JP3298334B2 - Gear noise evaluation device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for evaluating gear noise, and more particularly to an apparatus for absolutely evaluating gear noise of an automobile transmission.

[0002]

2. Description of the Related Art Gear noise has been conventionally evaluated by sensory evaluation by highly trained persons, and an example of such sensory evaluation is shown in Table 1 below.

[0003]

[Table 1]

That is, a gear noise that cannot be heard even by a highly trained inspector is evaluated as a score of 10, and a gear noise that all users feel as noise is evaluated as a score of 5. It has become. The score of 5 or more is an evaluation level that can be tolerated when the transmission is shipped, for example.

[0005] On the other hand, when all the users feel noisy, the score is evaluated as 4, and when it does not function as a gear at all, the score is evaluated as 1.

[0006] Such a sensory evaluation by a skilled person is possible for each gear position.

On the other hand, apart from the manual sensory evaluation described above, a relative evaluation method using an instrument such as an FFT analyzer (frequency analyzer) is also employed.

FIG. 7 shows a conventional gear noise evaluation apparatus using such an instrument. In the figure, reference numeral 1 denotes a transmission, and an input shaft 2 and an output shaft 3 are connected to the transmission 1. The input shaft 2 is connected to an engine (not shown), and the output shaft 3 is connected to wheels (not shown).

Reference numeral 4 denotes a rotation speed sensor provided for detecting the rotation speed of the output shaft 3 (or the input shaft 2), and reference numeral 5 denotes a microphone for detecting gear noise of the transmission 1.

The output signal of the rotation speed sensor 4 and the output signal of the microphone 5 are sent to the FFT analyzer 6, and the display unit displays the sound pressure corresponding to the rotation speed (frequency) at each gear stage as shown in FIG. It is analyzed and displayed every time, and overall (overall) noise is also displayed. This display example shows the rotational speed versus sound pressure characteristics at the primary meshing frequency.

The output signal of the rotation speed sensor 4 is also supplied to the control board 7 so that the sensory evaluator 8 can monitor the current rotation speed of the output shaft 3.

That is, the operator 8 sitting on the control table 7 does not need to be the above-mentioned sensory evaluator, and the operator 8 controls the number of rotations of the output shaft 3 indicated by the output signal from the number-of-rotations sensor 4. See the corresponding meter installed at 7, and set this rotation speed to, for example, 100 rpm.
Every time (see FIG. 8), and each time, the rotational speed from the rotational speed sensor 4 and the gear noise from the microphone 5 are expressed by F
Input to FT analyzer 6.

The FFT analyzer 6 analyzes the frequency of gear noise (sound pressure) for each order as shown in FIG. 8 in accordance with an increase in the rotation speed by these inputs, and outputs the result.

The operator 8 determines from the frequency analysis result whether the noise is large when the difference between the overall noise and other gear noises is small, or what the current gear noise is based on the waveform pattern. are doing.

[0015]

As described above, the evaluation by the sensory evaluator has a drawback that a person may vary, and the evaluation of the gear noise by the meter shown in FIG. 7 cannot quantitatively judge correctly. There's a problem.

Further, there is no correlation between the evaluation in the laboratory (bench) as shown in FIG. 7 and the measurement when actually mounted on a vehicle. That is, when mounted on a vehicle, all noise is mixed into the microphone, and the relevance of such a situation to the evaluation in a laboratory in a completely different environment is greatly impaired, making it impractical. was there.

Accordingly, an object of the present invention is to realize a gear noise evaluation apparatus in which the sensory evaluation and the evaluation by an instrument are matched.

[0018]

Means and Action for Solving the Problems

[1] As described above, the evaluation itself often differs between the sensory evaluation by an expert and the evaluation by an instrument. The reason is that the FFT analyzer can only analyze the sound pressure and the frequency, and cannot evaluate the "tone" which is another of the three elements of the sound.

The inventor of the present invention considers that the tone color is caused by the combination of the intensity of each spectrum of the sound having various frequencies and the attenuation value of each spectrum. We focused on realizing sensory evaluation on a computer by performing automatic evaluation using a neural network.

That is, the neurocomputer can automatically change a parameter (weighting coefficient) used internally so that an output signal corresponding to a certain input signal matches a reference teacher signal.

Therefore, a gear noise evaluation device according to the present invention comprises a rotation speed sensor for detecting the rotation speed of a shaft on which a gear is mounted, a microphone for detecting noise of the gear, and inputting the rotation speed and noise. a FFT analyzer that outputs a sound pressure value for a plurality of rotational speed for a given gear, a control stand evaluator for each speed is set an absolute sensory evaluation value as a teacher signal, or the FFT analyzer
Input each sound pressure value output from the
A neurocomputer that outputs value.
The computer sets the sensory evaluation value from the control table.
Learning in advance to match the absolute sensory evaluation value,
Learning when the difference between the two evaluation values is within the specified tolerance
To determine the function value and based on the function value
Output the actual sensory evaluation value from each actual sound pressure value
It is characterized by .

In the present invention, the FFT analyzer inputs the rotation speed of the shaft on which the gear is mounted from the rotation speed sensor and the noise of the gear from the microphone and outputs a plurality of signals for a predetermined gear position, as in the prior art. The sound pressure (see FIG. 8) corresponding to the rotation speed is output.

On the other hand, in the control table, when the evaluator changes the rotation speed of the shaft, the absolute sensory evaluation value is input and set as a teacher signal for each rotation speed.

A plurality of sound pressure values output from the FFT analyzer are input to a neurocomputer. In the neurocomputer, a predetermined number of absolute sensory evaluation values are set so that the plurality of inputs become respective absolute sensory evaluation values as teacher signals. , Learning is performed, and when the difference from the sensory evaluation value falls within the allowable error range, the learning is stopped and the function value (parameter) is determined.

When the gear noise evaluation device combining the neurocomputers whose function values have been determined in this way is used for actual gear noise evaluation, each output signal from the microphone and the rotation speed sensor is converted by the FFT analyzer 6 into each rotation speed. Is given as a sound pressure value to the neurocomputer, since a function value has already been set in this neurocomputer, a sensory evaluation value is output in accordance with the function value.

[2] Further, in the gear noise evaluation apparatus according to the present invention, in order to evaluate gear noise without using a neurocomputer as described above, a rotation speed sensor for detecting a rotation speed of a shaft on which a gear is mounted. A microphone for detecting noise of the gear; a conversion unit for converting the frequency of the rotation speed into a plurality of sideband frequencies; a sideband for each sideband frequency for a predetermined gear position by inputting the sideband frequency and noise. An FFT analyzer that outputs a sound pressure value, a control table on which an evaluator sets an absolute sensory evaluation value for each rotation speed, and a sound pressure other than the sideband sound pressure value by inputting each sideband sound pressure value The average value of the values is calculated, and the square root obtained by adding the value obtained by multiplying the square value of the deviation between each sideband sound pressure value and the average value by a coefficient becomes the absolute sensory evaluation value. A calculation unit for obtaining the the coefficient by simultaneous equations by performing a calculation of a plurality of rotation number of such can include a.

That is, the rotation speed of the shaft output from the rotation speed sensor is converted into a plurality of side bang frequencies by the conversion unit, and this is converted together with the gear noise into the FF as in the prior art.
Input to T analyzer.

As a result, the FFT analyzer outputs sound pressure values corresponding to the respective sideband frequencies. Therefore, the arithmetic unit inputs these sound pressure values, finds an average thereof, and calculates the average of each sound pressure value and the average value. Calculations for a plurality of rotations are performed so that the square root obtained by adding the value obtained by multiplying the square value of the deviation by the coefficient becomes each absolute sensory evaluation value given from the control table, and the simultaneous equations (undetermined coefficient Method) to obtain the above coefficient.

When the gear noise estimating apparatus thus manufactured is actually used for gear noise evaluation, since the operation coefficient has already been obtained in the operation section, the sound pressure value for each side band frequency output from the FFT analyzer is calculated. If given to the calculation unit, the calculation unit adds the value obtained by multiplying the square value of the deviation between the average value of each sound pressure value and each sound pressure value by the above-described coefficient, and obtains the square root thereof. Become.

[0030]

【Example】

[1] FIG. 1 shows a gear noise evaluation device (1) according to the present invention. In this embodiment, the sound pressure value A output from the FFT analyzer 6 is smaller than that of the conventional example shown in FIG. The difference is that a neurocomputer 9 for inputting an evaluation value (teacher signal) E by a sensory evaluator 8 from a control board 7 and displaying a sensory evaluation screen 10 is used.

In the operation of such an embodiment, the FF
The output signal from the rotation speed sensor 4 and the output signal from the microphone 5 are given to the T analyzer 6, and the analysis results for each order and each rotation speed for each gear speed as shown in FIG. ) Is output as in the conventional example of FIG.

The neurocomputer 9 to which such a large number of sound pressure values A have been input is set by the operator 8 as a sensory evaluator at each rotation speed when the rotation speed of the transmission 1 is sequentially changed from the control board 7 at the same time. Absolute sensory evaluation value E
Is input from the control board 7. The absolute sensory evaluation values are as shown in Table 1 above.

Then, in the neurocomputer 9, the function value is determined by an operation process as shown in FIG.

That is, in a neural network (back propagation) as shown in FIG. 2, an input pattern is output through an input layer, an intermediate layer, and an output layer, and this neural network outputs an output signal for a certain input signal. Is a kind of a mapping function that automatically changes parameters used internally so as to match the teacher signal to be referred to. The internal parameters are present in the neurons constituting the neural network, and are weights for multiplying a signal input to the neuron, a gradient of a sigmoid function which gives an output to the signal multiplied by the weight, and a value at which the output is effective. Indicates a threshold value for determining whether the threshold value has been reached. By automatically adjusting these parameters, an attempt is made to imitate the input / output relationship of the teacher signal.

Specifically, when three sensory evaluation values "5" to "7" are assigned in the output layer as shown in the figure, the input pattern causes all the coupling coefficients between the respective units to be calculated. One of the signals obtained in the output layer is set to “1”, and the rest is assigned to “0”.

For example, in the illustrated example, when the sensory evaluation value "6" is "1" and a plurality of sound pressure values are input from the FFT analyzer 6, all coupling coefficients between the units are determined.

As described above, the state of "1" in the output layer is sequentially shifted due to the change in the rotation speed, so that the coefficients of the neural network are learned and gradually converged. Usually, this learning involves 5
Zero or more teacher signal data is required.

Then, as a result of learning and conducting as described above,
When the error between the sensory evaluation value and the evaluation value indicated in the output layer is within an allowable range, the learning is stopped and such a function value,
That is, the coefficient is fixed.

Since the function value in the neurocomputer 9 has been determined in this way, the gear noise is evaluated using such a device.

That is, when the FFT analyzer 6 inputs the output signal from the rotation speed sensor 4 and the gear noise from the microphone 5 and gives the sound pressure value A as shown in FIG. Since the function value has already been determined, an operation is performed based on this function value, and the sensory evaluation screen 10 is obtained. This example indicates that the sensory evaluation value is “5”.

As described above, the conventional FFT analyzer 6 can add only the relationship between the sound pressure and the frequency to the evaluation. However, in the present invention, since the neurocomputer is used, the evaluation by the evaluator The evaluation is equivalent to the evaluation performed for "three elements of sound" including not only sound pressure and frequency but also "tone", and it is possible to always perform almost the same evaluation as sensory evaluation even with instruments. Becomes

Further, in addition to the evaluation in the laboratory as shown in FIG. 1, the present invention can evaluate the correct gear noise even when mounted on a vehicle as shown in FIG.

That is, as shown in the figure, the rotation speed signal from the rotation speed sensor 4 and the gear noise signal from the microphone 5 are taken into the digital audio tape 11 mounted on the vehicle. Note that when the digital audio tape 11 is loaded, data sampling has already been performed as shown in the figure.

Accordingly, when such a digital audio tape 11 is applied to the FFT analyzer 6, FIG.
In the same manner as described above, a plurality of sound pressure values A are provided from the FFT analyzer 6 to the neurocomputer 8 and a sensory evaluation screen 10 is provided.

In the case where gear noise is evaluated by being mounted on a vehicle as described above, various noises are mixed into the microphone 5, but the neurocomputer 9 uses the sensory evaluation value from the sensory evaluator 8 as shown in FIG. Since E is given in the same manner, finally, the neurocomputer 9 has a function value such that an absolute evaluation value can be obtained by the sensory evaluator. Become.

[2] In the above embodiment, the evaluation was performed using a neurocomputer. However, in the present invention, it is also possible to evaluate gear noise without using a neurocomputer.

FIG. 4 shows an embodiment in which such a neurocomputer is not used. Compared with the embodiment shown in FIG. 1, the output signal of the rotation speed sensor 4 is not directly input to the FFT analyzer 6. The difference is that it is provided via the conversion unit 12.

That is, in the conversion unit 12, the sound pressure is converted from the FFT analyzer 6 based on the sideband frequency (rotation speed) of the rotation speed of the shaft 3 (or 2) set by the operator 8 who is the sensory evaluator. It is intended to be output.

The side band frequency is defined as follows: when a pair of gears has the number of teeth Z ₁ on the drive side and the number of teeth on the driven side is Z ₂ , when only the gear noise on the drive side is considered, Gear center frequency C
_SO becomes C _SO = (Z ₁ × N) / 60. Then, the sound pressure at this time is A _0.

The number of gear teeth on the drive side is one, two,
The frequency when the number of teeth is virtually increased / decreased as the number of teeth, three,... Is the sideband frequency, and the conversion section 12 outputs this sideband frequency.

That is, for example, even if the rotation speed is actually 2000 rpm, the rotation speed given from the
Given as two rotation speeds near 2000 rpm, the sound pressure value at each sideband frequency is output from the FFT analyzer 6.

This can be expressed by the following equation (1).

[0053]

(Equation 1)

Note that the side band frequency C_s1, C_S2, C
_S3... produces two sideband frequencies each
However, for the sake of simplicity, here the sound pressure value is represented by one
A ₁, A_Two, A_Three,...

When the sound pressure value A is given to the calculation unit 13 based on the sideband frequency in this way, the calculation unit 13 performs calculation according to the following equation (2).

[0056]

(Equation 2)

That is, each side band sound pressure value A ₁ ,
A mean value X of sound pressure values other than A ₂ , A ₂ ,... Is obtained, and a deviation between the mean value X and each sound pressure value A ₁ , A ₂ , A ₃ ,. The noise component (DC component) of each sideband sound pressure value can be removed), and the square of each deviation is multiplied by coefficients m ₁ , m ₂ , and m ₃ , respectively, and added. By taking the square root, the sensory evaluation value of the gear noise is made to correspond to the sensory evaluation value E which is a representative characteristic value when the measurement is performed by an instrument.

When the operator 8 changes the number of revolutions, the sound pressure value shown in the equation (1) also changes, so that the coefficients m ₁ , m ₂ , m ₃ ,. If the data of the sound pressure values is obtained by a number, simultaneous equations regarding the coefficient m are established, and these coefficients are obtained.
These coefficients are referred to as “weighting modification coefficients”, for example, m ₁
= 1, m ₂ = 0.75, m ₃ = 0.75, and a value of 1 or less is obtained.

When the gear noise evaluation is actually performed by using the calculation unit 13 in which the weighted modification coefficient is obtained,
Since the coefficient has already been determined in the arithmetic unit 13, FF
The calculation unit 13 having received the sound pressure value A output from the T analyzer 6 can execute the calculation of the equation (2), and the value E of the square root of the equation (2) is equal to the absolute sensory evaluation value at this time. Become.

In the embodiment shown in FIG. 4, the sensory evaluation screen 10 is obtained in the arithmetic unit 13 in the same manner as in FIG.

In the above description, a vehicle transmission has been described as an example. However, the present invention is not limited to this, and can be applied to any location where gear noise can occur.

[0062]

As described above, according to the gear noise evaluation apparatus of the present invention, a plurality of sound pressure values from the FFT analyzer are supplied to the neurocomputer, and at this time, the evaluator instructs the evaluator for each rotation speed. Since the absolute sensory evaluation value is set at the same time as the signal, the neurocomputer learns so that each sound pressure value becomes each absolute sensory evaluation value, and stops learning when the difference between them is within the allowable error range. Determine the function value and, based on this function value,
Since the evaluation value is obtained, the function value is determined by the neurocomputer in the form including the three elements of the sound in the same manner as the sensory evaluation by the evaluator including the three elements of the sound. The sensory evaluation almost matches the sensory evaluation by the instrument, and the correct evaluation is automatically obtained.

FIG. 5 shows the judgment result (the number of neuro data samples = 145) between the evaluation using a neurocomputer and the sensory evaluation by the evaluator. As can be seen from this judgment result, the neuro evaluation score is It can be seen that the maximum value is within the allowable error range of 0.25 points from the sensory evaluation point.

Further, in the determination result shown in FIG. 6, as a result of increasing the number of neuro data samples to 183, it can be seen that the maximum evaluation difference is still only 0.25 in this case.

Further, according to the present invention, the gear noise is determined based on the sensory evaluation value not only in the laboratory but also in the vehicle, so that the correlation between the two becomes very strong.

Further, it is possible to obtain a modification coefficient that can be a sensory evaluation value by an evaluator by performing a predetermined operation using a concept of a plurality of sideband frequencies for each rotation speed without using a neurocomputer. And the same effect as above can be obtained.

[Brief description of the drawings]

FIG. 1 is a diagram showing an embodiment of a gear noise evaluation device according to the present invention.

FIG. 2 is a diagram for explaining the principle of a neurocomputer used in the gear noise evaluation device (1) according to the present invention.

FIG. 3 is a diagram when the gear noise evaluation device according to the present invention is mounted on a vehicle to perform an evaluation process.

FIG. 4 is a diagram showing an embodiment of a gear noise evaluation device (2) according to the present invention.

FIG. 5 is a graph showing a determination result (No. 1) of a neurocomputer by a gear noise evaluation device (1) according to the present invention.

FIG. 6 is a graph showing a determination result (No. 2) of the neurocomputer by the gear noise evaluation device (1) according to the present invention.

FIG. 7 is a diagram showing a conventional gear noise evaluation device.

FIG. 8 is a graph showing, for each gear, analysis results of rotational speed (frequency) versus sound pressure obtained by an FFT analyzer used in the present invention and a conventional gear noise evaluation device.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Transmission 2 Input shaft 3 Output shaft 4 Revolution speed sensor 5 Microphone 6 FFT analyzer 7 Control stand 8 Operator (sensory evaluator) 9 Neurocomputer 10 Sensory evaluation screen 12 Conversion part 13 Operation part Show the part.

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-1-260327 (JP, A) JP-A-6-241297 (JP, A) JP-A-2-272326 (JP, A) (58) Field (Int.Cl. ⁷ , DB name) G01H 3/00 G01M 13/02 JICST file (JOIS)

Claims (2)

(57) [Claims] 1. A rotational speed sensor for detecting a rotational speed of a shaft on which a gear is mounted, a microphone for detecting noise of the gear, a plurality of rotational speeds for a predetermined gear position by inputting the rotational speed and the noise. FFT that outputs sound pressure value for
An analyzer, a control table on which an evaluator sets an absolute sensory evaluation value as a teacher signal for each rotation speed, and the FFT analyzer
Input each sound pressure value output from the riser, and
And a neurocomputer that outputs sensory evaluation values
The neurocomputer calculates that the sensory evaluation value is
To match the absolute sensory evaluation value set from the table
Learning and the difference between the two evaluation values is within the specified tolerance
When learning is stopped, the function value is determined and the function
Output the actual sensory evaluation value from the actual sound pressure value based on the
A gear noise evaluation device characterized by applying force . 2. A rotation speed sensor for detecting a rotation speed of a shaft on which a gear is mounted, a microphone for detecting noise of the gear, and a converter for converting a frequency of the rotation speed to a plurality of sideband frequencies. A FFT analyzer that inputs the sideband frequency and the noise and outputs a sideband sound pressure value for each sideband frequency for a predetermined gear position, and a control table on which an evaluator sets an absolute sensory evaluation value for each rotation speed. Inputting each sideband sound pressure value, calculating an average value of sound pressure values other than the sideband sound pressure value, and multiplying a value obtained by multiplying a square value of a deviation between each sideband sound pressure value and the average value by a coefficient. A gear noise evaluation device, comprising: an operation unit that performs an operation for a plurality of rotation speeds so that the square root of the sum becomes each absolute sensory evaluation value and obtains the coefficient by a simultaneous equation.