A method for deploying active noise control sensor actuators in a helicopter cabin

Abstract

This invention discloses a method for deploying sensors and actuators for active noise control in helicopter cabins. It introduces machine learning into the sensor / actuator optimization problem in active noise control of helicopter cabins, proposing a BPNN-GA algorithm, which mainly includes two processes: sensor optimization based on neural networks and actuator optimization based on genetic algorithms. First, the radiation surface is divided into elements; then, machine learning is used to optimize the location and number of sensors, specifically by using a neural network to establish a high-precision roof vibration-radiated noise model and identifying the main sound radiation elements based on this model, which are then used as the sensor placement points; finally, a genetic algorithm is used to optimize the actuators, aiming to ensure that vibration at the control point is effectively suppressed while minimizing cabin noise levels. Compared with traditional genetic algorithms, this application significantly improves computational efficiency while maintaining control effectiveness, and its advantage in improving computational efficiency becomes more pronounced as the optimization scale increases.

Description

Technical Field

[0001] This invention belongs to the field of preferred methods for helicopter cabin noise control sensors / actuators, specifically relating to a preferred algorithm for active control of helicopter cabin noise sensors / actuators. Background Technology

[0002] Cockpit noise in helicopters is a serious problem. Severe cabin noise negatively impacts the physical and mental health of passengers and crew, and the structural vibrations caused by noise also affect the helicopter's performance and lifespan. Mid-to-high frequency harmonic noise generated by the meshing vibration of the main gearbox gears is of particular concern, with a frequency range of approximately 500-2000Hz, which is within the range sensitive to human hearing. Studies show that the vibration of the main gearbox is transmitted to the cabin walls through the supporting structure, exciting the wall panels to vibrate and radiating noise into the cabin. This structural sound transmission is the primary source of cabin noise in helicopters. Therefore, reducing cabin noise caused by this structural sound transmission is crucial.

[0003] Active and passive control technologies are two common noise control techniques in helicopter noise control. Passive control essentially involves installing specific energy dissipation devices at specific locations on the structure. These devices utilize vibration isolation and absorption techniques to convert energy, thereby reducing the energy absorbed by the main structure and achieving the goal of reducing vibration and noise. However, using traditional passive control technologies to control low- and mid-frequency noise results in large and heavy systems, leading to increased fuel consumption and adversely affecting helicopter performance. Furthermore, passive control technologies are often unable to adapt to changes in helicopter speed, making it difficult to achieve the desired noise suppression effect.

[0004] Active control technologies are divided into Active Sound Cancellation (ASC) and Active Structural Acoustic Control (ASAC). While ASC can address low-frequency cabin noise, achieving global noise control for high-frequency gear meshing noise is impractical due to the high modal density of the acoustic cavity. ASAC, through active control algorithms, uses actuators installed along the vibration transmission path to control mechanical vibrations, reducing radiated noise through vibration damping. For example, ASAC trials have been conducted on helicopters such as the EH-101, BK-117, and Bell-407, achieving effective cabin noise control through actuators mounted on the main suspension struts. However, for helicopters without main suspension struts, such as the S-72 and UH-60, ASAC technology based on active struts is no longer applicable.

[0005] As the primary transmission path and direct source of noise within the helicopter cabin, the wall panels' acoustic and vibration characteristics directly impact the cabin's acoustic environment. The main gearbox is directly connected to the ceiling panel via a support structure. The ceiling panel, as a critical forced vibration area, is a major source of radiated noise; reducing its vibration level can effectively decrease radiated noise and prevent vibration energy from being transmitted to other wall panels within the cabin. With the continuous development of smart materials technology, ASAC technology based on intelligent active wall panels has become the most valuable and promising helicopter cabin noise control technology. Summary of the Invention

[0006] To address the aforementioned shortcomings, this invention provides a method for deploying active noise control sensor actuators in helicopter cabins. The actuators and sensors are integrated with the helicopter roof panel to form an intelligent active wall panel that senses and controls the vibration of the roof panel in real time. Compared with traditional genetic algorithms, this method significantly reduces computation time while ensuring control effectiveness, greatly improving computational efficiency. Furthermore, as the optimization scale increases, the advantages of this algorithm in improving computational efficiency will become even more pronounced.

[0007] Technical solution:

[0008] A method for active noise control sensor and actuator deployment in helicopter cabins is proposed, which introduces machine learning into the sensor / actuator optimization problem in active noise control of helicopter cabins. A machine learning-based intelligent optimization algorithm—BPNN-GA (Back Propagation Neural Network-Genetic Algorithm)—is proposed. This algorithm mainly includes two processes: sensor optimization based on neural networks and actuator optimization based on genetic algorithms. Machine learning is used to optimize the location and number of sensors, specifically by using neural networks to establish a high-precision roof vibration-radiated noise model and identifying the main sound radiation elements based on this model, which are then used as sensor placement points. The genetic algorithm is used to optimize the actuators.

[0009] Specifically, it includes the following steps:

[0010] Step 1: Divide the helicopter cabin top plate into several radial surface elements based on the actuator specifications, and arrange an acceleration sensor and an MFC actuator at the center of the radial surface element.

[0011] Step 2: Taking the noise inside the helicopter cabin as the control object, establish a multi-line spectrum complex neural network model. Use the cabin wall vibration response obtained by the accelerometer and the cabin noise response obtained by the microphone as the input and output of the multi-line spectrum complex neural network model, respectively, and identify the main sound radiation surface elements as the preferred placement points of the accelerometer.

[0012] Step 3: Based on the determination of the preferred placement points of the accelerometer, with the goal of reducing the noise level of the monitoring points, a genetic algorithm is used to optimize the placement points of the actuator, and finally the preferred placement points of the MFC actuator are obtained.

[0013] Preferably, step 2 includes:

[0014] Step 2.1: Establish a multi-line spectrum complex neural network model. For each line frequency of the wall panel vibration, an independent sub-model is established. At the same time, the sub-model retains the amplitude and phase information of the vibration noise response.

[0015] Step 2.2: Correct the network weight coefficients based on the "backpropagation" algorithm;

[0016] Step 2.3: Repeat the gradient solution and weight coefficient update in Step 2.2 until the preset maximum number of iterations is reached. Based on the relationship between the input and output of the neural network, obtain the transfer matrix between the wall panel vibration and the radiated noise, so as to calculate the radiated noise contribution of each surface element and then identify the main sound radiation surface elements.

[0017] Preferably, in step 2.1, for a neural network containing an input layer, H hidden layers, and an output layer, the input vector of the BP neural network is X = [x1, x2, ..., x...]. M ] T It characterizes the vibration response of each radiating surface element of the top plate at ω = ω0, where M is the total number of radiating surface elements; for the m-th radiating surface element, its response is... and These represent the real and imaginary parts of the response, respectively, containing the amplitude and phase information of the vibration; the target vector of the BP neural network is Y = [y1, y2, ..., y]. N ] T That is, the noise response of N monitoring points inside the cabin at ω = ω0; similarly, for the nth monitoring point, its noise response is... Therefore, the relationship between the input and output of the neural network can be obtained as follows:

[0018]

[0019] In the formula, W is the actual output of the neural network. h Let h be the weight coefficient matrix of the h-th layer. This is the transfer matrix between panel vibration and radiated noise.

[0020] Preferably, in step 2.1, certain constraints are imposed on the bias matrix and activation function: all bias terms are set to 0, and the activation function is a linear activation function.

[0021] Preferably, in step 2.2, the chain rule is used for "backpropagation," and the mean squared error is used as the loss function. The gradient of the loss function with respect to each parameter is calculated. Based on the gradient descent method, the weight coefficients between each network layer are corrected along the negative gradient direction of the loss function. The real and imaginary parts are simplified by combining the applied constraints. At the same time, a learning rate η is given to obtain the updated weight coefficients, expressed as:

[0022]

[0023] In the formula, For u j conjugate, For the loss function E in u i The gradient at that point.

[0024] Preferably, in step 2.3, the transfer matrix between panel vibration and radiated noise is obtained based on the relationship between the neural network input and output in step 2.1. Calculate the radiated noise generated by M surface elements at the monitoring point. The radiated noise is converted onto a complex plane for marking and superposition to obtain the sum of the vibration responses of each radiating surface element, i.e., the total response vector. Based on the total response vector, the contribution of each surface element to the overall radiated noise is calculated, and the comprehensive contribution of each surface element is further determined.

[0025] Preferably, in step 2.3, the overall contribution of the m-th element is:

[0026]

[0027] In the formula, For frequency ω k The radiated noise contribution of the m-th surface element at the n-th monitoring point.

[0028] Phase weighting coefficients The symbolic function form is used to determine the contribution of radiated noise energy.

[0029] As a preferred method, to quantify the contribution of each surface element to the total radiated noise energy, the energy contribution rate of each radiating surface element is calculated based on the following formula:

[0030]

[0031] A threshold is set, and starting from the surface element with the largest contribution, the energy contribution rate of each surface element is accumulated sequentially until the cumulative contribution rate reaches or exceeds the set threshold. Surface elements whose energy contribution rate is accumulated to or exceeds the set threshold during this process are defined as the main sound radiation surface elements.

[0032] As a preferred embodiment, in step 3, firstly, a set of actuator deployment schemes is randomly generated as the initial population; then, based on the optimization objective and constraints, a fitness function is set to quantitatively evaluate the performance of each deployment scheme; finally, through selection, crossover, and mutation operations, the individuals in the population are iteratively optimized step by step until the termination condition is met.

[0033] As a preferred approach, a multi-line spectrum objective function is established to achieve multi-objective optimization. Specifically, the position and number of actuators are optimized, that is, q actuators are used to replace the original p actuators, where q < p. The error signal at each measuring point is:

[0034] e m =d m +G mq u q

[0035] In the formula, u q The active control quantity, i.e., the drive signal of the last q actuators, is a vector of q×1, G. mq The admittance matrix for the control channel;

[0036] Using the optimal control method, the optimal control quantity and the noise response at r monitoring points inside the cabin are obtained. The noise reduction effect after control is then calculated as the single-line spectrum objective function, expressed as:

[0037]

[0038] In the formula, The noise energy at the monitoring point under uncontrolled conditions. To control the noise energy at the monitoring points,

[0039] p0 is the reference sound pressure level;

[0040] The objective function for constructing a multi-line spectrum is expressed as:

[0041]

[0042] In the formula, J i w represents the noise reduction effect at the i-th line spectrum. i Let K be the weight of the objective function for the i-th line spectrum, and K be the number of line spectra.

[0043] The beneficial effects of this invention are:

[0044] (1) Compared with traditional genetic algorithms (TGA), the BPNN-GA algorithm (Back Propagation Neural Network-Genetic Algorithm) proposed in this invention significantly reduces the computation time while ensuring the control effect, greatly improving the computation efficiency. Moreover, as the optimization scale increases, the advantage of this algorithm in improving computation efficiency will become more obvious.

[0045] (2) Compared with empirical selection, the BPNN-GA algorithm proposed in this invention focuses on the relationship between the sensor actuator position and the noise at the monitoring point, resulting in better noise control at the monitoring point and in the cabin. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below.

[0047] Figure 1 This is an ASAC control strategy diagram according to an embodiment of the present invention;

[0048] Figure 2 This is a flowchart of the BPNN-GA algorithm according to an embodiment of the present invention;

[0049] Figure 3 This is a diagram of a multi-line spectrum complex neural network model according to an embodiment of the present invention;

[0050] Figure 4 This is a schematic diagram of the response vector of an embodiment of the present invention;

[0051] Figure 5 This is a diagram illustrating the actuator optimization process based on a genetic algorithm according to an embodiment of the present invention;

[0052] Figure 6 This is a diagram of an active control and noise monitoring system according to an embodiment of the present invention;

[0053] Figure 7 This is a schematic diagram of a sensor actuator layout according to an embodiment of the present invention;

[0054] Figure 8 This is an ear position plane noise level diagram according to an embodiment of the present invention, wherein (a) 360Hz is uncontrolled, (b) 360Hz is controlled, (c) 900Hz is uncontrolled, and (d) 900Hz is controlled. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the following detailed description, in conjunction with the accompanying drawings and embodiments, further illustrates the invention. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0056] During helicopter flight, the vibration of the main gearbox is transmitted to the cabin walls through the supporting structure, exciting the wall panels to vibrate and radiating noise into the cabin. This structural sound transmission is the main source of cabin noise in helicopters. The wall panels are the main transmission path and direct radiation source of cabin noise, and the roof panel, as a key forced vibration area, is the main source of radiated noise. Reducing its vibration level can effectively reduce radiated noise and prevent vibration energy from being transmitted to other wall panels in the cabin. By integrating actuators and sensors with the helicopter roof panel, an intelligent active wall panel is formed, which can sense and control the vibration of the roof panel in real time, thereby achieving the goal of controlling noise through vibration.

[0057] The Active Structural Acoustic Control (ASAC) of this invention is as follows: Figure 1 As shown, the system includes radiation surface element division and intelligent sensor / actuator optimization algorithms. The roof panel receives direct excitation from the main gearbox via the support structure, radiating noise into the cabin. Based on the specifications of the piezoelectric fiber composite (MFC) actuators, the helicopter cabin roof panel is divided into several radiation surface elements. An acceleration sensor is placed at the center of each radiation surface element, and an MFC actuator is positioned in the same location. At this point, the sensors and actuators on the roof panel constitute the input and output of the active control system. Subsequently, machine learning methods are used to optimize the position and number of sensors. A high-precision roof panel vibration-radiated noise model is constructed using a data-driven approach and a neural network. Several microphones are placed inside the cabin according to the actual positions of the crew to simulate the sound pressure at the crew's ears. Four concentrated load excitations are applied to the top of the cabin model to simulate the excitation load transmitted from the gearbox support / strut to the roof panel. Vibration was applied to the cabin model, and the vibration response of each radiating surface element was characterized by the acceleration value at its center. Acceleration response data and noise response data were used as the input and output of a neural network, respectively. Through multiple iterations, the relationship between panel vibration and radiated noise was accurately fitted. An intelligent optimization algorithm was used to identify the radiating surface elements that contributed significantly to cabin noise, i.e., the main sound radiation surface elements, and these were selected as the sensor placement points. After the sensor placement scheme was determined, with the goal of minimizing noise at the monitoring points, actuators were optimized using a genetic algorithm to ensure effective suppression of control point vibration while minimizing cabin noise levels.

[0058] Specifically, this application discloses a method for deploying active noise control sensor actuators in a helicopter cabin, comprising three steps:

[0059] Step 1: Based on the actuator specifications, the helicopter cabin roof is divided into several radiating elements, and an accelerometer and an MFC actuator are placed at the center of each radiating element. During helicopter flight, the vibration of the main gearbox causes the cabin wall panels to vibrate, radiating noise into the cabin. To reduce the noise level at the ears of the pilots and passengers, the accelerometer is used to acquire the cabin wall panel vibration response, and a microphone is used to acquire the cabin noise response. The data is then input into the active control and noise monitoring system.

[0060] Step 2: Sensor optimization based on neural networks. To reduce the control scale while ensuring noise reduction, it is necessary to identify and prioritize the control of major acoustic radiation surface elements that contribute significantly to cabin noise. Taking helicopter cabin noise as the control object, a multi-line spectrum complex neural network model is established. Based on this, acceleration response data and noise response data are used as the input and output of the neural network, respectively. The neural network model is then trained. Based on the "forward propagation" process of the BP neural network, the relationship between the hidden layers and the input and output is obtained. To optimize network performance, the weight coefficients of the network are corrected based on the "backward propagation" algorithm. Furthermore, based on the gradient descent method, the weight coefficients between each layer of the network are corrected in the negative gradient direction of the loss function. The above gradient calculation and weight coefficient update are repeated until the preset maximum number of iterations is reached, accurately fitting the relationship between panel vibration and radiated noise. Further, based on the radiated noise generated by the M surface elements at the monitoring points, the vibration responses of each radiation surface element are marked and superimposed on the complex plane to obtain the total response vector. The contribution of radiated noise is calculated, and the major acoustic radiation surface elements are identified as the sensor placement points.

[0061] Specifically,

[0062] Step 2.1 After dividing the radiating surface elements on the cabin roof, establish a system with the helicopter cabin noise as the control target. Figure 3 The multi-line spectrum complex neural network model shown is characterized by establishing an independent model for each line frequency while preserving the amplitude and phase information of the vibration noise response, thus ensuring that the model has accurate physical properties.

[0063] from Figure 3 It can be seen that in a multi-spectral complex neural network model, the number of models is equal to the number of feature lines. For a neural network containing an input layer, H hidden layers, and an output layer, the input vector of the BP neural network is X = [x1, x2, ..., x]. M ] TIt characterizes the vibration response of each radiating surface element of the top plate at ω = ω0, where M is the total number of radiating surface elements. For the m-th radiating surface element, its response is: and These represent the real and imaginary parts of the response, respectively, containing the amplitude and phase information of the vibration. The target vector of the BP neural network is Y = [y1, y2, ..., y]. N ] T This refers to the noise response of N monitoring points inside the cabin at ω = ω0. Similarly, for the nth monitoring point, its noise response is...

[0064] The relationship between the input and output of the neural network is thus obtained as follows:

[0065]

[0066] In the formula, W is the actual output of the neural network. h Let h be the weight coefficient matrix of the h-th layer. This is the transfer matrix between panel vibration and radiated noise.

[0067] To satisfy the physical characteristics of sound vibration transmission and improve the accuracy and reliability of the model, certain constraints are imposed on the bias matrix and activation function. (1) All bias terms are set to 0. In neural networks, bias terms usually represent a fixed constant term that is independent of the input. However, in the actual sound vibration transmission process, there is no constant term that is independent of vibration that affects the propagation of radiated noise. If arbitrary bias values ​​are added just to make the neural network converge faster, it will cause the model to produce large errors in prediction and violate the physical characteristics of sound vibration transmission. (2) The activation function is a linear activation function. In neural networks, the activation function determines the output of the neuron, and its selection is usually based on the nature of the problem being solved. Since this invention focuses on the linear transmission of sound vibration in the subsequent simulation, that is, nonlinear factors are not considered for the time being. Therefore, in order to ensure the accuracy of the model and prevent the distortion of the sound vibration transmission relationship, a linear activation function is selected.

[0068] Step 2.2: After the neural network's "forward propagation," the network's weight coefficients are adjusted based on the "backward propagation" algorithm to optimize network performance. This adjustment process requires defining a loss function that measures the difference between the network's actual output and its expected output. Here, mean squared error is chosen as the loss function to quantify the deviation between the predicted and true values, as follows:

[0069]

[0070] In the formula, the expected output y n Compared with the actual network output Both are plural numbers, * "" indicates conjugate. Based on gradient descent, the weight coefficients between network layers are corrected using the negative gradient direction of the loss function. Considering the influence of the real and imaginary parts, the gradient is represented based on the chain rule. Combining the applied constraints, namely, the bias is set to 0 and the activation function is a linear activation function, the real and imaginary parts are simplified. Given a learning rate η, the updated weight coefficients can be obtained as follows:

[0071]

[0072] In the formula, For u j conjugate, For the loss function E in u i The gradient at that point.

[0073] Step 2.3: Repeat the gradient calculation and weight coefficient update in step 2.2 until the preset maximum number of iterations is reached. Obtain the transfer matrix between panel vibration and radiated noise based on equation (1). The goal is to determine the noise contribution of each surface element and then identify the main sound radiating surface elements.

[0074] The radiated noise generated by the m-th surface element at the n-th monitoring point is derived from equation (1). Represented as:

[0075]

[0076] In the formula, For the transfer matrix The element in the nth row and mth column represents the acoustic-vibration transmission relationship between the vibration of the mth surface element and the noise at the nth monitoring point.

[0077] The overall contribution of the m-th element is:

[0078]

[0079] In the formula, For frequency ω k The radiated noise contribution of the m-th surface element at the n-th monitoring point is calculated and converted into energy by squaring. Phase weighting coefficients The sign function form is used to determine the contribution of radiated noise energy. When When it is the correct time, A value of 1 indicates that the element enhances the total noise energy, thus worsening the overall noise level. Conversely, when... When it is negative, A value of -1 indicates that the contribution of this element to the total noise energy is negative, meaning that it cancels out the total noise energy.

[0080] Analogous to the use of cumulative contribution rate as the criterion for principal component judgment in the field of machine learning, a selection criterion for the main sound radiation surface elements in active control of helicopter cabin noise is established. The specific implementation process is as follows: First, the comprehensive contribution of each radiation surface element is calculated based on equation (5). Second, in order to quantify the contribution of each surface element to the total radiated noise energy, the contribution rate of each radiation surface element is calculated based on equation (6):

[0081]

[0082] Furthermore, a threshold is set, typically ranging from 80% to 90%. Starting with the surface element with the largest contribution, the energy contribution rate of each surface element is accumulated sequentially until the cumulative contribution rate reaches or exceeds the set threshold. Surface elements whose energy contribution rate reaches or exceeds the set threshold during this process are defined as primary acoustic radiation surface elements, and sensors are placed on them as control points.

[0083] The selection of the threshold is a trade-off; an excessively high threshold may lead to an excessive number of selected primary acoustic radiation elements, thereby increasing control costs and complexity. Therefore, in this embodiment, the threshold is initially set to 80%, and the threshold is precisely adjusted using scree plot analysis from the field of machine learning. A scree plot is drawn with the total contribution of radiated noise as the ordinate, and the point where the contribution decreases and begins to level off, i.e., the "inflection point," is identified. The elements before these inflection points are usually those that contribute significantly to the total noise. By adjusting the threshold, it is ensured that the number of elements before the inflection point is not less than the number of elements whose cumulative energy contribution rate is greater than the threshold, thereby reducing control costs while satisfying control performance.

[0084] Step 3: Perform actuator optimization based on a genetic algorithm (including the number and location of actuators). Based on the determined sensor locations, and with the goal of reducing noise levels at monitoring points, a genetic algorithm is used to optimize the actuator selection. For example... Figure 5 As shown, firstly, a set of actuator placement schemes is randomly generated as the initial population. Then, based on the optimization objective and constraints, a fitness function is set to quantitatively evaluate the performance of each placement scheme. Further, through selection, crossover, and mutation operations, the individuals in the population are iteratively optimized until the termination condition is met. Finally, a set of optimized actuator placement schemes is obtained.

[0085] When using genetic algorithms for actuator optimization, the key lies in constructing the objective function. To effectively reduce cabin noise levels, this paper establishes a multi-line spectrum objective function, aiming to achieve multi-objective optimization and ensure that the noise reduction effect reaches a relatively optimal level for each line spectrum. When ω=ω0, the single-line spectrum objective function is established as follows:

[0086] Assuming there are m sensors and p actuators, the error signal measured by the sensors can be expressed as:

[0087] e m =d m +G mp u p (7)

[0088] In the formula, e m The error signal, i.e., the vibration response after control by m sensors, is an m×1 vector. d m For the uncontrolled response, it is also an m×1 vector. p The active control quantity, i.e., the drive signal of the actuator, is a vector of p×1. G mp Let be the admittance matrix of the control channel, which contains the one-to-one correspondence between m sensors and p actuators, and is an m×p matrix.

[0089] At this point, using the optimal control method in the frequency domain, the optimal control quantity can be obtained as:

[0090]

[0091] However, when a control system has m sensors and p actuators, it requires m×p controller channels. In engineering practice, an excessive number of actuators leads to a large control scale, complex control system, high control cost, and reduced stability. To reduce the control complexity and cost while ensuring control performance, this application considers optimizing the position and number of actuators (by using an optimization algorithm to reduce the number of actuators), that is, replacing the original p actuators with q actuators (q < p, and the optimal number of actuators is consistent with the number of sensors to ensure that the active control system is a completely deterministic system, thereby enhancing its stability), where p is an assumed value, corresponding to the number of actuators in the control system before optimization in actual calculations. At this point, the error signal at each measuring point can be expressed as:

[0092] e m =d m +G mq u q (9)

[0093] In the formula, u q The active control quantity, i.e., the drive signals of the selected q actuators, is a vector of q×1. Gmq Let be the admittance matrix of the control channel, containing the one-to-one correspondence between m sensors and q actuators, and be an m×q matrix. Using the optimal control method, the optimal control quantity and the noise response at r monitoring points within the cabin can be obtained, and the noise reduction effect after control can be calculated as the single-line spectrum objective function, as follows:

[0094]

[0095] In the formula, The noise energy at the monitoring point under uncontrolled conditions. The noise energy at the post-control monitoring point is p0, which is the reference sound pressure level with a magnitude of 2 × 10⁻⁶. -5 Pa.

[0096] To comprehensively evaluate the noise reduction performance of the actuator layout scheme under multiple line spectra, a multi-line spectrum objective function is constructed based on equation (10), as follows:

[0097]

[0098] In the formula, J i The noise reduction effect under the i-th line spectrum can be calculated by equation (10). i represents the weight of the objective function for the i-th line spectrum, used to adjust the relative importance of different line spectrum noise reduction effects in the overall objective function, with an initial value of 1. K is the number of line spectra.

[0099] To verify the feasibility and effectiveness of the BRNN-GA intelligent optimization algorithm proposed in this invention, the inventors constructed an active control and noise monitoring system and conducted simulation analysis. The helicopter cabin model's ceiling was divided into 63 acoustic radiation elements. An accelerometer was placed at the center of each radiation element to characterize its vibration response and serve as feedback signals for active control. An MFC actuator was also placed in the same location. Thus, the 63 actuators and 63 sensors on the ceiling constitute the input and output of the active control system. To monitor the cabin noise level, an cabin noise monitoring system was implemented, with eight monitoring points (e.g., eight microphones) placed inside the cabin to simulate the sound pressure level at the ears of the occupants. The active control and noise monitoring system is as follows: Figure 6 As shown. Taking the established helicopter cabin model as the research object, the sensor actuators are optimized based on the BPNN-GA algorithm to obtain the optimal scheme, namely the sensor actuator layout scheme as shown. Figure 7As shown in the table below. The optimization results of the intelligent optimization algorithm were then verified through simulation. Table 1 shows the simulation results of noise reduction inside the helicopter cabin after sensor / actuator optimization. It can be seen that at 360Hz, the average sound pressure level at the monitoring point decreased by 18.0dB. At 900Hz, the average sound pressure level at the monitoring point decreased by 22.3dB. The cabin cavity noise level and cabin wall vibration level were also suppressed. Figure 8 Simulation results (a)-(d) show the noise level in the plane at the human ear position before and after applying control. It can be seen that after applying active control, the sound pressure level in the plane at the human ear position is significantly attenuated, especially in the central region of the plane, where high noise is effectively suppressed, forming a clear "quiet zone," thus achieving the goal of reducing the noise level at the driver's ear. The sensor actuator scheme optimized based on the BPNN-GA algorithm achieves good control performance during active control, effectively suppressing the noise level at each monitoring point while also exhibiting certain global vibration reduction and global noise reduction effects, thereby verifying the feasibility and effectiveness of the algorithm.

[0100] Table 1

[0101]

[0102]

[0103] The contents not described in detail in this specification are common knowledge to those skilled in the art.

[0104] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for deploying active noise control sensor actuators in a helicopter cabin, characterized in that, Includes the following steps: Step 1: Divide the helicopter cabin top plate into several radial surface elements based on the actuator specifications, and arrange an acceleration sensor and an MFC actuator at the center of the radial surface element. Step 2: Taking helicopter cabin noise as the control object, establish a multi-line spectrum complex neural network model. Use the cabin wall vibration response obtained from accelerometers and the cabin noise response obtained from microphones as the input and output of the multi-line spectrum complex neural network model, respectively. Identify the main sound radiation surface elements as the optimal placement points for the accelerometers. Specifically, this includes: Step 2.1: Establish a multi-line spectrum complex neural network model. For each spectral frequency of the wall panel vibration, an independent sub-model is established. Simultaneously, the sub-model retains the amplitude and phase information of the vibration noise response. For a neural network containing an input layer, H hidden layers, and an output layer, the input vector of the BP neural network is X = [x1, x2, ..., x...]. M ] T It characterizes the vibration response of each radiating surface element of the top plate at ω = ω0, where M is the total number of radiating surface elements; for the m-th radiating surface element, its response is: and These represent the real and imaginary parts of the response, respectively, containing the amplitude and phase information of the vibration; the target vector of the BP neural network is Y = [y1, y2, ..., y]. N ] T That is, the noise response of N monitoring points inside the cabin at ω = ω0; similarly, for the nth monitoring point, its noise response is... Therefore, the relationship between the input and output of the neural network can be obtained as follows: In the formula, W is the actual output of the neural network. h Let h be the weight coefficient matrix of the h-th layer. The transfer matrix between panel vibration and radiated noise is defined; certain constraints are imposed on the bias matrix and activation function: all bias terms are set to 0, and the activation function is a linear activation function. Step 2.2: The chain rule is used for "backpropagation" to correct the network weight coefficients. The mean squared error is used as the loss function, and the gradient of the loss function with respect to each parameter is calculated. Based on gradient descent, the weight coefficients between each network layer are corrected along the negative gradient direction of the loss function. The real and imaginary parts are simplified by combining the applied constraints. A learning rate η is given, resulting in the updated weight coefficients, expressed as: In the formula, For u j conjugate, For the loss function E in u i gradient at; Step 2.3: Repeat the gradient calculation and weight coefficient update in Step 2.2 until the preset maximum number of iterations is reached. Based on the relationship between the neural network input and output in Step 2.1, obtain the transfer matrix between the panel vibration and radiated noise. Calculate the radiated noise generated by M surface elements at the monitoring point. The radiated noise is converted onto a complex plane for marking and superposition to obtain the sum of the vibration responses of each radiating surface element, i.e., the total response vector. Based on the total response vector, the contribution of each surface element to the overall radiated noise is calculated, and the comprehensive contribution of each surface element is further determined. Step 3: Based on the determined optimal placement points of the accelerometers, and with the goal of reducing the noise level at the monitoring points, a genetic algorithm is used to optimize the actuator placement points. First, a set of actuator placement schemes is randomly generated as the initial population. Then, based on the optimization objective and constraints, a fitness function is set to quantitatively evaluate the performance of each placement scheme. Specifically, a multi-line spectrum objective function is established to optimize the position and number of actuators, that is, q actuators are used to replace the original p actuators, where q < p. The error signal at each measurement point is: e m =d m +G mq u q , In the formula, u q The active control quantity, i.e., the drive signal of the last q actuators, is a vector of q×1, G. mq The admittance matrix for the control channel; Using the optimal control method, the optimal control quantity and the noise response at r monitoring points inside the cabin are obtained. The noise reduction effect after control is then calculated as the single-line spectrum objective function, expressed as: In the formula, The noise energy at the monitoring point under uncontrolled conditions. p0 represents the noise energy at the post-control monitoring point, and p0 represents the reference sound pressure. The objective function for constructing a multi-line spectrum is expressed as: In the formula, J i w represents the noise reduction effect at the i-th line spectrum. i The weights of the objective function for the i-th line spectrum are given by K, where K is the number of line spectra. Finally, through selection, crossover, and mutation operations, the individuals in the population are iteratively optimized until the termination condition is met, and the optimal placement point of the MFC actuator is finally obtained.

2. The deployment method according to claim 1, characterized in that, In step 2.3, the overall contribution of the m-th element is: In the formula, For frequency ω k The radiated noise contribution of the m-th surface element at the n-th monitoring point. Phase weighting coefficients The symbolic function form is used to determine the contribution of radiated noise energy.

3. The deployment method according to claim 1, characterized in that, To quantify the contribution of each surface element to the total radiated noise energy, the energy contribution rate of each radiating surface element is calculated based on the following formula: A threshold is set, and starting from the surface element with the largest contribution, the energy contribution rate of each surface element is accumulated sequentially until the cumulative contribution rate reaches or exceeds the set threshold. The surface element that is accumulated during this process is defined as the main sound radiation surface element.

Images