JPH0953460A - Noise reducer of envelope type engine - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve in noise reducing performance by installing a vacuum fan in a ventilating part and providing a system identification means, seeking an electroacoustic characteristic on basis of a corrective learning identification process, and driving an counteracting noise generating means for generating an offset sound to cancel the noise at a final exhaust port. SOLUTION: An envelope type engine generator 1 installs each partition wall 5 in the upper side and the sidepiece of a diesel engine 3, and in this constitution, a radiator 6 is attached to these partition walls 5, through which the inside of a soundproof case 2 is partitioned off to an engine generating chamber 7, an exhaust passage 8 and a side space 9. When engine 3 is started, a radiator fan 11 is also rotated and thereby cooling air M is produced there, but at this time, it gets a reference noise signal from a reference noise microphone 22, while it inputs a residual noise signal from a residual noise microphone 19. In succession, on the basis of the reference noise signal, the residual noise signal and an electroacoustic characteristic to be secured a system identification process, a counteracting sound generating means is driven and controlled so as to generate a counteracting sound to cancel any noise at the final exhaust outlet.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a noise reduction device for an enclosed engine, and more particularly to a noise reduction device for an enclosed engine that reduces noise of a noise source by active noise control.

[0002]

2. Description of the Related Art Conventionally, as a technique for reducing noise in an engine and an exhaust duct, a passive method of providing a sound absorbing material and a vibration damping material has been proposed. On the other hand, in recent years, active noise control (active noise control, hereinafter referred to as ANC) that emits a canceling sound of the same amplitude and opposite phase to noise to reduce the noise by the interference effect of sound waves has been put to practical use. Has been actively researched in various fields. According to ANC, it is difficult to reduce with a sound absorbing material around 100 Hz to 50 Hz.
Since it also has a reducing effect on low frequency noise near 0 Hz, it is used to reduce engine exhaust noise, car interior noise, etc. (Japanese Patent Publication No. 2-503219).
JP-A-3-204354, etc.).

For example, a reference noise microphone 101 is provided on the upstream side of the duct 100 shown in FIG.
There is an ANC device in which a residual noise microphone 102 is disposed nearby and an anti-noise speaker 103 is disposed at a predetermined position downstream of the reference noise microphone 101 and upstream of the duct port 100a. In order to realize ANC control in this device, the reference noise microphone 101 and the anti-noise speaker are used in order to investigate in advance the transfer characteristics of the electroacoustic system H of the duct 100 to be silenced, using substantially the same algorithm as when performing ANC control. It is necessary to accurately obtain the electro-acoustic characteristic B between 103 and the electro-acoustic characteristic C between the residual noise microphone 102 and the anti-noise speaker 103. Such processing is called system identification. The filter 107 compensates for the electroacoustic characteristic B (in FIG. 20, <B> is the characteristic B).
Is used to prevent howling between the reference noise microphone 101 and the anti-noise speaker 103, and the filter 108 compensates the electroacoustic characteristic C to muffle the residual noise microphone 102. It is something to do.

To identify the electroacoustic characteristics B and C, specifically, M-sequence random noise is output from the loudspeaker 103, passed through the respective electroacoustic characteristics, and obtained from the reference noise microphone 101 and the residual noise microphone 102. It is performed by applying an adaptive digital filter to the signal to be processed by ANC control. That is, the M-sequence random noise is output from the loudspeaker 103, and the input / output relationship of the transfer function is specified for sounds of all frequencies. In FIG. 20, reference numeral 104 is an ANC processing device, and reference numeral 105 is an adaptive digital filter for generating an antiphase waveform which becomes a control sound, and a residual noise signal based on the interference result so as to always obtain the maximum volume. The filter coefficient is updated according to. This adaptive digital filter 10
The adaptive algorithm unit 106 of No. 5 uses a residual noise signal based on the reference input signal Xn and the residual noise signal en from the residual noise microphone so that the canceling sound emitted from the speaker 103 reduces the noise of the surrounding engine. The filter coefficient of the adaptive digital filter 105 is updated so that en is minimized.

As a method of making the filter coefficient of the adaptive digital filter 105 an optimum value so that the residual noise signal en is minimized, for example, B. The "Filtered-xLMS algorithm" proposed by Widrow et al. Is well known. In this algorithm, for example, the filter coefficient of the adaptive digital filter 105 is recursively updated by the following formula. W _{n + 1} = W _n +2 μe _n · X _n, where Wn: adaptive digital filter coefficient (W [0], W
[1], ..., W [L-1]) Xn: Reference noise signal (X [n], X [n-1], ..., X
[N−L]) en: residual noise signal n: discrete time μ: convergence coefficient L: filter order.

Here, the convergence coefficient μ is a positive scalar and is a parameter for controlling the magnitude of the correction amount in each iteration, and is also called a step size parameter. According to the formula, when the reference noise signal Xn and the residual noise signal en are given, the root mean square error (MSE) reaches the minimum value only by appropriately setting the convergence coefficient μ, and the filter coefficient Wn
(Tap weight) has an optimum value. How to set the value of the convergence coefficient μ is an extremely important parameter that determines the performance of the adaptive system of ANC. However, no theoretical solution has been made regarding the setting of the convergence coefficient μ, and in the present ANC control, 0.01,
At present, the value is set to about 0.001 by trial and error.

[0007]

Therefore, if the convergence coefficient is improperly set, the optimum value cannot be reached no matter how many calculations are repeated, and as a result, noise cannot be reduced or the optimum value cannot be obtained. There is a problem that it takes a long time to reach the value, and as a result, the expected noise reduction level cannot be obtained easily. FIG.
As shown in Fig. 7, the noise of the enclosed engine has a complicated spectrum including the fundamental frequency and its harmonics caused by the rotation speed of the crankshaft and the cooling fan shaft, the wind noise of the fan, the vibration noise of the soundproof case, etc. There is. Therefore, unlike the case of reducing periodic noise, the noise uncertainties are large, and in order to actually reduce the noise of the enclosed engine, it is necessary to consider an ANC system that matches the actual configuration of the enclosed engine. . Further, in the enclosed engine, it is essential to generate cooling air to cool the inside of the enclosure, and in addition to engine noise and vibration, the effect of cooling air must be considered. Specifically, the cooling air flow and A
It is necessary to consider a duct configuration that allows NC muffling control.

[0008]

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and when ANC is applied to an enclosed engine,
(EN) Provided is a noise reduction device for an enclosed engine, which enables quick and appropriate setting of the convergence coefficient, can perform system identification processing work in a short time, and can improve noise reduction performance of ANC processing operation. The purpose is to A second object of the present invention is to provide a specific configuration of a practical noise reduction device for an enclosed engine, which can improve the sound deadening effect when the ANC is applied to the enclosed engine.

[0009]

In order to achieve the above object, a noise reduction device according to claim 1 surrounds an engine with a soundproof wall, and an outside air intake port and a final discharge port are provided at predetermined positions of the soundproof wall. Provide a ventilation part that takes in outside air from the outside air intake port and exhausts ventilation air from the final exhaust port.In the ventilation part, a suction fan that sucks out the hot air of the engine from the engine to the final exhaust port side is provided, and a predetermined position of the ventilation part. Is provided with reference noise detecting means for detecting noise including engine sound and fan wind noise, residual noise detecting means is provided near the final outlet of the ventilation part, and the downstream side of the ventilation part position where the reference noise detecting means is provided. Thus, a canceling noise generating means is provided at a predetermined position on the upstream side of the ventilation portion position where the residual noise detecting means is disposed, and at least the electroacoustic characteristics between the canceling noise generating means and the residual noise detecting means are set. A system identification means for determining the electroacoustic characteristics of the ventilation part based on the modified learning identification method is provided, the reference noise signal of the reference noise detection means and the residual noise signal of the residual noise detection means are detected, and the LMS method is used. , ANC processing means for driving the canceling sound generating means so as to generate canceling sound for canceling noise at the final outlet based on the detected two noise signals and the electroacoustic characteristics obtained by the system identification processing It is characterized by

A noise reducing apparatus according to a second aspect of the present invention has convergence coefficient setting means for setting the convergence coefficient μ in the modified learning identification method and the LMS method based on the sum of squares of filter coefficients. The noise reduction device according to claim 3 repeatedly adds the square sum of the filter coefficients by an integer multiple of the filter order, obtains the difference between the integer multiple addition value and the previous integer multiple addition value, and based on the difference. It has a convergence coefficient setting means for setting the convergence coefficient.
According to a fourth aspect of the present invention, in the noise reduction device, a muffler of the engine is provided on the downstream side of the suction fan at the position of the ventilation part where the suction air blown by the suction fan blows, and the reference noise detection means is provided for the engine sound, fan wind noise, and muffler chassis It is provided at the position of the ventilation part where the transmitted noise can be detected.

According to a fifth aspect of the present invention, there is provided a noise reducing device in which an engine is housed in an enclosure, at least an intake port and an exhaust port are provided at predetermined locations in the enclosure, and ambient air is taken in from the intake port and exhausted from the exhaust port. In order to reduce the noise of the exhaust, at least the exhaust side duct communicating with the exhaust port is formed around the enclosure, and the primary noise detection microphone that detects the noise near the exhaust port and the noise near the duct port of the exhaust side duct are detected. A residual noise detecting microphone, a canceling sound generating means arranged in the exhaust side duct, and a canceling sound generating means for generating a canceling sound for canceling noise at the duct mouth based on the primary noise signal and the residual noise signal. The exhaust side duct includes a bag-shaped air guide plate having an exhaust port on the lower side so as to cover the exhaust port from the upper side, and an air guide plate on the wall surface of the enclosure having the exhaust port. It is composed of a rectangular bowl-shaped exhaust duct case provided with an opening in the upper wall so as to be housed, and a horizontal duct that is partitioned from the exhaust duct case by the upper wall and communicates with the opening and extends in the horizontal direction. , The exhaust air guide chamber where the cooling air from the exhaust port is surrounded by the air guide plate, the exhaust side upward air passage formed by the air guide plate and the exhaust duct case, the horizontal air passage in the horizontal duct, and the duct opening And a convergence coefficient setting means for setting the convergence coefficient μ of the ANC processing means based on the sum of squares of the filter coefficients. The noise reduction device according to claim 6 stores the filter coefficient in a memory, searches for the maximum value and the minimum value of the filter coefficient, detects a state in which the storage addresses of the maximum value and the minimum value are fixed, and converges. It is characterized in that the evaluation processing for determining whether or not the coefficient switching is necessary is performed thereafter. According to a seventh aspect of the present invention, in the noise reduction device, a square value of a maximum value and a square value of a minimum value are respectively obtained, and two square values are added to obtain a sum of squares of the filter coefficient.

[0012]

According to the noise reducing device of the present invention, by providing the suction fan in the ventilation section, a region where the canceling sound is emitted between the final exhaust port on the exhaust side and the position where the suction fan is disposed. Can be formed, so that an advantageous configuration can be obtained in performing ANC control. Also,
By providing the system identification means for obtaining the electroacoustic characteristics of the ventilation part based on the modified learning identification method, the convergence is fast and the time required for system identification can be shortened. Further, by providing the ANC processing means using the LMS method, which is easy to calculate, the order of the adaptive filter can be expanded and the noise reduction performance can be improved.

Further, by providing the reference noise detecting means for detecting the noise including the engine sound and the fan wind noise at a predetermined position of the ventilation part, the engine noise and the fan wind noise can be silenced, and the whole surrounding type engine can be suppressed. The noise level can be reduced.

According to the noise reducing apparatus of claim 2, at the time of system identification, at the first stage of adaptive control during ANC control,
Since each filter coefficient (tap weight) determined by the update formula of the filter coefficient of each ANC control algorithm varies, when the sum of squares of the filter coefficients is calculated, it greatly varies in each update loop. However, at the final stage of adaptive control, each filter coefficient hardly changes, so that the value of the sum of squares becomes stable. By evaluating the change of this state by some evaluation standard, it can be used as a standard for setting the convergence coefficient. Also, the filter coefficient of 2
By taking a power, it is possible to perform processing only by addition without performing subtraction, and it is possible to set optimal convergence coefficient at high speed in system identification processing and ANC processing. According to the noise reduction device of claim 3, the square sum of the filter coefficients is repeatedly added by an integral multiple of the filter order, and the difference between the integral multiple addition value and the previous integral multiple addition value is obtained, and the difference is calculated. By providing the convergence coefficient setting means for setting the convergence coefficient on the basis of the convergence coefficient, the convergence can be improved in the system identification processing and the ANC processing.

According to the noise reducing device of the fourth aspect, the muffler can be cooled by providing the muffler of the engine at the position of the ventilation part on which the suction air from the suction fan blows, on the downstream side of the suction fan, so that the muffler chassis can be cooled. The frequency change range of the transmitted noise (hereinafter referred to as muffler transmitted noise) can be narrowed, and the control in the ANC processing can be facilitated. Further, since the reference noise detecting means is provided at the position of the ventilation part capable of detecting the engine sound, the fan wind noise, and the muffler transmission noise, not only the engine noise and the fan wind noise but also the muffler transmission noise can be silenced. According to the noise reduction device of claim 5, by configuring the exhaust duct as described above, the cooling air that has exited the exhaust port bypasses the air guide plate and is discharged from the lower side where the exhaust port is provided, so that the exhaust side rising wind. It will be discharged from the duct opening of the horizontal wind passage through the passage. Since the cooling air bypasses the baffle plate here, it forms an exhaust side upward air path that rises from the lower side to the upper side in the exhaust duct case, so that the heated cooling air does not stagnate well. It can be released to the outside. Also,
Since the muffling duct can be taken up and down by the vertical length of the exhaust duct case, it is possible to secure a space that can cancel the sound in the low frequency region. Further, by providing a suction type ventilation fan inside the exhaust port of the enclosure, the flow of cooling air can be improved and the ventilation performance can be improved. Further, since the convergence coefficient is set based on the sum of squares of the filter coefficient as described above, it is possible to set the convergence coefficient by switching to an appropriate value according to the progress degree of the adaptive control at that time. According to the noise reduction device of claim 6, it is necessary to search the maximum value and the minimum value of the filter coefficient, detect the state where the storage address of the maximum value and the minimum value is fixed, and switch the convergence coefficient. Since the evaluation processing of whether or not there is is, the rough convergence can be determined only by the storage address search processing, and the value of the convergence coefficient is not changed even if the number of stages of the adaptive filter increases. This makes it possible to perform the determination processing in a short time. The noise reduction device according to claim 7 obtains the square value of the maximum value and the minimum value, respectively, and adds the two square values to obtain the sum of squares of the filter coefficients. The calculation time can be reduced as compared with the case where the multiplication value is calculated.

[0016]

As described above, according to the first aspect of the present invention, the following specific effects can be obtained. (A) By providing the system identification means for obtaining the electroacoustic characteristics of the ventilation part based on the modified learning identification method, the convergence is accelerated, and the time required for system identification can be shortened. In particular, the enclosed engine is often used in a place where the environment changes drastically, such as outdoors, and it is preferable to perform system identification in the environment (temperature, etc.) at that time when the engine is restarted. The ability to accelerate the convergence of adaptation can improve usability on the user side.

(B) A using the simple LMS method of operation
By providing the NC processing means, the filter coefficient can be calculated by a simpler calculation within the program processing time within the sampling frequency, so that the filter order of the adaptive filter (the number of stages of product-sum calculation) can be increased. You can
The noise reduction performance can be improved. (C) As described above, in the enclosed engine, it is possible to provide a configuration capable of comprehensively reducing engine noise and fan wind noise that leak from the exhaust port.

According to the invention of claim 2, in addition to the effect of claim 1, (d) since the optimum convergence coefficient can be set more finely according to the degree of convergence, the convergence of the adaptive filter can be performed in each of the system identification processing and the ANC processing. You can reduce the time required,
In the system identification process, accurate electroacoustic characteristics can be obtained quickly, and in the ANC process, actual noise reduction performance can be enhanced, which is a unique effect. According to the invention of claim 3, in addition to the effect of claim 2, the convergence performance of the adaptive filter can be further improved, and the effect of claim 2 can be enhanced.

The invention of claim 4 has the following unique effect in addition to the effect of claim 1. (E) The muffler can be cooled by the cooling air, the frequency change range of the muffler transmission noise can be narrowed, and control in the ANC processing can be facilitated. (F) Not only engine noise and fan wind noise but also muffler transmission noise can be suppressed. The invention of claim 5 is
There is a unique effect that the convergence performance of the adaptive filter can be improved and the exhaust duct can be configured to be suitable for ANC control. According to the invention of claim 6, the rough determination of the convergence of the filter coefficient is performed first, and then it is determined whether or not the switching of the convergence coefficient is necessary. Therefore, the wasteful evaluation process can be omitted. , The unique effect that the degree of convergence of the adaptive control can be identified remarkably quickly. According to the invention of claim 7, it is possible to perform the determination processing of whether or not to change the value of the convergence coefficient in a shorter time than to calculate and calculate the squared values of all the filter coefficients, and to reduce the calculation time. As a result, the peculiar effect that the followability of ANC control is improved is obtained.

[0020]

BEST MODE FOR CARRYING OUT THE INVENTION A detailed description will be given below with reference to the accompanying drawings showing an embodiment. FIG. 1 is a diagram showing an embodiment of a noise reduction device for an enclosed engine according to the present invention.
1 is a vertical cross-sectional view of a surrounding engine generator, and FIG.
It is a BB line cross-sectional view of (A). In this enclosure type engine generator 1, a diesel engine 3 is arranged in a substantially central portion of a soundproof case 2, and a generator 4 is arranged on the right side in the drawing. A partition wall 5 is provided on the upper side and the side surface of the diesel engine 3, and a radiator 6 is attached to the left side surface of the partition wall 5, so that an engine power generation chamber 7 in which the engine 3 and the generator 4 are housed in the soundproof case 2 is provided. Exhaust passage 8
And the side space 9 are divided.

The exhaust passage 8 is composed of an ascending passage 8a provided so as to face the radiator 6 and an exhaust passage 8b communicating with an exhaust port 10 provided at the upper wall 2a of the soundproof case 2. A suction type radiator fan 11 is attached to the radiator 6, and the radiator fan 11 is driven by a crankshaft of the engine 3 by a fan belt and a pulley. A cylindrical muffler 12 extending in the horizontal direction is provided in the ascending passage 8a facing the radiator 6, and the suction air of the radiator fan 11 flows through the side surface of the cylindrical muffler 12 into the exhaust passage 8b.
The exhaust port of the engine is an exhaust pipe 2 that bypasses the side space 9.
The exhaust gas from the muffler 12 is communicated with the outside by passing through the left wall of the soundproof case by the exhaust pipe 20b.

On the upper wall 2a of the soundproof case 2, a substantially rectangular duct 13 having a base end portion on the radiator 6 side and a duct outlet 13a on the opposite side is fixed, and a duct portion above the discharge port 10 is fixed. Is provided with a shielding wall 14 that horizontally divides the duct 13 into two rooms. The outlet 10 is covered by the shielding wall 14 by about half. A loudspeaker 16 having a directivity is disposed in an upper room (referred to as an upper duct room 15) partitioned by the shielding wall 14 with its speaker surface facing the duct outlet 13a. In addition, the lower room (lower duct room 17
Has a function of guiding cooling air passing through the ascending passage 8a and the exhaust passage 8b toward the duct outlet 13a. The loudspeaker 16 is surrounded by a unit case 18 so that it can be detachably replaced, and by fitting the loudspeaker 16 with the upper duct chamber 15 without leaving a gap, the duct 1 of the loudspeaker 16 is removed.
It can be easily attached to the 3. A residual noise microphone 19 for detecting residual noise near the duct outlet 13a is provided at a predetermined position near the duct outlet 13a.

Further, the ascending passage 8 on the downstream side of the muffler 12
A reference noise microphone 22 is provided in a, and reference noise in a state in which engine noise, wind noise of the radiator fan 11 and muffler transmission noise are mixed is detected and input to the ANC controller 23. On the other hand, the side wall 2 b of the soundproof case 2 on the generator 4 side is provided with an intake port 24 for sucking the outside air for ventilating the inside of the soundproof case 2.
A fuel tank 25 supported by the partition wall 5 is provided above. An air passage that extends from the intake port 24 to the duct port 13a through the upper portion of the power generation chamber 7, the exhaust passage 8, and the duct 13 forms the ventilation unit 21. Further, the fuel supply port 26 of the fuel tank 25 is provided in the upper part of the soundproof case 2 on the front side of the drawing in FIG. An air cleaner 27 is provided at a substantially central position of the engine power generation chamber 7, and supplies the outside air sucked from the intake port 24 to the intake port through the intake pipe.

As shown in FIG. 2, the noise reduction device comprises a reference noise microphone 22, a directional loudspeaker 16, a residual noise microphone 19, and an ANC controller 23. The ANC controller 23 is shown in FIG. A)
As shown in FIG. 5, the soundproof case 2 is disposed at a lower position so as to be surrounded by the heatproof member 28 so as not to be affected by the heat of the engine. Further, on the inner wall of the soundproof case 2, the inner wall of the exhaust passage 8 and the inner wall of the duct 13, a sound deadening material such as glass wool is attached to prevent sound in a high frequency region of 1 kHz or more from leaking to the outside.

FIG. 2 shows the ANC during system identification processing.
4 is a block diagram showing a configuration of a system identification unit 31 configured as software in the controller 23. FIG. The system identification unit 31 includes a white noise generation unit 32 that generates white noise such as M-sequence random noise, and a “correction learning identification method”.
Based on the residual noise signal, a correction learning unit 34 that updates the filter coefficient of the adaptive digital filter 33 every moment according to the residual noise signal, and the output of the adaptive digital filter 33 and the noise microphones 22 and 19 are detected. The difference calculation unit 30 for obtaining the difference from the signal and the convergence coefficient setting unit 42 for performing the processing of the evaluation function routine described later are included. The amplifier section and the A / D conversion section are omitted in the drawing.

The configuration of the system identification section 31 shown in FIG. 2 will be briefly described. The reference noise microphone 2 emits the input signal Xn composed of M-sequence random noise from the speaker 16.
2. A noise signal is detected from the residual noise microphone 19. The difference calculation unit 30 takes the difference between the signal obtained by passing the input signal Xn through the adaptive digital filter 33 and the noise signal from the residual noise microphone 19, for example. This difference signal is treated as an error signal en in the system identification processing. The correction learning unit 34 determines the residual noise signal e
Adaptive digital filter 3 to minimize the power of n
Since the filter coefficient Wn of 3 is sequentially updated, the adaptive digital filter 33 is set so that the residual noise signal en becomes 0.
Determining the filter coefficient Wn of is the same as reproducing the electroacoustic characteristic B between the speaker 16 and the residual noise microphone 19 on the adaptive digital filter 33. Similarly, the electroacoustic characteristic C between the speaker 16 and the reference noise microphone 22 is also identified. The modified learning identification method is an algorithm operated by the mathematical formulas shown in the following formulas 1 and 2.

[0027]

[Equation 1]

[0028]

[Equation 2]

However, Wn: Adaptive filter coefficient (W [0], W [1], ..., W
[L]) Xn: Input signal (X [n], X [n-1], ..., X [n
-L]) en: Error signal n: Discrete time a, μ: Convergence coefficient (however, a is fixed at 0.5 and treated as a constant in the embodiment) L: Filter order σn ² : Square of filter input 2 (in the configuration shown in FIG. 2, the average power of the input signal Xn is shown. In an actual application, such a method is used because the average power is easy to use). Regarding the modified learning identification method and the LMS method,
There is a detailed explanation in the basics of the adaptive filter (2), Journal of Acoustical Society of Japan, Vol.

FIG. 3 shows an ANC processing unit 37 configured as software in the ANC controller 23 during ANC processing.
FIG. 3 is a block diagram showing the configuration of FIG. The ANC processing unit 37
An adder 70, an adaptive digital filter 38 that generates an antiphase waveform that serves as a control sound, an LMS control unit 39, a compensation filter 40 between the reference noise microphone 22 and the loudspeaker 16, a residual noise microphone 19, and a loudspeaker. It includes a compensation filter 41 between 16 and a convergence coefficient setting unit 42 that performs processing of an evaluation function routine described later. The electroacoustic characteristic B determined by system identification becomes the fixed filter value of the compensation filter 40, and the electroacoustic characteristic C becomes the fixed filter value of the compensation filter 41. The noise signal detected from the reference noise microphone is added to the adder 70.
The output signal from the compensation filter 40 is added to the adder 70, and the adder 70 adds these two signals to obtain a reference noise signal Xn. The reference noise signal Xn is output to the compensation filter 41 and the adaptive digital filter 38. The output signal Yn of the adaptive digital filter 38 becomes the drive signal for the speaker. The LMS control unit 39 uses the reference noise signal Xn and the residual noise microphone 19 so that the canceling sound emitted from the speaker reduces the noise of the surrounding engine.
The filter coefficient of the adaptive digital filter 38 is updated so that the residual noise signal en is minimized based on the residual noise signal en. The LMS method is an algorithm operated by the mathematical formula shown in the following Expression 3.

[0031]

(Equation 3)

The operation of the noise reducing device for the enclosed engine will be described with reference to FIGS. 1 to 3 and the flow chart shown below. First, in the basic flow chart shown in FIG. 4, the system identification unit 31 based on the modified learning method performs system identification in step SP1, and the ANC processing unit 3 based on the LMS method in step SP2.
At 7, ANC processing for canceling the actual sound is performed.

FIG. 5 is a flow chart for explaining the system identification processing shown in step SP1 in FIG. 4. First, in step SP1a, the system identification unit 31 is generated (initialized) by software in the ANC controller 23. In step SP1b, the white noise generation unit 32 performs an operation for generating M-sequence random noise to generate M-sequence random noise from the loudspeaker 16, and in step SP1c, a filter value corresponding to the electroacoustic characteristic C in FIG. Obtained by the modified learning identification method, it is determined whether or not the <C> filter has converged in step SP1d. If it is determined that the filter has not converged, the M-series random noise in step SP1b is continued to be generated, and then converged. If it is determined that there is, proceed to step SP1e

In step SP1e as well, M-sequence random noise is generated from the loudspeaker 16, and step SP1
In f, the filter value corresponding to the electroacoustic characteristic B in FIG. 2 is obtained by the modified learning identification method, and step SP1g
In <B>, it is determined whether or not the filter has converged. If it is determined that the filter has not converged, M-sequence random noise in step SP1e is continuously generated, and if it is determined that the filter has converged, system identification is performed. Finish the process, A
Shift to NC processing. The modified learning identification method is adopted in step SP1c and step SP1d because the modified learning identification method converges faster than the LMS method and the system identification processing can be performed in a short time. In the flowchart shown in FIG. 5, the method of converging the <C> filter and then the <B> filter is adopted. However, if a processor having a high processing speed is used, the same M-sequence random noise It is also possible to perform the convergence calculation for both the C> filter and the <B> filter.

FIG. 6 is a flow chart for explaining the details of the ANC processing shown in step SP2 in FIG. 4. First, in step SP2a, the ANC processing unit 37 is generated (initial setting) by software in the ANC controller 23. , The reference noise microphone 22 in step SP2b
To the reference signal Xn and the residual noise signal en from the residual noise microphone 19, and the compensation value <C> is calculated by the LMS method using the filter value C obtained by the system identification processing as a fixed filter value in step SP2c. In SP2d, the filter value B obtained by the system identification processing is used as a fixed filter value, and the compensation filter <B> is calculated by the LMS method.
The adaptive filter coefficient W based on the LMS method in P2e
Is calculated, the filter coefficient of the adaptive filter is corrected to the optimum value based on the LMS method in step SP2f, the canceling sound is generated from the loudspeaker 16 in step SP2g, and the process returns to step SP2b. Note that step S
The LMS method is adopted in P2e because the LMS method is simpler to calculate than the modified learning identification method, the order of the adaptive filter can be expanded, and the noise reduction performance can be improved.

FIG. 7 is a diagram showing an evaluation function routine executed in the calculation of the filter coefficient B and the filter coefficient C in the system identification processing and the calculation of the compensation filter coefficient B, the compensation filter coefficient C and the adaptive filter coefficient W in the ANC processing. . First, in step SP1, a filter coefficient is calculated by the modified learning identification method or the LMS method, and in step SP2, the sum of squares S (n) of filter coefficients in one loop, that is, S (n) = W ₁ ² + W ₂ ² + ... + W _n-1 ² + W _n ² is calculated, and the deviation β = S (n) −S ′ (from the sum of squares S ′ (n) obtained in the previous loop in step SP3. n) is calculated, and the deviation β is evaluated by comparing the deviation β with the reference value in step SP4, and the step S4 is performed.
In P5, it is determined whether or not the convergence coefficient needs to be switched. If it is determined that the convergence coefficient does not need to be switched, the convergence coefficient used in the previous loop is used as it is, and the evaluation process ends.

On the other hand, if it is determined in step SP5 that the convergence coefficient needs to be switched, step SP6.
In the above, the convergence coefficient for switching is appropriately switched based on the evaluation, the convergence coefficient having a value different from the convergence coefficient of the previous loop is set, and the evaluation routine processing is ended. The convergence coefficient switching process in FIG. 7 will be further described based on FIGS. 8 to 11. FIG. 8 is a diagram showing the weight value at each tap of the digital filter. The horizontal axis shows the number of taps of the digital filter and the vertical axis shows the weight value of the digital filter. As shown in FIG. 8, the weight value takes both positive and negative values. Assuming that the weight value at each tap at a certain point is as shown in FIG. 8, the weight value of each tap greatly varies (indicated by the swing width D) at the first stage of adaptive control, but at the final stage of adaptive control. Then, the weight value of each tap hardly changes (indicated by the swing width d).

In this convergence state, the sum ΣW of the filter coefficients of the digital filter becomes a certain constant value p as shown in FIG. However, since the weight value can take positive or negative values, the value fluctuates up and down around the convergence value p as shown by the solid line 60 in FIG. 9 in the process of convergence. In addition, when the sum of the filter coefficients is calculated, subtraction must be performed, which requires time for calculation, and when a CPU with low computing capacity is adopted, controllability and accuracy of control deteriorate. Therefore, considering the square value of the filter coefficient of the digital filter, each filter coefficient becomes a positive value as shown in FIG. 10, and the sum of squares of the filter coefficient also rises and falls above and below the convergence value q as shown in FIG. It will converge without swinging to. Thus, by taking the square sum of the filter coefficients, the convergence can be made faster than by simply taking the sum of the filter coefficients.
Since the subtraction does not have to be performed in the calculation of the total sum, the time required for the calculation can be shortened.

Further, the deviation β of the sum of squares S (n) of the filter coefficients of the digital filter at a certain point of time and the sum of squares S ′ (n) of the filter coefficients of the digital filters of the preceding loops is taken, and its magnitude is taken. The current state of convergence can be known by investigating. Therefore, if the deviation β is large, the convergence coefficient is increased to accelerate the convergence, and if the deviation β is within a predetermined range, the convergence coefficient is decreased to search for the optimum value.
It is possible to solve the problem that the LMS algorithm takes a long time to converge. In the calculation of the difference, in order to suppress the variation of β and obtain the integration effect, as shown in the evaluation function routine of FIG. 12, for example, the sum of products is summed over a cycle of an integer i times the filter order k. Sk (n) (shown in equation 4 and step SP2 of FIG. 12)

[0040]

(Equation 4)

And the sum Sk ′ (n) in the previous loop
It is also possible to employ a method of setting the convergence coefficient based on the difference β ′ between the two, that is, β ′ = Sk (n) −Sk ′ (n). The integer i times the filter order k means, for example, if the number of stages of the filter is 50, 50 × i (1, 2, 3, 4,
, N), which means that the sum of the loops of 50, 100, 150, 200, ... Is taken. In this way, it has been experimentally confirmed that the convergence is improved by taking an integer i times the filter order k. The reason for this is expected as follows. That is, the adaptive filter is FI
It is based on the R digital filter. The FIR digital filter takes the history of the input signal and shifts it to perform convolution integration with the filter coefficient. Paying attention to the history of this input signal, when the integer multiple of the filter order is taken, the sum of the integer multiples of the periodic signal is exactly the zero crossing point of the signal, and in the periodic signal, the integer multiple of the period is the whole. It seems that it is due to entering into the history of.

FIG. 13 is a diagram showing the configuration of the convergence coefficient setting unit 42 which sets the convergence coefficient by the evaluation function routine shown in FIG. 7 when the filter coefficient B, the filter coefficient C and the adaptive filter coefficient W are obtained by the LMS method. is there. The convergence coefficient setting unit 42 calculates the sum of squares of the filter coefficients of the digital filter unit 38, a holding unit 44 that holds the calculated sum of squares, and a difference that calculates the difference between the sums of squares. A calculation unit 45, a reference value storage unit 46 in which reference values are stored, a convergence coefficient determination unit 47 that determines a convergence coefficient based on the difference between the sum of squares calculated by the difference calculation unit 45 and the reference value,
It has a convergence coefficient control unit 48 which sends a command to extract the value of the filter coefficient of the digital filter unit 38 by the predetermined control signal and cause the square sum calculation unit 43 to calculate the square sum.

In the convergence coefficient setting unit 42 in the first loop, the square sum calculation unit 43 calculates the square sum, the holding unit 44 holds the value, and the difference calculation unit 45 calculates the difference in the second loop. That is, β = S (n) −S ′ (n) is calculated, and the convergence coefficient determination unit 47 converges the convergence coefficient (corresponding to a and μ in Expressions 1 to 3) based on the difference β. The value is determined to be an easy value, and adaptive control is performed based on the LMS method. When the evaluation function routine having the integration effect shown in FIG. 12 is adopted, the square sum calculation unit 43 and the convergence coefficient control unit 48 in FIG.
The sum-of-products addition unit and the convergence-coefficient control unit may be changed so as to perform cumulative addition in a loop of a multiple of i.

Further, the AN of the enclosed engine of this embodiment
The C process will be described with reference to FIG. When the engine is started, the radiator fan 11 also rotates and the cooling air indicated by the symbol M in FIG. 1 is generated. Further, a reference noise signal including engine sound, muffler transmission noise, fan wind noise, etc. is obtained from the reference noise microphone 22, and the residual noise signal is input from the residual noise microphone 19. And
A canceling sound is generated from the loudspeaker 16 to reduce all noises including the engine sound, the muffler transmission noise, and the fan wind noise in the duct 13. Here, as shown in FIG. 1, the shielding wall 14 is provided, the directional loudspeaker 16 is adopted, and the sound from the loudspeaker 16 is configured to be hard to enter the reference noise microphone 22. The amount of calculation of can be extremely reduced or omitted, and the adaptive speed and accuracy in silencing can be improved.

Since the reference noise microphone 22 is located downstream of the cylindrical muffler 12 and the radiator fan 6 in the flow of cooling air, not only engine noise but also
The muffler transmission noise and the wind noise of the radiator fan 11 can also be detected, and the muffler transmission noise and the wind noise of the radiator fan 11 can also be silenced by the loudspeaker 16.
Furthermore, since the cylindrical muffler 12 is provided on the downstream side of the radiator fan 11 at a position facing the suction air discharge surface, cooling air flows through the upper and lower side surfaces of the cylindrical muffler 12 to cool the muffler. . Cooling of this muffler reduces the frequency variation of muffler transmitted noise.
In the NC control, there is an advantage that the sound is easily muted.

Further, since the muffler 12 is not provided in the engine power generation chamber 7 but is provided in the ascending passage 8a, the temperature rise in the engine power generation chamber 7 can be suppressed. Further, since the engine 3, the generator 4, and the radiator fan 11 housed in the engine power generation chamber 7 are surrounded by the double soundproof wall, it is possible to reduce the leakage of noise to the outside. The noise at the outlet 10 where noise easily leaks can be almost completely silenced by ANC. Further, since the duct 13 for silencing the noise leaking from the outlet 10 is provided on the upper wall 2a of the soundproof case 2, there is an advantage that the width direction of the engine generator can be made compact.

[0047]

[Embodiment 2] FIG. 14 is a view showing a second embodiment of the present invention, and FIG. 14 (A) is a schematic vertical sectional view showing the structure of a noise reduction device for an enclosed engine generator, and FIG. 14 (B). FIG. 14C is a left side view, and FIG. 14C is a right side view with the exhaust duct case removed. The characteristic point of the fourth embodiment as compared with the first embodiment is that the structure of the enclosed engine is changed. In this enclosure type engine generator 51, a diesel engine 3 is arranged on the right side in the figure of a soundproof case 2 composed of six rectangular parallelepiped soundproof walls, and an engine generator 4 is arranged on the left side in the figure. is there. Further, the discharge port 10 is opened at a substantially central position of the right soundproof wall 2b, and the left soundproof wall 2
An intake port 24 is opened at a lower position of c.

An air guide plate 53 having an air outlet 52 is attached to the right soundproof wall 2b so as to cover the outlet 10 from above, and an air exhaust duct case 54 is provided on the right side so as to accommodate the air guide plate 53. It is attached to the soundproof wall 2b. A duct outlet 55 is provided on the upper right side of the exhaust duct case 54, and a wire mesh having a predetermined exhaust area is stretched over the duct outlet 55. Inside the baffle plate 53, a cylindrical muffler 12 extending in the horizontal direction is provided.
The intake port and the muffler 12 are communicated with each other by the first exhaust pipe 20a, and the exhaust gas passing through the muffler 12 is discharged to the outside from the side surface of the exhaust duct case 34 by the second exhaust pipe 20b. A wire mesh having a predetermined exhaust area is also stretched on the intake port 24.

The radiator 6 is disposed so as to face the discharge port 10, and the radiator 6 is provided with a radiator fan 11 which is driven by a fan belt in conjunction with the crankshaft. The radiator fan 11 is a fan that sucks out the air inside the soundproof case 2. Also,
Above the engine 3, an air cleaner 2 that communicates with an intake pipe
7 are provided. The enclosed engine generator 51 according to this embodiment includes an exhaust-side noise reduction device that reduces noise from the duct outlet 55. The exhaust-side noise reduction device includes a reference noise microphone 22 provided in an exhaust air guide chamber 56, which is a space surrounded by a baffle plate 53, on the downstream side of the cylindrical muffler 12, and an exhaust duct case 54. The loudspeaker 16 provided at the upper position of the soundproof wall 2b and the duct outlet 55
A residual noise microphone 19 provided at an upper end position;
It is composed of an NC controller (not shown).

Since the engine 3 and the generator 4 are housed in the soundproof case 2 via a vibration-proof rubber 57 arranged at a predetermined position on the bottom wall 2d of the soundproof case 2, the engine 3
Also, the vibration of the generator 4 is reduced from being transmitted to the soundproof case 2. In addition, the air intake 2 at the upper left of the soundproof case 2
A fuel tank 25 is provided at a position where the fuel cell 4 is not closed, and fuel can be supplied from a fuel supply port 26 provided in the soundproof case upper wall 2a. The operation of the engine generator noise reduction device configured as described above will be described. The system identification process is performed in the same manner as in the first embodiment, the compensation filter <C> and the compensation filter <B> are identified, and then the actual ANC process is performed. When the engine 3 of the engine generator 51 is operated and power generation is started, the radiator fan 11 that is interlocked with the crankshaft of the engine is driven to suck the outside air from the intake port 24. The sucked outside air flows through the soundproof case 2, cools the generator 4 and the engine 3, then flows through the exhaust air guide chamber 56 and the exhaust rising air passage 58 in the exhaust duct case 54, and is exhausted from the duct outlet 55. (See arrow M in the figure). The passage through which the ventilation air flows constitutes the ventilation section 21.

When the engine 3 starts operating, the exhaust ANC
The controller 23 uses the reference noise microphone to output the engine sound,
Noise mixed with fan wind noise and muffler transmitted noise is detected in the exhaust air guide chamber 56, the residual noise microphone 19 detects the residual noise, and the ANC controller determines the duct outlet 55 based on the reference noise signal and the residual noise signal. The loudspeaker 16 is driven to generate a canceling sound so that the noise of the noise is minimized. In the system identification process and the ANC process, since the wind guide plate 53 electrically and acoustically isolates the loudspeaker 16 and the reference noise microphone 22, howling correction during system identification and compensation during ANC process are performed as necessary. The calculation of the filter <B> can be omitted, the system identification can be simplified, and the noise reduction effect can be improved.

The muffler 12 is a radiator fan 11
Since it is provided on the further downstream side so as to face the radiator fan 11 and is not provided in the soundproof case 2, the muffler transmission noise can be silenced by the canceling sound of the loudspeaker 16, and the muffler 12 can be cooled. Muffler 12
It is possible to suppress the temperature rise in the soundproof case 2 due to the heat of. Further, since the duct for absorbing the noise leaking from the exhaust port 10 by riding on the cooling air is formed on the side wall of the soundproof case 2, the height of the engine generator can be made compact.

[0053]

Third Embodiment FIGS. 15 and 16 are views for explaining a third embodiment of the present invention. The third embodiment is different from the first embodiment only in that the evaluation function routine of FIG. 7 is replaced with the evaluation function routine shown in FIG. 15 and the configuration of the enclosed engine is replaced with the configuration of FIG. 19 described later. Is. This embodiment is characterized in that when the convergence coefficient is switched, the sum of the filter coefficients of all the taps is not calculated, but a part of the filter coefficient is representatively examined for the degree of convergence. FIG. 15 is a diagram showing an evaluation function routine performed in the calculation of the filter coefficient B and the filter coefficient C in the system identification processing and the calculation of the compensation filter coefficient B, the compensation filter coefficient C and the adaptive filter coefficient W in the ANC processing in this embodiment. is there.

First, in step SP1, the filter coefficient is calculated by the modified learning identification method or the LMS method, and it is determined in step SP2 whether or not the rough convergence flag is "1", and the rough convergence flag is "1". If not, the maximum value Wmax and the minimum value Wmin are searched in the memory where the filter coefficient is stored in step SP3, and Wmax and Wmin are searched in step SP4.
Whether the storage addresses of Wmax and Wmin have been fixed in step SP5. If it is determined that the storage addresses of Wmax and Wmin have been fixed, In step SP6, the general convergence flag is set to "1", and then step SP
Go to 7.

On the other hand, in step SP5, Wmax, Wm
If it is determined that the storage address of in is not fixed, the process returns to step SP1 to perform the filter operation by the adaptive algorithm. If it is determined in step SP2 that the roughly convergent flag is "1", then Wma
Since the storage addresses of x and Wmin are fixed, Wma
Since it is not necessary to search for x and Wmin, the process jumps to step SP7. Step In SP7 Wmax, the square sum Sf of Wmin, ^{i.e., Sf = W 2 max + W} 2 min to calculate a deviation between the square sum Sf 'sought in the previous loop at Step SP8 gamma = Sf-Sf 'Is calculated and the deviation γ is evaluated by comparing the deviation γ with a reference value in step SP9, and the step S
In P10, it is determined whether or not the convergence coefficient needs to be switched. If it is determined that the convergence coefficient does not need to be switched, the convergence coefficient used in the previous loop is used as it is, and the evaluation process ends. On the other hand, if it is determined in step SP10 that the convergence coefficient needs to be switched, the convergence coefficient for switching is appropriately switched based on the evaluation in step SP11 so that the convergence coefficient has a value different from that of the previous loop. Is set, and the evaluation routine process ends.

By checking the flag in step SP2, it is possible to detect whether or not the filter coefficients have roughly converged. When rough convergence is achieved, that is, the mutual addresses of Wmax and Wmin are Only when it is confirmed to some extent, subsequent search routine processing (step S
P7 to step SP11) can be performed. Also,
To determine whether the filter coefficients have roughly converged, use Wma
Since it is sufficient to search the storage address positions of x and Wmin,
It is possible to significantly reduce the calculation time of the substantial evaluation function.

The convergence coefficient switching process in FIG. 15 will be further described with reference to FIG. FIG. 16 is a diagram showing the weight value at each tap of the <B> filter 40 in FIG. 3, where the horizontal axis represents the tap position of the digital filter.
The vertical axis shows the weight value of the digital filter. However, the horizontal axis is the storage address on the memory. FIG.
As shown in, the weight value takes both positive and negative values. If the weight value at each tap at a certain point is as shown in FIG. 16, the weight value at each tap varies greatly at the first stage of adaptive control, and Wmax
The address position A1 of Wmin and the address position A2 of Wmin change greatly. In this state, it is meaningless to determine the convergence by taking the deviation β as described in the first embodiment. Therefore, after the storage addresses of Wmax and Wmin are fixed to some extent, the deviation γ is taken and the convergence coefficient is switched by evaluating the deviation γ.

That is, in this embodiment, when the respective filters of the <B> filter, the <C> filter, and the <W> filter converge to a certain degree, the shapes of the respective filter coefficients are almost fixed. By providing a process (processing of steps SP1 to SP6 in FIG. 15) in which the CPU determines whether or not the filter coefficients have converged to some extent, wasteful calculation time is reduced and substantial calculation time of the evaluation function is reduced. I try to reduce it. Also, instead of the sum of squares of the sum of squares of all filter coefficients, the sum of squares of Wmax and Wmin Sf = W ² max + W ² min
Since is used as the standard for the deviation γ of the evaluation function, the calculation time can be reduced.

[0059]

[Fourth Embodiment] FIG. 18 is a flow chart showing an evaluation function routine of an enclosed engine generator according to a fourth embodiment of the present invention. The fourth embodiment is different from the third embodiment only in that the evaluation function routine shown in FIG. 15 is replaced with the evaluation function routine shown in FIG. The characteristic point of this evaluation function routine is that in the configuration shown in FIG.
Although it is determined whether or not the shape of the filter coefficient is substantially determined based only on two values of max and Wmin, in the fourth embodiment, the maximum value and the minimum value are obtained for each zero-cross point region. The point is that it is determined whether or not the filter coefficient is stabilized in a certain shape, based on the degree of fixing the storage addresses of the maximum value and the minimum value. FIG.
8, the processing different from the flowchart shown in FIG. 15 is step SP3a, step SP4a, step SP
5a only.

The evaluation function routine of the second embodiment will be further described with reference to FIG. Figure 17 is <B
In the filter, it is a graph in which the horizontal axis represents the storage address and the vertical axis represents the weight value of the filter coefficient. As described above, the shape of the filter coefficient becomes substantially constant as the degree of convergence increases. Where the zero-cross point Z
, Z2, Z3, ... Zn regions E1, E2 ,.
In En, at least one of the maximum and minimum values is determined. Specifically, it takes the maximum value in the areas E1, E3, E5 ... And takes the minimum value in the areas E2, E4, E6. Area maximum value of each of these areas,
Whether the areas have converged roughly by searching the area minimum value and checking a plurality of storage addresses including the maximum value Wmax and the minimum value Wmin to determine whether or not the plurality of storage addresses are fixed. Whether or not it is determined. The number of areas in which the storage address is checked may be set appropriately in each device.
With this configuration, it takes more time to determine whether or not there is a substantial convergence, as compared with the above-described embodiment, but the accuracy of detecting the roughly converged state increases. When the number of stages of the filter is small, the time for determining whether or not the convergence is substantially completed can be shortened, and therefore the configuration of this embodiment is also advantageous depending on the configuration of the noise reduction device.

[0061]

Fifth Embodiment FIG. 19 is a diagram showing a specific configuration of an enclosed engine to which the ANC system of the third and fourth embodiments is applied, and FIG. 19 (A) shows an enclosed engine generator. 1 is a schematic vertical sectional view showing the configuration of a noise reduction device.
9 (B) is a perspective view of an essential part thereof. As shown in FIG. 19, in this enclosure type engine generator, a diesel engine 3 is arranged on the right side of the figure of a soundproof case 2 composed of six rectangular parallelepiped soundproof walls, and an engine generator 4 is arranged on the left side of the figure. Is provided. Further, the discharge port 10 is opened at a substantially central position of the right soundproof wall 2b, and the intake port 24 is opened at a lower position of the left soundproof wall 2c. Further, on the inner wall of the soundproof case 2, the exhaust duct case 71, the air intake duct case 67, and the inner wall of the horizontal duct 75, which will be described later, a silencer is attached on one surface to prevent high-frequency sound of 1 kHz or more from leaking to the outside. There is. As such a sound deadening material, a known lightweight and cheap sound deadening material such as glass wool is used.

The radiator 6 is provided in the soundproof case 2 with the discharge port 1
The radiator 6 is provided with a suction type radiator fan 11 that is driven by a fan belt in conjunction with the crankshaft. Further, a bag-shaped air guide plate 73 having an air outlet 72 is attached so as to cover the outlet 10 of the right soundproof wall 2b from above. The baffle plate 73 employs a curved surface portion 73a having an upper wall with a predetermined radius. A substantially rectangular bowl-shaped exhaust duct case 71 having an opening 74 of a predetermined size is attached to the upper wall 71a of the soundproof case right wall 2b, and a horizontal duct 75 is provided so as to cover the opening 74. 71a is attached in the horizontal direction. Lower wall 71c of the exhaust duct case
And the right wall 71b, the upper wall 75 of the horizontal duct 75
The corners connecting a and the right wall 75b are curved surface portions 71d and 75c having a smooth predetermined radius. Horizontal duct 7
At the duct outlet 35 of No. 5, a wire net 76 for preventing foreign matter from entering is stretched.

The exhaust duct case 71 is attached to the soundproof case right wall 2
By attaching to b, the noise emitted from the exhaust port 72 of the exhaust air guide chamber 79 in the baffle plate 73 causes the exhaust side ascending air passage 81 composed of the baffle plate 73 and the exhaust duct case right wall 71b. After passing through the opening 74 and the horizontal air passage 82 in the horizontal duct 75, the wire net 76 is discharged to the outside from the duct outlet 35 in which the wire net 76 is stretched (see arrow H in FIG. 19).
The horizontal duct 75 has a curved surface portion 75c that is not perpendicular to the extending direction of the exhaust side upward air passage 81, and can prevent a standing wave from being generated in the exhaust side upward air passage 81. Further, the reference noise microphone 22 is provided at the lower end of the wind guide plate 73, and the residual noise microphone 19 is attached to the horizontal duct 75.
It is provided at the outer upper end of the duct outlet 35. When the residual noise microphone 19 is provided at the outer upper end portion, the microphone surface is parallel to the extending direction of the horizontal duct 75 and does not protrude from the extension line of the horizontal duct 75 as shown in FIG. 19B. It is set. With this configuration, the cooling air discharged from the duct outlet 35 can be prevented from hitting directly, and the vibration of the residual noise microphone 19 can be prevented, and good noise detection can be performed.

Further, the position of the exhaust side speaker 16 is set on the upper wall 71a of the exhaust duct case 71, and the speaker surface of the speaker 16 is directed toward the curved surface portion 75c. As shown in FIG. 19, in the structure in which the residual noise microphone 19 is arranged outside the duct outlet 35, the structure in which the speaker 16 is arranged above the air guide plate 73 (see FIG. 14A) is If the speaker 47 is provided on the upper wall 71a,
It has been confirmed that the noise reduction performance of ANC is enhanced. The residual noise microphone 19 is attached to the outer upper end of the horizontal duct 75 via a shock absorbing material. Further, the length L in the depth direction of the exhaust duct case 71 in which the reference noise microphone 22 and the residual noise microphone 19 are arranged, and the horizontal duct 75 are the same (see FIG. 19B), and both are symmetrical. It is arranged in the shape. This configuration is necessary to enhance the sound coherence of both microphones 22 and 19. An intake side air guide plate 68 is provided on the intake side so as to cover the intake port 24 from below, and the intake duct case 67 is provided.
Is provided.

The ANC noise reduction device comprises the reference noise microphone 22, the speaker 16, the residual noise microphone 19 provided near the duct outlet 35, and an ANC controller (not shown). The detailed configuration of the ANC controller is as described in FIG. In the experiment using the actual machine, the noise level on the exhaust port 10 side was 85 dB when no countermeasure was taken, and after the countermeasure was taken, the noise absorption on the inner wall and the ANC control were combined. It was confirmed that it could be reduced to 75 dB.

The present invention is not limited to the above embodiments, and various design changes can be made within the scope of the present invention. Hereinafter, such an embodiment will be described. (1) In the above-described embodiment, the description has been made assuming that there are one reference noise microphone and one residual noise microphone, but one anti-noise speaker is used even when there are a plurality of reference noise microphones and residual noise microphones. The electroacoustic characteristics of and can be identified. Further, when the ANC silencer has a plurality of anti-noise speakers, it is possible to sequentially identify the electroacoustic characteristics by driving the anti-noise speakers one by one. In that case as well, the evaluation function of the present invention can be used. Routines can be adopted. (2) The present invention can be applied not only to the ANC algorithm in the time domain but also to the ANC algorithm in the frequency domain. (3) In the first and second embodiments, the muffler transmission noise is configured to be reduced by the ANC. However, if the exhaust gas of the muffler 12 is also discharged to the duct in the ventilation part,
Not only the muffler transmission noise but also the muffler sound emitted from the exhaust pipe 22b can be reduced by the ANC.

[Brief description of drawings]

1 is a diagram showing a configuration of a noise reduction device for an enclosed engine generator according to the present invention, FIG. 1 (A) is a vertical sectional view,
FIG. 1B is a cross-sectional view taken along the line BB of FIG.

FIG. 2 is a block diagram showing a configuration of a system identification unit.

FIG. 3 is a block diagram showing a configuration of an ANC processing unit.

FIG. 4 is a flowchart showing a basic process of the noise reduction device.

FIG. 5 is a flowchart showing a system identification process.

FIG. 6 is a flowchart showing ANC processing.

FIG. 7 is a flowchart showing an evaluation function routine.

FIG. 8 is a diagram showing filter coefficients of a digital filter.

FIG. 9 is a diagram showing how the addition value of filter coefficients converges.

FIG. 10 is a diagram showing a squared value of a filter coefficient of a digital filter.

FIG. 11 is a diagram showing how the sum of squared filter coefficients converges.

FIG. 12 is a flowchart showing an evaluation function routine.

FIG. 13 is a block diagram showing a detailed configuration of a convergence coefficient setting unit.

14 is a diagram showing a second embodiment of the present invention, and FIG.
4 (A) is a schematic vertical sectional view, FIG. 14 (B) is a left side view, and FIG. 14 (C) is a right side view with an exhaust duct case removed.

FIG. 15 is a flowchart illustrating a third embodiment of the present invention.

FIG. 16 is a diagram showing filter coefficients of a digital filter for explaining a third embodiment of the present invention.

FIG. 17 is a diagram showing filter coefficients of a digital filter for explaining a fourth embodiment of the present invention.

FIG. 18 is a flow chart for explaining the fourth embodiment of the present invention.

FIG. 19 is a diagram showing a specific configuration of an enclosed engine to which the ANC system of the third and fourth embodiments is applied, and FIG. 19 (A) is a noise of the enclosed engine generator. FIG. 19 is a schematic vertical sectional view showing the configuration of the reduction device.
(B) is a perspective view of the main part.

FIG. 20 is a block diagram showing an example of a conventional active silencer.

FIG. 21 is a diagram showing a noise spectrum of the enclosed engine.

[Explanation of symbols]

2 ... Soundproof case, 2b ... Soundproof case wall surface, 3 ... Engine, 5 ... Partition wall, 10 ... Discharge port, 11 ... Radiator fan, 12 ... Cylindrical muffler, 13a ... Duct outlet, 16 ...
Loudspeaker, 19 ... Residual noise microphone, 21 ... Ventilation part, 22 ... Reference noise microphone, 24 ... Intake port, 31 ... System identification part, 35 ... Duct outlet, 37 ... ANC processing part,
42 ... Convergence coefficient setting part, 53 ... Air guide plate, 54 ... Exhaust duct case, 55 ... Duct outlet, 71 ... Exhaust duct case, 73 ... Air guide plate, 75 ... Horizontal duct, 79 ... Exhaust air guide chamber, 81 ... Exhaust side upwind passage, 82 ... Horizontal air passage, μ ... Convergence coefficient.

Claims (7)

[Claims] 1. A soundproof wall (2,5) (2,53,54) surrounds the engine (3), and an outside air intake port is provided at a predetermined position of the soundproof wall (2,5,) (2,53,54). (24) and final outlets (13a) (55) are provided, and the outside air is taken in through the outside air intake (24), and the final outlets (13a) (5
A ventilation part (21) for exhausting ventilation air from the (5) is provided, and the final outlet (13a) from the engine (3) in the ventilation part (21).
A suction fan that sucks out the hot air of the engine (3) to the (55) side
(11) is provided, reference noise detection means (22) for detecting noise including engine sound and fan wind noise is provided at a predetermined position of the ventilation section (21), and the final discharge port (13a) of the ventilation section (21) ( 55) Ventilation with the residual noise detecting means (19) installed near the ventilation part (21) with the reference noise detecting means (22) provided downstream of the position of the ventilation part (21) Department (2
1) A canceling sound generating means (16) is provided at a predetermined position upstream of the position, and at least the canceling sound generating means (16) and the residual noise detecting means (19)
System identification means (31) for obtaining electroacoustic characteristics of the ventilation part (21) including electroacoustic characteristics based on the modified learning identification method
Is provided to detect the reference noise signal of the reference noise detecting means (22) and the residual noise signal of the residual noise detecting means (19), and to use the LMS method, and to detect the two detected noise signals and the system identification processing. ANC processing means for driving the canceling sound generating means (16) so as to generate canceling sound for canceling noise at the final outlets (13a) (55) based on the electroacoustic characteristics obtained by
(37) is provided, the noise reduction device for an enclosed engine. 2. The noise of the enclosed engine according to claim 1, further comprising convergence coefficient setting means (42) for setting the convergence coefficient (μ) in the modified learning identification method and the LMS method based on the sum of squared filter coefficients. Reduction device. 3. The sum of squares of filter coefficients is repeatedly added by an integer multiple of the filter order, the difference between the integer multiple added value and the previous integer multiple added value is obtained, and the convergence coefficient is set based on the difference. The noise reduction device for an enclosed engine according to claim 2, further comprising: convergence coefficient setting means. 4. A ventilation part (2) on the downstream side of the suction fan (11), to which suction air from the suction fan (11) is blown.
The muffler (12) of the engine (3) is installed at the position of (1), and the reference noise detecting means (22) is used to set the engine sound, the fan wind noise, and the muffler (1
The noise reduction device for an enclosed engine according to claim 1, wherein the noise reduction device is provided at the position of the ventilation part (21) where the transmitted noise from the chassis of (2) can be detected. 5. An engine (3) is housed in the enclosure (2), and at least a suction port (2) is provided at a predetermined position of the enclosure (2).
4) and an exhaust port (10) are provided, and at least the exhaust port (10) is provided in order to reduce the noise of the enclosed engine that takes in the outside air from the intake port (24) and exhausts it from the exhaust port (10).
The exhaust side duct (71, 75) communicating with the enclosure (2)
A primary noise detection microphone (22) that is formed around and detects noise near the exhaust port (10), and an exhaust side duct (71,
75) residual noise detection microphone (19) for detecting noise close to the duct opening (35), and exhaust side ducts (71, 75)
And a canceling sound generating means (16) disposed therein, and a canceling sound generating means (16) for generating a canceling sound for canceling noise at the duct opening (35) based on the primary noise signal and the residual noise signal. The exhaust side duct (71, 75) has a bag-shaped air guide plate (73) having an exhaust port (72) on the lower side so as to cover the exhaust port (10) from the upper side. ) And a rectangular parallelepiped having an opening (74) in an upper wall (71a) provided to accommodate a baffle plate (73) on a wall surface (2b) of an enclosure (2) having an exhaust port (10). It is composed of a bowl-shaped exhaust duct case (71) and a horizontal duct (75) which is partitioned by the upper wall (71a) from the exhaust duct case (71) and communicates with the opening (74) and which extends in the horizontal direction. The cooling air from the exhaust port (10) is guided by the baffle plate (73).
The exhaust side air duct (8) formed by the exhaust air guide chamber (79), the air guide plate (73) and the exhaust duct case (71) surrounded by
1), it is configured to be discharged from the duct port (35) through the horizontal air passage (82) in the horizontal duct (75), and the convergence coefficient (μ) of the ANC processing means (23) is set to the filter coefficient of 2 A noise reducing device for an enclosed engine, comprising: convergence coefficient setting means for setting based on a sum of multiplications. 6. A filter coefficient is stored in a memory, a maximum value and a minimum value of the filter coefficient are searched, a state in which the storage address of the maximum value and the minimum value is fixed is detected, and the convergence coefficient is switched. The noise reduction device for an enclosed engine according to claim 5, which is subsequent to the evaluation process of whether or not it is necessary. 7. The squared value of the maximum value and the minimum value is obtained respectively, and the two squared values are added to obtain a sum of squares of the filter coefficient. Surrounding engine noise reduction device.