JPH07104770A - Active vibration controller - Google Patents

Abstract

PURPOSE:To provide an active vibration controller rapidly converging even to a random noise and with excellent estimation precision of a transfer function. CONSTITUTION:A reference signal u(n) detected by a noise senser 2 is supplied to an adaptive digital filter 3 and an FIR filter 4. The output b(n) of the FIR filter 4 is supplied to an LSL processing part 5. The output a (n) of the adaptive digital filter 3 is supplied to a speaker 6, and is supplied to the FIR filter 7. The output from the speaker 6 is interferred with a noise, and the interference result is detected as an error signal e (n) by an error microphone 8 mounted on a control point as an offset error to be supplied to an adder 9, and is added to the output yLAMBDAM(n) of the FIR filter 7 supplied to others by the adder 9, and the addition result yM(n) is supplied to the LSL processing part 5. The operation is performed using the supplied b(n) and yM(n) by the LSL processing part 5, and the filter coefficient of the adaptive digital filter 3 is updated.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an active vibration control system for actively controlling and reducing vibrations caused by running of a vehicle and noises caused by these vibrations.

[0002]

2. Description of the Related Art The objects to be controlled in the present invention are "vibration" and "noise", but for convenience of description, they are limited to "noise", and "vibration" includes "noise".

Conventionally, this type of active vibration control device is generally designed to reduce noise by attenuating noise generated from a noise source by canceling noise. Since the canceling signal for generating the canceling sound must be generated in real time, L is an adaptive identification algorithm from the viewpoint of the amount of calculation.
The MS (Least Mean Square) method is used. The basic principle is that the transfer function from the noise source to the control point is FIR
By approximating with a model and performing system identification by the LMS method, vibration of opposite phase is generated from an actuator such as a speaker with respect to noise, and noise is suppressed.

When actually performing active vibration control using this LMS method, from the speaker used as the secondary sound source (sound source for canceling sound output) to the error microphone (microphone for detecting canceling error) installed at the control point. Since it is necessary to consider the transfer function including the sound field characteristics, the Filtered-x LMS method proposed by B. Widrow is often used. Hereinafter, an active vibration control device using the filtered X LMS method will be described with reference to FIG. In addition, 1 input-1 output-1 point control (Single Input Single Output Sin
gle Point-control).

FIG. 8 is a block diagram of an application example in which the filtered X LMS method is applied to an active vibration control device.

In FIG. 8, noise from a noise source 101 is detected by a noise sensor 102 and supplied as a reference signal u (n) to a filter 103 and a filter 104 having a transfer function C ^ (z). Generally, FIR filters are often used as the filters 103 and 104. The output b (n) of the filter 104, that is, the filtering result of the reference signal u (n) by the filter 104 is supplied to the LMS processing unit 105 that updates the filter coefficient of the filter 103 by the LMS algorithm. Further, the reference signal u (n) is supplied to the speaker 106 as the canceling signal a (n) by the filter 103, and the canceling sound generated from the speaker 106 and the noise from the noise source 101 interfere with each other,
The result (cancellation error) is detected by the error microphone 107, and the LMS processing unit 10 outputs the error signal e (n).
5 is supplied. In the LMS processing unit 105, the signal b
Using (n) and the error signal e (n), the optimum filter coefficient of the filter 103 is calculated by the LMS method.

The transfer function C ^ (z) identifies the transfer characteristic from the speaker 106 to the error microphone 107. The reason why the FIR filter having this transfer function C ^ (z) is inserted between the noise sensor 102 and the LMS processing unit 105 is as follows.

In order to effectively perform active vibration control, the transfer function H (z) of the filter 103 is P (z) / C (z).
Need to be designed so that That is, when the filter 103 is designed in this way, the transfer function P (z) from the noise source 101 to the error microphone 107 and the transfer function (H (z )
This is because × C (z) = P (z)) is the same. However, since the transfer function C (z) is generally a non-minimum phase system, it is not possible to simply obtain a transfer function of 1 / C (z). So the filtered X LMS method
The transfer function C (z) is previously identified (the identification result is C ^ (z)
By using this, it is possible to sequentially estimate 1 / C (z) as H (z) = P (z) / C (z) without independently obtaining it.

[0009]

However, when the conventional active vibration control system using the filtered X LMS method is applied to, for example, an automobile, the conventional active vibration control system is effective against pseudo periodic noise such as engine muffled noise. Was very effective, but with respect to random noise such as road noise, there was a case where convergence was slow and sufficient noise reduction was not always possible.

The present invention has been made in view of the above problems, and it is an object of the present invention to provide an active vibration control device which converges quickly even against random noise and is excellent in transfer function estimation accuracy.

[0011]

In order to achieve the above object, the present invention inputs a signal having a high correlation with vibration from a vibration source as a reference signal, and filters the reference signal using an adaptive filter. Control means for generating a canceling signal by the canceling vibration generating means for generating a canceling vibration based on the canceling signal generated by the controlling means, and a canceling vibration generated by the canceling vibration generating means and a vibration from the vibration source. Error signal detecting means for detecting a canceling error between the canceling vibration generating means, the canceling vibration generating means and the error signal detecting means, first filter means for filtering the reference signal, and the canceling vibration generating means. To the error signal detection means, second transfer means for filtering the cancellation signal from the control means, and the second transfer means are identified. And a summing means for adding the result filtered by the filter means to the error signal from the error signal detecting means and inputting the addition result to the control means, wherein the control means is the least squares as an algorithm for generating the cancellation signal. It is characterized by using the lattice algorithm.

[0012]

The canceling signal from the control means is filtered by the second filter means, and the filtering result is added to the error signal from the error signal detecting means by the adding means. Further, the reference signal is filtered by the first filter means, and the filtering result and the addition result are input to the control means. In the control means, the canceling signal is generated by the least square lattice algorithm based on the input two signals, and the canceling vibration is generated by the canceling vibration generating means.

[0013]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing a schematic configuration of an application example in which an active vibration control device according to an embodiment of the present invention is applied to vehicle interior noise control. The main purpose of this embodiment is to reduce road noise.

In FIG. 1, reference numeral 1 denotes a noise input source for generating road noise such as a car suspension, which is installed in the vicinity of the noise input source 1 by a sensor 2 for detecting a signal highly correlated with the road noise. The detected noise signal is digitally converted by an A / D converter (not shown), and F having a transfer function H (z) is provided as a reference signal u (n).
IR type adaptive digital filter 3 and transfer function C ^
FIR digital filter having (z) (hereinafter referred to as “F
"IR filter"). The output b (n) of the FIR filter 4, that is, the reference signal u (n) filtered by the FIR filter 4 will be described in detail later, but a least-squares lattice (Least Squares Lattice: hereinafter referred to as “LSL”) algorithm. LSL processing unit 5 which determines the filter coefficient of adaptive digital filter 3 by
And the filter coefficient of the adaptive digital filter 3 is updated by the calculation of the LSL processing unit 5.

Further, the cancellation signal a (n) which is the output of the adaptive digital filter 3 is branched and one is a D / A not shown.
It is converted into an analog signal through a converter and is supplied to a speaker 6 for outputting a canceling sound in accordance with this signal, and the other one is a FI having a transfer function C ^ (z).
It is supplied to the R filter 7. The canceling sound output from the speaker 6 interferes with the noise from the noise source 1, and the interference result is detected as a canceling error by the error microphone 8 installed at the control point, and the detected signal is not shown. The error signal e (n) is digitally converted by the A / D converter. The error signal e (n) is 1 of the adder 9
Is supplied to one input terminal, and the output y ^ M (n) of the FIR filter 7 is supplied to the other input terminal of the adder 9. Both signals e (n) and y ^ M (n) are added to the adder. Added by 9,
The addition result yM (n) is supplied to the LSL processing unit 5. The LSL processing unit 5 outputs the output b of the FIR filter 4.
Based on (n) and the addition result yM (n), the optimum filter coefficient is calculated by the LSL algorithm, and the filter coefficient of the adaptive digital filter 3 is updated. The transfer function C ^ (Z) identifies the transfer characteristic from the speaker 6 to the error microphone 8.

Here, the LSL algorithm has the advantages of both the least squares method and the lattice algorithm as described below.

(1) Fast convergence (effects of both least square method and lattice algorithm).

(2) There is no variation around the true value of the estimated value (effect of the least squares method).

However, the noise added to the observed output is white noise.

(3) Strong against rounding error (effect of lattice algorithm).

Then, this LSL algorithm is used for FIR
By using the model as an algorithm for estimating,
It has become possible to apply the filtered X LSL method in which the MS method algorithm is replaced with the LSL algorithm to the active vibration control device.

The reason for this will be described below with reference to a block diagram equivalent to the block diagram of FIG. 1 shown in FIG.

The LSL algorithm is based on the unknown plant 10
It requires an output signal yp (n) (assuming this transfer function is P (z)). Here, the unknown plant 10 refers to the noise input source 1 to the error microphone 8 as shown in FIG. 1, and the transfer function P (z) represents its transfer characteristic. On the other hand, the observed signal is the error microphone 8
Is the error signal e (n) detected by the error signal e (n), and it is necessary to estimate the output signal yp (n) using this error signal e (n).

The error signal e (n) is expressed by the following equation.

E (n) = yp (n) -y ^ p (n) Here, in y ^ p (n), the cancellation signal a (n) is the transfer function C (z).
Is a signal when it reaches the error microphone via.

Therefore, the filter 7 generates a signal y ^ M (n) equivalent to the signal y ^ p (n), and the adder 9 produces the error signal e (n) and the signal y ^ M (n). ) And, the signal yM (n) equivalent to the signal yp (n) can be obtained.

Next, the calculation procedure by the LSL algorithm will be described. The calculation procedure described below is based on a paper previously published by the present inventors (Adachi, Ishida, Sano: Least-squares adaptive lattice equalizer with automatic adjustment of weighting coefficient, IEICE Transactions, J68-A-9,992 / 995 (1985)), the detailed description of the LSL algorithm will be omitted.

First, the symbols in the formulas of the LSL algorithm described later will be explained.

[0030]

[Equation 1] Here, “n” indicates that it is sampled at time n (n = 0, 1, ...) And “m” indicates FIR.
It is shown that the transfer function H (z) described in the model is at the m-th stage of the order N.

Next, initial values (m = 0, ..., N-1) are set as follows.

[0032]

[Equation 2] However, δ is a small positive number.

Then, at each time n (n = 0, 1, ...), the following equation is calculated for the orders m = 0 ,.

[0034]

[Equation 3]

[0035]

[Equation 4] FIG. 3 and FIG. 4 are actual configurations of the algorithms represented by the above formulas (1) to (15).

The results of a simulation experiment for comparing the convergence speed and the identification accuracy of the LMS method and the LSL method are shown in FIGS.

FIG. 5 shows the results of simulations using a normal white signal as an input signal (reference signal), and (a) and (b) show the output error and the parameter estimation result at that time, respectively. Shows. FIGS. 6 and 7 show simulation results using an artificially generated colored signal and a suspension acceleration signal when traveling on a rough pavement where road noise is generated as an input signal, respectively. , (B) respectively show the output error and the parameter estimation result at that time. As can be seen from these results, the LSL method has a faster convergence speed than the LMS method even when the input signal has strong chromaticity.
Excellent identification accuracy. The details of this simulation experiment are described in the paper presented at the academic conference of the present application, and thus the description thereof is omitted.

[0038]

As described above, according to the present invention,
Control means for inputting a signal having a high correlation with vibration from a vibration source as a reference signal, and for generating a canceling signal by filtering the reference signal using an adaptive filter, and a canceling signal generated by the controlling means A canceling vibration generating means for generating a canceling vibration based on the above, an error signal detecting means for detecting a canceling error between the canceling vibration generated by the canceling vibration generating means and the vibration from the vibration source, and the canceling vibration generating means The transfer characteristic up to the error signal detecting means is identified, the transfer characteristic between the first filter means for filtering the reference signal and the canceling vibration generating means to the error signal detecting means is identified, and the transfer characteristic from the control means is determined. Second filter means for filtering the cancellation signal, and the error signal detection result obtained by filtering by the second filter means. And an addition means for adding the addition result to the control means and inputting the addition result to the control means, and the control means uses the least-squares lattice algorithm as an algorithm for generating the cancellation signal. Also, the effect is that convergence is fast and control with excellent estimation accuracy can be performed.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of an application example in which an active vibration control device according to an embodiment of the present invention is applied to vehicle interior noise control of an automobile.

FIG. 2 is a diagram showing a block diagram equivalent to the block diagram of FIG.

FIG. 3 is a circuit diagram which constitutes an LSL algorithm of the present embodiment.

4 is a circuit diagram showing a detailed configuration of a stage m in the circuit of FIG.

FIG. 5 is a diagram showing a result of simulation using a normal white signal as an input signal.

FIG. 6 is a diagram showing a result of simulation using a color signal artificially generated as an input signal.

FIG. 7 is a diagram showing a result of simulation performed by using a suspension acceleration signal during traveling on a rough paved road where road noise is generated as an input signal.

FIG. 8 is a block diagram of an application example in which a conventional filtered X LMS method is applied to an active vibration control device.

[Explanation of symbols]

 3 Adaptive Digital Filter (Control Means) 4 FIR Filter (First Filter Means) 5 LSL Processing Unit (Control Means) 6 Speaker (Cancellation Vibration Generation Means) 7 FIR Filter (Second Filter Means) 8 Error Microphone (Error Signal) Detecting means) 9 Adder (adding means)

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Hideki Kasuya 2753 Ishii-cho, Utsunomiya City, Tochigi Prefecture Utsunomiya University Faculty of Engineering (72) Inventor Shuichi Adachi 2753 Ishii-cho, Utsunomiya City Tochigi Prefecture Utsunomiya University Faculty of Engineering

Claims (1)

[Claims] 1. A control means for inputting a signal having a high correlation with a vibration from a vibration source as a reference signal and filtering the reference signal with an adaptive filter to generate a cancellation signal, and the control means. Canceling vibration generating means for generating canceling vibration based on the generated canceling signal, error signal detecting means for detecting a canceling error between the canceling vibration generated by the canceling vibration generating means and the vibration from the vibration source, A transfer characteristic from the canceling vibration generating means to the error signal detecting means is identified, a first filter means for filtering the reference signal, and a transfer characteristic from the canceling vibration generating means to the error signal detecting means are identified, Second filtering means for filtering the cancellation signal from the control means, and a result of filtering by the second filtering means And an addition means for adding the addition result to the error signal from the error signal detection means and inputting the addition result to the control means, wherein the control means uses a least square lattice algorithm as an algorithm for generating the cancellation signal. A characteristic active vibration control device.