random time series signal generating means (302) for generating a target random time series signal Tg [t] of non-normal distribution from the probability density distribution function Dist (x),
Target spectrum generating means (303) for converting the target random time-series signal Tg [t] into a target spectrum Tf that is a frequency-sequence signal to be a control target in a frequency space;
Frequency series correction that calculates a transfer function and an inverse transfer function from the response time series signal Rt and the drive time series signal Dt, corrects the amplitude and phase in the frequency space of the target spectrum Tf, and generates a drive Fourier spectrum Dr Control means (304);
Time-series signal decoding means (305) for generating a smooth drive time-series signal Dt from the drive Fourier spectrum Dr by an inverse fast Fourier transform and overlap synthesis method;
With
A vibration test control device that performs a control in which the target random time series signal Tg [t] and the response time series signal Rt match to realize a vibration test using a non-normally distributed random signal. Probability density distribution function generation means (301)
Calculating an rms value from a given PSD pattern and obtaining an average PSD from the rms value (R01);
Generating a normally distributed time series signal F (t) satisfying a flat spectrum of the average PSD (R02);
A step (R04) of determining the variance σ ² and the kurtosis Kw by the step (R03) of obtaining the variance σ ² in the normal distribution time series signal F (t);
The vibration test control device according to claim 1, further comprising a step (R05) of obtaining a probability density distribution function Dist (x) from the variance σ ² and the kurtosis Kw. Probability density distribution function generation means (301)
A step (D01) of setting an initial normal distribution function g ₀ (x) having a specified standard deviation σ;
Further, a step (D02) for newly defining a standard deviation σw and a standard deviation σs to narrow the initial standard deviation σ, and
Generate a normal distribution function (Gaussian distribution) with standard deviation σw and standard deviation σs, add three normal distribution functions (Gaussian distribution) with standard deviation σ, standard deviation σw, and standard deviation σs, and calculate the average Process (D03) to be gi (x),
Calculating the kurtosis Kg of gi (x) while simultaneously changing the standard deviation σw and the standard deviation σs (D04);
A step (D06) of obtaining a non-normal distribution function gi (x) = g (x) of kurtosis Kw by repeating the calculation until the kurtosis Kg of gi (x) reaches a predetermined kurtosis Kw (D02 to D05) When,
a step of calculating a standard deviation σg of g (x) (D07),
a step of creating g (x) obtained by laterally deforming the distribution function by σ / σg (D08),
Dividing g (x) by the maximum value Max _g of _g (x) to obtain a probability density distribution function Dist (x) of a non-normal distribution with a maximum value of 1.0 and a kurtosis of Kw (D09) and The vibration test control device according to claim 1, further comprising a vibration test control device. The target random time series signal generating means (302)
As shown in Equation (7) from X_Min to X_Max by Rand function, white random number RandX is generated and probability density Probab in RandX is obtained from probability density distribution function Dist (x), and at the same time, white random number RandC from 0.0 to 1.0 And generating
(Formula 7)
X_Min = −βσ
X_Max = + βσ
However, β is a value determined from the maximum excitation force of the vibration exciter device (308).

If RandC value <Probab, RandX is used as the output value. Otherwise, discard processing is repeated, and the adopted RandX is stored in FIFO (first-in first-out) memory, resulting in non-normal distribution probability density. A step (R06) in which a random time series signal Rn (t) having a distribution function Dist (x) is generated;
The generated random time series signal Rn (t) is subjected to an FFT transform process to obtain a Fourier spectrum fft (i) and a power spectrum density PSD (i) (R07),
Since PSD (i) is a flat distribution pattern with minute fluctuations around a certain value, in order to obtain a stable flat distribution, PSD (i) is exponentially averaged to obtain Ave PSD (i) (R08),
In order to match the PSD pattern to the target PSD pattern, Factor (i) = Target PSD (i) / Ave PSD (i) is calculated, and Factor (i) is multiplied by Fourier spectrum fft (i) (R09),
A process of generating a target random time series signal Tg [t] having an intended Kw, σ and PSD pattern by performing an inverse FFT transform on the Fourier spectrum fft (i) multiplied by Factor (i) (R10) The vibration test control device according to claim 1, 2 or 3, further comprising:

Images