阅读说明：本技术 一种用于民航客机客舱噪声的主动控制方法 (Active control method for passenger cabin noise of civil aviation passenger plane ) 是由 何法江 任浩杰 于 2021-06-25 设计创作，主要内容包括：本发明属于自动控制的技术领域,公开了一种用于民航客机客舱噪声的主动控制方法,其特征在于：包括通过设置于客舱内不同位置的拾音器采集客舱内的初级声源；以飞机本身产生的噪声作为输入、对应客舱位置的噪声作为输出对BP神经网络进行训练,获取BP神经网络模型；以民航客机的噪声控制方法产生的次级声源作为训练好的BP神经网络模型的输入,产生补偿噪声信号；以所述初级声源和补偿噪声信号的差值作为噪声控制方法的输入,对客舱内人员所在位置进行主动降噪。(The invention belongs to the technical field of automatic control, and discloses an active control method for passenger cabin noise of a civil aviation passenger plane, which is characterized by comprising the following steps of: collecting primary sound sources in a passenger cabin through sound collectors arranged at different positions in the passenger cabin; training a BP neural network by taking noise generated by the airplane as input and noise corresponding to a cabin position as output to obtain a BP neural network model; a secondary sound source generated by a noise control method of a civil aviation passenger plane is used as the input of a trained BP neural network model to generate a compensation noise signal; and taking the difference value of the primary sound source and the compensation noise signal as the input of a noise control method, and actively reducing the noise of the position of the personnel in the passenger cabin.)

1. An active control method for passenger cabin noise of civil aircraft, characterized by comprising:

collecting primary sound sources in a passenger cabin through sound pickups arranged at different positions in the passenger cabin;

training a BP neural network by taking noise generated by the airplane as input and noise corresponding to a cabin position as output to obtain a BP neural network model;

a secondary sound source generated by a noise control method of a civil aviation passenger plane is used as the input of a trained BP neural network model to generate a compensation noise signal;

and taking the difference value of the primary sound source and the compensation noise signal as the input of a noise control method, and actively reducing the noise of the position of the personnel in the passenger cabin.

2. Active control method for the cabin noise of civil airliners, as in claim 1, characterized in that: the noise control method is set as an active noise control method using the FXLMS algorithm.

3. Active control method for the cabin noise of civil airliners, as in claim 2, characterized in that: recording the primary sound source as X (n), filtering the primary sound source by a trained BP neural network to obtain an input signal as X' (n), and acquiring an error signal as e (n) by an error signal acquisition device;

the weights of the adjustment function W (z) at the nth time and the input signal are expressed in the form of vectors, i.e.

W(n)＝[w₁(n)，w₂(n)，...，w_L(n)]^T

X′(n)＝[x(n)，x(n-1)，...，x(n-L-1)]^T

And (3) adjusting the weight:

W(n+1)＝W(n)+μe(n)X(n)

where μ is the step size factor of the FXLMS algorithm,a represents a constant, b represents a gain coefficient, μ₀Representing the dc component of the step factor.

4. Active control method for the cabin noise of civil airliners, as in claim 3, characterized in that: the noise generated by the aircraft itself includes engine noise, turbulent boundary layer noise, and cabin interior air conditioning noise.

Technical Field

The invention belongs to the technical field of automatic control, and particularly relates to an active control method for passenger cabin noise of a civil aviation passenger plane.

Background

The rapid development of national economy drives the rapid development of the civil aviation industry, and more people select airplanes as travel vehicles. In addition to speed, people pay more and more attention to comfort when going out, but the noise of the passenger cabin directly influences the comfort of passengers in the cabin, and is one of the problems which plague the development of the civil aviation industry. The passenger cabin has good noise active control, the competitiveness is greatly improved, and how to effectively control the noise in the cabin is still the problem to be solved in the design process of the airplane.

The noise in the passenger cabin mainly comprises engine noise, turbulent boundary layer noise, air conditioning noise in the passenger cabin and the like, and the engine noise mainly comprises fan noise, combustion chamber noise, turbine noise and jet noise. Currently, aircraft mainly adopt a passive noise control method: the inner wall of the passenger cabin adopts sound insulation and absorption structure and material; optimizing the aerodynamic shape of the airplane; the passive noise control method has a certain effect on high-frequency noise, but has an unobvious control effect on medium-low frequency noise generated by an engine and the like, and the noise in a passenger cabin is still strong. Meanwhile, under different working conditions, the generated noise characteristic changes greatly, and the action effect of the passive noise control method is very limited. In the prior art, if the noise in the passenger cabin is to be controlled continuously, a huge cost is generated, and the noise control effect is limited.

Although some aircraft design houses start to perform active cabin noise control, a feedforward ANC system is generally adopted, and a secondary path effect exists, namely, a cancellation sound wave emitted by a secondary sound source is propagated to a reference microphone through air, so that a signal received by the reference microphone contains not only a noise source signal but also a secondary sound source signal. Although the identification of the secondary path is realized based on the FX-LMS algorithm, the algorithm assumes that the cabin interior environment is approximately linear acoustic environment, but the actual cabin interior noise environment is not linear and directly affects the noise control effect.

Disclosure of Invention

The invention provides an active control method for passenger cabin noise of a civil aviation passenger plane, which adopts a BP neural network model to replace an adaptive filter of a secondary sound source path in the traditional FXLMS algorithm, can accurately estimate the actual noise condition of the secondary sound source transmitted to a passenger cabin seat, better conforms to the nonlinear state of actual noise, and has higher prediction accuracy and better noise reduction effect.

The invention can be realized by the following technical scheme:

an active control method for passenger cabin noise of civil aircraft, comprising:

collecting primary sound sources in a passenger cabin through sound pickups arranged at different positions in the passenger cabin;

training a BP neural network by taking noise generated by the airplane as input and noise corresponding to a cabin position as output to obtain a BP neural network model;

a secondary sound source generated by a noise control method of a civil aviation passenger plane is used as the input of a trained BP neural network model to generate a compensation noise signal;

and taking the difference value of the primary sound source and the compensation noise signal as the input of a noise control method, and actively reducing the noise of the position of the personnel in the passenger cabin.

Further, the noise control method is set as an active noise control method using the FXLMS algorithm.

Further, the primary sound source is recorded as X (n), an input signal obtained by filtering through a trained BP neural network is X' (n), and an error signal obtained by acquiring through an error signal acquirer is e (n);

the weights of the adjustment function W (z) at the nth time and the input signal are expressed in the form of vectors, i.e.

W(n)＝[w₁(n)，w₂(n)，，w_L(n)]^T

X′(n)＝[x(n)，x(n-1)，...，x(n-L-1)]^T

And (3) adjusting the weight:

W(n+1)＝W(n)+μe(n)X(n)

where μ is the step size factor of the FXLMS algorithm,a denotes a constant, b denotes a gain coefficient, and μ 0 denotes a direct current component of the step factor.

Further, the noise generated by the aircraft itself includes engine noise, turbulent boundary layer noise, and cabin interior air conditioning noise.

The beneficial technical effects of the invention are as follows:

compared with the traditional FXLMS adaptive algorithm based on a linear predictor, the FXLMS adaptive algorithm based on the BP neural network has the advantages that the prediction of a secondary sound source is more accurate, the noise reduction effect is better, the control of engine noise, turbulent boundary layer noise and air conditioner noise in a passenger cabin can be better finished, the personnel communication is not influenced, meanwhile, the Sigmoid function is introduced to improve the LMS algorithm, the anti-interference capability is stronger, and the convergence is easier.

Drawings

FIG. 1 is a schematic overview of the process of the present invention.

Detailed Description

The following detailed description of the preferred embodiments will be made with reference to the accompanying drawings.

With the improvement of living standard and industrialization standard, the noise problem gradually becomes a key problem concerned in daily life of people. The control of noise is mainly divided into traditional noise control and active noise control, the effect of the commonly used traditional noise control method is obviously reduced in a low-frequency noise environment, and the problem is better solved due to the active noise control method.

A sound field (secondary sound field) with the same intensity and opposite phase with the sound to be eliminated is established in the sound field area to be controlled, and the destructive interference of the sound field is artificially caused by utilizing the wave interference principle, so that the noise is eliminated. After the 70 s, with the rapid development of acoustic theory and control theory, people gradually had a deeper understanding of the mechanism of active noise reduction control. In view of the fact that most of common noise sources have strong time-varying characteristics, after the adaptive theory becomes mature, people begin to explore a new problem of adaptive active noise cancellation control (AANC) which has great application value, and the AANC automatically adjusts the secondary sound signal by using a proper adaptive algorithm to ensure that the secondary sound signal can effectively track and cancel the noise signal, thereby achieving the purpose of eliminating the noise.

In practical application scenarios, since the secondary sound source signal propagates back through the secondary path, the pollution noise source signal cannot be directly acquired by the reference input microphone, which means that the LMS algorithm cannot be directly applied to the ANC system. The filtering X-LMS algorithm, FX-LMS, is currently used to model the propagation process of the sound waves generated by the secondary sound source to the noise collection to eliminate the influence of the secondary path effect. The method is realized on the basis that the cabin noise environment is approximately regarded as a linear acoustic environment, but the actual cabin noise environment is nonlinear, so that the method uses the BP neural network to replace an adaptive algorithm for realizing the secondary path identification function, the prediction of the secondary sound source is more accurate, and the noise reduction effect is better.

As shown in fig. 1, the present invention provides an active control method for cabin noise of a civil aircraft, comprising the following steps:

collecting primary sound sources in a passenger cabin through sound pickups arranged at different positions in the passenger cabin;

according to actual needs, placing noise source sound pickups and error sound pickups at various positions of a passenger cabin, wherein the noise source sound pickups are used for collecting primary sound sources x (n) in the aircraft cabin, including human sounds and aircraft noises, wherein the aircraft noises mainly include engine noises, turbulent boundary layer noises, air conditioner noises inside the passenger cabin and the like under different working conditions, and x (n) is a matrix formed by sampling main noises at n discrete times, and the acoustic properties of the matrix include amplitude, phase and frequency; the error pickup is used for acquiring a noise residual signal e (n) after the primary sound source and the secondary sound source are cancelled, namely, noise residual after passenger cabin noise control, namely noise which can be felt by passengers in a seat, wherein e (n) is a matrix formed by sampling the residual noise at n discrete times, and acoustic properties of the matrix comprise amplitude, phase and frequency. In the noise reduction method, the sound pickup is arranged in an area needing to control the noise level in the cabin in the aircraft cabin and is used for monitoring the sound source characteristics in the aircraft cabin and feeding the result back to the active noise reduction controller.

Training a BP neural network by taking the noise generated by the airplane as input and the noise corresponding to the cabin position as output to obtain a BP neural network model;

because the secondary sound source is calculated according to the noise generated by the airplane, the secondary sound source is used for offsetting the noise generated by the airplane, and therefore, the ideal state is that the two are equal, but the actual situation is difficult to achieve, when a BP neural network is established, noise signals generated by the airplane under different working conditions, such as engine noise, turbulent boundary layer noise, cabin interior air conditioning noise and the like, can be collected by a reference microphone and processed by a digital-to-analog converter as input, and the noise signals collected at the passenger cabin seat are processed by the digital-to-analog converter as output, so that the BP neural network is trained by taking the noise signals as a training data set, specifically as follows:

(1) initializing neural network parameters w0, b 0;

(2) the collected engine noise, turbulence boundary layer noise and air conditioner noise inside the passenger cabin form a training set input matrix, and the collected passenger cabin seat noise forms a training expected output matrix Y0;

(3) calculating a prediction output matrix Y in a forward propagation process;

(4) calculating a loss value L according to the loss function, and judging whether the loss value meets convergence;

(5) if the result converges, the training is finished, if the result does not converge, the back propagation is carried out to adjust the neural network parameters w and b, and the steps (3), (4) and (5) are repeatedly executed.

Step three, taking a secondary sound source generated by a noise control method of the civil aircraft as the input of a trained BP neural network model to generate a compensation noise signal;

and step four, taking the difference value of the primary sound source and the compensation noise signal as the input of the noise control method, and actively reducing the noise of the positions of the personnel in the passenger cabin.

The FXLMS algorithm is adopted as an active noise control method, wherein the conventional LMS algorithm is weak in anti-interference capability and difficult to converge, and the Sigmoid function is insensitive to larger input, so that the FXLMS algorithm introduces the Sigmoid function and carries out certain transformation to control the step size factor in the LMS algorithm, and the method specifically comprises the following steps:

marking a primary sound source as X (n), filtering the primary sound source by a trained BP neural network to obtain an input signal as X' (n), and acquiring an error signal as e (n) by an error signal acquisition device such as an error sound pickup;

the weights of the adjustment function W (z) at the nth time and the input signal are expressed in the form of vectors, i.e.

W(n)＝[w₁(n)，W₂(n)，...，w_L(n)]^T

X′(n)＝[x(n)，x(n-1)，...，x(n-L-1)]^T

And (3) adjusting the weight:

W(n+1)＝W(n)+μe(n)X(n)

where μ is the step size factor of the FXLMS algorithm,a denotes a constant, b denotes a gain coefficient, and μ 0 denotes a direct current component of the step factor.

Parameters a, b and mu 0 are introduced, and the step-size factors are controlled through the parameters to improve the anti-interference capability and promote convergence, wherein the parameter a is used for reducing steady state imbalance, and a is properly reduced when the system imbalance is severe; the parameter b is used for scaling the error e (n), and the step size factor is not sensitive to the change of the error e (n) and can be properly adjusted to be larger than b; mu 0 is a direct current component of the step factor, and the function is to prevent the step factor from being infinitely close to 0 and losing the adjusting effect when the error e (n) is close to 0, namely, the output signal y (n) of the active noise control system sends a cancellation signal to the passenger cabin seat through the secondary sound source to cancel the noise transmitted to the seat, thereby completing the noise reduction processing.

Although specific embodiments of the present invention have been described above, it will be appreciated by those skilled in the art that these are merely examples and that many variations or modifications may be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is therefore defined by the appended claims.