Secondary sound source parameter optimization system and cooperative control method based on adjustable holder and fixed array

Abstract

A secondary sound source parameter optimization system based on an adjustable holder and a fixed array and a cooperative control method are used for solving the problem that an existing noise reduction system does not solve sound source directivity optimization and cannot adapt to a dynamic noise source. The optimization system comprises a fixed infrasound source array, an adjustable holder module, a multi-mode sensor network and a control unit, wherein the fixed infrasound source array is used for generating reverse sound waves opposite to noise; the adjustable holder module is used for adjusting the azimuth angle and the pitch angle of the loudspeaker of the fixed infrasound source array; the multi-mode sensor network is used for detecting noise, environment humidity and vibration acceleration data; the control unit adjusts the parameters of the secondary sound source according to the detection data of the multi-mode sensor network, optimizes the noise reduction parameters and improves the noise reduction effect. According to the invention, the noise reduction range is expanded, the method adapts to a transient noise scene, and 50Hz-1kHz full-band suppression is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer noise reduction, and in particular to a secondary sound source parameter optimization system and a collaborative control method based on an adjustable pan-tilt platform and a fixed array. Background Art

[0002] When operating in urban areas, power transformers generate significant low-frequency electromagnetic noise and high-frequency vibration noise due to the magnetostriction of the core and the electromagnetic force of the windings. The main frequency of low-frequency electromagnetic noise is 100Hz-200Hz, and the main frequency of high-frequency vibration noise is >500Hz, affecting residents' lives and posing a certain threat to the normal operation of equipment. Traditional noise reduction technologies mainly rely on passive sound insulation materials or fixed active noise reduction systems, which have the following drawbacks:

[0003] 1. Limitations of fixed arrays.

[0004] Insufficient spatial coverage: The fixed infrasound source array of traditional noise reduction technology cannot flexibly adjust the radiation direction, resulting in poor noise reduction effect in acoustic shadow areas, such as the back of the transformer box.

[0005] Weak high-frequency suppression capability: Fixed speaker units are limited by physical size and cannot effectively cover high-frequency noise.

[0006] Redundant energy consumption: To cover the entire frequency band, a large number of sound sources must be deployed, resulting in high energy consumption and maintenance costs.

[0007] 2. Dynamic noise environment challenges.

[0008] Transient noise response lag: For example, the pulse noise caused by the transformer closing inrush current cannot be tracked in real time by traditional algorithms due to calculation delays.

[0009] Complex electromagnetic interference: DC bias and harmonic interference in substations cause sensor signal distortion and affect control accuracy.

[0010] Existing noise reduction systems based on adaptive filtering fail to optimize the directionality of sound sources. Beamforming technology improves directivity, but relies on a static array layout and cannot adapt to dynamic noise sources. Therefore, an intelligent noise reduction system integrating dynamic directional adjustment, broadband coverage, and environmental adaptability is urgently needed to address the noise control needs of urban power equipment. Summary of the Invention

[0011] The purpose of the present invention is to provide a secondary sound source parameter optimization system and a collaborative control method based on an adjustable pan-tilt head and a fixed array, which are used to solve the problems that the existing noise reduction system fails to solve the sound source directionality optimization and cannot adapt to dynamic noise sources.

[0012] The technical solution adopted by the present invention to solve its technical problem is: a secondary sound source parameter optimization system based on an adjustable pan-tilt platform and a fixed array, including a fixed subsound source array, an adjustable pan-tilt platform module, a multimodal sensor network and a control unit.

[0013] The fixed infrasound source array includes a loudspeaker configured to generate an anti-phase sound wave opposite to the transformer noise.

[0014] The adjustable pan-tilt platform module includes a two-degree-of-freedom pan-tilt platform, a stepper motor and a harmonic reducer. The speaker is arranged on the two-degree-of-freedom pan-tilt platform. The harmonic reducer is located between the output end of the stepper motor and the two-degree-of-freedom pan-tilt platform. After starting the stepper motor, the azimuth and pitch angles of the two-degree-of-freedom pan-tilt platform are adjusted through the harmonic reducer.

[0015] The multimodal sensor network includes a microphone and an environmental sensor. The microphone is used to collect / detect noise spectrum. The environmental sensor includes a temperature and humidity sensor and a vibration accelerometer. The temperature and humidity sensor is used to collect temperature and humidity data of the environment. The vibration accelerometer is used to collect vibration acceleration data of the transformer.

[0016] The control unit adjusts the speaker parameters according to the detection data of the multimodal sensor network.

[0017] Furthermore, the loudspeakers include six groups of full-range loudspeakers and two groups of high-frequency directional loudspeakers.

[0018] Furthermore, six groups of full-range loudspeakers with a frequency of 50 Hz-500 Hz are arranged in a circular manner on the surface of the transformer box, and the distance between the full-range loudspeakers is 0.5 m.

[0019] Furthermore, two groups of high-frequency directional speakers with a frequency of 500Hz-1kHz are located on a two-degree-of-freedom pan-tilt platform. The azimuth angle of the high-frequency directional speakers is 0°-360°, and the pitch angle is -15°-45°, covering sensitive directions of residential areas.

[0020] Furthermore, the microphone adopts a 16-channel MEMS microphone with a signal-to-noise ratio SNR≥70dB.

[0021] The present invention also provides a dynamic collaborative control method for a secondary sound source parameter optimization system based on an adjustable pan-tilt platform and a fixed array, comprising the following steps:

[0022] S1. Noise detection.

[0023] The transformer noise is collected by a microphone.

[0024] S2. Dynamically allocate computing resources.

[0025] The time-sharing attention mechanism TSA is used to dynamically allocate computing resources.

[0026] S3. Dynamic beamforming.

[0027] Optimize the beam weight vector w, the formula is: (1).

[0028] Where R is the noise covariance matrix with dimension N×N, which represents the spatial correlation of the noise.

[0029] a(θ) is the steering vector, dimension N×1, which represents the phase response of each unit when the sound wave is incident from the direction θ. H (θ) represents the conjugate transpose of the steering vector a(θ).

[0030] S4. Optimization of secondary sound source parameters.

[0031] The physical information reinforcement learning (PIRL) algorithm is used to dynamically optimize the secondary sound source parameters, including direction, frequency, amplitude, and phase.

[0032] S5. Generate control instructions.

[0033] According to the optimization results of the secondary sound source parameters, control instructions for the loudspeaker are generated.

[0034] S6. Noise suppression.

[0035] The control instruction generated in step S5 is transmitted to the fixed infrasound source array and the adjustable platform module, and noise suppression is performed by adjusting the azimuth and pitch angles of the two-degree-of-freedom platform and the speaker power.

[0036] S7. Feedback and status updates.

[0037] Perform residual noise detection. If the residual noise fails to meet the requirements, upload the execution results of parameter optimization to the physical information reinforcement learning (PIRL) algorithm and repeat step S4.

[0038] Furthermore, the radiation pattern of the secondary sound source is determined by the two-degree-of-freedom pan-tilt angle and the fixed subsound source array layout, and its directivity function is (2). Where θ is the vertical pitch angle of the main lobe of the sound beam, which ranges from ±30°; φ is the horizontal azimuth angle, which ranges from 0 to 360°; n is the directivity index, which controls the vertical radiation range of the sound source; n = 2 means the sound beam is narrower and more directive. N is the number of speakers in the fixed array; d is the distance between speakers; K is the wave number, k = ;λ is the wavelength of the sound wave; is the target direction angle.

[0039] Furthermore, by adjusting the phase difference Control the sound beam coverage D, that is (3). Where D is the desired sound beam coverage.

[0040] Furthermore, the control unit has a built-in physical information reinforcement learning (PIRL) algorithm. The quantitative reward mechanism of the PIRL algorithm includes:

[0041] State space: Fusion of the noise spectrum S(f) and the sound field distribution P(x,y,z).

[0042] Action space: pan / tilt angle (θ, φ), sound source frequency f, amplitude A, and phase.

[0043] The reward function is (4). Among them, is the noise reduction amount, and its calculation formula is L p,初始 -L p,残余 ;P consumed is the real-time power consumption, which is composed of the energy consumption of the source and the energy consumption of the pan / tilt motor drive; P max is the maximum allowable power consumption of the system; Stability(θ,φ) is the gimbal angle stability index, defined as ;in, is the gimbal angle fluctuation variance, is the maximum allowable fluctuation angle; 、 、 is the weight coefficient.

[0044] Furthermore, by minimizing the noise power w H Rw, suppresses the interference corresponding to the noise covariance matrix while keeping the target direction signal undistorted. Each speaker optimizes the beam weight vector w and phase difference Output anti-phase sound waves to achieve optimal noise reduction; where w H is the conjugate transpose of w in the complex field.

[0045] The beneficial effects of the present invention are: (1) Improved noise reduction range: Through the dynamic steering of the two-degree-of-freedom pan-tilt, the effective coverage angle is increased to ±150°, and the blind area is reduced by 60%. (2) Energy efficiency optimization: High-frequency noise is suppressed by the directional pan-tilt focusing, and dynamic resource allocation reduces redundant power consumption, effectively improving the overall energy efficiency ratio and reducing energy consumption by 30%. (3) Enhanced real-time performance: The two-degree-of-freedom pan-tilt response time is <80ms, adapting to transient noise scenarios. (4) Broadband coverage: The fixed infrasound source array and the adjustable pan-tilt module work together to achieve full-band suppression of 50Hz-1kHz. (5) Environmental robustness: The anti-electromagnetic interference design and compact structure adapt to the complex environment of urban areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1Flow chart of the method of the present invention.

[0047] Figure 2 This is a system architecture diagram of the present invention. DETAILED DESCRIPTION

[0048] The present invention's secondary sound source parameter optimization system based on an adjustable pan-tilt and fixed array includes a fixed infrasound source array, an adjustable pan-tilt module, a multimodal sensor network, and a control unit. The fixed infrasound source array is used to generate anti-phase sound waves that are opposite to the noise; the adjustable pan-tilt module is used to adjust the azimuth and pitch angles of the speakers in the fixed infrasound source array; the multimodal sensor network is used to detect noise, ambient humidity, and vibration acceleration data; and the control unit adjusts the secondary sound source parameters based on the detection data from the multimodal sensor network, optimizing the noise reduction parameters and improving the noise reduction effect. The structure and control method of the present invention are described below with reference to the accompanying drawings.

[0049] like Figure 2 As shown in Figure 1, the secondary sound source parameter optimization system based on an adjustable pan-tilt and fixed array consists of four parts: a fixed subsound source array, an adjustable pan-tilt module, a multimodal sensor network, and a control unit. Each structure is described below.

[0050] 1. Fixed Infrasound Source Array

[0051] The fixed infrasound source array includes six sets of full-range speakers and two sets of high-frequency directional speakers. The speakers are used to generate anti-phase sound waves opposite to the transformer noise, thereby achieving noise reduction.

[0052] Six full-range speakers with a frequency range of 50Hz-500Hz are arranged in a circular pattern on the surface of the transformer enclosure, fixed to the enclosure. Based on acoustic topology analysis, the spacing between the full-range speakers in the fixed infrasound source array is 0.5m to avoid spatial aliasing and acoustic interference. The full-range speakers are 8 inches in size and have a power output of 100W.

[0053] Two sets of high-frequency directional speakers with a frequency of 500Hz-1kHz are mounted on a two-degree-of-freedom pan-tilt platform. The azimuth angle of the high-frequency directional speakers is 0°-360°, the pitch angle is -15°-45°, and the response time is <50ms, covering sensitive directions of residential areas.

[0054] 2. Adjustable PTZ Module

[0055] The adjustable pan-tilt module consists of a two-degree-of-freedom pan-tilt, a stepper motor, and a harmonic reducer. The two-degree-of-freedom pan-tilt is used to support the high-frequency directional speaker. Specifically, the high-frequency directional speaker is mounted on the two-degree-of-freedom pan-tilt, and the two-degree-of-freedom pan-tilt adjusts the azimuth and pitch angles of the high-frequency directional speaker through its own movement. The stepper motor has a resolution of 0.01°. The output of the stepper motor is connected to the input of the harmonic reducer, and the output of the harmonic reducer is connected to the two-degree-of-freedom pan-tilt. After the stepper motor is started, the harmonic reducer decelerates the rotational motion output by the stepper motor and then transmits it to the two-degree-of-freedom pan-tilt. The adjustable pan-tilt module controls the azimuth and pitch angles of the two-degree-of-freedom pan-tilt through the stepper motor and harmonic reducer, and combines it with the MVDR beamforming algorithm to optimize the sound field coverage.

[0056] The radiation pattern of the secondary sound source is determined by the two-degree-of-freedom pan-tilt angle and the fixed subsound source array layout. Its directivity function is (1).

[0057] Where θ is the pitch angle of the main lobe of the sound beam in the vertical direction, and the range is ±30°.

[0058] φ is the horizontal azimuth angle, ranging from 0 to 360°.

[0059] n is the directivity index, which controls the vertical radiation range of the sound source; n=2 means the sound beam is narrower and more directive.

[0060] N is the number of array elements, that is, the number of loudspeakers in a fixed array.

[0061] d is the distance between the loudspeakers, satisfying , to avoid spatial aliasing. min is the wavelength corresponding to the highest frequency.

[0062] k is the wave number, k= .

[0063] λ is the wavelength of the sound wave.

[0064] is the target direction angle.

[0065] In the above formula, the first term Used to describe the vertical sound pressure attenuation characteristics. The larger n is, the narrower the vertical coverage range is. It is used to describe the horizontal beamforming characteristics. By adjusting the spacing and number of loudspeakers, the main lobe of the sound beam is optimized to align with the noise source.

[0066] The present invention adjusts the phase difference Control the sound beam coverage D, that is (2).

[0067] Where D is the desired sound beam coverage, in meters, indicating the radius of the noise reduction area.

[0068] λ is the wavelength of sound waves, unit: meter.

[0069] θ is the pitch angle of the main lobe of the sound beam, expressed in radians.

[0070] is the array phase difference, expressed in radians.

[0071] By adjusting the phase difference Control the sound beam coverage so that the reverse phase sound field accurately covers the target area. Combined with the MVDR algorithm, the beam weight vector w is optimized. The formula is: (3).

[0072] Where R is the noise covariance matrix with dimension N×N, which represents the spatial correlation of the noise.

[0073] a(θ) is a steering vector with dimension N×1, which represents the phase response of each loudspeaker when the sound wave is incident from the direction θ.

[0074] a H (θ) represents the conjugate transpose of the steering vector a(θ), which is used to calculate the gain of the beamformer output in the target direction to ensure that the anti-phase sound wave accurately covers the noise source.

[0075] w is the optimized beam weight vector with dimension N×1, which is used to weight the output of each speaker.

[0076] 3. Multimodal Sensor Networks

[0077] The multimodal sensor network includes microphones and environmental sensors, and the microphones are arranged in an array.

[0078] The microphone uses a 16-channel MEMS microphone with a signal-to-noise ratio (SNR) of ≥70dB. The microphone is used to collect / detect the noise spectrum.

[0079] Environmental sensors include a temperature and humidity sensor and a vibration accelerometer. The temperature and humidity sensor collects ambient temperature and humidity data with an accuracy of ±1% RH. The vibration accelerometer collects transformer vibration acceleration data with a bandwidth of 0.1-10 kHz.

[0080] 4. Control Unit

[0081] The control unit incorporates a built-in physical-information reinforcement learning (PIRL) algorithm, a hybrid algorithm that combines physical models with reinforcement learning. Its core goal is to leverage known physical laws, such as the sound wave propagation equation, to constrain the reinforcement learning exploration process, reduce ineffective search space, and improve the model's generalization and interpretability. In the transformer noise reduction system, the PIRL algorithm achieves efficient noise suppression by dynamically optimizing secondary sound source parameters, such as direction, frequency, amplitude, and phase. The PIRL algorithm's quantitative reward mechanism includes:

[0082] State space: Fusion of the noise spectrum S(f) and the sound field distribution P(x,y,z).

[0083] Action space: pan / tilt angle (θ, φ), sound source frequency f, amplitude A, and phase.

[0084] The reward function is (4).

[0085] in, is the noise reduction amount, and its calculation formula is L p,初始 -L p,残余 .

[0086] P consumed It is the real-time power consumption, which is composed of the energy consumption of the source and the energy consumption of the pan-tilt motor drive.

[0087] P max is the maximum allowable power consumption of the system, such as P max =500W.

[0088] Stability (θ, φ) is the gimbal angle stability index, defined as ;in, is the gimbal angle fluctuation variance, The maximum allowable fluctuation angle, such as ±5°.

[0089] 、 、 is the weight coefficient, which needs to be tuned through experiments, such as , , The physical meanings of the three weight coefficients are as follows.

[0090] To ensure noise suppression is the primary goal.

[0091] Prevent excessive power consumption to extend device life.

[0092] Used to restrain the frequent swing of the gimbal and avoid mechanical wear.

[0093] The present invention minimizes the noise power w H Rw, suppresses the interference corresponding to the noise covariance matrix while keeping the target direction signal undistorted. Each speaker optimizes the beam weight vector w and phase difference Outputs inverted sound waves to achieve optimal noise reduction. H is the conjugate transpose of w in the complex field.

[0094] The state encoding of the physical information reinforcement learning (PIRL) algorithm is as follows: embed the sound field physics equation, namely the Helmholtz equation, into the neural network and constrain the strategy search space. The formula is: (5).

[0095] Where P is the sound pressure field, unit: Pa.

[0096] k is the wave number, which is related to the wavelength.

[0097] is the Laplace operator, which represents the propagation characteristics of sound waves.

[0098] In the reinforcement learning policy network training, add physical constraints: (6).

[0099] The role of formula (6) is to ensure that the actions output by the policy network, such as the gimbal angle, conform to physical laws in reinforcement learning, thus avoiding unrealistic optimization.

[0100] The sound field distribution p(x, y, z) is used as part of the input features to form a state space encoding, ensuring that the actions output by the policy network, such as the two-degree-of-freedom pan-tilt angle, conform to the laws of sound wave propagation.

[0101] The control unit has a built-in time-sharing attention (TSA) mechanism, which dynamically allocates computing resources. On resource-constrained embedded devices, computing resources are dynamically allocated to key frequency bands, such as the 100Hz iron core vibration frequency. The noise spectrum is extracted to calculate the energy contribution of each frequency band. The computing resources of the fully connected layer are dynamically adjusted based on the energy Ei of the i-th frequency band, with higher attention weight allocated to high-energy bands.

[0102] The following introduces the network structure of the physical information reinforcement learning PIRL algorithm.

[0103] The network structure of the physical information reinforcement learning (PIRL) algorithm is that the encoder extracts noise spectrum characteristics, captures timing dependencies, and focuses on key frequency bands, such as the 100Hz main frequency of the iron core vibration. Finally, the actuator outputs the pan-tilt angle, frequency, and amplitude.

[0104] (1) Encoder.

[0105] 1D-CNN: The input is the noise spectrum S(f) and the output is the frequency domain feature vector.

[0106] LSTM: Captures temporal dependencies and predicts noise change trends.

[0107] (2) Attention layer.

[0108] TSA module: Dynamically weights features according to frequency band importance.

[0109] (3) Actuator.

[0110] Fully connected layer: outputs pan / tilt angle (θ, φ), sound source frequency f, and amplitude A.

[0111] (4) Evaluator.

[0112] Fully connected layer: Evaluates the value of the current state and guides strategy updates.

[0113] The training process of the secondary sound source parameter optimization system based on the adjustable pan-tilt head and fixed array of the present invention includes the following aspects:

[0114] Offline simulation: Generate multi-condition acoustic field data based on COMSOL Multiphysics, such as 100,000 sets of samples. Simulate the acoustic field distribution under different operating conditions, such as transformer load changes, and collect samples covering the 50Hz-1kHz frequency range.

[0115] Transfer learning: Train the PIRL model on simulated data to optimize the policy network parameters. After deployment on an embedded platform, the model is continuously updated using real-world sensor data, such as microphone arrays and vibration accelerometers.

[0116] The secondary sound source parameter optimization system and collaborative control method based on an adjustable pan-tilt head and a fixed array of the present invention are described below with reference to specific examples.

[0117] 1. Hardware structure of the secondary sound source parameter optimization system based on adjustable pan-tilt and fixed array.

[0118] (1) Fixed infrasound source array.

[0119] It uses 8-inch full-range speakers with a frequency response range of 50Hz-500Hz and a power of 100W. Six sets of full-range speakers are arranged in a circular manner on the surface of the transformer box.

[0120] Layout optimization: Based on acoustic topology analysis, the array spacing is minimized to 0.5m to avoid spatial aliasing.

[0121] (2) Adjustable pan / tilt module.

[0122] Drive module: uses a stepper motor with a resolution of 0.01° and a harmonic reducer, with a response time of <50ms.

[0123] High-frequency sound source: 4-inch compression driver with a frequency response range of 500Hz-1kHz and a sound pressure level ≥110dB.

[0124] (3) Multimodal sensor networks.

[0125] Microphone array: 16-channel MEMS microphones with a signal-to-noise ratio (SNR) ≥ 70dB, distributed in a spherical grid.

[0126] Environmental sensor: A temperature and humidity sensor with an accuracy of ±1%RH and a vibration accelerometer with a bandwidth of 0.1-10kHz are used.

[0127] 2. Software for optimizing secondary sound source parameters based on an adjustable pan / tilt and fixed array.

[0128] (1) Control algorithm.

[0129] The real-time operating system ROS2 is used to implement multi-threaded task scheduling.

[0130] (2) Human-computer interaction interface.

[0131] Use a web-based dashboard to display noise reduction effects, energy consumption statistics, and device status in real time.

[0132] Fault diagnosis: An LSTM-based anomaly detection model is used to provide early warning of 2DOF gimbal jamming or sound source failure.

[0133] The following is combined with Figure 1 , the dynamic collaborative control method of the secondary sound source parameter optimization system based on the adjustable pan-tilt platform and the fixed array of the present invention is described:

[0134] S1. Noise detection.

[0135] The transformer noise is collected by a microphone.

[0136] S2. Dynamically allocate computing resources.

[0137] The time-sharing attention mechanism TSA is used to dynamically allocate computing resources.

[0138] S3. Dynamic beamforming.

[0139] Optimize the beam weight vector w, the formula is: (7).

[0140] Where R is the noise covariance matrix with dimension N×N, which represents the spatial correlation of the noise.

[0141] a(θ) is the steering vector with dimension N×1, which represents the phase response of each unit when the sound wave is incident from the direction θ.

[0142] S4. Optimization of secondary sound source parameters.

[0143] The physical information reinforcement learning (PIRL) algorithm is used to dynamically optimize the secondary sound source parameters, such as direction, frequency, amplitude, and phase, to achieve efficient noise suppression.

[0144] S5. Generate control instructions.

[0145] According to the optimization results of the secondary sound source parameters, control instructions for the loudspeaker are generated.

[0146] S6. Noise suppression.

[0147] The control instruction generated in step S5 is transmitted to the fixed infrasound source array and the adjustable platform module, and noise suppression is performed by adjusting the azimuth and pitch angles of the two-degree-of-freedom platform and the speaker power.

[0148] S7. Feedback and status updates.

[0149] Perform residual noise detection. If the residual noise fails to meet the requirements, upload the execution effect of parameter optimization to the physical information reinforcement learning (PIRL) algorithm of the control unit and repeat step S4.

[0150] Typical application scenarios for this invention include transient noise suppression and broadband noise coverage. For transient noise suppression, for example, a two-degree-of-freedom pan / tilt system can rapidly pivot to the noise source, mitigating the impulse noise of a transformer closing. For broadband noise coverage, for example, a fixed infrasound source array can suppress low-frequency noise, while an adjustable pan / tilt system can directional cancel high-frequency components. Table 1 shows the advantages of this invention over traditional noise reduction technologies in terms of noise reduction, high-frequency coverage, response time, energy consumption, and electromagnetic compatibility.

[0151] Table 1 Comparison of the noise reduction amount, high frequency coverage, response time, energy consumption and electromagnetic compatibility between the present invention and traditional noise reduction technology

[0152]

[0153] From the above comparative analysis, it can be seen that the present invention has significant advantages over traditional noise reduction technologies in the following aspects.

[0154] (1) Improved noise reduction range: Through the dynamic steering of the two-degree-of-freedom gimbal, the effective coverage angle is increased to ±150° and the blind spot is reduced by 60%.

[0155] (2) Energy efficiency optimization: High-frequency noise is suppressed by directional gimbal focusing, and dynamic resource allocation reduces redundant power consumption, effectively improving the overall energy efficiency ratio and reducing energy consumption by 30%.

[0156] (3) Enhanced real-time performance: The response time of the two-degree-of-freedom gimbal is <80ms, which can adapt to transient noise scenarios.

[0157] (4) Broadband coverage: The fixed infrasound source array and the adjustable pan-tilt module work together to achieve full-band suppression of 50Hz-1kHz.

[0158] (5) Environmental robustness: Anti-electromagnetic interference design and compact structure are suitable for complex urban environments.

Claims

1. A secondary sound source parameter optimization system based on an adjustable pan-tilt head and a fixed array, characterized in that: It includes a fixed infrasound source array, an adjustable pan-tilt module, a multimodal sensor network and a control unit; The fixed infrasound source array includes a loudspeaker configured to generate an anti-phase sound wave opposite to the transformer noise; The adjustable pan-tilt platform module includes a two-degree-of-freedom pan-tilt platform, a stepper motor, and a harmonic reducer. The speaker is arranged on the two-degree-of-freedom pan-tilt platform. The harmonic reducer is located between the output end of the stepper motor and the two-degree-of-freedom pan-tilt platform. After the stepper motor is started, the azimuth and pitch angles of the two-degree-of-freedom pan-tilt platform are adjusted through the harmonic reducer. The multimodal sensor network includes a microphone and an environmental sensor, wherein the microphone is used to collect / detect a noise spectrum, the environmental sensor includes a temperature and humidity sensor and a vibration accelerometer, the temperature and humidity sensor is used to collect temperature and humidity data of the environment, and the vibration accelerometer is used to collect vibration acceleration data of the transformer; The control unit adjusts the speaker parameters according to the detection data of the multimodal sensor network.

2. The secondary sound source parameter optimization system based on an adjustable pan-tilt head and a fixed array according to claim 1, characterized in that: The loudspeakers include six groups of full-range loudspeakers and two groups of high-frequency directional loudspeakers.

3. The secondary sound source parameter optimization system based on an adjustable pan-tilt head and a fixed array according to claim 2, characterized in that: Six groups of full-range loudspeakers with a frequency of 50 Hz-500 Hz are arranged in a circular manner on the surface of the transformer box, and the distance between the full-range loudspeakers is 0.5 m.

4. The secondary sound source parameter optimization system based on an adjustable pan-tilt head and a fixed array according to claim 3, characterized in that: Two groups of high-frequency directional speakers with a frequency of 500Hz-1kHz are located on a two-degree-of-freedom pan-tilt platform. The azimuth angle of the high-frequency directional speakers is 0°-360°, and the pitch angle is -15°-45°.

5. The secondary sound source parameter optimization system based on an adjustable pan-tilt head and a fixed array according to claim 4, characterized in that: The microphone adopts a 16-channel MEMS microphone with a signal-to-noise ratio SNR≥70dB.

6. The collaborative control method of the secondary sound source parameter optimization system based on an adjustable pan-tilt head and a fixed array according to claim 5, characterized in that: The following steps are involved: S1. Noise detection Collect transformer noise through microphone; S2. Dynamic allocation of computing resources Use the time-sharing attention mechanism TSA to dynamically allocate computing resources; S3, Dynamic Beamforming Calculate the optimized beam weight vector w, the formula is: (1); Where R is the noise covariance matrix with dimension N×N, which represents the spatial correlation of the noise; a(θ) is the steering vector, dimension N×1, which represents the phase response of each unit when the sound wave is incident from the direction θ; a H (θ) represents the conjugate transpose of the steering vector a(θ); S4. Secondary sound source parameter optimization Use the physical information reinforcement learning (PIRL) algorithm to dynamically optimize the secondary sound source parameters, including direction, frequency, amplitude, and phase; S5. Generate control instructions Generate control instructions for the loudspeaker according to the optimization results of the secondary sound source parameters; S6, Noise Suppression The control instruction generated in step S5 is transmitted to the fixed infrasound source array and the adjustable platform module, and noise suppression is performed by adjusting the azimuth and pitch angles of the two-degree-of-freedom platform and the speaker power; S7. Feedback and Status Updates Perform residual noise detection. If the residual noise fails to meet the requirements, upload the execution results of parameter optimization to the physical information reinforcement learning (PIRL) algorithm and repeat step S4.

7. The method for collaboratively controlling secondary sound source parameters based on an adjustable pan-tilt head and a fixed array according to claim 6, characterized in that: The radiation pattern of the secondary sound source is determined by the two-degree-of-freedom pan-tilt angle and the fixed subsound source array layout, and its directivity function is (2); where θ is the vertical pitch angle of the main lobe of the sound beam, ranging from ±30°; φ is the horizontal azimuth angle, ranging from 0 to 360°; n is the directivity index, which controls the vertical radiation range of the sound source; N is the number of speakers in the fixed array; d is the distance between the speakers; k is the wave number, k= ;λ is the wavelength of the sound wave; is the target direction angle.

8. The method for collaboratively controlling secondary sound source parameters based on an adjustable pan-tilt head and a fixed array according to claim 7, characterized in that: By adjusting the phase difference Control the sound beam coverage D, that is (3); where D is the desired sound beam coverage.

9. The method for collaboratively controlling secondary sound source parameters based on an adjustable pan-tilt head and a fixed array according to claim 8, characterized in that: The control unit has a built-in physical information reinforcement learning (PIRL) algorithm. The quantitative reward mechanism of the PIRL algorithm includes: State space: Fusion of noise spectrum S(f) and sound field distribution P(x,y,z); Action space: pan / tilt angle (θ, φ), sound source frequency f, amplitude A, and phase; The reward function is (4); among them, is the noise reduction amount, and its calculation formula is L p,初始 -L p,残余 ;P consumed is the real-time power consumption, which is composed of the energy consumption of the source and the energy consumption of the pan / tilt motor drive; P max is the maximum allowable power consumption of the system; Stability(θ,φ) is the gimbal angle stability index, defined as ;in, is the gimbal angle fluctuation variance, is the maximum allowable fluctuation angle; 、 、 is the weight coefficient.

10. The method for collaboratively controlling secondary sound source parameters based on an adjustable pan-tilt head and a fixed array according to claim 9, characterized in that: By minimizing the noise power w H Rw, suppresses the interference corresponding to the noise covariance matrix while keeping the target direction signal undistorted. Each speaker optimizes the beam weight vector w and phase difference Output anti-phase sound waves to achieve optimal noise reduction; H is the conjugate transpose of w in the complex field.

Images