JP2013011697A - Active type vibration noise suppression device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an active type vibration noise suppression device which can rapidly converge vibration or noise.SOLUTION: In an active type vibration noise suppression device, a sine-wave control signal yis constituted of a frequency of a source of vibration or noise, an amplitude filter coefficient aas an adaptive filter coefficient, and a phase filter coefficient φ, and includes a value multiplying the phase filter coefficient φby a value larger than 1, as a phase component in the sine-wave control signal y. Hence, the phase filter coefficient φcan be rapidly converged, and as a result, the vibration or the noise can be rapidly converged.

Description

  The present invention relates to an active vibration noise suppression apparatus that can actively suppress vibration or noise using adaptive control.

  Conventionally, there are devices described in Patent Documents 1 to 3 as devices that actively suppress vibration or noise using adaptive control. Patent Documents 1 to 3 describe a control method to which an LMS algorithm is applied as an adaptive control algorithm. In particular, Patent Documents 1 and 2 describe the DXHS algorithm in the Filtered-X LMS algorithm.

JP-A-8-44377 JP-A-8-272378 Japanese Patent Laid-Open No. 5-61483

  It is an object of the present invention to provide an active vibration noise suppression device that can converge vibration or noise earlier.

An active vibration noise suppression apparatus according to the present invention is an active vibration noise suppression apparatus that outputs control vibration or control sound and actively suppresses vibration or noise at an evaluation point. A control signal generating unit that generates a sine wave control signal y _(n) composed of a frequency, an amplitude filter coefficient as an adaptive filter coefficient, and a phase filter coefficient, and the control vibration corresponding to the sine wave control signal y _(n) Or a control vibration control sound generator that outputs a control sound, and a residual signal detector that detects a residual signal e _(n) due to interference between the vibration or noise generated by the generation source and the control vibration or control sound at the evaluation point; An amplitude filter for calculating an amplitude update term Δa _{(n + 1)} for adding and subtracting and updating the previously updated amplitude filter coefficient a _(n) so that the residual signal e _(n) becomes small A coefficient update unit and a phase update term Δφ _{(n + 1)} for adding and subtracting and updating the phase filter coefficient φ _(n) updated last time so that the residual signal e _(n) is reduced. And a sine wave control signal y _(n) including a value obtained by multiplying the phase filter coefficient φ _(n) by a value larger than 1 as a phase component.

According to the present invention, the phase component of the sine wave control signal y _(n), instead of using the updated phase filter coefficients phi _(n) as the value in the phase filter coefficient updating unit, phase filter coefficients phi _{(n )} Multiplied by a value greater than 1. That is, a value proportional to the updated phase filter coefficient φ _(n) is used. Thereby, at the evaluation point, the phase of the control vibration or the control sound can be converged so as to be in an early phase opposite to the phase of the vibration or noise transmitted from the generation source. As a result, the vibration or noise itself at the evaluation point can be converged early.

Further, the sine wave control signal y _(n) may be expressed by the equation (1). Thereby, the said effect can be show | played reliably.

The amplitude update term Δa _{(n + 1)} may include a term of equation (2), and the phase update term Δφ _{(n + 1)} may include a term of equation (3).

Thereby, in addition to the effects described above, that is, the vibration or noise itself at the evaluation point can be converged at an early stage, the following effects can be obtained. That is, when the sine wave control signal y _(n) is transmitted to the evaluation point via the transfer function G, the control vibration or control sound is related to the vibration or noise transmitted from the vibration source to the evaluation point. Even in such a phase, the control can be finally converged without diverging. The stabilization coefficient m may be 1 or a value larger than 1.

Then, the amplitude coefficient a1 and the phase coefficient φ1 in Expression (2) and the amplitude coefficient a2 and phase coefficient φ2 in Expression (3) are coefficients that do not depend on the transfer function G between the control signal generator and the evaluation point. can do. Even if the phase component of the periodic function in the amplitude update term Δa _{(n + 1)} and the phase update term Δφ _{(n + 1)} is set to a value irrelevant to the transfer function G, the control can be finally converged. Therefore, it is not necessary to identify the transfer function G, and it is not necessary to pursue identification accuracy. Therefore, the arithmetic processing can be simplified, and the arithmetic processing load can be reduced.

  Here, when the amplitude coefficient a2 and the phase coefficient φ2 are coefficients that do not depend on the transfer function G, Expression (2) is Expression (4), and Expression (3) is preferably Expression (5). . Thereby, control can be reliably converged.

On the other hand, the amplitude update term Δa _{(n + 1)} and the phase update term Δφ _{(n + 1)} may be terms using the transfer function G. In this case, the active vibration noise suppression device further includes an estimated transfer function storage unit that stores in advance an estimated value of a transfer function between the control signal generation unit and the evaluation point, and the amplitude coefficient in Expression (2) At least one of a1 and phase coefficient φ1 and at least one of amplitude coefficient a2 and phase coefficient φ2 in equation (3) are preferably coefficients obtained based on the estimated value of the transfer function.

In this case, the transfer function as the phase component [Phi _G and the phase component .phi.H _G of the estimated transfer function Gh becomes out of the range of -90 ° to 90 ° of G, control is eventually converge without diverging be able to. For example, even if the phase difference is 180 °, it can be converged. However, if the transfer function identification accuracy is high, it can be converged more quickly.

  When the estimated transfer function is used, equation (2) is equation (6), and equation (3) is preferably equation (7). Thereby, control can be reliably converged.

The stabilization coefficient m may be set to a value larger than 1, and the phase multiplication coefficient q may be set to a value larger than the stabilization coefficient m. Here, when q ≦ m, the amplitude filter coefficient a _(n) may be larger than the convergence value. That is, the amplitude filter coefficient a _(n) may overshoot. However, since the phase can be converged early by making the phase multiplication coefficient q larger than the stabilization coefficient m, the occurrence of overshoot can be suppressed.

Further, the initial value φ ₍₀₎ of the phase filter coefficient may be set based on the phase filter coefficient φ _(last) at the time of previous convergence. Here, the initial value φ ₍₀₎ of the phase filter coefficient will be described. The active vibration noise suppression apparatus performs adaptive control by updating the adaptive filter. If the vibration or noise of the generation source is constant, the control gradually converges by executing adaptive control. Then, the state where the vibration or noise at the evaluation point is small is maintained. At this time, the adaptive filter of the sine wave control signal y _(n) is also converged. However, in actual control, since the adaptive filter slightly fluctuates, it may not be a completely constant value. Therefore, it can be determined that the adaptive filter has converged when the adaptive filter is continuously included in the predetermined range in consideration of the variation error.

  In some cases, the sampling time n becomes 0 at the start of control, or the sampling time n becomes 0 in the calculation process. It is necessary to set the initial value of the adaptive filter coefficient when n = 0 as some value. Here, conventionally, generally, the initial value of the adaptive filter coefficient is set to zero, and the process of gradually adapting is performed.

On the other hand, as described above, the initial value φ ₍₀₎ of the phase filter coefficient as the adaptive filter coefficient is not simply set to zero, but based on the phase filter coefficient φ _(last) at the previous convergence. It is set. At this time, it is considered that the phase filter coefficient φ _(last) at the previous convergence is a value that takes into account the phase of the transfer function G at the previous convergence. That is, it can be said that the initial value φ ₍₀₎ of the phase filter coefficient follows the secular change of the current transfer function G. Therefore, it is possible to prevent the control from diverging and finally converge the control.

The initial value φ ₍₀₎ of the phase filter coefficient may be the phase filter coefficient φ _{(last) when} it has converged last time. Thereby, the initial value φ ₍₀₎ of the phase filter coefficient can be easily determined, and the above-described effect can be obtained.

The initial value φ ₍₀₎ of the phase filter coefficient is set based on the phase filter coefficient φ _{(last) at} the previous convergence, the frequency f at the previous convergence, and the current frequency f. May be. Thereby, even if the frequency f at the time of the previous convergence and the current frequency f change, the initial value φ ₍₀₎ of the phase filter coefficient according to the frequency f can be set. As a result, it can be converged early.

In this case, the initial value a ₍₀₎ of the amplitude filter coefficient may be set based on the current frequency f. Thereby, it can be made to converge at an early stage.

Further, the step size parameter of the amplitude update term Δa _{(n + 1)} may be set based on the amplitude filter coefficient a _(last) at the time of previous convergence. Conventionally, the step size parameter of the amplitude update term Δa _{(n + 1)} has conventionally been generally set to a constant value. On the other hand, according to the present invention, the step size parameter of the amplitude update term Δa _{(n + 1)} is set based on the amplitude filter coefficient a _(last) at the time of previous convergence, thereby allowing early convergence. Can do.

Also, the amplitude of the transfer function G may change over time, as with the phase. The amplitude filter coefficient a _(last) at the time of the previous convergence is considered to be a value considering the amplitude of the transfer function G at the time of the previous convergence. That is, it can be said that the step size parameter of the amplitude update term Δa _{(n + 1)} follows the current aging of the transfer function G. Therefore, it is possible to reliably converge at an early stage.

Further, the active vibration noise suppression device is applied to a vehicle having an engine, and the step size parameter of the amplitude update term Δa _{(n + 1)} is the amplitude filter coefficient a _(last) at the previous convergence, It may be set to a proportional value of a value (divided value) divided by the engine driving torque fluctuation amount trq _(last) .

Here, when the engine is used as a source of vibration or noise, the vibration or noise transmitted from the engine to the evaluation point is proportional to the engine driving torque fluctuation amount trq. That is, in the state where the control is converged, the amplitude of the transfer function G from the output of the control signal to the evaluation point is a value obtained by dividing the amplitude filter coefficient a _(last) by the drive torque fluctuation amount trq _(last). Corresponding to As described above, the amplitude filter coefficient a _(last) at the previous convergence is divided by the drive torque fluctuation amount trq _(last) at the previous convergence in the step size parameter of the amplitude update term Δa _{(n + 1).} By setting a proportional value of the obtained value, the step size parameter corresponds to the amplitude of the transfer function G when it converged last time. Therefore, it is possible to reliably converge at an early stage.

1 is a control block diagram showing an active vibration and noise suppression device. In the equation (48), when q = p = 3 and (3Φ _G −Φh _G ) = 0, the range of the phase filter coefficient φ (t) at each of n = 0 and 1 is shown. In the equation (48), when q = p = 3, a stable region and an unstable region when n = -1, 0, 1 are shown. The horizontal axis is (3Φ _G −Φh _G ), and the vertical axis is φ. An analysis result when the phase difference between the phase component Φ _G of the first transfer function _G and the phase component Φh _G of the first estimated transfer function Gh is 150 ° is shown. (A) shows a residual signal, (b) shows a sine wave control signal, (c) shows an amplitude filter coefficient, and (d) shows a phase filter coefficient. 2nd embodiment: It is a control block diagram which shows an active vibration noise suppression apparatus.

<First embodiment>
(1. Overview of active vibration and noise suppression device)
An outline of the active vibration noise suppression device 100 will be described. The active vibration noise suppression device 100 is a sine wave control signal for actively suppressing vibration or noise to be suppressed at a desired position (evaluation point) when various sources generate vibration or noise. It is a device that generates control vibration or control sound according to y _(n) . That is, by combining the control vibration or control sound with the vibration or noise to be suppressed, the control vibration or control sound acts to cancel the vibration or noise to be suppressed at a predetermined position (evaluation point). As a result, the vibration or noise to be suppressed is suppressed at the evaluation point.

  Here, an automobile will be described as an example. In automobiles, it is desirable that an engine (internal combustion engine) is a source of vibration noise, so that vibrations and noise generated by the engine are not transmitted to the passenger compartment. Therefore, in order to actively suppress vibration or noise generated by the engine, control vibration or control sound is generated by the generator. In the following description, the active vibration noise suppression device is applied to an automobile and described as an example of a device that suppresses vibration or noise generated by an engine, but is not limited thereto. The present invention can be applied to anything that generates vibration and noise to be suppressed.

The adaptive control algorithm by the active vibration noise suppression apparatus uses an adaptive least mean square filter (Filtered-X LMS), particularly the DXHS algorithm. In other words, the apparatus, an amplitude filter coefficients as the adaptive filter coefficients W _(n) a _(n), calculates the phase filter coefficients phi _(n), the filter coefficients a _(n), phi a and the engine _(n) Using the angular frequency ω of vibration or noise, a sine wave control signal y _(n) is generated, and a control vibration or control sound corresponding to the sine wave control signal y _(n) is output. Is a device that actively suppresses.

(2. Detailed configuration of active vibration and noise suppression device)
A detailed configuration of the active vibration noise suppression device 100 will be described with reference to FIG. As described above, the active vibration noise suppression apparatus 100 applies adaptive control using the DXHS algorithm. As shown in FIG. 1, the active vibration noise suppression device 100 is generated by the engine 10 when the engine 10 (described as “E / G” in FIG. 1) is driven by the engine control unit 30. When the vibration or noise to be suppressed is transmitted to the evaluation point 20 via the second transfer function H, the vibration or noise at the evaluation point 20 is reduced.

  The active vibration noise suppression device 100 includes a frequency calculation unit 110, a control signal generation unit 120, a generation device 130, a residual signal detection unit 140, and a first estimated transfer function setting unit (hereinafter referred to as “Gh data setting unit”). 150), an adaptive filter coefficient updating unit 160, and a parameter setting unit 170. In addition, although “^” meaning an estimated value is used in the mathematical expressions and drawings, the estimated value “hat (^)” is described as “h” in the text of the specification for convenience of description. Below, each structure of the active vibration noise suppression apparatus 100 is demonstrated.

  The frequency calculation unit 110 receives a periodic pulse signal from a rotation detector (not shown) that detects the number of revolutions of the engine 10, and based on the pulse signal, generates vibration or noise (suppression target) generated by the engine 10. The frequency f of the main component of vibration etc. is calculated. Note that the value obtained by multiplying the frequency f by 2π is the angular frequency ω. That is, the frequency calculation unit 110 can also calculate the angular frequency ω.

The control signal generation unit 120 generates a sine wave control signal y _(n) . The sine wave control signal y _(n) is expressed as shown in Equation (8). Here, the subscript _(n) is a subscript representing the sampling number (time step). In other words, as is clear from the equations (8) and (9), the sine wave control signal y _(n) includes the angular frequency ω, the amplitude filter coefficient a _(n) and the phase as the adaptive filter coefficient W _(n). This is a signal at time t _(n) including the filter coefficient φ _(n) as a component.

Here, the angular frequency ω in Equation (8) is a value calculated based on the frequency f calculated by the frequency calculation unit 110 or a value calculated by the frequency calculation unit 110. Therefore, the sine wave control signal y _(n) has a value corresponding to the frequency f of the main component of vibration or noise caused by the engine 10. Further, as shown in Expression (8), the value obtained by multiplying the phase filter coefficient φ (n) by the phase multiplication coefficient q larger than 1 is added to ωt _(n) as the phase of the periodic function (sin function). To do. Further, the amplitude filter coefficient a _(n) and the phase filter coefficient φ _(n) in Expression (8) are adaptive filter coefficients W _(n) in the DXHS algorithm as shown in Expression (9), and are adaptively updated. Is done.

The update formula of the adaptive filter coefficient W _(n) is expressed by formula (10). Thus, the adaptive filter coefficient W _{(n + 1)} is updated by adding / subtracting the update term ΔW _{(n + 1)} to the previous value W _(n) . The update term ΔW _{(n + 1)} is adaptively determined by the adaptive filter coefficient update unit 160 described later.

When this expression (10) is expressed by each of the amplitude filter coefficient a _(n) and the phase filter coefficient φ _(n) , the expressions (11) and (12) are obtained. That is, as shown in Expression (11), the amplitude filter coefficient a _{(n + 1)} is updated by adding / subtracting the amplitude update term Δa _{(n + 1)} to / from the previously updated amplitude filter coefficient a _(n) . Is done. Further, as shown in Expression (12), the phase filter coefficient φ _{(n + 1)} is updated by adding / subtracting the phase update term Δφ _{(n + 1)} to / from the previously updated phase filter coefficient φ _(n) . Is done.

Here, the initial value a ₍₀₎ of the amplitude filter coefficient and the phase filter coefficient φ ₍₀₎ are set by a parameter setting unit 170 described later. Here, the sampling time n may be reset by starting the engine 10. In this case, the case where the sampling time n becomes zero (0) is when the engine 10 is started. In addition, n = 0 when n reaches the maximum value n _(max) that n can take in the arithmetic processing.

The generator 130 is a device that actually generates vibrations and sounds. The generator 130 is driven based on the sine wave control signal y _(n) generated by the control signal generator 120. For example, the generation device 130 that generates control vibration is a vibration generation device that is disposed in a subframe such as a frame or a suspension member that is connected to a drive system of a vehicle. Moreover, the generator 130 which generates a control sound is a speaker etc. which are arrange | positioned in a vehicle interior, for example. In the case where the generator 130 is a device that generates a control vibration or control sound using magnetic force, such as a solenoid or a voice coil, for example, the current, voltage, or power supplied to the coil (not shown) is supplied at each time t. _By driving so as to correspond to the sine wave control signal y _(n) in _(n) , the generator 130 generates control vibration or control sound corresponding to the sine wave control signal y _(n) .

Then, at the evaluation point 20, the control vibration or control sound Z _(n) transmitted through the transmission system G _< b _> 1 by the control vibration or control sound generated by the generator 130 is the suppression target generated by the engine 10. Vibration or noise interferes with vibration noise X _(n) transmitted via the second transfer function H.

Therefore, the residual signal detection unit 140 is disposed at the evaluation point 20 and detects residual vibration or residual noise (corresponding to “residual signal” in the present invention) e _{(n) due} to interference at the evaluation point 20. This residual signal e _(n) is expressed by equation (13). For example, an acceleration sensor or the like can be applied as the residual signal detection unit 140 that detects residual vibration. In addition, a sound absorbing microphone or the like can be applied as the residual signal detection unit 140 that detects residual sound. The ideal state is that the residual signal e _(n) detected by the residual signal detector 140 becomes zero.

Here, the first transfer function G is a transfer function of the transfer system between the control signal generation unit 120 and the evaluation point 20. That is, the first transfer function G includes the transfer function of the generator 130 itself and the transfer function of the transfer system G1 between the generator 130 and the evaluation point 20. The first transfer function G is represented by an amplitude component A _G and a phase component Φ _G corresponding to the frequency f. The second transfer function H is a transfer function of the transfer system between the engine 10 and the evaluation point 20. That is, the second transfer function H is represented by the amplitude component A _H and the phase component Φ _H. If it does so, Formula (13) will be represented like Formula (14).

The Gh data setting unit 150 stores a first estimated transfer function Gh (estimated value of the first transfer function G) calculated by a known transfer function identification process. The first transfer function G is represented by an amplitude component A _G and a phase component Φ _G corresponding to the frequency f. Therefore, as shown in Expression (15), the first estimated transfer function Gh is represented by an amplitude component Ah _G and a phase component Φh _G corresponding to the frequency f. In Equation (15), the first estimated transfer function Gh, the amplitude component Ah _G, and the phase component Φh _G are in accordance with the frequency f. Gh (f), Ah _G (f) and Φh _G (f) are described.

Then, the Gh data setting unit 150 selects the first estimated transfer function Gh corresponding to the frequency f calculated by the frequency calculating unit 110 from the stored first estimated transfer function Gh. That is, the Gh data setting unit 150 determines the amplitude component Ah _G and the phase component Φh _G corresponding to the frequency f calculated by the frequency calculation unit 110.

The adaptive filter coefficient updating unit 160 (corresponding to the “amplitude filter coefficient updating unit” and the “phase filter coefficient updating unit” of the present invention ₎ controls the control signal generation unit 120 to reduce the residual signal e _(n) . An update term ΔW _{(n + 1)} for adding and subtracting the generated adaptive filter coefficient W _{(n )} is calculated. As described above, the adaptive filter coefficient W _(n) is composed of the amplitude filter coefficient a _(n) and the phase filter coefficient φ _(n) . That is, the adaptive filter coefficient updating unit 160 performs addition / subtraction for the amplitude update term Δa _{(n + 1)} for addition / subtraction to the amplitude filter coefficient a _(n) and the phase filter coefficient φ _(n) . The phase update term Δφ _{(n + 1)} is calculated.

The adaptive filter coefficient updating unit 160 calculates an update term ΔW _{(n + 1)} of the adaptive filter coefficient W _(n) so as to minimize the evaluation function J _(n) set based on the residual signal e _(n). To do. In order to apply the DXHS algorithm, the adaptive filter coefficient updating unit 160 calculates the frequency f (or angular frequency ω) calculated by the frequency calculating unit 110 and the Gh data when calculating the update term ΔW _{(n + 1).} The amplitude component Ah _G and the phase component Φh _G of the first estimated transfer function Gh determined by the setting unit 150 are used.

Hereinafter, how to derive the update term ΔW _{(n + 1)} of the adaptive filter coefficient W _(n) in the adaptive filter coefficient update unit 160 will be described. The evaluation function J _(n) is defined as in Expression (16). That is, the evaluation function J _(n) is the square of the residual signal e _(n) detected by the residual signal detection unit 140. That is, a sine wave control signal y _(n) that minimizes the evaluation function J _(n) is obtained.

Next, the gradient vector ▽ _(n) in the LMS algorithm is calculated according to the equation (17). The gradient vector ▽ _(n) is obtained by partial differentiation of the evaluation function J _(n) with the adaptive filter coefficient W _(n) . Then, the gradient vector ▽ _(n) is expressed as the right side of Expression (17).

Then, as shown in Expression (18), a term obtained by multiplying the calculated gradient vector ▽ _(n) by the step size parameter μ is set as an update term ΔW _{(n + 1)} .

From Expression (10) and Expression (18), the update expression of the adaptive filter coefficient W _(n) is expressed as Expression (19).

Here, the adaptive filter coefficient W _(n) is composed of the amplitude filter coefficient a _(n) and the phase filter coefficient φ _(n) as described above. Then, the amplitude component A _G and the phase component Φ _G of the first transfer function G in the equation (14) are replaced with the amplitude component Ah _G and the phase component Φh _G of the first estimated transfer function Gh, respectively, and the equation (19) Is calculated. Then, the amplitude update term Δa _{(n + 1 )} for the amplitude filter coefficient a _(n) is expressed as shown in Equation (20), and the phase update term Δφ _{(n + 1)} for the phase filter coefficient φ _{(n )} is It is expressed as equation (21). Here, (1 / Ah _G ) in Expression (20) is obtained by adding a normalization process to the update of the amplitude filter coefficient a _(n) .

Here, when the update term according to the equations (20) and (21) is applied, the difference between the phase component Φ _G of the first transfer function _G and the phase component Φh _{G of} the first estimated transfer function Gh is −90. The control can be converged if it is in the range of ° to 90 °. However, if the phase shift exceeds the range of −90 ° to 90 °, the control becomes unstable and the control may diverge. Therefore, Expression (20) and Expression (21) are replaced with Expression (22) and Expression (23). In other words, Equations (22) and (23) are obtained by deleting the phase multiplication coefficient q from the update terms of Equations (20) and (21). However, the phase multiplication coefficient q in the sine wave control signal y _(n) shown in the equation (8) is left as it is. By doing in this way, control can be converged without diverging. This theory will be described later.

Further, in order to ensure convergence, the expressions (22) and (23) are replaced with the expressions (24) and (25). That is, in each update term equation, the previously updated phase filter coefficient φ _(n) is divided by a stabilization coefficient m that is a value greater than one. However, the phase multiplication coefficient q is larger than the stabilization coefficient m. Further, in each update term equation, the phase component Φh _{G of} the first estimated transfer function Gh is divided by an arbitrary coefficient p. This coefficient p may include a number other than 1 including 1. By doing in this way, it can be made to converge reliably.

Then, the update formula for the amplitude filter coefficient a _(n) is expressed as shown in Expression (26), and the update expression for the phase filter coefficient φ _(n) is expressed as shown in Expression (27).

In this way, the adaptive filter coefficient updating unit 160 performs the amplitude update term Δa _{(n + 1)} for the amplitude filter coefficient a _(n) and the phase update term Δφ _{(n + 1)} for the phase filter coefficient φ _{(n ).} Is calculated. That is, the control signal generation unit 120 can adaptively update the sine wave control signal y _(n) using the update terms Δa _{(n + 1)} and Δφ _{(n + 1)} .

Parameter setting section 170 sets initial value W ₍₀₎ as adaptive filter coefficient W _(n) of sine wave control signal y ₍₀₎ in control signal generation section 120. Specifically, the parameter setting unit 170 sets the initial value a ₍₀₎ of the amplitude filter coefficient and the initial value φ ₍₀₎ of the phase filter coefficient. Here, the sampling time n = 0 at the initial values a ₍₀₎ and φ ₍₀₎ is, for example, when the engine 10 is started as described above. In this case, the sampling time n increases from when the engine 10 is started until the engine 10 is stopped. In addition, n = 0 when n reaches the maximum value n _(max) that n can take in the arithmetic processing.

First, the initial value a ₍₀₎ of the amplitude filter coefficient is set to ₀ as shown in Expression (28). The initial value φ ₍₀₎ of the phase filter coefficient is the phase filter coefficient φ _(last) at the time of previous convergence as shown in the equation (29). In Expression (29), the initial value φ ₍₀₎ is the phase filter coefficient φ _(last) at the previous convergence, but it may be a proportional value of the phase filter coefficient φ _(last) .

Here, the parameter setting unit 170 acquires and stores the phase filter coefficient φ _(last) used as the initial value φ ₍₀₎ as follows. The residual signal e _(n) detected by the residual signal detector 140 is always acquired, and it is determined whether or not the absolute value of the residual signal e _(n) is equal to or less than a set value close to zero, and the control is converged. Determine whether or not. Whether the control has converged is determined by the amplitude update term Δa _{(n + 1)} of the amplitude filter coefficient a _{(n )} or the phase update term of the phase filter coefficient φ _(n) in addition to the residual signal e _(n). It can also be determined by whether or not the absolute value of Δφ _{(n + 1)} is equal to or less than a set value close to zero. Then, the parameter setting unit 170 acquires the phase filter coefficient φ _(last) from the control signal generation unit 120 when it is determined that the control has converged. The parameter setting unit 170 stores the phase filter coefficient φ _(last) at the time of the previous convergence obtained in this way.

Here, in the above equation (28), the initial value a ₍₀₎ of the amplitude filter coefficient is always 0. In addition, the initial value a ₍₀₎ of the amplitude filter coefficient can be set to a value corresponding to the frequency f. In this case, for example, the parameter setting unit 170 stores a map related to the initial value a ₍₀₎ of the amplitude filter coefficient corresponding to the frequency f, and based on the current frequency f and information stored in the map. To set the initial value a ₍₀₎ of the amplitude filter coefficient. Thus, by setting the initial value a ₍₀₎ of the amplitude filter coefficient corresponding to the current frequency f, early convergence can be realized.

Further, in the above equation (29), the initial value φ ₍₀₎ of the phase filter coefficient is the phase filter coefficient φ _(last) at the previous convergence. In addition to this, the initial value φ ₍₀₎ of the phase filter coefficient can be set to a value corresponding to the frequency f. In this case, the initial value φ ₍₀₎ of the phase filter coefficient stores the phase filter coefficient φ _(last) at the previous convergence and the frequency f at the previous convergence, and the frequency f at the previous convergence is the current value f It is good also as a value which multiplied the phase filter coefficient ₍ phi _{) (last)} when it converged last time with the coefficient for converting into the frequency f of ₍₂₎ . For example, the parameter setting unit 170 stores a map related to the conversion coefficient (proportional coefficient) of the phase filter coefficient according to the frequency f, and multiplies the phase filter coefficient φ _(last) at the previous convergence by the conversion coefficient. Let the value be the initial value φ ₍₀₎ of the phase filter coefficient. Thus, by setting the initial value φ ₍₀₎ of the phase filter coefficient corresponding to the current frequency f, early convergence can be realized.

  As described above, when the parameter setting unit 170 sets each initial value based on the frequency f, information is transmitted from the frequency calculation unit 110 to the parameter setting unit 170 in FIG. That is, in FIG. 1, it is represented by a dashed arrow from the frequency calculation unit 110 to the parameter setting unit 170. In the embodiment in which the initial value not depending on the frequency f is set as described above, information is not transmitted from the frequency calculation unit 110 to the parameter setting unit 170. did.

(2. Explanation of theory)
Next, the theory for updating the amplitude update term Δa _{(n + 1 )} for the amplitude filter coefficient a _(n) and the phase update term Δφ _{(n + 1)} for the phase filter coefficient φ _(n) as described above. Give an explanation. The basic expression of the DXHS algorithm described above is expressed as Expression (30) when Expression (14), Expression (22), and Expression (23) are expressed in continuous time. However, as described above, in reality, not the expressions (22) and (23) but the expressions (24) and (25) are applied. In the following, (t) represents a function at time t. Further, it is defined as φ _imag (t) = q × φ (t), and μ _φ11 = q × μ _φ1 .

Here, in order to simplify the expression (30), the amplitude components A _G and Ah _G are omitted, and the update term da / dt is returned to before the normalization. Then, Expression (30) is expressed as Expression (31).

Considering the case where the control is stable, it is sufficient that e ² → 0 when t → ∞. That is, when t is sufficiently large, if the expression (32) is satisfied, the control is guaranteed to be stable.

  Therefore, from the first equation of equation (31), it is represented as equation (33).

  When the second and third expressions of Expression (31) are substituted into Expression (33) and expanded, Expression (34) is obtained.

  Here, when calculation is performed by substituting the first expression of Expression (31) into the first term on the right side of Expression (34), Expression (35) is obtained. Further, when the first expression of Expression (31) is substituted into the third term on the right side of Expression (34) and calculated, Expression (36) is obtained.

  Then, when Expression (35) and Expression (36) are added, Expression (37) is obtained.

  Here, when the second term on the right side of Expression (34) is calculated, Expression (38) is obtained. Further, when the fourth term on the right side of Expression (34) is calculated, Expression (39) is obtained.

  Then, when Expression (38) and Expression (39) are added, Expression (40) is obtained.

  From Expression (37) and Expression (40), Expression (34) is expressed as Expression (41).

  Here, in the equation (41), all except the fifth item are 2ω periodic functions. Since the control is converged when t → ∞, e (t), a (t), and φ (t) converge to a constant value when t → ∞. Then, when t → ∞, the sum of the periodic functions from the first term to the fourth term of Equation (41) must be a constant value. However, the constant value is not limited to zero, and may be a value other than zero.

  And in order to derive | lead-out said fixed value, the average value of each periodic function is calculated. The periodic functions from the first term to the fourth term in the equation (41) are all functions of 2ω, and assuming that time has passed sufficiently, the average value of each periodic function is zero. This can be derived from equation (42). Here, T represents a period.

  Therefore, the convergence value of the sum of the periodic functions from the first term to the fourth term of Equation (41) can be estimated as zero. Then, it is sufficient to consider only the fifth term in the equation (41). Therefore, from the fifth term of Expression (32) and Expression (41), in order to ensure stable control, it is necessary to satisfy at least the condition of Expression (43).

  Here, from the equation (44), the condition of the equation (45) is required.

  From Equation (45), the condition of Equation (46) is required. However, n is an integer.

  Since q> 1 and p is an arbitrary coefficient, Expression (46) can be expanded as Expression (47). That is, when t → ∞, by setting the phase filter coefficient φ (t) that satisfies the equation (47), the control does not diverge.

Here, in the case of a general DXHS algorithm different from the present embodiment, q = 1 in Expression (45). At this time, if the phase shift between the actual phase component Φ _G of the first transfer function _G and the phase component Φh _{G of} the first estimated transfer function Gh is in the range of −90 ° to 90 °, the equation is set to q = 1. The condition (45) is satisfied. That is, stability is guaranteed when the phase shift is in the range of −90 ° to 90 °. However, in the case of q = 1, control is diverged when it is outside the range of −90 ° to 90 °. On the other hand, in this embodiment, since q> 1, as described above, the control converges when Expression (47) is satisfied. For example, when q = 3 and p = 3, equation (47) becomes equation (48).

In Expression (48), when the transfer function phase difference (3Φ _G −Φh _G ) = 0, the range of the phase filter coefficient φ (t) is −135 ° to + 135 ° when n = 0, and n = 1. At + 405 ° to + 675 °, that is, + 45 ° to −45 °. When this range is shown in the range of −180 ° to + 180 °, it is as shown in FIG. That is, when n = 0, 1, the total phase filter coefficient φ (t) in the two cases of n = 0 and n = 1 can be selected over the entire 360 ° range. Further, when n = −1, it is −675 ° to −405 °, that is, + 45 ° to −45 °. That is, the same applies to the case where n = 0 and −1.

Next, consider a case where the transfer function phase difference (3Φ _G −Φh _G ) is other than zero. In this case, when n = -1, 0, 1 in equation (48), the respective stable regions and unstable regions are as shown in FIG. Based on this relationship, the range in which the phase filter coefficient φ (t) can be set is examined.

When q = p = 3, the phase filter coefficient φ (t) of n = 0, 1 can be selected anywhere of 360 °, and the transfer function phase difference −720 ° <(3Φ _G −Φh _G ) <+ 720 °. That is, when n = 0 and (3Φ _G −Φh _G ) = − 720 °, + 225 ° <φ (t) <+ 495 °. When n = 1 and (3Φ _G −Φh _G ) = − 720 °, + 765 ° <φ (t) <+ 1035 °. When n = 0 and (3Φ _G −Φh _G ) = + 720 °, −495 ° <φ (t) <− 225 °. When n = 1 and (3Φ _G −Φh _G ) = + 720 °, 45 ° <φ (t) <315 °. From these ranges, if −495 ° <φ (t) <+ 1035 °, the control does not diverge, that is, the control can converge. Further, when q = p = 3 and n = 0, −1, the control does not diverge if -1035 ° <φ (t) <+ 495 ° is set based on the same idea as described above.

(analysis)
Next, in the above embodiment, the phase difference between the phase component Φ _G of the first transfer function _G and the phase component Φh _{G of} the first estimated transfer function Gh is 180 °, and the phase multiplication coefficient q = 3, m Analysis was performed for the case of = 2. The analysis result is shown in FIG. As shown in FIG. 4A, it can be seen that the residual signal e has converged. The sine wave control signal y at this time is as shown in FIG. Then, as shown in FIGS. 4C and 4D, it can be seen that the amplitude filter coefficient a and the phase filter coefficient φ converge. In particular, as shown in FIG. 4D, it can be seen that the phase filter coefficient φ converges earlier than the amplitude filter coefficient a.

  Thus, even if the phase difference is 180 °, the control can be converged at an early stage without causing the control to diverge. Therefore, even if the phase component of the first transfer function G and the phase component of the first estimated transfer function Gh are shifted due to the change of the first transfer function G with temperature or aging, the control can be converged early. . Further, by using the first estimated transfer function Gh, when the accuracy of the first estimated transfer function Gh is high, it is possible to converge quickly. Further, the amplitude filter coefficient a converges without becoming a value larger than the convergence value. That is, the amplitude filter coefficient a does not overshoot. Therefore, it can suppress that a vibration becomes large until it converges.

Further, as described above, the initial value φ ₍₀₎ of the phase filter coefficient is not simply set to zero, but is set based on the phase filter coefficient φ _(last) at the time of previous convergence. At this time, it is considered that the phase filter coefficient φ _(last) at the previous convergence is a value that takes into account the phase of the transfer function G at the previous convergence. That is, it can be said that the initial value φ ₍₀₎ of the phase filter coefficient follows the secular change of the current transfer function G. Therefore, it is possible to prevent the control from diverging and finally converge the control.

Here, in the above embodiment, the initial value φ ₍₀₎ of the phase filter coefficient may be simply set to zero. Even if it does in this way, the effect that control can be converged without diverging can be produced. However, in this case, the effect of the early convergence is inferior to the case where the initial value φ ₍₀₎ of the phase filter coefficient is set based on the phase filter coefficient φ _(last) when the previous convergence is performed.

<Second embodiment>
The active vibration noise suppression apparatus 200 of this embodiment will be described with reference to FIG. In the first embodiment, the estimated value Gh of the first transfer function G is used in the update formula of the adaptive filter coefficient W. On the other hand, the active vibration noise suppression apparatus 200 of the present embodiment does not use the estimated value Gh of the first transfer function G in the update formula of the adaptive filter coefficient W.

  The active vibration noise suppression apparatus 200 includes a frequency calculation unit 110, a control signal generation unit 120, a generation device 130, a residual signal detection unit 140, an adaptive filter coefficient update unit 260, a torque fluctuation amount calculation unit 280, A parameter setting unit 270 is provided. Here, about the same structure as 1st embodiment, the same code | symbol is attached | subjected and description is abbreviate | omitted. That is, the active vibration noise suppression apparatus 200 of the present embodiment is different from the first embodiment in an adaptive filter coefficient update unit 260, a torque fluctuation amount calculation unit 280, and a parameter setting unit 270.

The adaptive filter coefficient update unit 260 (corresponding to the “amplitude filter coefficient update unit” and “phase filter coefficient update unit” of the present invention ₎ reduces the residual signal e _(n) as in the first embodiment. Then, an update term ΔW _{(n + 1)} for adding / subtracting to the adaptive filter coefficient W _(n) generated by the control signal generation unit 120 is calculated. The adaptive filter coefficient updating unit 260 adds an amplitude update term Δa _{(n + 1)} for addition / subtraction to the amplitude filter coefficient a _(n) and a phase for addition / subtraction to the phase filter coefficient φ _(n) . An update term Δφ _{(n + 1)} is calculated.

Here, the amplitude update term Δa _{(n + 1)} in the first embodiment is as shown in Expression (24). In Expression (24), (μ _a1 / Ah _G ) is set to a constant amplitude step size parameter μ _a2 that does not depend on the amplitude component Ah _G of the first estimated transfer function Gh, and (Φh _G / p) is set to zero. To do. Then, the amplitude update term Δa _{(n + 1)} is expressed as shown in Equation (49). That is, the amplitude update term Δa _{(n + 1)} represented by the equation (49) is an update term that does not depend on the first estimated transfer function Gh. Here, as will be described later, the amplitude step size parameter μ _a2 is changed by the parameter setting unit 270 using the amplitude filter coefficient a _(last) at the previous convergence.

Further, the phase update term Δφ _{(n + 1)} in the first embodiment is as shown in Expression (25). In Expression (25), (Φh _G / p) is set to zero. Then, the phase update term Δφ _{(n + 1)} is expressed as in Expression (50). That is, the phase update term Δφ _{(n + 1)} represented by the equation (50) is an update term that does not depend on the first estimated transfer function Gh.

Torque change amount calculation unit 280, the engine control unit 30 receives information about driving torque fluctuation amount trq of the engine 10 _(n), calculates the drive torque fluctuation trq of the engine 10 _(n). For example, the information regarding the drive torque fluctuation amount trq _(n) includes the drive torque fluctuation amount trq _(n) itself, the change amount of the accelerator opening, and the like.

As in the first embodiment, the parameter setting unit 270 sets the initial value a ₍₀₎ of the amplitude filter coefficient and the initial value φ ₍₀₎ of the phase filter coefficient as shown in equations (28) and (29). . Further, the parameter setting unit 270 uses the amplitude filter coefficient a _(last) at the previous convergence and the drive torque fluctuation amount trq _(last) at the previous convergence, and the amplitude step size according to the equation (51). Set the parameter μ _a2 . As shown in the equation (51), the amplitude step size parameter μ _a2 is a value obtained by dividing the amplitude filter coefficient a _(last) at the previous convergence by the drive torque fluctuation amount trq _(last). A value obtained by multiplying a constant step size parameter μ _a3 .

Here, as described in the first embodiment, the phase filter coefficient φ _(n) can converge control even if the entire 360 ° range is selected. Therefore, in the equations (24) and (25), even if the phase component Φh _G of the first estimated transfer function Gh is not used in the phase component of the periodic function (sin function or cos function), the phase filter coefficient φ _(n) Control does not diverge depending on the value. Therefore, even if an update equation that does not use the phase component Φh _G is used as in equations (49) and (50), control can be converged.

In the equation (49), the amplitude component Ah _G of the first estimated transfer function Gh is not used, but the amplitude filter when the amplitude step size parameter μ _{a2 of} the amplitude update term Δa _{(n + 1)} has converged last time is used. Based on the coefficient a _(last) . Here, it is considered that the amplitude filter coefficient a _(last) at the time of the previous convergence is a value considering the amplitude of the transfer function G at the time of the previous convergence. That is, it can be said that the amplitude step size parameter μ _a2 of the amplitude update term Δa _{(n + 1)} follows the aging of the current transfer function G.

In particular, as shown in the equation (51), the amplitude step size parameter μ _a2 is a proportional value of a value obtained by dividing the amplitude filter coefficient a _(last) at the previous convergence by the drive torque fluctuation amount trq _(last). I am trying to set it. Here, when the engine 10 is used as a vibration or noise generation source, the vibration or noise transmitted from the engine 10 to the evaluation point 20 is proportional to the drive torque fluctuation amount trq of the engine 10. That is, in the state where the control is converged, the amplitude of the first transfer function G corresponds to a value obtained by dividing the amplitude filter coefficient a _(last) by the drive torque fluctuation amount trq _(last) .

Then, as described above, the amplitude filter coefficient a _(last) when the previous convergence is performed on the amplitude step size parameter μ _a2 of the amplitude update term Δa _{(n + 1)} , the drive torque fluctuation amount trq ₍ By setting a proportional value of the value divided by “ _last” , the amplitude step size parameter μ _a2 corresponds to the amplitude of the first transfer function G when it converged last time. Therefore, even if Formula (49) is applied, the amplitude tracking performance is not inferior, and it can be surely converged early.

  As described above, even if the adaptive filter coefficient W is updated without using the first estimated transfer function Gh as in the second embodiment, the control can be reliably and quickly converged as in the first embodiment. it can. Furthermore, since the first estimated transfer function Gh is not used, it is not necessary to perform identification processing of the first transfer function G, and it is not necessary to pursue identification accuracy. Therefore, the arithmetic processing can be simplified, and the arithmetic processing load can be reduced.

Here, in the above embodiment, the amplitude step size parameter μ _a2 is set based on the amplitude filter coefficient a _(last) at the time of the previous convergence, but may be set to a constant value. Even if it does in this way, the effect that control can be converged without diverging can be produced. However, in this case, compared with the case where the amplitude step size parameter μ _a2 is set based on the amplitude filter coefficient a _(last) at the previous convergence, the amplitude tracking performance is inferior and the effect of early convergence is inferior. .

<Third embodiment>
In the first embodiment, the processing was performed by replacing the equations (22) and (23) with the equations (24) and (25). In addition, the update term Δa _{(n + 1)} of the amplitude filter coefficient a _{(n + 1)} and the update term Δφ _{(n + 1)} of the phase filter coefficient φ _(n) are _expressed by the equations (22) and (23) themselves. May be used. That is, this corresponds to the case where m = 1 in the equations (24) and (25). Also in this case, the control can be converged.

<Others>
In the above embodiment, the amplitude update term Δa _{(n + 1)} and the phase update term Δφ _{(n + 1)} are the update equations shown in equations (24), (25), and equations (49), (50). In addition, the amplitude update term Δa _{(n + 1)} and the phase update term Δφ _{(n + 1)} increase the convergence stability to the terms shown in the equations (24), (25) or (49), (50). For this reason, an expression for adding or subtracting another term may be used. That is, at least the amplitude update term Δa _{(n + 1)} and the phase update term Δφ _{(n + 1)} are update equations including the terms shown in the equations (24), (25) or (49), (50). The above effects can be exhibited.

10: vibration noise generation source (engine), 20: evaluation points 100, 200: active vibration noise suppression device 110: frequency calculation unit, 120: control signal generation unit, 130: generation device 140: residual signal detection unit, 150: First estimated transfer function setting unit 160, 260: adaptive filter coefficient updating unit, 170, 270: parameter setting unit 180: torque fluctuation amount calculating unit

Claims (14)

An active vibration noise suppression device that outputs control vibration or control sound and actively suppresses vibration or noise at an evaluation point,
A control signal generation unit that generates a sine wave control signal y _(n) composed of a frequency of a vibration or noise source, an amplitude filter coefficient as an adaptive filter coefficient, and a phase filter coefficient;
A control vibration control sound generator for outputting the control vibration or control sound according to the sine wave control signal y _(n) ;
A residual signal detector for detecting a residual signal e _(n) due to interference between the vibration or noise caused by the source and the control vibration or control sound at the evaluation point;
An amplitude filter coefficient for calculating an amplitude update term Δa _{(n + 1)} for adding and subtracting and updating the previously updated amplitude filter coefficient a _(n) so that the residual signal e _(n) becomes small. An update section;
Phase filter coefficient for calculating a phase update term Δφ _{(n + 1)} for adding and subtracting and updating the phase filter coefficient φ _(n) updated last time so that the residual signal e _(n) becomes small An update section;
With
The sine wave control signal y _(n) is an active vibration noise suppression device including a value obtained by multiplying the phase filter coefficient φ _(n) by a value larger than 1 as a phase component. In claim 1,
The sine wave control signal y _(n) is an active vibration noise suppression device represented by the formula (1).

In claim 1 or 2,
The amplitude update term Δa _{(n + 1)} includes the term of Equation (2),
The phase update term Δφ _{(n + 1)} is an active vibration noise suppression device including the term of the formula (3).

In claim 3,
The amplitude coefficient a1 and phase coefficient φ1 in equation (2) and the amplitude coefficient a2 and phase coefficient φ2 in equation (3) are coefficients that do not depend on the transfer function between the control signal generator and the evaluation point. Vibration noise suppression device. In claim 4,
Formula (2) is Formula (4),
Formula (3) is the active vibration noise suppression apparatus which is Formula (5).

In claim 3,
The active vibration noise suppression device further includes an estimated transfer function storage unit that stores in advance an estimated value of a transfer function between the control signal generation unit and the evaluation point,
At least one of the amplitude coefficient a1 and the phase coefficient φ1 in Expression (2) and at least one of the amplitude coefficient a2 and the phase coefficient φ2 in Expression (3) are active coefficients that are obtained based on the estimated value of the transfer function. Type vibration noise suppression device. In claim 6,
Formula (2) is Formula (6),
Formula (3) is the active vibration noise suppression apparatus which is Formula (7).

In any one of Claims 3-7,
The stabilization factor m is set to a value greater than 1,
The active vibration noise suppression device in which the phase multiplication coefficient q is set to a value larger than the stabilization coefficient m. In any one of Claims 1-8,
The active vibration noise suppression device in which the initial value φ ₍₀₎ of the phase filter coefficient is set based on the phase filter coefficient φ _(last) when it converged last time. In claim 9,
The active vibration noise suppression device in which the initial value φ ₍₀₎ of the phase filter coefficient is the phase filter coefficient φ _(last) when it has converged last time. In claim 9,
The initial value φ ₍₀₎ of the phase filter coefficient is the active vibration noise set based on the phase filter coefficient φ _{(last) at} the previous convergence, the frequency f at the previous convergence, and the current frequency f. Suppression device. In any one of Claims 9-11,
The active vibration noise suppression device in which the initial value a ₍₀₎ of the amplitude filter coefficient is set based on the current frequency f. In any one of Claims 9-12,
The active vibration noise suppression device in which the step size parameter of the amplitude update term Δa _{(n + 1)} is set based on the amplitude filter coefficient a _(last) when it converged last time. In claim 13,
The active vibration noise suppression device is applied to a vehicle having an engine,
The step size parameter of the amplitude update term Δa _{(n + 1)} is a proportional value obtained by dividing the amplitude filter coefficient a _(last) at the previous convergence by the engine driving torque fluctuation amount trq _(last). Active vibration noise suppression device to be set.

Images