JP2007057460A - Device and method for analyzing vibration/sound pressure propagation characteristic - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To calculate a vibration propagation characteristic between a vibration source and a response point and a sound pressure propagation characteristic between a sound source and the response point. <P>SOLUTION: A vibration detection signal is output from an acceleration sensor 52(i) to a vibration normalizing section 96a via an FFT 54(i) and a matrix forming means 56, and is normalized by the vibration normalizing section 96a. While, a sound pressure detection signal is output from a sound pressure detecting means 90(k) to a sound pressure normalizing section 96b via an FFT 92(k) and a matrix forming means 94, and is normalized by the sound pressure normalizing section 96b. The normalized vibration detection signal and sound pressure detection signal are output to a singular value decomposition section 66 via a matrix forming section 57, and is decomposed with a singular value in the singular value decomposition section 66. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a vibration / sound pressure transmission characteristic analyzing apparatus and method for analyzing a vibration transmission characteristic between a vibration source and a response point and a sound pressure transmission characteristic between a sound source and the response point.

  Conventionally, when a vibration generated from a vibration source such as an engine in a vehicle and propagating through a chassis, a body, etc. changes to sound (hereinafter also referred to as solid-propagating sound) and propagates as noise in the vehicle interior, the vehicle A noise control device for reducing indoor noise is disclosed in Patent Document 1.

  In the noise control device disclosed in Patent Document 1, the noise between the vibration source and the passenger compartment is related to the noise detected in the passenger compartment and the acceleration detected in the vicinity of the vibration source. Noise reduction control in the passenger compartment is performed by calculating a transfer characteristic and outputting a control sound that cancels the noise from a control sound source based on the calculated transfer characteristic.

JP-A-8-166788 (paragraphs [0005], [0057] to [0063], [0067] to [0081], see FIG. 1)

  As described above, in Patent Document 1, transfer characteristics relating to vehicle interior noise (hereinafter also referred to as vehicle interior noise) are calculated, but the vehicle interior sound during acceleration passes through the chassis, body, etc., and enters the vehicle interior. In addition to the propagating solid-propagating sound, so-called air-propagating sound in which engine-generated sound using the engine as a sound source and sound using the intake system, exhaust system, etc. as a sound source propagates in the air and propagates to the passenger compartment. Many are included. Therefore, in actuality, a transmission characteristic related to the solid propagation sound (hereinafter also referred to as vibration transmission characteristic) and a transmission characteristic related to the air propagation sound (hereinafter also referred to as sound pressure transmission characteristic) are calculated and calculated. It is necessary to perform noise reduction control based on the vibration transmission characteristic and the sound pressure transmission characteristic.

  However, when trying to calculate the vibration transmission characteristics between the vibration source and the passenger compartment and the sound pressure transmission characteristics between the sound source and the passenger compartment, As a result, crosstalk between the vibration transfer characteristic and the sound pressure transfer characteristic and noise are superimposed, so that each transfer characteristic cannot be calculated accurately. There is a problem that it cannot be provided.

  In addition, it is assumed that if each of the transmission characteristics related to the in-vehicle sound is obtained and the vibration transmission characteristics and the sound pressure transmission characteristics are separated from the transmission characteristics, it is possible to accurately calculate the transmission characteristics. Since the transmission path of the solid propagation sound and the air propagation sound are in a relationship of mutual coupling, it is difficult to separate these transmission characteristics.

Furthermore, since the vehicle interior sound is a synthesized sound of a solid propagation sound and an air propagation sound, in order to analyze the contribution of the solid propagation sound and the air propagation sound to the vehicle interior sound, the solid propagation sound and the air propagation sound are analyzed. It is necessary to measure the air propagation sound at the same time and calculate the vibration transmission characteristic and the sound pressure transmission characteristic based on the measurement result. However, as described above, it is difficult to calculate these transfer characteristics with high accuracy, and the unit of vibration as a solid propagation sound is [m / s ² ]. Since the unit of sound pressure is [N / m ² ], even if these propagation sounds can be measured, the conventional technology cannot separate the components of two propagation sounds having different units. .

  The present invention has been made to solve the above-described problem, and calculates a vibration transmission characteristic between a vibration source and a response point and a sound pressure transmission characteristic between a sound source and the response point, respectively. It is an object of the present invention to provide a vibration / sound pressure transmission characteristic analysis apparatus and method that can perform the above-mentioned.

  The vibration / sound pressure transmission characteristic analyzing apparatus according to the present invention detects a vibration from a vibration source and outputs a vibration detection signal, and detects a sound pressure from a sound source and outputs a sound pressure detection signal. Sound pressure detection means, vibration / sound pressure detection means for detecting vibration or sound pressure at a predetermined response point with respect to the vibration source and the sound source and outputting a vibration / sound pressure detection signal, the vibration detection signal, the sound A frequency analysis means for frequency-analyzing the pressure detection signal and the vibration / sound pressure detection signal, and the vibration source based on the vibration detection signal, the sound pressure detection signal and the vibration / sound pressure detection signal subjected to frequency analysis; Principal component regression analysis means for calculating vibration transfer characteristics between the response points and sound pressure transfer characteristics between the sound source and the response points, and the principal component regression analysis means performs frequency analysis. Normalize the vibration detection signal A vibration normalization unit, a sound pressure normalization unit that normalizes the sound pressure detection signal subjected to frequency analysis, and a singular value decomposition unit that decomposes the normalized vibration detection signal and the sound pressure detection signal into singular values And the vibration source, the sound source, and the response point by performing regression analysis between the vibration detection signal and the sound pressure detection signal subjected to singular value decomposition and the vibration / sound pressure detection signal subjected to frequency analysis. A regression analysis unit that calculates vibration / sound pressure transfer characteristics between the vibration and sound pressure transfer characteristics, and the vibration / sound pressure transfer characteristics that are calculated by the vibration transfer characteristics that are the vibration components and the sound pressure transfer characteristics that are the sound pressure components A transfer characteristic distributing unit that distributes the sound, a vibration transfer characteristic restoring unit that restores the distributed vibration transfer characteristic, and a sound pressure transfer characteristic restoring unit that restores the distributed sound pressure transfer characteristic. To do.

  The vibration / sound pressure transfer characteristic analysis method according to the present invention includes a step of detecting vibration from a vibration source by a vibration detection unit and outputting a vibration detection signal, and a sound pressure detection unit detecting a sound pressure from a sound source. Outputting a sound pressure detection signal, and detecting a vibration or sound pressure at a predetermined response point with respect to the vibration source and the sound source by a vibration / sound pressure detection means and outputting a vibration / sound pressure detection signal; Analyzing the frequency of the vibration detection signal, the sound pressure detection signal, and the vibration / sound pressure detection signal by frequency analysis means; and analyzing the frequency of the vibration detection signal, the sound pressure detection signal, and the vibration / sound pressure detection. Calculating a vibration transmission characteristic between the vibration source and the response point and a sound pressure transmission characteristic between the sound source and the response point by a principal component regression analysis unit based on a signal; The principal component regression analysis means normalizes the vibration detection signal and the sound pressure detection signal subjected to frequency analysis, decomposes the normalized vibration detection signal and the sound pressure detection signal by a singular value, Vibration between the vibration source and the sound source and the response point is performed by performing regression analysis between the vibration detection signal and the sound pressure detection signal subjected to the value decomposition and the vibration / sound pressure detection signal subjected to frequency analysis. Sound pressure transmission characteristics are calculated, and the calculated vibration / sound pressure transmission characteristics are distributed to the vibration transmission characteristics that are the vibration components and the sound pressure transmission characteristics that are the sound pressure components. The vibration transmission characteristic and the sound pressure transmission characteristic are restored respectively.

  According to the above configuration, the vibration detection signal subjected to frequency analysis is normalized in the vibration normalization unit, and the sound pressure detection signal subjected to frequency analysis is normalized in the sound pressure normalization unit. Different vibration detection signals and sound pressure detection signals can be handled in the same way. As a result, the vibration from the vibration source and the sound pressure from the sound source are simultaneously measured, and the principal component analysis is simultaneously performed on the vibration detection signal and the sound pressure detection signal in the principal component regression analysis means. It becomes possible.

  The normalized vibration detection signal and the sound pressure detection signal are subjected to singular value decomposition in the singular value decomposition unit, and the vibration subjected to frequency analysis with the vibration detection signal and the sound pressure detection signal subjected to singular value decomposition. The vibration / sound pressure transfer characteristic is calculated by performing a regression analysis on the sound pressure detection signal in the regression analysis unit, and the calculated vibration / sound pressure transfer characteristic is converted into the vibration in the transfer characteristic distribution unit. Since it is divided into the transmission characteristic and the sound pressure transmission characteristic, it is possible to treat the solid propagation sound due to the vibration and the air propagation sound due to the sound pressure in the same way, and the vibration transmission characteristic and the sound pressure transmission characteristic Can be calculated respectively. Furthermore, it is possible to clearly grasp the magnitude of the contribution of the solid propagation sound and the air propagation sound to the sound detected at the response point.

  Furthermore, since the principal component regression analysis means includes the vibration transfer characteristic restoration unit and the sound pressure transfer characteristic restoration unit, it is possible to restore each normalized transfer characteristic to the original transfer characteristic. Become.

  Therefore, in the present invention, compared to the prior art, each transfer characteristic can be calculated by running in a running test apparatus that can be mounted on a vehicle and can absorb a load imitating actual running, It is possible to greatly reduce the time from the measurement of vibration and the sound pressure to the calculation of each transfer characteristic, and the calculation accuracy of each transfer characteristic is greatly improved. It is possible to accurately grasp vibration characteristics and sound pressure characteristics of the sound source.

  Here, the vibration normalization unit calculates a standard deviation at a predetermined frequency among the vibration detection signals subjected to frequency analysis, normalizes the vibration detection signal at the predetermined frequency using the calculated standard deviation, and The sound pressure normalization unit calculates a standard deviation at a predetermined frequency among the sound pressure detection signals subjected to frequency analysis, and normalizes the sound pressure detection signal at the predetermined frequency using the calculated standard deviation. The detection signals are normalized using the standard deviations, and the principal component analysis is performed on the normalized detection signals so that the vibrations and the sound pressures are equally handled. Can do.

  In this case, the vibration transfer characteristic restoring unit restores the vibration transfer characteristic using the standard deviation calculated by the vibration normalizing unit, and the sound pressure transfer characteristic restoring unit is calculated by the sound pressure normalizing unit. The sound pressure transfer characteristic is restored using the standard deviation. This makes it possible to restore the transfer characteristics with high accuracy and output them to the outside.

  The principal component regression analysis unit synthesizes the normalized vibration detection signal and the sound pressure detection signal, and outputs the combined vibration detection signal and the sound pressure detection signal to the singular value decomposition unit. It further has a vibration / sound pressure synthesis unit. Thereby, the synthesized vibration detection signal and sound pressure detection signal can be efficiently decomposed into singular values by the singular value decomposition unit.

  Further, the principal component regression analysis means removes a noise component included in the vibration detection signal and the sound pressure detection signal subjected to singular value decomposition, and the vibration detection signal and the sound pressure detection from which the noise component has been removed. A noise removing unit for outputting a signal to the regression analysis unit; As a result, it is possible to improve both the calculation accuracy of the vibration / sound pressure transmission characteristics in the regression analysis unit and shorten the calculation time.

  According to the present invention, the vibration detection signal having different units is normalized by normalizing the vibration detection signal subjected to frequency analysis in the vibration normalization unit and normalizing the sound pressure detection signal subjected to frequency analysis in the sound pressure normalization unit. And the sound pressure detection signal can be handled in the same way. As a result, it is possible to simultaneously measure the vibration from the vibration source and the sound pressure from the sound source, and simultaneously perform the principal component analysis on the vibration detection signal and the sound pressure detection signal in the principal component regression analysis means. .

  In addition, the normalized vibration detection signal and the sound pressure detection signal are subjected to singular value decomposition in a singular value decomposition unit, and the vibration detection signal and the sound pressure detection signal subjected to singular value decomposition are subjected to frequency analysis. The vibration / sound pressure transfer characteristic is calculated by performing regression analysis on the pressure detection signal with the regression analysis unit, and the calculated vibration / sound pressure transfer characteristic is converted into the vibration transfer characteristic and the sound pressure transfer by the transfer characteristic distribution unit. Therefore, it is possible to treat the solid propagation sound due to the vibration and the air propagation sound due to the sound pressure in the same way, and the vibration transmission characteristic and the sound pressure transmission characteristic can be calculated respectively. . Furthermore, it is possible to clearly grasp the magnitude of the contribution of the solid propagation sound and the air propagation sound to the sound detected at the response point.

  Furthermore, since the principal component regression analysis means includes the vibration transfer characteristic restoration unit and the sound pressure transfer characteristic restoration unit, it is possible to restore each normalized transfer characteristic to the original transfer characteristic. Become.

  Therefore, in the present invention, compared to the prior art, each transfer characteristic can be calculated by running in a running test apparatus that can be mounted on a vehicle and can absorb a load imitating actual running, It is possible to greatly reduce the time from the measurement of vibration and the sound pressure to the calculation of each transfer characteristic, and the calculation accuracy of each transfer characteristic is greatly improved. It is possible to accurately grasp vibration characteristics and sound pressure characteristics of the sound source.

  The vibration / sound pressure transmission characteristic analysis apparatus according to the present invention will be described in detail below with reference to the accompanying drawings by giving a preferred embodiment applied to a vehicle in relation to the analysis method.

  In this embodiment, (A) 1st Example: Calculation method of vibration transfer function in vehicle based on excitation experiment with hammer (FIGS. 1 to 10), (B) 2nd Example: Calculation of (A) Vibration / sound pressure transmission characteristic analysis apparatus 50A using the method and its method (FIGS. 11-14C), (C) Third embodiment: vibration / sound that can handle solid propagation sound and air propagation sound equally The pressure transfer characteristic analyzer 50B and the method thereof (FIGS. 15 to 28) will be described in detail below.

  Here, prior to the description of the first example, solid propagation sound and air propagation sound in the vehicle to be analyzed in this embodiment (first to third examples) will be described with reference to FIG. .

  When there is an occupant 12 in the passenger compartment 10 shown in FIG. 1 and the vehicle is running (operating), the vibration source of the solid propagation sound is, for example, the engine 14 or the exhaust system 20. The vibration of the exhaust system 20 propagates through the engine mount 16 and the body 18 and changes to a solid-propagating sound in the passenger compartment 10, while the vibration of the exhaust system 20 propagates through the muffler mount 22 and the body 18 in the passenger compartment 10. Changes to solid-borne sound. The sound source of the air propagation sound is, for example, the exhaust port of the engine 14 or the exhaust system 20, and the sound of the intake port of the engine 14 or the radiated sound of the engine 14 is propagated into the vehicle compartment 10 as the air propagation sound, On the other hand, the exhaust sound at the exhaust port propagates into the passenger compartment 10 as air propagation sound.

  About these solid propagation sound and air propagation sound which propagated in the vehicle interior 10, it is necessary to provide as the optimal acoustic environment with respect to the passenger | crew 12. FIG.

  Therefore, in the present embodiment, a transmission path from the plurality of vibration sources to the interior of the passenger compartment 10 and a transmission path from the plurality of sound sources to the interior of the passenger compartment 10 from the plurality of sound sources. Separately, the part where the sound quality of the vehicle interior sound is lowered and the respective transmission paths in the vehicle interior 10 are quantitatively grasped, and the sound quality inhibiting factor of the vehicle interior sound is quantitatively analyzed.

  That is, in this embodiment, as shown in FIG. 2, basically, vibration from a vibration source (not shown) is detected by a vibration detection means, and vibration detection signals A1 to An are output. The sound pressure is calculated by multiplying the transfer function between the response point to the source (in FIG. 1, the positions of both ears of the occupant 12) and the vibration detection signals A1 to An, and the calculated sound pressures are calculated. The synthesized sound (solid propagation sound) P related to the vibration source, which is the synthesized sound pressure, is calculated. Further, sound pressure from a sound source (not shown) is detected by sound pressure detecting means and sound pressure detection signals A1 to An are output, a transfer function between the sound source and the response point for the sound source, and the sound pressure detection The sound pressure is calculated by multiplying the signals A1 to An, and the calculated sound pressures are synthesized to calculate the synthesized sound (air propagation sound) P related to the sound source as the synthesized sound pressure.

  In this transmission path analysis, the contribution of the air propagation sound and the solid propagation sound to the vehicle interior sound felt by the occupant 12 in the passenger compartment 10 is analyzed for each transmission path.

Here, when analyzing the transmission path of the in-vehicle sound by the solid propagation sound, the total number of vibration sources is n, the vibration detection signal is Ai (i = 1 to n), and the transfer function of each solid propagation sound (vibration detection signal Ai). Is Hi (i = 1 to n) and the synthesized sound pressure is Pout, the synthesized sound pressure Pout is expressed by the following equation (1).

    Pout = A1H1 + A2H2 + A3H3 + ... + AnHn (1)

  Since the solid propagation sound changes according to the frequency f, when performing the transmission path analysis, the equation (1) is constructed for each predetermined frequency.

  The above is description about the solid propagation sound and air propagation sound in the vehicle used as the analysis object of this embodiment (1st-3rd Example).

(A) 1st Example: Calculation method of vibration transfer function in vehicle based on vibration experiment with hammer (FIGS. 1 to 10)
Next, as a premise technique for analyzing the transmission path of in-vehicle sound in the passenger compartment 10 in the first embodiment, the vibration transfer function between the vibration source and the response point is represented by the inverse matrix [A] ⁻¹ of the vibration detection signal. A method of using the calculation will be described with reference to FIGS.

  3A is a cross-sectional view showing an outline of a vibration test in which the frames 30 and 32 extending from the trimmed body 23 are vibrated a predetermined number of times by the hammer 38, and FIG. 3B is obtained by the vibration test in FIG. 3A. 4A is a graph showing the frequency characteristics of the vibration transfer function. FIG. 4A is a cross-sectional view showing an outline of a vibration test in which the frames 30 and 32 are continuously vibrated with a vibrator for a predetermined time, and FIG. It is the block diagram explaining the procedure for comparing the measured sound and the synthetic sound in the trimmed body 23 based on the result of the vibration test of FIG. 3A, FIG. 3B, and FIG. 4A.

  Here, the trimmed body 23 is obtained by removing the engine, interior parts, and the like from the vehicle and configuring the vehicle compartment 10 (see FIG. 1) with the chassis and the body. A dummy head 24 schematically representing the upper half of the twelve body is disposed, and microphones (vibration / sound pressure detecting means) 26 and 28 are disposed at both ears of the dummy head 24. Therefore, in FIG. 3A, the arrangement positions of the microphones 26 and 28 are the response points. The side on which the microphone 26 is disposed is the right ear side of the dummy head 24, and the side on which the microphone 28 is disposed is the left ear side of the dummy head 24.

  On the side of the trimmed body 23, two frames 30, 32 extend in parallel toward the outside. These frames 30 and 32 are side frames of the vehicle, and acceleration sensors (vibration detecting means) 34 and 36 are arranged in the vicinity of the vibration location of the hammer 38 in the frames 30 and 32.

  Here, when the vicinity of the acceleration sensors 34, 36 of the frames 30, 32 is hit with a hammer 38, the acceleration sensors 34, 36 detect vibrations generated in the frames 30, 32 due to striking (vibration) by the hammer 38. The vibration detection signal Ai (i = 1 to n) is output. In this case, the acceleration sensors 34 and 36 can detect acceleration components (X component, Y component, and Z component) in the three-dimensional direction of the vibration.

  Accordingly, when the frames 30 and 32 are hit with the hammer 38 from the X direction, the Y direction, or the Z direction, the acceleration sensors 34 and 36 have a total acceleration component of any one of the X direction, the Y direction, and the Z direction. Six vibration detection signals Ai (i = 1 to 6) can be output. Further, the vibration is transmitted to the inside of the outer walls of the frames 30 and 32 and the trimmed body 23 and is changed to solid propagation sound. As a result, the microphones 26 and 28 detect sound pressure of the solid propagation sound and detect sound. A pressure detection signal is output.

  In the vibration test shown in FIG. 3A, as described above, the frames 30 and 32 are struck 6 times in total by the hammer 38 from the X direction, the Y direction, and the Z direction, and are caused by vibrations generated by these vibrations. The solid propagation sound is detected by the microphones 26 and 28. In this case, six inputs (vibration detection signal Ai due to vibration in the X direction, Y direction and Z direction of the frame 30, and vibration detection signal Ai due to vibration in the X direction, Y direction and Z direction of the frame 32) and two outputs ( From the sound pressure signals from the microphones 26 and 28, twelve vibration transfer functions shown in FIG. 3B are calculated.

  Then, in order to compare the actually measured sound pressure detected by the microphones 26 and 28 in the trimmed body 23 with the synthesized sound pressure Pout calculated using the vibration transfer functions, as shown in FIG. A vibration test for continuously vibrating the frames 30 and 32 using the vibrator 40 is performed.

  This vibration test simulates the generation of noise in the passenger compartment 10 (see FIG. 1) while the vehicle is running. The vibration exciter 40 applies vibration to the frames 30 and 32 only in the Z direction. However, in the frames 30 and 32, the vibrations in the X direction and the Y direction are slightly vibrated by the vibration in the Z direction. As a result, the vibrations in the three-dimensional direction are generated in the frames 30 and 32. It occurs at the same time.

  In the vibration test shown in FIG. 4A, six vibration detection signals Ai (vibration detection signals having an X direction component, a Y direction component, or a Z direction component, and an X direction component and a Y direction component detected by the acceleration sensors 34 and 36 are used. Or a vibration detection signal having a Z-direction component) and the twelve transfer functions obtained from the vibration tests of FIGS. 3A and 3B, respectively, to calculate the sound pressure, and calculate each sound pressure by (1 ) Based on the equation (1) to calculate the synthesized sound pressure Pout, and the calculated synthesized sound pressure Pout is compared with the sound pressure of the actually measured sound detected by the microphones 26 and 28 in the vibration test of FIG. 4A.

  FIG. 5A is a frequency characteristic comparing the sound pressure of the actually measured sound detected by the left ear microphone 28 and the synthesized sound pressure Pout (synthesized sound), and FIG. 5B is detected by the right ear microphone 26. This is a frequency characteristic comparing the sound pressure of the measured sound and the synthesized sound pressure Pout (synthesized sound).

  In this case, it is understood that the synthesized sound is larger than the actually measured sound in the frequency band (0 [Hz] to 2000 [Hz]) targeted for each excitation test. As shown in FIG. 6, this is because vibrations from two vibration sources are detected by the acceleration sensors 34 and 36 and vibration detection signals (input 1 and input 2) are output. When detecting a solid propagation sound with the microphones 26 and 28 and outputting a sound pressure signal (response), a solid propagation sound transfer function H1 corresponding to the input 1 and a solid propagation sound transfer function H2 corresponding to the input 2 As a result, the sound pressure detected by the microphones 26 and 28 and the synthesized sound pressure Pout described above cannot be calculated due to the crosstalk with the noise or the noise superimposed on the input 1 or the input 2. This is because they do not match.

  In other words, (input 1) × H1 + (input 2) × H2 ≠ (response), and even if the synthesized sound pressure Pout is calculated based on the equation (1), the reproducibility with respect to the actually measured sound detected by the microphones 26 and 28 is high. Low. That is, even if a plurality of vibration detection signals Ai are detected, it is difficult for all transfer functions Hi to maintain high coherence due to the influence of the above-described crosstalk and noise. In order to improve reproducibility, it is necessary to eliminate the measurement error of each transfer function Hi by eliminating the influence of the crosstalk and the noise.

  Here, in FIG. 7A and FIG. 7B, the number of vibrations of n vibration sources (not shown) is m, vibrations from the vibration sources are detected by the acceleration sensors, and vibration detection signals Aij (i = 1 to 1). n, j = 1 to m), and the sum of outputs obtained by multiplying the vibration detection signal Aij and the vibration transfer function H between each acceleration sensor and the response point is defined as a synthesized sound pressure Pout.

  In this case, if the matrix of the synthesized sound pressure Poutj is a sound pressure matrix [Pout], the matrix of the vibration detection signal Aij (vibration detection signal matrix) [A] and the matrix of the transfer function H (transfer function matrix) [H] From the above, it is expressed by the following equation (2).

      [Pout] = [A] [H] (2)

  Here, the sound pressure matrix [Pout] is represented by the following equation (3), the vibration detection signal matrix [A] is represented by the following equation (4), and the transfer function matrix [H] is represented by the following ( 5) It is expressed by the formula.

  Here, Aij (i = 1 to n, j = 1 to m) is a component of the vibration detection signal matrix [A]. As shown in FIGS. 7A and 7B, the jth acceleration sensor is the i-th acceleration sensor. It is a vibration detection signal that is output when vibration is detected. Further, each time each vibration source is vibrated, vibration is detected by an acceleration sensor corresponding to each vibration source and a vibration detection signal Aij is output. Therefore, the number of vibrations of each vibration source and each acceleration sensor This is consistent with the number of measurements.

  Then, the transfer function matrix [H] is derived from the following equation (6) from the equation (2).

[H] = [A] ⁻¹ [Pout] (6)

Here, [A] ⁻¹ is an inverse matrix of the vibration detection signal matrix [A].

  As described above, the synthesized sound pressure Pout and the sound pressure of the actually measured sound detected by the microphones 26 and 28 vary depending on the frequency. Therefore, as shown in FIG. ) To calculate the vibration transfer function Hi.

  8A shows a sound pressure (measured sound) detected by the microphone 28 on the left ear side and a synthesized sound pressure Pout (synthesized sound) calculated based on the vibration transfer function Hi derived based on the equation (6). 8B is a frequency characteristic in which the sound pressure (measured sound) detected by the microphone 26 on the right ear side and the synthesized sound pressure Pout (synthesized sound) are compared.

In this case, in the frequency range of 0 [Hz] to 1500 [Hz], the actually measured sound and the synthesized sound substantially coincide with each other, and vibration is generated using an inverse matrix [A] ⁻¹ of the vibration detection signal matrix [A]. By calculating the transfer function Hi, the influence of crosstalk is eliminated from the vibration transfer function Hi. However, in the frequency region exceeding 1500 [Hz], the synthesized sound is larger than the actually measured sound. This is because the vibration detection signal Aij output from each acceleration sensor includes an error component due to the measurement noise of each angular velocity sensor, and as a result, the calculation accuracy of the vibration transfer function Hi is greatly reduced. . As described above, also in the method of calculating the vibration transfer function Hi using the inverse matrix [A] ⁻¹ , components of crosstalk and measurement noise are removed from the vibration transfer function Hi in a high frequency region of 1500 [Hz] or higher. There is an inconvenience that cannot be done.

The above is the description of the method for calculating the vibration transfer function using the inverse matrix [A] ⁻¹ of the vibration detection signal, which is the prerequisite technology of the first embodiment.

  Next, a method for calculating the vibration transfer function according to the first embodiment will be described with reference to FIGS.

  In this calculation method, as shown in FIGS. 9A and 9B, the vibration transfer function Hi is calculated by calculating the vibration transfer function Hi from the vibration detection signal Aj detected by each acceleration sensor based on the principal component regression method. Thus, the influence of measurement noise and crosstalk components (hereinafter also referred to as noise components) of the acceleration sensor is eliminated.

  That is, there are n vibration sources and acceleration sensors. As shown in FIG. 9A, vibrations generated from the vibration sources are detected by the acceleration sensors and vibration detection signals Aj (j = 1 to m) are output. In this case, the principal components Tk {k = 1 to (m−1)} are calculated from the vibration detection signals Aj to make the principal components Tk uncorrelated with each other, and based on the uncorrelated principal components Tk. The transfer function matrix [H] and the synthesized sound pressure Pout at the response point are calculated.

  In this case, a principal component matrix having each principal component Tk as a component is [T], a singular value matrix as a diagonal matrix constituting the principal component matrix [T] is [S], and the principal component matrix [T] If the orthogonal matrices for determining the direction are [U] and [V], the vibration detection signal matrix [A] is expressed by the following equation (7), and the principal component matrix [T] is expressed by the following (8) ).

[A] = [T] [V] ^T = [U] [S] [V] ^T (7)

      [T] = [U] [S] (8)

Here, [V] ^T is a transposed matrix of the orthogonal matrix [V]. In the equation (7), the vibration detection signal matrix [A] is singularly decomposed into a singular value matrix [S].

The orthogonal matrix [U], singular value matrix [S], and transposed matrix [V] ^T are expressed by the following equations (9) to (11).

Here, Uji (i = 1 to n, j = 1 to m) is a component of the orthogonal matrix [U], and Vij is a component of the transposed matrix [V] ^T.

  Sji is a component of the singular value matrix [S] and is called a singular value. In this case, the value of the singular value Sji of the singular value matrix [S] decreases in the order of S11, S22,..., Smn, and these singular values are sequentially converted into the first principal component, the second principal component,. (N) Also referred to as a main component. Further, since the singular value matrix [S] is an eigenvector (eigenvalue matrix) for the vibration detection signal matrix [A], each singular value Sji is an eigenvalue for each vibration detection signal Aij.

  Using the equations (7) and (8), the sound pressure matrix [Pout] represented by the equation (2) is transformed into the following equation (12), and the transfer function matrix [H] represented by the equation (6) is used. Is transformed into the following equation (13).

[Pout] = [A] [H] = [U] [S] [V] ^T [H]
= [T] [V] ^T [H] (12)

[H] = [A] ⁻¹ [Pout]
= [V] [S] ^-1 [U] ^T [Pout] (13)

  Here, as shown in FIG. 9B, for example, the vibration detection signals output from the two acceleration sensors are plotted when plotted on a two-dimensional plane with the magnitudes of the signals of the acceleration sensors as coordinate axes. Further, a signal component axis (first principal component axis) having the highest correlation with each vibration detection signal (circled in FIG. 9B) and a noise component axis (second principal component axis) orthogonal to the signal component axis are defined. Formed on the two-dimensional plane, the vibration detection signals are coordinate-converted from the coordinates of the sizes of the acceleration sensors to the coordinates of the first and second principal component axes. As a result, the component of the first principal component axis in each vibration detection signal is larger than the component of the second principal component axis, and therefore information on each vibration detection signal is aggregated in the first principal component axis. Thus, the information greatly contributes to the synthesized sound pressure Pout. On the other hand, the component of the second principal component axis can be removed as a noise component.

  In this way, by using the principal component analysis method to singularly decompose the vibration detection signal, the vibration detection signal originally composed of components expressed by two coordinate axes (two acceleration sensors) Singular value decomposition is performed on two principal components that are uncorrelated with each other (first principal component and second principal component), and the information of each vibration detection signal is collected on one coordinate axis (first principal component axis). By removing the component axis component as a noise component, the vibration transfer function can be calculated based on the aggregation result (singular value) of the first principal component axis.

  Accordingly, as shown in FIG. 10, even when the number of excitations at each vibration source is m and the number of acceleration sensors is n, each vibration detection signal Aij is decomposed by singular value decomposition using the principal component regression method. When the value Sij is calculated and a singular value (for example, Smn) smaller than a predetermined threshold is removed as a noise component from the calculated singular value Sji, the singular value Smn is calculated when the transfer function matrix [H] is calculated. The row and column of the singular value matrix [S] can be deleted, and the column of the orthogonal matrix [U] and the column of the orthogonal matrix [V] corresponding to the singular value Smn can be deleted. The calculation time of the matrix [H] can be shortened.

  That is, in the calculation method of the vibration transfer function according to the first embodiment, the transfer function matrix [H], in other words, the vibration transfer characteristic (function) is obtained using the vibration detection signal Aij obtained by the vibration experiment with the hammer. It is possible to calculate with high accuracy and in a short time.

(B) Second Example: Vibration / Sound Pressure Transfer Characteristic Analysis Device 50A Using the Calculation Method of (A) and its Method (FIGS. 11-14C)
FIG. 11 is a block diagram showing a configuration of a vibration / sound pressure transfer characteristic analyzer 50A for calculating the transfer function matrix [H] using the above-described principal component regression method, and FIG. It is a block diagram which shows the flow of data in the pressure transfer characteristic analyzer 50A.

  The vibration / sound pressure transfer characteristic analyzing apparatus 50A is applicable to a vehicle (not shown), detects vibration generated from a vibration source in the vehicle, and generates vibration detection signals Ai (t) (i = 1 to n). A plurality of acceleration sensors 52 (i) for output, frequency analysis means (FFT) 54 (i) for frequency analysis of the vibration detection signal Ai (t) to output the vibration detection signal Ai (f), and each vibration detection signal Matrix forming means 56 that constructs a vibration detection signal matrix [A] for each frequency based on Ai (f), and detects a sound pressure or vibration at a response point and outputs a vibration / sound pressure detection signal Pout (t). Vibration / sound pressure detection means 62, frequency analysis means (FFT) 64 for outputting vibration / sound pressure detection signal Pout (t) and vibration / sound pressure detection signal Pout (f), vibration detection signal matrix [A] and Vibration / sound pressure detection signal Pout (f The principal component regression analysis means 58 for calculating the transfer function matrix [H] by performing the principal component analysis on the transfer function, and the transfer function distribution for distributing the vibration transfer function Hi for each transfer path from the calculated transfer function matrix [H] Means 60.

  The principal component regression analysis means 58 performs singular value decomposition on the vibration detection signal matrix [A], calculates a singular value (principal component) Sji that is uncorrelated with each other, and the calculated singular value. Among the noise removal unit 68 that removes a component determined as a noise component, the vibration detection signal matrix [A] subjected to singular value decomposition and noise removal, and the vibration / sound pressure detection signal Pout (f). The regression analysis unit 70 calculates a vibration transfer characteristic between each vibration source and the response point, that is, a transfer function matrix [H], by performing regression analysis.

  11 and 12, the vibration source is the engine 14 or the exhaust system 20 as shown in FIG. 1, and each acceleration sensor 52 (i) is a predetermined part in the vicinity of the engine 14 or the exhaust system 20 inside the vehicle. Placed in position. Further, the vibration / sound pressure detection means 62 is arranged with a predetermined position in the passenger compartment 10 as a response point.

  Here, assuming that the vehicle is traveling or operating, and the number of times of vibration when the engine 14 or the exhaust system 20 vibrates for a predetermined time is m, each acceleration sensor 52 (i) Each time the system 20 is vibrated, vibration due to the vibration is detected and a vibration detection signal Ai (t) is output to the FFT 54 (i).

  More specifically, the acceleration sensor 52 (i) uses the vibration detection signal matrix [Ai (t) with the number of excitations m as a row and the number l of time series data of the vibration detection signal Ai (t) as a column. ], And the constructed vibration detection signal matrix [Ai (t)] is output to the FFT 54 (i). The vibration detection signal matrix [Ai (t)] is expressed by the following equation (14).

  Here, Aji (o) (i = 1 to n, j = 1 to m, o = 1 to 1) is a component of the vibration detection signal matrix [Ai (t)], and the i-th acceleration sensor 52 ( In i), it is the o-th time-series data among the vibration detection signals for the vibration detected in the j-th measurement.

  On the other hand, the vibration / sound pressure detecting means 62 detects the solid propagation sound caused by the vibration as the in-vehicle sound (noise) and outputs the vibration / sound pressure detection signal Pout (t) to the FFT 64.

  In this case as well, the vibration / sound pressure detection means 62 uses the vibration / sound pressure detection signal matrix [[m] as the row and the number l of time-series data of the vibration / sound pressure detection signal Pout (t) as a column. Pout (t)] is constructed, and the constructed vibration / sound pressure detection signal matrix [Pout (t)] is output to the FFT 64. The vibration / sound pressure detection signal matrix [Pout (t)] is expressed by the following equation (15).

  Here, Poutj (o) (j = 1 to m, o = 1 to 1) is a component of the vibration / sound pressure detection signal matrix [Pout (t)], and is detected by the j-th measurement. This shows the o-th time-series data among the vibration / sound pressure detection signals for the sound.

  The FFT 54 (i) performs frequency analysis on the input vibration detection signal Ai (t) {vibration detection signal matrix [Ai (t)]} and generates a matrix [Ai (f)] of the vibration detection signal Ai (f). On the other hand, the FFT 64 outputs the vibration / sound pressure detection signal Pout (t) {vibration / sound pressure detection signal matrix [Pout (t)]} to the vibration / sound pressure. The matrix [Pout (f)] of the detection signal Pout (f) is output to the regression analysis unit 70.

  In this case, the vibration detection signal matrix [Ai (f)] is expressed by the following equation (16), while the vibration / sound pressure detection signal matrix [Pout (f)] is expressed by the following equation (17). It is represented by

  Here, Aji (f) is a component of the vibration detection signal matrix [Ai (f)], and is obtained by converting the vibration detection signal Aji (t) into a component of frequency f. On the other hand, Poutj (f) is a component of the vibration / sound pressure detection signal matrix [Poutj (f)], and is obtained by converting the vibration / sound pressure detection signal Poutj (t) into a component of frequency f.

  The matrix forming means 56 constructs a vibration detection signal matrix [A] for each frequency f from the vibration detection signal matrix [Ai (f)] input from each FFT 54 (i), and constructs each vibration detection signal matrix [A ] Is output to the singular value decomposition unit 66. In this case, the vibration detection signal matrix [A] is a matrix of m rows and n columns expressed by Equation (4).

The singular value decomposition unit 66 performs singular value decomposition on the input vibration detection signal matrix [A] based on the above equation (7), and the singular value decomposed vibration detection signal matrix [A], that is, The singular value matrix [S], the orthogonal matrix [U], and the transposed matrix [V] ^T calculated by performing the singular value decomposition on the vibration detection signal matrix [A] are output to the noise removing unit 68.

  The noise removing unit 68 determines whether or not each singular value Sji constituting the singular value matrix [S] is smaller than a predetermined threshold. If there is a singular value Sji smaller than the threshold, the singular value Sji Is a noise component, the row and column related to the singular value Sji in the singular value matrix [S] are removed, and the singular value Sji in the orthogonal matrix [U] and the orthogonal matrix [V] Remove a column.

  FIG. 13 is a graph showing the relationship between the principal component j (i) and the singular value Sji. When the noise component is not superimposed on the singular value Sji (dotted line in FIG. 13), the singular value Sji decreases as the number of principal components increases, and the singular value Sji is substantially zero when the predetermined number of principal components is exceeded. Become. However, if a noise component is superimposed on the singular value Sji, the singular value Sji does not become 0 even when the predetermined number of principal components is reached, and the singular value Sji has a characteristic that a straight line is bent at the number of principal components. Become. In this case, if the value of the singular value Sji at the inflection point at which the straight line bends is defined as the threshold value, the singular value Sji is all a noise component in the main component below the threshold value.

  Therefore, the noise removing unit 68 compares each singular value Sji constituting the singular value matrix [S] with the threshold value based on the graph of FIG. 13, and if there is a singular value Sji lower than the threshold value, It is determined that the singular value Sji includes a noise component, and based on the determination result, the removal of the row and column in the singular value matrix [S] described above, and the column of the orthogonal matrix [U] and the orthogonal matrix [V] are determined. Removal.

  However, in the noise removing unit 68, the rows and columns in the singular value matrix [S] described above are actually based on the contribution rate (dist) q shown in the following equation (18) using the graph of FIG. And removal of columns in the orthogonal matrix [U] and the orthogonal matrix [V].

  Here, q is an arbitrary number of principal components, and Sqq is a singular value in the q-th principal component. In this case, as shown in FIG. 13, since the singular value Sii decreases as the number of principal components i increases, the contribution rate (dist) q decreases as the number of principal components q increases. The higher the contribution rate (dist) q is, the more the singular value Sqq is determined to be collected from the information related to the vibration detection signal Ai (f). Can be determined to be included.

  Therefore, when the contribution rate (dist) q in the q-th principal component is lower than the contribution rate corresponding to the threshold, the noise removal unit 68 determines that the singular value Sqq of the q-th principal component is composed of a noise component. Thus, the removal of the row and the column relating to the singular value Sqq in the singular value matrix [S] and the removal of the column corresponding to the singular value Sqq in the orthogonal matrix [U] and the orthogonal matrix [V] are performed.

In the regression analysis unit 70, the singular value matrix [S], the orthogonal matrix [U] and the transposed matrix [V] ^T input from the noise removal unit 68, and the vibration / sound pressure detection signal matrix [Pout ( f)] and a regression analysis based on the equation (13) is performed to calculate a transfer function matrix [H (f)] for each frequency f, and the calculated transfer function matrix [H (f)] is transferred to the transfer function. Output to means 60.

  In this case, since the regression analysis unit 70 calculates vibration transfer functions Hj (f) (j = 1 to n) having n measurement points corresponding to the number of times of vibration m, the transfer function matrix [H (f)] Is represented by the following equation (19).

  The transfer function distribution unit 60 converts the input transfer function matrix [H (f)] for each frequency f into a matrix [Hi (f) of the vibration transfer function Hi (f) of each vibration source {for each acceleration sensor 52 (i)}. f)]. In this case, the transfer function matrix [Hi (f)] is expressed by the following equation (20).

[Hi (f)] = [Hi (1) Hi (2)... Hi (r)]
(20)

  Here, Hi (g) (g = 1 to r) is a vibration transfer function at the g-th frequency f with respect to the i-th vibration source or the acceleration sensor 52 (i).

  14A is calculated from the sound pressure (measured sound) detected by the left-ear microphone 28 (see FIGS. 3A and 4A) and the vibration transfer function Hi (k) calculated based on the equation (19). FIG. 14B compares the sound pressure (measured sound) detected by the right ear microphone 26 and the synthesized sound pressure Pout (synthesized sound). Frequency characteristics.

  In this case, in the frequency range of 0 [Hz] to 2000 [Hz], the actually measured sound and the synthesized sound are substantially the same, and the influence of the noise component is excluded from the vibration transfer function Hi (k). It is understood.

  Thus, in the vibration / sound pressure transfer characteristic analyzing apparatus 50A according to the second embodiment, the principal component regression analysis unit 58 receives the vibration detection signal matrix [A] input from the matrix forming unit 56 and the FFT 64. Since the vibration transfer function matrix [H (f)] is calculated based on the principal component regression method using the vibration / sound pressure detection signal matrix [Pout (f)], the transfer function matrix related to the solid propagation sound is calculated. It is possible to eliminate the influence of the noise component from [H (f)], and as a result, the calculation accuracy of the vibration transfer function Hi can be improved.

  In this case, the singular value decomposition unit 66 of the principal component regression analysis unit 58 performs singular value decomposition on the vibration detection signal matrix [A] to calculate uncorrelated principal components (singular values Sji), and a noise removal unit 68. When the calculated value of each singular value Sji is equal to or smaller than a predetermined threshold value, it is determined that the principal component of the singular value Sji is a noise component, and the singular value matrix [S] related to the determined singular value Sji The rows and columns are deleted, and the columns of the orthogonal matrix [U] and the columns of the orthogonal matrix [V] corresponding to the singular value Smn are deleted. Thereby, the calculation time of the transfer function matrix [Ha] in the regression analysis unit 70 can be shortened.

  Further, since the noise removal unit 68 performs the above-described determination using the contribution rate (dist) q corresponding to the threshold value, the row and column deletion in the above-described singular value matrix [S], and corresponding to this It is possible to efficiently delete the columns of the orthogonal matrix [U] and the columns of the orthogonal matrix [V].

  In the vibration / sound pressure transfer characteristic analyzing apparatus 50A according to the second embodiment, the sound pressure is detected by the microphones 26 and 28 as vibration / sound pressure detection means at the response point, and the vibration / sound pressure detection signal is generated. However, instead of this configuration, it is possible to detect a vibration at the response point using a vibration sensor (not shown) and output a vibration / sound pressure detection signal.

(C) Third embodiment: Vibration / sound pressure transmission characteristic analyzer 50B and method (FIGS. 15 to 28) capable of handling solid propagation sound and air propagation sound equally.
Next, a vibration / sound pressure transmission characteristic analyzing apparatus 50B and method according to the third embodiment will be described with reference to FIGS. The same components as those of the vibration / sound pressure transmission characteristic analyzer 50A (see FIGS. 11 and 12) according to the second embodiment are denoted by the same reference numerals, and the same applies hereinafter.

  Prior to the description of the vibration / sound pressure transfer characteristic analyzing apparatus 50B and its method, problems and the like in the calculation of the transfer function related to the vibration transfer function and the air-borne sound (hereinafter also referred to as sound pressure transfer function) will be described.

  As described with reference to FIG. 1, the interior sound of the passenger compartment 10 is transmitted through the air by the solid propagation sound caused by the vibration of the engine 14 and the exhaust system 20 and the sound generated from the engine 14 and the exhaust system 20. This is a sound generated by synthesizing the air propagating sound caused by the vehicle interior 10 in the passenger compartment 10.

  The relationship between the propagation sound and the vehicle interior sound will be described in detail with reference to FIG. 15. The vibration caused by the solid propagation sound is, for example, the vibration on the engine 14 side of the engine mount 16 (see FIG. 1). The vibration of the undercarriage system of the vehicle and the vibration of the exhaust system 20 on the body 18 side, on the other hand, the air-propagating sound includes the radiated sound of the engine 14 and the sound of the intake port and the exhaust port of the exhaust system 20. There is an exhaust sound.

  In this case, the vibration on the engine 14 side in the engine mount 16 is transmitted to the body 18 side through the engine mount 16. In other words, the vibration on the engine 14 side and the vibration between the engine 14 and the body 18 are transmitted. The product of the transfer function is the vibration on the body 18 side. Then, the product of the vibration on the body 18 side and the vibration transfer function of the engine mount 16 is propagated into the passenger compartment 10 as a solid propagation sound related to the engine mount 16.

  Similarly, the product of the vibration of the undercarriage system and the vibration transfer function of the undercarriage system is propagated into the passenger compartment 10 as a solid propagation sound of the undercarriage system. A product obtained by multiplying the vibration on the body 18 side and the vibration transfer function of the exhaust system 20 is propagated into the passenger compartment 10 as a solid propagation sound of the exhaust system 20.

  Further, a product obtained by multiplying the radiated sound of the engine 14 and the sound pressure transfer function related to the radiated sound of the engine 14 is transmitted into the passenger compartment 10 as an engine transmitted sound (air propagation sound), and the intake port Is multiplied by the sound pressure transfer function related to the sound at the intake port and is transmitted through the vehicle interior 10 as a transmitted sound (air propagation sound) of the intake system. A product obtained by multiplying the sound pressure transfer function related to the exhaust sound at the exhaust port is transmitted through the exhaust system 20 (air propagation sound) and propagated into the passenger compartment 10.

  Then, in the passenger compartment 10, each of the above-mentioned solid propagation sounds and each of the air propagation sounds are synthesized, and this synthesized sound is detected as a vehicle interior sound at a response point. Therefore, when detecting the in-vehicle sound, it is necessary to accurately calculate the sound pressure transfer function related to the air propagation sound in addition to the vibration transfer function related to the solid propagation sound.

  Here, when the vibration transfer function between the predetermined location and the response point is calculated based on the vibration generated by hitting the predetermined location of the vehicle in a stopped state with a hammer, in actuality, solid propagation sound and air Since the synthesized sound with the propagation sound becomes the interior sound in the passenger compartment 10, in order to analyze the contribution of the solid propagation sound and the air propagation sound to the interior sound, the solid propagation sound, the air propagation sound, Are simultaneously measured, and a vibration transfer function based on the solid propagation sound and a sound pressure transfer function based on the air propagation sound need to be calculated based on the measurement result.

Further, the unit of vibration as the solid propagation sound is [m / s ² ], while the unit of sound pressure as the air propagation sound is [N / m ² ], and therefore the unit is different in the conventional technology. It is difficult to measure two propagation sounds at the same time and calculate the transfer functions for these propagation sounds.

  Further, when the vibration transfer function is calculated, it is necessary to perform many vibration experiments on the vehicle in order to increase the accuracy of the vibration transfer function. As a result, the calculation time of the vibration transfer function increases. Moreover, if the hammer's excitation force with respect to the predetermined location of the vehicle is insufficient, the S / N ratio is lowered, and the transmission loss of the solid propagation sound and the vibration is generated.

  The above explanation is a problem in the calculation of the vibration transfer function and the sound pressure transfer function.

  Next, a preliminary test performed to calculate the sound pressure transfer function will be described with reference to FIGS.

  16A and 16B, the dummy head 24 having the microphones 26 and 28 is disposed in the anechoic chamber 80 simulating the passenger compartment 10 (see FIG. 1), and the sound from the speaker 84 is transmitted to the microphones 26, 28 and 82. It is sectional drawing which shows the outline | summary of the preliminary test detected by (1) -82 (4).

  In FIG. 16A, the arrangement positions 86 of the speakers 84 are set at five places, and the microphones 82 (1) to 82 (4) are arranged in parallel with respect to these five places. In this case, in the preliminary test, sound is generated from the speaker 84 at a position shown in the figure for a predetermined time and detected by the microphones 26, 28, 82 (1) to 82 (4), and then the speaker 84 is moved to another arrangement position 86. To repeat the above-mentioned sound generation. The microphones 26 and 28 detect the sound and output a vibration / sound pressure detection signal, and the microphones 82 (1) to (4) detect the sound and output a sound pressure detection signal. Note that FIG. 16A simulates detection of a solid propagation sound, and the sound from the speaker 84 at each arrangement position 86 corresponds to the solid propagation sound from a vibration source (not shown).

  In FIG. 16B, the speaker 84 is fixed at one place, and the microphones 82 (1) to 82 (4) detect the sound generated from the speaker 84 and output a vibration / sound pressure detection signal or a sound pressure detection signal. . In FIG. 16B, the sound generated from the speaker 84 corresponds to the air propagation sound.

  FIG. 17 shows the synthesized sound based on the sound pressure transfer function calculated using the sound pressure detection signal and the vibration / sound pressure detection signal obtained from the preliminary tests, and the microphones 26 and 28 detected in FIG. 16B. FIG. 11 is a flowchart for comparing with the actually measured sound. Basically, the action of the vibration / sound pressure transfer characteristic analyzing apparatus 50A (see FIGS. 11 and 12) according to the second embodiment is expressed by the sound pressure transfer function. It is replaced with calculation.

  First, in FIG. 16A, a sound (solid propagation sound) is generated a predetermined number of times from the speaker 84 corresponding to the vibration source, and the microphones 26 and 28 output vibration / sound pressure detection signals each time the sounds are detected, The microphones 82 (1) to (4) output a sound pressure detection signal every time the sounds are detected (step S1 in FIG. 17). In this case, the speaker 84 repeatedly generates the above sound while changing the arrangement position 86.

  Next, in FIG. 16B, the microphones 26 and 28 detect the sound from the speaker 84 corresponding to the air propagation sound from the sound source and output the vibration / sound pressure detection signal, and the microphones 82 (1) to (4) The sound is detected and a sound pressure detection signal is output (step S2 in FIG. 17).

  Next, each sound pressure detection signal obtained in step S1 is subjected to frequency analysis, and for each sound pressure detection signal subjected to frequency analysis, the number of times of sound generation is set as a row, and microphones 82 (1) to 82 (4) are used. A sound pressure detection signal matrix having the number as a column is constructed (step S3).

  Next, singular value decomposition is performed on the sound pressure detection signal matrix to calculate a singular value matrix, an orthogonal matrix and a transpose matrix, and each singular value constituting the obtained singular value matrix is a predetermined value. It is determined whether or not it is equal to or less than a threshold value. If there is a singular value less than or equal to the threshold value, it is determined that the singular value is made up of a noise component, the row and column relating to the singular value in the singular value matrix are deleted, and the singular value and the transposed matrix are in the singular value matrix. The column related to the value is deleted (step S4).

  Next, regression analysis is performed using the singular value matrix, the orthogonal matrix, and the transpose matrix from which the singular values composed of the noise components have been removed (step S5), and the microphones 82 (1) to 82 (4) and the microphones are analyzed. Each sound pressure transfer function between 26 and 28 is calculated (step S6).

  Subsequently, in step S2, the sound pressure detection signals output from the microphones 82 (1) to 82 (4) are multiplied by the sound pressure transfer functions to calculate a synthesized sound pressure (step S7). The sound pressure of the vibration / sound pressure detection signal output from the microphones 26 and 28 is set as the sound pressure of the actually measured sound (step S8).

  Finally, the synthesized sound pressure calculated in step S7 is compared with the sound pressure of the actually measured sound in step S8 to evaluate whether or not each sound pressure transfer function has been calculated with high precision (step S9). .

  FIG. 18 shows frequency characteristics comparing the sound pressure (measured sound) detected by the left-ear microphone 28 and the synthesized sound pressure Pout (synthesized sound).

  In this case, in the frequency band (0 [Hz] to 2000 [Hz]) targeted for the preliminary test, the actually measured sound and the synthesized sound do not match at all, and the synthesized sound is determined based on steps S1 to S9. Sound reproducibility is not obtained at all even if calculated.

  As shown in FIG. 19A, in the case of a solid propagation sound, when the solid object 88 is hit with the hammer 38 and the vibration in the solid object 88 is detected by the acceleration sensors 34 and 36, the vibration is the solid sound. Propagates inside the object 88. In other words, no matter which part of the solid object 88 is hit with the hammer 38, the vibration, that is, the solid propagation sound propagates through a predetermined path.

  On the other hand, as shown in FIG. 19B, the sound (air propagation sound) from the speaker 84 propagates radially from the speaker 84 as a sound source, and the microphones 26, 28, The transmission path of the air propagation sound with respect to 82 (1) to 82 (4) changes. Therefore, when the position of the sound source changes, the transmission path is not uniform. Therefore, the sound pressure transfer function cannot be calculated unless the position of the speaker 84 as a sound source is fixed. In other words, the position of the sound source, the position of the sound pressure detecting means for detecting the sound pressure generated from the sound source, and the position of the vibration / sound pressure detecting means for detecting the sound pressure at the response point are fixed. Thus, it is possible to calculate the sound pressure transfer function.

  FIG. 20 shows the sound pressure detection signal and the vibration / sound pressure detection signal obtained from the preliminary test when the arrangement positions of the sound source, sound pressure detection unit and vibration / sound pressure detection unit are fixed based on the above conclusion. 5 is a flowchart for comparing the synthesized sound based on the sound pressure transfer function calculated using the sound and the actually measured sound.

  In this case, as shown in FIG. 16B, the arrangement positions of the speaker 84 as a sound source, the microphones 82 (1) to 82 (4) as sound pressure detection means, and the microphones 26 and 28 as vibration / sound pressure detection means are fixed. In this state, the microphones 82 (1) to 82 (4) detect the air propagation sound from the speaker 84 and output a sound pressure detection signal, and the microphones 26 and 28 detect the air propagation sound and vibrate. A sound pressure detection signal is output (step S10).

  Next, similarly to the processing in steps S3 to S6, a sound pressure detection signal matrix is constructed from the sound pressure detection signal (step S11), and singular value decomposition and noise component removal are performed on the sound pressure detection signal matrix. (Step S12), a regression analysis is performed using the sound pressure detection signal matrix subjected to singular value decomposition and noise removal and the vibration / sound pressure detection signal (step S13), and a sound pressure transfer function is calculated. (Step S14).

  Next, similarly to the processing in step S7, the sound pressure detection signal and the sound pressure transfer function are multiplied to calculate a synthesized sound pressure at the response point (step S15), and the sound pressure of the vibration / sound pressure detection signal is calculated. Is measured sound pressure at the response point (step S16), and the synthetic sound pressure is compared with the actually measured sound pressure to evaluate whether or not the sound pressure transfer function is accurately calculated (step S17). ).

  FIG. 21 shows frequency characteristics comparing the sound pressure (measured sound) detected by the left-ear microphone 28 and the synthesized sound pressure Pout (synthesized sound).

  In this case, in the frequency band (0 [Hz] to 2000 [Hz]) targeted for the preliminary test, the actually measured sound and the synthesized sound substantially coincide with each other, and the synthesized sound is calculated based on steps S10 to S17. This shows that the air-borne sound at the response point can be accurately reproduced.

By the way, as shown in FIG.1 and FIG.16, the vehicle interior sound in the compartment 10 is comprised from a solid propagation sound and an air propagation sound. Therefore, in order to accurately reproduce the vehicle interior sound at the response point in the passenger compartment 10, it is necessary to reproduce the vehicle interior sound using the vibration transfer function related to the solid propagation sound and the sound pressure transfer function related to the air propagation sound. There is. However, as described above, the unit of the vibration as the solid propagation sound is [m / s ² ], while the unit of the sound pressure of the air propagation sound is [N / m ² ], which is a different unit. It is difficult to handle two propagation sounds in the same way.

  Therefore, as shown in FIG. 22, the vibration detection signals A1 and A2 detected by the respective acceleration sensors (vibration detection means) and the sound pressure detection signals Pin1 and Pin2 detected by the respective sound pressure detection means are respectively normalized. With respect to the normalized vibration detection signals NA1 and NA2 and the sound pressure detection signals NPin1 and NPin2, principal components T1 to T3 are calculated based on the principal component regression method, and the synthesized sound is calculated based on the calculated principal components T1 to T3. By calculating the pressure Pout, it is possible to handle the vibration detection signals A1 and A2 and the sound pressure detection signals Pin1 and Pin2 in the same manner to calculate the vibration transfer function Hi and the sound pressure transfer function Hp.

  That is, the number of sound sources and the number of sound pressure detection means corresponding to the sound sources are k, the sound pressure detection signal matrix having a plurality of sound pressure detection signals Pin as components, [Pin], the sound pressure detection means and the response points. If the transfer function matrix having the sound pressure transfer function Hp between and Hp as a component is [Hp] and the matrix of the vibration transfer function Ha is [Ha], the matrix [Pout] of the synthesized sound pressure (in-vehicle sound pressure) Pout is Is expressed by the following equation (21).

      [Pout] = [A] [Ha] + [Pin] [Hp] (21)

  Here, the transfer function matrix [Ha], the sound pressure detection signal matrix [Pin], and the transfer function matrix [Hp] are expressed by the following equations (22) to (24).

  Then, the normalized vibration detection signal matrix [NA], the diagonal matrix (normalization matrix) for normalizing the vibration detection signal matrix [A] to [NA] [Nσa], and the normalized sound If the pressure detection signal matrix is [NPin] and the diagonal matrix (normalization matrix) for normalizing the sound pressure detection signal matrix [Pin] to [NPin] is [Nσp], the vibration detection signal matrix [A] The sound pressure detection signal matrix [Pin] is expressed by the following equations (25) and (26).

      [A] = [NA] [Nσa] (25)

      [Pin] = [NPin] [Nσp] (26)

  When the expressions (25) and (26) are substituted into the expression (21), the sound pressure matrix [Pout] is expressed by the following expression (27).

[Pout] = [NA] [Nσa] [Ha]
+ [NPin] [Nσp] [Hp]
= [NA, NPin] [Nσa, Nσp] [Ha, Hp]
(27)

  Here, [NA, NPin] is a composite matrix of the normalized vibration detection signal matrix [NA] and the normalized sound pressure detection signal matrix [NPin], and [Nσa, Nσp] is normalized The matrix [Nσa] and the normalization matrix [Nσp] are combined, and [Ha, Hp] is the combined matrix (vibration / sound pressure transfer function matrix) of the transfer function matrix [Ha] and the transfer function matrix [Hp]. ). When principal component analysis is performed on the vibration / sound pressure transfer function matrix [Ha, Hp], [Ha, Hp] is expressed by the following equation (28).

[Ha, Hp] = [Nσa, Nσp] ⁻¹ [V] [S] ⁻¹ [U] ^T
× [Pout] (28)

  Therefore, by calculating the vibration / sound pressure transfer function matrix [Ha, Hp] based on the equation (28), it is possible to calculate the vibration transfer function matrix Ha and the sound pressure transfer function matrix Hp constituting this matrix. It becomes.

  The vibration detection signal matrix [NA] and the normalization matrix [Nσa] are expressed by the following equations (29) and (30).

  Here, nΣa is the total sum of the standard deviations σai of all components constituting the i-th column of the vibration detection signal matrix [NA], NA1i to NAmi, and is expressed by the following equation (31).

  On the other hand, the sound pressure detection signal matrix [NPin] and the normalization matrix [Nσp] are expressed by the following equations (32) and (33).

  Here, kΣp is the total sum of standard deviations σpi of all components NPin1i to NPinmi constituting the i-th column of the sound pressure detection signal matrix [NPin], and is expressed by the following equation (34).

  Each component of the vibration detection signal matrix [A] indicates the amplitude of the vibration detection signal Aij, while each component of the sound pressure detection signal matrix [Pin] indicates the amplitude of the sound pressure detection signal Pin. The above normalization is normalization of the amplitudes of the vibration detection signal Aij and the sound pressure detection signal Pin.

  Further, the normalized vibration / sound pressure detection signal matrix [NA, NPin] is expressed by the following equation (35) from the equations (29) and (32).

Furthermore, in the equation (28), the singular value matrix [S] and the transposed matrix [V] ^T are expressed by the following equations (36) and (37).

  FIG. 23 shows a flowchart for calculating the synthesized sound pressure Pout when the vehicle is running or operating by applying the normalization method shown in FIG. 22 to the reproduction of the vehicle interior sound.

  In this case, each acceleration sensor 52 (i) is arranged in the vicinity of the vibration source of the vehicle (the engine 14 and the exhaust system 20 shown in FIG. 1), and the sound source (the intake port and the exhaust system 20 of the engine 14 shown in FIG. 1). The sound pressure detecting means (not shown) is disposed in the vicinity of the exhaust port), and the vibration / sound pressure detecting means 62 is disposed at the response point in the passenger compartment 10.

  Here, each sound pressure detection means detects air propagation sound from the sound source and outputs a sound pressure detection signal Pin, and each acceleration sensor 52 (i) detects vibration from a vibration source and vibrates. The detection signal Ai is output, and the vibration / sound pressure detection means 62 detects the vehicle interior sound at the response point and outputs the vibration / sound pressure detection signal Pout (step S18).

  Next, similarly to the processing of steps S3 and S11, the vibration detection signal matrix [Ai] based on the equation (4) is constructed from the vibration detection signal Ai, while the sound based on the equation (22) is constructed from the sound pressure detection signal Pin. A pressure detection signal matrix [Pin] is constructed (step S19).

  Next, the vibration detection signal matrix [Ai] and the sound pressure detection signal matrix [Pin] are normalized based on the equations (25) and (26) (step S20), and normalized as in steps S4 and S12. Singular value decomposition and noise component removal are performed on the vibration detection signal matrix [NA] and the sound pressure detection signal matrix [NPin] (step S21).

  Next, as in steps S5 and S13, the vibration detection signal matrix [NA] and the sound pressure detection signal matrix [NPin] that have undergone singular value decomposition and noise removal are between the vibration / sound pressure detection signal Pout. The regression analysis is performed (step S22), and the vibration transfer function Ha and the sound pressure transfer function Hp are calculated based on the equation (28) (step S23).

  Next, similarly to the processing of steps S7 and S15, the sound pressure is calculated by multiplying the sound pressure detection signal Pin and the sound pressure transfer function Hp based on the equation (21), while the vibration detection signal Ai and the vibration are calculated. The sound pressure is calculated by multiplying the transfer function Ha, and the synthesized sound pressure Pout is calculated by synthesizing these sound pressures (step S24).

  FIG. 24 is a block diagram showing the configuration of a vibration / sound pressure transfer characteristic analyzer 50B according to the third embodiment for calculating the vibration transfer function Ha and the sound pressure transfer function Hp using the normalization method described above. FIG. 25 is a block diagram showing a data flow in the vibration / sound pressure transfer characteristic analyzing apparatus 50B.

  This vibration / sound pressure transfer characteristic analysis device 50B detects air propagation sound from a sound source (not shown) as compared with the vibration / sound pressure transfer characteristic analysis device 50A (see FIGS. 11 and 12) according to the second embodiment. And the principal component regression analysis means 58 normalizes the vibration detection signal and the sound pressure detection signal, while restoring the transfer function calculated based on the normalized vibration detection signal and the sound pressure detection signal. It is different.

  That is, the vibration / sound pressure transfer characteristic analysis device 50B is mounted on a vehicle (not shown), detects air propagation sound generated from a sound source (not shown), and outputs a sound pressure detection signal Pk (t) (k = 1 to d). A plurality of sound pressure detection means 90 (k), a frequency analysis means (FFT) 92 (k) for frequency-analyzing the sound pressure detection signal Pk (t) and outputting a sound pressure detection signal Pk (f), A matrix forming means 94 for constructing a sound pressure detection signal matrix [P] for each frequency based on the sound pressure detection signal Pk (f), and a transfer function matrix [Ha] output from the principal component regression analysis means 58 are transferred. Transfer function distribution means 60b for distributing the vibration transfer function Hai for each transfer function, and transfer function distribution means 60c for distributing the transfer function matrix [Hp] output from the principal component regression analysis means 58 to the sound pressure transfer function Hpi for each transfer path. And further.

  The principal component regression analysis means 58 includes a vibration normalization unit 96a that normalizes the vibration detection signal matrix [A], a sound pressure normalization unit 96b that normalizes the sound pressure detection signal matrix [P], and normalization. Matrix formation unit (vibration / sound pressure synthesis unit) that synthesizes the detected vibration detection signal matrix [NA] and the sound pressure detection signal matrix [NP] into one matrix [NA, NP] and outputs it to the singular value decomposition unit 66 57 and the normalized vibration / sound pressure transfer function [NHa, NHp] output from the regression analysis unit 70 into a normalized vibration transfer function [NHa] and a normalized sound pressure transfer function [NHp]. The transfer function distributing unit 60a for distributing, the vibration restoring unit 98a for restoring the normalized vibration transfer function [NHa] to the original vibration transfer function [Ha], and the normalized sound pressure transfer function [NHp] Sound pressure restoration unit 9 for restoring the sound pressure transfer function [Hp] of Further comprising a and b.

  24 and 25, when the engine 14 or the exhaust system 20 as a vibration source vibrates for a predetermined time during running or operation of the vehicle and the number of times of analysis is m, each acceleration sensor 52 (i) A vibration detection signal matrix [Ai (t)] is constructed from the vibration detection signal Ai (t), and the constructed vibration detection signal matrix [Ai (t)] is output to the FFT 54 (i). The FFT 54 (i) The vibration detection signal matrix [Ai (t)] is frequency-analyzed and the vibration detection signal matrix [Ai (f)] is output to the matrix forming unit 56. The matrix forming unit 56 receives the vibration detection signal matrix [Ai (f)]. Thus, a vibration detection signal matrix [A] for each frequency f is constructed and output to the vibration normalization unit 96a.

  On the other hand, the sound pressure detection means 90 (k) detects the air propagation sound generated from the sound source every time the engine 14 or the exhaust system 20 vibrates, and in other words, the predetermined number of analysis times m, in other words, the predetermined A sound pressure detection signal matrix [Pin (t)] having the number m of air propagation sound occurrences in time as a row and the number l of time-series data of the sound pressure detection signal Pin (t) as a column is constructed. The pressure detection signal matrix [Pin (t)] is output to the FFT 92 (k). The FFT 92 (k) performs frequency analysis on the input sound pressure detection signal matrix [Pin (t)] and outputs the matrix [Pin (f)] of the sound pressure detection signal Pin (f) to the matrix forming means 94. The matrix forming means 94 constructs a sound pressure detection signal matrix [Pin] for each frequency f from the sound pressure detection signal matrix [Pin (f)] input from each FFT 92 (k), and constructs each sound pressure detection signal thus constructed. The matrix [Pin] is output to the sound pressure normalization unit 96b.

  Further, the vibration / sound pressure detecting means 62 detects the vehicle interior sound at the response point in the passenger compartment 10 and outputs the vibration / sound pressure detection signal matrix [Pout (t)] to the FFT 64. The FFT 64 is inputted. The frequency analysis of the vibration / sound pressure detection signal matrix [Pout (t)] is performed, and the vibration / sound pressure detection signal matrix [Pout (f)] is output to the regression analysis unit 70.

  Based on the input vibration detection signal matrix [A], the vibration normalization unit 96a calculates the sum nΣa of the standard deviations σai for each frequency f based on the equation (31), and the normalized matrix [Nσa ], And the vibration detection signal matrix [A] is normalized to [NA] based on the normalization matrix [Nσa] and the equation (25). Next, the vibration normalizing unit 96a outputs the normalized vibration detection signal matrix [NA] to the matrix forming unit 57 and outputs the normalized matrix [Nσa] to the vibration restoring unit 98a.

  On the other hand, the sound pressure normalization unit 96b calculates the sum kΣp of the standard deviation σpi for each frequency f based on the equation (34) from the input sound pressure detection signal matrix [Pin], and from the calculated sum total kΣp. A normalization matrix [Nσp] is constructed, and the sound pressure detection signal matrix [Pin] is normalized to [NPin] based on the normalization matrix [Nσp] and the equation (26). Next, the sound pressure normalizing unit 96b outputs the normalized sound pressure detection signal matrix [NPin] to the matrix forming unit 57 and outputs the normalized matrix [Nσp] to the sound pressure restoring unit 98b.

  The matrix formation unit 57 synthesizes the vibration detection signal matrix [NA] input from the vibration normalization unit 96a and the sound pressure detection signal matrix [NPin] input from the sound pressure normalization unit 96b for each frequency f. The vibration / sound pressure signal matrix [NA, NPin] is constructed based on the equation (35), and the constructed vibration / sound pressure signal matrix [NA, NPin] is output to the singular value decomposition unit 66. It should be noted that the vibration / sound pressure signal matrix [NA, NPin] is a matrix of m rows (n + k) columns as shown by the equation (35).

The singular value decomposition unit 66 performs singular value decomposition on the input vibration / sound pressure signal matrix [NA, NPin], and the singular value decomposed vibration / sound pressure signal matrix [NA, NPin], that is, the vibration / sound Singular value matrix [S] shown in equation (36), orthogonal matrix [U] and transposed matrix [V] ^T shown in equation (37) calculated by singular value decomposition of the sound pressure signal matrix [NA, NPin]. Are output to the noise removing unit 68.

In this case, the noise removing unit 68 removes rows and columns related to singular values determined as noise components from the singular values See (e = 1 to c, c = n + k) constituting the singular value matrix [S]. In addition, the column related to the singular value See in the orthogonal matrix [U] and the orthogonal matrix [V] is removed, and the singular value matrix [S], the orthogonal matrix [U], and the transposed matrix [V] ^T for which the removal operation is completed. Is output to the regression analysis unit 70.

  In this case, the noise removing unit 68 performs the matrix removal operation based on the cumulative contribution rate instead of the contribution rate (dist) q expressed by the equation (18). Here, the cumulative contribution rate in the q-th principal component whose singular value is Sqq is (cumulative contribution rate) = (sum of singular values See from the first principal component to the q-th principal component) / (all singular values). Sum)) × 100 [%].

  FIG. 27 shows the relationship between the number of principal components and the cumulative contribution rate at a predetermined frequency (f = 50 [Hz], 100 [Hz], 200 [Hz], 500 [Hz], 1000 [Hz]). In this case, at any frequency, the cumulative contribution rate increases rapidly in a region where the number of principal components is small (for example, the number of principal components: 1 to 10), but a region where the number of principal components is large (the number of principal components: 11 or more). ), The cumulative contribution rate slightly increases with an increase in the number of principal components. As described with reference to FIG. 13, in the region where the number of principal components q is small and the singular value Sqq is large, information on the vibration detection signal Ai (f) and the sound pressure detection signal Pin (f) is collected. On the other hand, the noise component is considered to be dominant in the region where the number of principal components q is large and the singular value Sqq is low.

  In addition, the straight line in FIG. 27 theoretically indicates the number of principal components and the accumulation when all data of the vibration detection signal Ai (f) and the sound pressure detection signal Pin (f) are uncorrelated with each other and have substantially the same amplitude. The relationship with the contribution rate is shown.

  FIG. 28 is a graph in which the difference between the cumulative contribution rate in the straight line and the cumulative contribution rate at the predetermined frequency (difference cumulative contribution rate) is plotted, and the differential cumulative contribution rate increases sharply as the number of principal components increases. However, when it exceeds a predetermined number of main components, it decreases linearly. In this case, a region where the difference cumulative contribution rate decreases linearly is a region where the noise component is dominant, and the difference cumulative contribution rate corresponding to the predetermined number of principal components corresponds to the threshold value.

  Therefore, the noise removal unit 68 performs the matrix removal operation based on the maximum value of the cumulative difference contribution ratio instead of the threshold value. That is, the noise removing unit 68 calculates the difference cumulative contribution rate in each principal component e from each singular value Seee in the singular value matrix [S], and selects the number of principal components that becomes the maximum value in these difference cumulative contribution rates. To do. Next, the noise removing unit 68 determines that the singular values of the principal components exceeding the selected number of the principal components are singular values composed of noise components, and is determined to be the noise components in the singular value matrix [S]. The rows and columns related to the singular values are deleted, and the columns related to the singular values in the orthogonal matrix [U] and the orthogonal matrix [V] are deleted.

In the regression analysis unit 70, the singular value matrix [S], the orthogonal matrix [U] and the transposed matrix [V] ^T input from the noise removal unit 68, and the vibration / sound pressure detection signal matrix [Pout ( f)] and performing a regression analysis based on the equation (28) to calculate the vibration / sound pressure transfer function matrix [NHa, NHp] for each frequency f, and the calculated vibration / sound pressure transfer function matrix [NHa, NHp] is output to the transfer function distributor 60a. The normalized vibration / sound pressure transfer function matrix [NHa, NHp] is a matrix obtained by normalizing the vibration / sound pressure transfer function matrix [Ha, Hp] shown in the equations (27) and (28). is there.

  The transfer function distribution unit 60a distributes the input vibration / sound pressure transfer function matrix [NHa, NHp] to the normalized transfer function matrices [NHa], [NHp], and the distributed transfer function matrix [NHa]. While outputting to the vibration restoration unit 98a, the distributed transfer function matrix [NHp] is outputted to the sound pressure restoration unit 98p.

In the vibration restoration unit 98a, based on the transfer function matrix [NHa] input from the transfer function distribution unit 60a and the inverse matrix [Nσa] ^{−1 of} the normalization matrix [Nσa] input from the vibration normalization unit 96a. The normalized transfer function matrix [NHa] is restored to the original transfer function matrix [Ha], and the restored transfer function matrix [Ha] is output to the transfer function distribution means 60b. The transfer function distribution unit 60b distributes the input transfer function matrix [Ha] to the vibration transfer function Hai for each transfer path.

On the other hand, in the sound pressure restoration unit 98b, the transfer function matrix [NHp] input from the transfer function distribution unit 60a and the inverse matrix [Nσp] ^{−1 of} the normalization matrix [Nσp] input from the sound pressure normalization unit 96b. The transfer function matrix [NHp] normalized based on the above is restored to the original transfer function matrix [Hp], and the restored transfer function matrix [Hp] is output to the transfer function distribution means 60c. The transfer function distribution unit 60c distributes the input transfer function matrix [Hp] to the vibration transfer function Hpk for each transfer path.

  As described above, in the vibration / sound pressure transfer characteristic analyzing apparatus 50B and the method according to the third embodiment, the vibration detection signal A (f) {vibration detection signal matrix [A (f)]} subjected to frequency analysis is converted into a vibration normal signal. The sound pressure detection signal Pin (f) {sound pressure detection signal matrix [Pin (f)]} normalized and frequency-analyzed by the normalizing unit 96a is normalized by the sound pressure normalizing unit 96b, so that vibrations having different units are obtained. The detection signal A (f) and the sound pressure detection signal Pin (f) can be handled in the same way. As a result, the vibration from the vibration source and the sound pressure from the sound source are simultaneously measured, and the principal component analysis is performed on the vibration detection signal A (f) and the sound pressure detection signal Pin (f) in the principal component regression analysis means 58. It can be performed simultaneously.

  Further, the normalized vibration detection signal NA (f) {vibration detection signal matrix [NA (f)]} and the sound pressure detection signal NPin (f) {sound pressure detection signal matrix [NPin (f)]} are singular values. Singular value decomposition is performed in the decomposition unit 66, and the regression analysis unit 70 performs regression using the vibration / sound pressure signal matrix [NA, NPin] subjected to singular value decomposition and the vibration / sound pressure detection signal Pout (f) subjected to frequency analysis. The vibration / sound pressure transfer function matrix [NHa, NHp] is calculated by performing the analysis, and the calculated vibration / sound pressure transfer function matrix [NHa, NHp] is normalized in the transfer function distribution unit 60a. Since it is distributed to [NHa] and [NHp], it is possible to handle the solid propagation sound due to the vibration and the air propagation sound due to the sound pressure in the same way, and the vibration transfer function Ha and the sound can be handled without performing an excitation experiment. Pressure transfer function May each calculate the hp. Furthermore, it is possible to clearly grasp the magnitude of the contribution of the solid propagation sound and the air propagation sound to the synthesized sound pressure Pout at the response point.

  Therefore, in the vibration / sound pressure transfer characteristic analyzing apparatus 50B, compared to the conventional technique, the vibration transfer function Ha and the sound pressure transfer function Hp are installed in the vehicle and are actually run or run on the simulation line of the actual run. Therefore, the time from the measurement of vibration and sound pressure to the calculation of the vibration transfer function Ha and the sound pressure transfer function Hp can be greatly reduced, and the vibration transfer function Ha and The calculation accuracy of the sound pressure transfer function Hp is greatly improved, and the vibration characteristics of the vibration source and the sound pressure characteristics of the sound source can be accurately grasped.

  Also, each component (amplitude) of the vibration detection signal matrix [A (f)] and each component (amplitude) of the sound pressure detection signal matrix [Pin (f)] using the standard deviations σai, σpi and their sums nΣa, kΣp. By normalizing, it is possible to handle the vibration and sound pressure amounts equally, and the principal component analysis in the principal component regression analysis means 58 can be easily performed.

  Further, the matrix forming unit 57 combines the normalized vibration detection signal matrix [NA (f)] and the sound pressure detection signal matrix [NPin (f)] to generate the vibration / sound pressure signal matrix [NA, NPin]. Since this is constructed and this vibration / sound pressure signal matrix [NA, NPin] is output to the singular value decomposition unit 66, the singular value decomposition unit 66 can efficiently perform the singular value decomposition.

  Furthermore, since the principal component regression analysis means 58 includes the vibration restoration unit 98a and the sound pressure restoration unit 98b, the normalized transfer function matrices [NHa] and [NHp] are converted into the original transfer function matrix [Ha], It becomes possible to restore to [Hp].

  In this case, the vibration restoration unit 98a uses the normalized matrix [Nσa] to restore the normalized transfer function matrix [NHa] to the original transfer function matrix [Ha], while the sound pressure restoration unit 98b Since the normalized transfer function matrix [NHp] is restored to the original transfer function matrix [Hp] using the normalized matrix [Nσp], each transfer function matrix [Ha], [Hp] is accurately restored. Thus, it is possible to output to the transfer function distribution means 60b and 60c.

  Of course, the vibration / sound pressure transmission characteristic analyzing apparatus and method according to the present invention are not limited to the above-described embodiments, and various configurations can be adopted without departing from the gist of the present invention.

It is sectional drawing which shows the transmission path | route of the solid propagation sound and air propagation sound in a vehicle. It is a block diagram which shows the basic processing content in this embodiment. FIG. 3A is a cross-sectional view showing an outline of a vibration test in which each frame of the trimmed body is vibrated with a hammer, and FIG. 3B shows the frequency characteristic of the transfer function obtained by the vibration test in FIG. 3A. It is a graph to show. 4A is a cross-sectional view showing an outline of a vibration test in which vibration is continuously applied to each frame of the trimmed body by a vibration exciter, and FIG. 4B is a result of the vibration test of FIGS. 3A to 4A. It is a block diagram explaining the procedure for comparing an actual measurement sound and a synthetic sound based on this. FIG. 5A is a graph of frequency characteristics comparing the measured sound detected by the left ear microphone and the synthesized sound, and FIG. 5B compares the measured sound detected by the right ear microphone and the synthesized sound. It is the graph of the frequency characteristic made. It is a block diagram explaining crosstalk. FIG. 7A is a block diagram illustrating construction of a sound pressure matrix for each frequency, and FIG. 7B is a block diagram illustrating construction of a sound pressure matrix and a transfer function matrix for each frequency. FIG. 8A is a graph of frequency characteristics comparing the measured sound detected by the left ear microphone and the synthesized sound, and FIG. 8B compares the measured sound detected by the right ear microphone and the synthesized sound. It is the graph of the frequency characteristic made. FIG. 9A is a block diagram showing that the principal component analysis is performed on the vibration detection signal using the principal component regression method, and FIG. 9B is a graph for explaining the meaning of the principal component analysis. It is a block diagram which shows performing singular value decomposition | disassembly, noise removal, and regression analysis with respect to a vibration detection signal for every frequency. It is a block diagram which shows the structure of the vibration and sound pressure transmission characteristic analyzer based on 2nd Example. FIG. 12 is a block diagram showing a data flow in the vibration / sound pressure transmission characteristic analyzer of FIG. 11. 12 is a graph showing the relationship between the number of principal components and singular values used in the noise removing unit in FIG. 11. FIG. 14A is a graph of frequency characteristics comparing the measured sound detected by the left ear microphone and the synthesized sound, and FIG. 14B compares the measured sound detected by the right ear microphone and the synthesized sound. It is the graph of the frequency characteristic made. It is a block diagram which shows the relationship between the solid propagation sound and air propagation sound of FIG. 1, and a vehicle interior sound. FIG. 16A is a cross-sectional view showing an outline of a preliminary test in which sound from a speaker arranged in an anechoic chamber is detected by a microphone, and FIG. 16B is a diagram illustrating a state in which the speaker arrangement position of FIG. It is sectional drawing which shows the outline | summary of the preliminary test which detects a sound with a microphone. FIG. 17 is a flowchart for calculating a synthesized sound based on the sound pressure detection signal and the vibration / sound pressure detection signal obtained from the preliminary test of FIGS. 16A and 16B, and comparing the calculated synthesized sound with the actually measured sound. It is. FIG. 18 is a graph of frequency characteristics comparing the measured sound detected by the left ear microphone and the synthesized sound obtained by executing the flowchart of FIG. FIG. 19A is a perspective view showing transmission of vibration (solid propagation sound) in a solid object, and FIG. 19B is a perspective view showing transmission of air propagation sound from a speaker. FIG. 16B is a flowchart for comparing the synthesized sound and the actually measured sound when a preliminary test is performed with the arrangement positions of the sound source and each microphone in FIG. 16B fixed. It is a graph of the frequency characteristic which compared the measured sound and synthetic sound which were detected with the microphone of the left ear side obtained by performing the flowchart of FIG. It is a block diagram which shows performing a principal component analysis with respect to the normalized vibration detection signal and sound pressure detection signal by normalizing a vibration detection signal and a sound pressure detection signal. It is a flowchart for performing the normalization method and principal component regression method of FIG. It is a block diagram which shows the structure of the vibration and sound pressure transmission characteristic analyzer based on 3rd Example. FIG. 25 is a block diagram showing a data flow in the vibration / sound pressure transmission characteristic analyzer of FIG. 24. FIG. 26 is a block diagram showing the flow of data inside the principal component regression analysis means of FIGS. 24 and 25. It is a graph which shows the relationship between the main component number utilized in the noise removal part of FIG. 24, and a cumulative contribution rate. It is a graph which shows the relationship with the number of main components when the difference of the cumulative contribution rate of FIG. 27 and a theoretical line is made into a difference cumulative contribution rate.

Explanation of symbols

50A, 50B ... vibration / sound pressure transmission characteristic analyzer 52 (i) ... acceleration sensors 54 (i), 64, 92 (k) ... frequency analysis means 56, 94 ... matrix formation means 57 ... matrix formation part 58 ... main component Regression analysis means 60, 60b, 60c ... transfer function distribution means 60a ... transfer function distribution section 62 ... vibration / sound pressure detection means 66 ... singular value decomposition section 68 ... noise removal section 70 ... regression analysis section 90 (k) ... sound pressure Detection means 96a ... vibration normalization unit 96b ... sound pressure normalization unit 98a ... vibration restoration unit 98b ... sound pressure restoration unit

Claims (6)

Vibration detection means for detecting vibration from a vibration source and outputting a vibration detection signal;
Sound pressure detection means for detecting sound pressure from a sound source and outputting a sound pressure detection signal;
Vibration / sound pressure detecting means for detecting vibration or sound pressure at a predetermined response point with respect to the vibration source and the sound source and outputting a vibration / sound pressure detection signal;
Frequency analysis means for analyzing the frequency of the vibration detection signal, the sound pressure detection signal and the vibration / sound pressure detection signal;
Based on the vibration detection signal, the sound pressure detection signal, and the vibration / sound pressure detection signal subjected to frequency analysis, vibration transfer characteristics between the vibration source and the response point, and between the sound source and the response point. A principal component regression analysis means for calculating the sound pressure transmission characteristics of
Have
The principal component regression analysis means is normalized by a vibration normalization unit that normalizes the frequency-analyzed vibration detection signal, and a sound pressure normalization unit that normalizes the frequency-analyzed sound pressure detection signal. A singular value decomposition unit that singularly decomposes the vibration detection signal and the sound pressure detection signal, and the vibration detection signal and the sound pressure detection signal that have been subjected to singular value decomposition and the vibration / sound pressure detection signal that has been subjected to frequency analysis. A regression analysis unit that calculates vibration / sound pressure transfer characteristics between the vibration source and the sound source and the response point by performing a regression analysis between the vibration source and the calculated vibration / sound pressure transfer characteristic A transmission characteristic distributing unit that distributes the vibration transmission characteristic that is a component and the sound pressure transmission characteristic that is a sound pressure component; a vibration transmission characteristic restoring unit that restores the distributed vibration transmission characteristic; and the distributed sound Sound pressure transmission to restore pressure transmission characteristics Vibration and sound pressure transmission characteristic analyzing apparatus characterized by having a characteristic recovery unit. In the vibration / sound pressure transmission characteristic analyzing apparatus according to claim 1,
The vibration normalization unit calculates a standard deviation at a predetermined frequency among the vibration detection signals subjected to frequency analysis, normalizes the vibration detection signal at the predetermined frequency using the calculated standard deviation,
The sound pressure normalization unit calculates a standard deviation at a predetermined frequency among the sound pressure detection signals subjected to frequency analysis, and normalizes the sound pressure detection signal at the predetermined frequency using the calculated standard deviation. Vibration / sound pressure transmission characteristic analysis device characterized by In the vibration / sound pressure transmission characteristic analyzing apparatus according to claim 2,
The vibration transfer characteristic restoration unit restores the vibration transfer characteristic using the standard deviation calculated in the vibration normalization unit,
The sound pressure transfer characteristic restoring unit restores the sound pressure transfer characteristic by using the standard deviation calculated by the sound pressure normalizing unit. In the vibration / sound pressure transmission characteristic analyzing apparatus according to any one of claims 1 to 3,
The principal component regression analysis unit synthesizes the normalized vibration detection signal and the sound pressure detection signal, and outputs the combined vibration detection signal and the sound pressure detection signal to the singular value decomposition unit. A vibration / sound pressure transmission characteristic analyzer further comprising a sound pressure synthesis unit. In the vibration / sound pressure transmission characteristic analyzing apparatus according to any one of claims 1 to 4,
The principal component regression analysis means removes a noise component included in the vibration detection signal and the sound pressure detection signal subjected to singular value decomposition, and outputs the vibration detection signal and the sound pressure detection signal from which the noise component has been removed. A vibration / sound pressure transfer characteristic analyzing apparatus further comprising a noise removing unit that outputs the regression analysis unit. Detecting vibration from a vibration source by means of vibration detection means and outputting a vibration detection signal;
Detecting the sound pressure from the sound source by the sound pressure detecting means and outputting a sound pressure detection signal;
Detecting vibration or sound pressure at a predetermined response point with respect to the vibration source and the sound source by a vibration / sound pressure detection means and outputting a vibration / sound pressure detection signal;
Analyzing the frequency of the vibration detection signal, the sound pressure detection signal and the vibration / sound pressure detection signal by frequency analysis means;
Based on the vibration detection signal, the sound pressure detection signal, and the vibration / sound pressure detection signal subjected to frequency analysis, vibration transfer characteristics between the vibration source and the response point and between the sound source and the response point. Calculating each of the sound pressure transfer characteristics by the principal component regression analysis means;
Have
The principal component regression analysis means normalizes the vibration detection signal and the sound pressure detection signal subjected to frequency analysis, performs singular value decomposition on the normalized vibration detection signal and sound pressure detection signal, and performs singular value decomposition. Vibration / sound pressure between the vibration source and the sound source and the response point by performing a regression analysis between the vibration detection signal and the sound pressure detection signal and the vibration / sound pressure detection signal subjected to frequency analysis. The transmission characteristic is calculated, the calculated vibration / sound pressure transmission characteristic is distributed to the vibration transmission characteristic that is the vibration component and the sound pressure transmission characteristic that is the sound pressure component, and the distributed vibration transmission characteristic And the sound pressure transfer characteristic are restored respectively.