Noise suppression method and system for rear main reducer assembly of automobile

Abstract

The invention discloses an automobile rear main reducer assembly noise suppression method and system, and belongs to the technical field of automobile part noise control. Comprising the steps that noise data and vibration data are processed through a trained recurrent neural network, and a noise prediction signal at the next moment is acquired; the recurrent neural network is deployed in the FPGA chip to accelerate the convolution operation of the weight matrix; acquiring real-time working condition data of the automobile, constructing a real-time feature vector representing noise and working conditions according to the real-time working condition data and the noise prediction signal, performing fuzzy reasoning by using a fuzzy rule base according to the real-time feature vector, and generating a noise control strategy so as to perform noise suppression through corresponding noise reverse sound waves; constructing and dynamically updating a fuzzy rule base by using an improved fuzzy C-means clustering method; noise reverse sound waves are generated by a sound system including no less than two sound field generating units. The stability of the noise reduction result can be improved, and the problem that the noise control effect of an existing rear main reducer assembly is not ideal is solved.

Description

Technical Field

[0001] The invention relates to the technical field of automobile component noise control, and in particular to a method and system for suppressing noise of an automobile rear main reducer assembly. Background Art

[0002] The statements in this section merely mention background art related to the present invention and do not necessarily constitute prior art.

[0003] With the rapid development of the automobile industry and the continuous improvement of people's living standards, people have higher and higher requirements for automobile riding comfort. NVH (Noise, Vibration, and Harshness) is one of the main considerations.

[0004] During the operation of the RDU (Rear Drive Unit) of a car, due to the existence of its own dynamic imbalance, the unstable exciting force it is subjected to and the transmission error during the meshing process, a medium-high frequency noise called gear whine will be caused. At the same time, the exciting force generated will cause torsional and bending vibrations centered on the transmission shaft. This vibration centered on the transmission shaft is transmitted to the body through the connection point between the suspension longitudinal arm and the body, thereby generating noise that can be felt by the passengers. These noises will not only affect the riding experience of the passengers in the car, but may also have a negative impact on the overall performance of the car.

[0005] Some traditional passive noise reduction measures are difficult to effectively suppress noise in a wide frequency band. At the same time, optimizing gear design parameters and improving the vibration transmission path of the rear main reducer assembly face problems such as high cost and complex structure. In order to solve the noise problem, the specific methods currently adopted by the mainstream are as follows:

[0006] (1) Install sound insulation panels or sound insulation cotton on the vehicle body floor; isolate and absorb the noise path; (2) Press-fit the torsional vibration damper at the RDU flange to optimize vibration and reduce noise; (3) Modify the software strategy in the controller to disconnect and reduce torque of the four-wheel drive system at risk speeds; (4) Reduce the suspension stiffness of the subframe RDU connection point and absorb vibration through bushings. However, the above methods still have certain limitations, and the specific problems are as follows:

[0007] 1. The vehicle chassis is used in a harsh environment. Chassis sound insulation products will fail due to stone impact and water immersion, and the sound insulation products in the cabin will compress the effective space for passengers, increase production costs, and have limited effect in dealing with low-frequency noise. Low-frequency noise has a longer wavelength and can easily bypass the barrier of traditional sound insulation materials and enter the car.

[0008] 2. Elastic elements (such as springs) and damping elements (such as rubber, etc.) inside the torsional damper will experience problems such as wear and aging during long-term use. For example, rubber parts may lose their elasticity due to the relatively harsh temperature around the exhaust system, resulting in performance degradation, which in turn affects the normal operation of the torsional damper. Moreover, as the usage time increases, the damping effect of the torsional damper will gradually decrease, and it may need to be replaced regularly, which also increases the usage cost and the complexity of maintenance.

[0009] 3. Affect the four-wheel drive performance: At the same time, the disconnection of the four-wheel drive system will cause the passengers in the car to feel a sense of interrupted power output and a delay in the subsequent power recovery; Disconnecting and reducing the torque of the four-wheel drive system at the risk speed will cause the transmission components to bear a large impact. This sudden torque change will cause the force conditions between components such as the drive shaft and differential to change, easily resulting in increased wear of the transmission components. If this continues for a long time, it may shorten the service life of these components and increase the maintenance cost.

[0010] 4. Adjusting the bushing stiffness of the subframe results in a larger dynamic envelope of the RDU, reducing the safety clearance with surrounding parts and increasing the vehicle layout risk.

[0011] The existing methods for suppressing the noise of the rear main reducer in related technologies do not consider these influencing factors, resulting in an unsatisfactory control effect on the rear main reducer assembly, low control accuracy, and easy to cause passenger complaints. Summary of the Invention

[0012] To solve the deficiencies of the existing technology, the present invention provides a method, system, electronic device, computer-readable storage medium, and computer program product for suppressing the noise of an automotive rear main reducer assembly, which can accurately predict the change trend of the noise and perform active noise reduction in combination with the real-time working conditions of the vehicle, improving the noise suppression effect of the rear main reducer assembly.

[0013] In the first aspect, the present invention provides a method for suppressing the noise of an automotive rear main reducer assembly;

[0014] A method for suppressing the noise of an automotive rear main reducer assembly includes:

[0015] Obtain the noise data inside the cockpit and the vibration data of the rear main reducer assembly, and calculate the differences between the noise data and the vibration data and the vehicle calibration data;

[0016] If the difference is greater than a preset threshold, process the noise data and the vibration data through a trained recurrent neural network to obtain the noise prediction signal at the next moment; The recurrent neural network is deployed on an FPGA chip to accelerate the convolution operation of the weight matrix;

[0017] Obtain the real-time operating condition data of the vehicle, construct a real-time feature vector characterizing the noise and the operating condition according to the real-time operating condition data and the noise prediction signal, and perform fuzzy reasoning using the fuzzy rule base according to the real-time feature vector to generate a noise control strategy, so as to suppress the noise through the corresponding reverse acoustic wave of the noise;

[0018] Among them, an improved fuzzy C-means clustering method is used to construct and dynamically update the fuzzy rule base; the reverse acoustic wave of the noise is generated by an audio system including no less than two sound field generating units, the phase of the reverse acoustic wave of the noise is opposite to the noise prediction signal, and the amplitude of the reverse acoustic wave of the noise is dynamically adjusted according to the noise control strategy.

[0019] In some embodiments, before processing the noise data and the vibration data through a trained recurrent neural network, it further includes:

[0020] The noise data and the vibration data are sequentially processed through an FPGA chip to generate corresponding noise frequency domain signals.

[0021] In some embodiments, the recurrent neural network is trained with the goal of minimizing the mean square error of the time-frequency characteristics between the noise-free reference signal and the actual operating condition vibration signal.

[0022] In some embodiments, according to the real-time operating condition data and the noise prediction signal, constructing a real-time feature vector characterizing the noise and the operating condition includes:

[0023] Extract the time-domain features of the noise prediction signal to obtain the corresponding root mean square value; perform band-pass filtering and time-frequency conversion on the noise prediction signal in sequence, obtain the noise prediction frequency domain signal and perform feature extraction to determine the meshing frequency and the sideband amplitude; perform normalization processing on the real-time operating condition data to obtain the standard rotational speed and the standard torque;

[0024] Combine the root mean square value, the meshing frequency, the sideband amplitude, the standard rotational speed and the standard torque to generate a real-time feature vector.

[0025] In some embodiments, the step of performing fuzzy reasoning using the fuzzy rule base according to the real-time feature vector to generate a noise control strategy is specifically: according to the membership degrees of each rule in the fuzzy rule base corresponding to the real-time feature vector, calculate the real-time control parameter by weighting and output the corresponding noise control strategy.

[0026] In some embodiments, using the improved fuzzy C-means clustering method to construct and dynamically update the fuzzy rule base includes:

[0027] Obtain the vibration signal, the noise signal and the operating condition parameters, perform preprocessing and feature extraction, and obtain the feature vector;

[0028] Cluster the feature vectors by the fuzzy C - means clustering method, divide the fuzzy rule nodes, and store them and the corresponding fuzzy rules in the fuzzy rule base;

[0029] Among them, the clustering centers and widths of the membership functions are dynamically adjusted according to the real - time feature vectors.

[0030] In a second aspect, the present invention provides a noise suppression system for the rear main reducer assembly of an automobile;

[0031] A noise suppression system for the rear main reducer assembly of an automobile includes:

[0032] A comparison module configured to: obtain the noise data inside the cockpit and the vibration data of the rear main reducer assembly, and calculate the differences between the noise data and the vibration data and the vehicle calibration data;

[0033] A noise suppression module configured to: if the difference is greater than a preset threshold, process the noise data and the vibration data through a trained recurrent neural network to obtain the noise prediction signal for the next moment;

[0034] Obtain the real - time working condition data of the automobile, construct a real - time feature vector representing the noise and the working condition according to the real - time working condition data and the noise prediction signal, and perform fuzzy inference using the fuzzy rule base according to the real - time feature vector to generate a noise control strategy, so as to suppress the noise through the corresponding reverse acoustic wave of the noise;

[0035] Among them, the fuzzy rule base online updates the membership function parameters and the control strategy through self - organizing learning, and dynamically learns the mapping relationship between the features and the noise control by using the fuzzy clustering method; the reverse acoustic wave of the noise is jointly generated by the whole - cabin speakers and the headrest speakers, the phase of the reverse acoustic wave of the noise is opposite to the noise prediction signal, and the amplitude of the reverse acoustic wave of the noise is dynamically adjusted according to the noise control strategy.

[0036] In a third aspect, the present invention provides an electronic device;

[0037] An electronic device includes a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the steps of the above - mentioned noise suppression method for the rear main reducer assembly of an automobile.

[0038] In a fourth aspect, the present invention provides a computer - readable storage medium;

[0039] A computer - readable storage medium, on which a computer program / instructions are stored, and when the computer program / instructions are executed by a processor, the steps of the above - mentioned noise suppression method for the rear main reducer assembly of an automobile are implemented.

[0040] In a fifth aspect, the present invention provides a computer program product;

[0041] A computer program product includes a computer program / instructions, and when the computer program / instructions are executed by a processor, the steps of the above-mentioned method for suppressing the noise of the rear main reducer assembly of the vehicle are implemented.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] 1. For the technical solution provided by the present invention, a recurrent neural network is used to model the time series of the noise signal, and it is deployed on an FPGA chip. The high-speed parallel processing ability of the FPGA can accelerate the learning and analysis process of the RNN for the noise signal; when dealing with complex non-stationary noise of the RDU, the RNN can utilize the computing power of the FPGA to quickly learn the time series characteristics of the noise, so as to more accurately predict the change trend of the noise and provide a more effective basis for active noise reduction.

[0044] 2. For the technical solution provided by the present invention, the RNN can continuously adjust its own model parameters according to the input noise signal to achieve adaptive processing of the RDU signal; combined with the reconfigurability of the FPGA, the entire system can more flexibly adapt to different noise environments; when the type or intensity of the environmental noise suddenly changes, the RNN-FPGA system can quickly adjust its noise reduction strategy, re-optimize the parameters of the digital filter or adjust the generated anti-phase noise signal to maintain a good noise reduction effect.

[0045] 3. For the technical solution provided by the present invention, the FPGA provides efficient computing resources at the hardware level, while the RNN provides powerful sequence data processing capabilities at the software algorithm level. The combination of the two can improve the overall performance of the present invention in terms of power consumption, computing speed, noise reduction effect, etc.; compared with using traditional noise reduction methods alone or relying solely on software-implemented RNN noise reduction, this combination method can achieve faster and more accurate noise reduction operations at lower power consumption.

[0046] 4. For the technical solution provided by the present invention, the noise reverse sound waves are generated jointly by the whole-cabin speakers and the headrest speakers. The vehicle's audio system can process the noise in the entire carriage space and macroscopically reduce the overall noise level inside the vehicle. The headrest speaker system can, on the basis of the overall noise reduction of the whole vehicle, provide personalized noise reduction supplements for each seat area.

[0047] 5. For the technical solution provided by the present invention, when generating the noise control strategy, the fuzzy rule base is adaptively updated by combining the real-time working condition data of the vehicle and the noise prediction data of the next moment to adapt to the changing real-time state of the vehicle, improve the accuracy of noise control, and ensure the noise reduction effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The accompanying drawings forming a part of this invention are used to provide a further understanding of the invention. The schematic embodiments and descriptions thereof of the invention are used to explain the invention and shall not unduly limit the invention.

[0049] Figure 1 It is a schematic flow chart of the method for suppressing the noise of the rear main reducer assembly of an automobile provided by an embodiment of the present invention;

[0050] Figure 2 It is a schematic logic diagram of the method for suppressing the noise of the rear main reducer assembly of an automobile provided by an embodiment of the present invention;

[0051] Figure 3 It is a schematic structural diagram of a headrest provided by an embodiment of the present invention;

[0052] Figure 4 It is a schematic calculation structure diagram provided by an embodiment of the present invention;

[0053] Figure 5 It is a schematic architecture diagram of the noise suppression system for the rear main reducer assembly of an automobile provided by an embodiment of the present invention. Detailed Embodiments

[0054] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used in the present invention have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0055] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0056] In the case of no conflict, the embodiments in the present invention and the features in the embodiments may be combined with each other.

[0057] Embodiment 1

[0058] Existing noise control methods for automotive rear main reduction assemblies mainly focus on improving the vehicle's structure, which affects the service life and maintenance cost of automotive components, and the noise control effect is not ideal, affecting the passenger experience. Therefore, the present invention provides a method for suppressing noise in an automotive rear main reduction assembly, deploying a recurrent neural network on an FPGA chip, using its superior computing performance to improve data processing efficiency, and adaptively updating in combination with real-time operating condition parameters of the vehicle to improve the accuracy of noise suppression.

[0059] Next, in combination with Figures 1-4 , a method for suppressing noise in an automotive rear main reduction assembly disclosed in this embodiment will be described in detail. The method for suppressing noise in an automotive rear main reduction assembly includes the following steps:

[0060] S1. In the four-wheel drive mode of the vehicle, obtain the noise data inside the cockpit and the vibration data of the rear main reduction assembly, calculate the differences between the noise data and the vibration data and the vehicle calibration data. If the difference is greater than the preset threshold, execute S2; otherwise, continue to obtain the noise data inside the cockpit and the vibration data of the rear main reduction assembly.

[0061] Here, the noise data inside the cockpit is collected by a microphone and can be represented as a noise signal sequence X N , and the vibration data of the rear main reduction assembly is collected by a vibration sensor installed on the rear main reduction assembly and can be represented as a vibration load spectrum Y N . The vibration load spectrum can determine the frequency and amplitude information of the noise source during the noise analysis process.

[0062] The above sensors should have high sensitivity and high fidelity to ensure that the noise signals generated by the rear main reduction assembly can be accurately captured.

[0063] Specifically, perform a vector subtraction on the noise signal sequence X N and the vibration load spectrum Y N with the calibration signal sequence provided by the vehicle manufacturer to obtain the absolute value K of the signal difference. When K is greater than the given noise threshold, execute S2.

[0064] In this embodiment, the noise threshold is 50 dB.

[0065] Furthermore, before executing S2, a pop-up window is displayed on the touch screen to remind whether to execute noise control. If the user does not intervene, S2 will be automatically executed.

[0066] S2. Process the noise data and the vibration data through a trained recurrent neural network to obtain the noise prediction signal for the next moment.

[0067] Further, before executing S2, digital signal processing is performed on the noise data and vibration data through an FPGA chip. Specifically, first, the XY composed of the noise data and vibration data is processed successively through the Cooley-Tukey algorithm (butterfly algorithm) and Fourier transform, so that the signal is transformed from a high number of points to multiple low numbers of points, and the noise frequency domain signal sequence XY is output. N T .

[0068] It should be noted that the entire Fourier transform can be composed of the Fourier transform of base 2 and base 4. The 2k FFT (Fast Fourier Transform) can be realized through 5 base 4 and 1 base 2 transforms, the 4k FFT transform can be realized through 6 base 4 transforms, and the 8k FFT can be realized through 6 base 4 and 1 base 2 transforms. The butterfly operation unit is the base 2 / 4 module.

[0069] In this embodiment, the RAM of the FPGA chip is used to store the input data, the intermediate results during the operation, and the data after the operation is completed, and the ROM is used to store the rotation factor table; the algorithm and database can be upgraded through the NB-LOT module integrated in the FPGA chip.

[0070] The control module of the FPGA chip generates control timing and address signals through state machine logic to coordinate the operation processes of each sub-module inside the FPGA, specifically including:

[0071] 1. Timing control: Define the phased clock beats of subsequent FFT, RNN prediction, and fuzzy inference.

[0072] 2. Address management: Dynamically allocate the read and write addresses of the RAM according to the operation stage (such as the original data storage address 0x0000 - 0x0FFF, the intermediate result address 0x1000 - 0x1FFF), and trigger the reading of the rotation factor table in the ROM.

[0073] 3. Result output: Write the finally generated reverse acoustic control instruction into the specified RAM area for the speaker drive module to call.

[0074] ​FPGA chips have the characteristics of high parallel processing ability and low latency. They can quickly process the input noisy signals. FPGAs can easily implement various digital filter structures, such as FIR (Finite Impulse Response filter) and IIR (Infinite Impulse Response filter), so as to customize the frequency characteristics of the RDU. Digital signal noise reduction in FPGAs is a key technology to optimize signal quality and improve system stability. Through effective noise reduction, the signal processing accuracy can be improved, errors can be reduced, the characteristics of noise signals can be more accurately identified, so as to generate a more appropriate inverted noise signal to cancel the original noise, improve the stability of the noise reduction effect, and avoid large fluctuations in the noise reduction effect. And FPGAs can perform resource allocation and customization according to specific signal processing requirements. In this application, signals of multiple channels need to be processed (such as environmental noise collected by multiple microphones), and resources such as logic gates and storage units of FPGAs can be flexibly allocated to meet the processing requirements of different channel signals and improve the overall noise reduction efficiency.

[0075] In this embodiment, the FPGA chip can be of the Xilinx Zynq-7000 series, and the number of logic units ≥ 85k.

[0076] As an implementation method, S2 is specifically: the noise frequency domain signal sequence XY T is input into the recurrent neural network RNN for processing. The noise frequency domain signal sequence XY T is unfolded in time to obtain a neuron structure, and after passing through the tanh activation function and the sigmoid activation function, a vector matrix XY K is output. The vector matrix XY K is subjected to an inverse Fourier transform to convert the signal back to the time domain, and the final noise prediction result is output to provide accurate basic data for the future time period for subsequent noise suppression.

[0077] Moreover, through the processing of the recurrent neural network, the eigenvalue of the input data is captured, making the output result lighter.

[0078] Specifically, combined with Figure 4 , the weight from the input layer to the hidden layer is represented by U, which abstracts the original input as the input of the hidden layer; the weight W from the hidden layer to the hidden layer is responsible for scheduling memory and performing network memory control; the weight V from the hidden layer to the output layer performs a secondary abstraction from the hidden layer and serves as the final output.

[0079] Figure 4 In, S tis the "memory" at time t; f() represents a non-linear transformation function, such as an activation function like tanh, S t represents a neuron that has two outputs; o t represents the output at time t (the output result after neuron activation). For example, if it is to predict the next signal, it may be the probability of each signal value output by sigmoid (softmax), O t = softamx(VS t ), where Softmax represents a normalized exponential function, with each element between 0 and 1 and the sum equal to 1.

[0080] Furthermore, when training a recurrent neural network, first, select a recurrent neural network, determine hyperparameters such as the number of layers of the network and the number of neurons in each layer, and construct the model structure; then, based on the previously collected and preprocessed data, with the minimization of the mean square error of the time-frequency characteristics of the noise-free reference signal and the actual working condition vibration signal as the training objective, train the recurrent neural network to optimize the parameters of the recurrent neural network.

[0081] Furthermore, to avoid the problem of gradient loss caused by too large a gradient, the memory gradient W is intercepted by an algorithm to prevent the problem of signal explosion.

[0082] In this embodiment, the algorithm can be truncation by value and truncation by norm. For example, when training a recurrent neural network, if the gradient exceeds a certain threshold, it is limited within that threshold to avoid excessive parameter updates.

[0083] S3. Obtain the real-time working condition data of the vehicle, construct a real-time feature vector representing the noise and the working condition according to the real-time working condition data and the noise prediction signal, and use the fuzzy rule base for fuzzy inference based on the real-time feature vector to generate a noise control strategy to suppress noise through the corresponding noise reverse sound wave. Specifically, it includes:

[0084] S301. Extract the time-domain features of the noise prediction signal to obtain the corresponding root mean square value; perform band-pass filtering on the noise prediction signal, perform time-frequency conversion on the band-pass filtered noise prediction signal through FFT to obtain the noise prediction frequency-domain signal and perform feature extraction to determine the meshing frequency f m and the sideband amplitude A side ; perform normalization processing on the real-time working condition data to obtain the standard rotational speed and the standard torque; combine the root mean square value, the meshing frequency, the sideband amplitude, the standard rotational speed, and the standard torque to generate a real-time feature vector

[0085] Among them, the meshing frequency is expressed as:

[0086] f m= f 1 ·z 1 = f 2 ·z 2 = (n 1 / 60)·z 1 = (n 2 / 60)·z 2 ;

[0087] In the formula, f 1 represents the rotational frequency of the driving wheel, f 2 represents the rotational frequency of the driven wheel, z 1 represents the number of teeth of the driving wheel, z 2 represents the number of teeth of the driven wheel, n 1 represents the rotational speed of the driving wheel, n 2 represents the rotational speed of the driven wheel.

[0088] The root mean square value is expressed as:

[0089]

[0090] Band-pass filtering is expressed as:

[0091] x(t) = Butterworth(x(t), f low , f high );

[0092] The characteristic frequency band of the main reduction gear meshing is retained through band-pass filtering (such as 500 kHz - 2 kHz).

[0093] The standard rotational speed is expressed as:

[0094]

[0095] The standard torque is expressed as:

[0096]

[0097] Here, in the spectrogram, the sidebands are frequency components symmetrically distributed on both sides centered on the meshing frequency, and the sideband amplitudes can be directly read through existing software. In addition, the relative ratio of the sideband amplitude to the meshing frequency amplitude can be used as an index to judge the severity of gear faults. As the gear fault develops, the sideband amplitude may increase relative to the meshing frequency amplitude.

[0098] S302. According to the membership degree u of each rule in the fuzzy rule base corresponding to the real-time feature vector ij, calculate the real-time control parameters by weighted calculation, and screen the corresponding phase compensation and damping gain in the fuzzy rule base according to the real-time control parameters as the corresponding noise control strategy, so as to input to the cabin speakers and headrest speakers to generate noise reverse sound waves cooperatively. The phase of the noise reverse sound wave is opposite to that of the noise prediction signal, and the amplitude of the noise reverse sound wave is dynamically adjusted according to the inference result.

[0099] In noise control, the opposite phase of the noise reverse sound wave to the noise prediction signal is the basis for achieving destructive interference. And the degree of phase compensation will indirectly affect the amplitude adjustment. For example, if the phase compensation is inaccurate, in order to achieve a better noise cancellation effect, it may be necessary to adjust the amplitude. Assuming there is an ideal cancellation point, when the phase deviates from this ideal value, it may be necessary to increase or decrease the amplitude of the reverse sound wave to make up for the insufficient cancellation effect.

[0100] Among them, the real-time control parameters are expressed as:

[0101]

[0102] Furthermore, before executing S3, it also includes: constructing the fuzzy rule base by using the improved fuzzy C-means clustering method, and dynamically updating the fuzzy rule base by using real-time data; the specific process is as follows:

[0103] Step 1, obtain the original vibration signal, original noise signal and working condition parameters, perform preprocessing and feature extraction, and obtain multiple feature vector sequences representing noise and working conditions

[0104]

[0105] The data processing method here is the same as that in S301 and will not be elaborated here.

[0106] Step 2, use the fuzzy C-means method to cluster the feature vector sequence F N and divide the fuzzy rule nodes R j .

[0107] Among them, the membership function corresponding to the fuzzy rule node R j is expressed as:

[0108]

[0109] In the formula, c j represents the clustering center, m represents the fuzzy index, and C represents the number of rules.

[0110] Furthermore, in each fuzzy inference, update the clustering center c j and width σ j of the membership function according to the real-time feature vector.

[0111] By adjusting the mathematical representation of the fuzzy rule base in real time, the robustness and accuracy of noise suppression are improved; essentially, the fuzzy rule base is updated through online self-organizing learning to meet the high-precision noise reduction under complex working conditions.

[0112] Exemplarily, the updated cluster center is expressed as:

[0113]

[0114] The updated width is expressed as:

[0115]

[0116] In the formula, η represents the learning rate, which controls the convergence speed.

[0117] Based on this, the cluster center and width can be continuously optimized according to new data, so that the fuzzy rule base can be adaptively adjusted according to the vehicle state changes and is suitable for different working conditions.

[0118] Step 3: Each cluster center corresponds to a fuzzy rule, which is stored in the fuzzy rule base and expressed as:

[0119] IFF belongs to R j , THEN control parameter is phase compensation, K j is the damping gain.

[0120] To sum up, the method of generating a noise suppression strategy through fuzzy inference in this step has strong adaptability. Through self-organizing learning, it can cope with complex working conditions (such as sudden changes in speed and load changes). According to the vibration, noise, and working condition data collected in real time, the control strategy is dynamically adjusted, significantly reducing the noise of the rear main reducer assembly; the fuzzy inference calculation amount is low, meeting the real-time requirements of noise control during the driving of four-wheel drive vehicles; and the fuzzy rule base is stored in the form of "IF-THEN", which is convenient for engineers to debug and optimize.

[0121] Furthermore, in some embodiments, a shift button can be pre-installed on the side of the car seat. The user can divide the noise reduction depth into three gears according to their own noise reduction needs through the pre-installed shift button and the program OTA (Over The Air wireless download technology). The noise reduction depths are 40%, 70%, and 100% of the maximum noise reduction depth respectively.

[0122] It has been experimentally proven that after being processed by the method for suppressing the noise of the rear main reducer assembly of the vehicle described in this embodiment, the 1 / 3 octave band sound pressure level attenuation ≥ 15 dB(A), the speech intelligibility index is improved by ≥ 0.25, the transmission loss in the 20 - 2000 Hz frequency band ≥ 18 dB, and the phase delay ≤ 0.15 ms.

[0123] Embodiment 2

[0124] Combined with Figure 5 , this embodiment discloses a noise suppression system for the rear main reducer assembly of an automobile, including:

[0125] A comparison module, configured to: obtain the noise data inside the cockpit and the vibration data of the rear main reducer assembly, and calculate the differences between the noise data and the vibration data and the vehicle calibration data;

[0126] A noise suppression module, configured to: if the difference is greater than a preset threshold, process the noise data and the vibration data through a trained recurrent neural network to obtain the noise prediction signal for the next moment;

[0127] Obtain the real-time working condition data of the automobile, construct a real-time feature vector representing the noise and the working condition according to the real-time working condition data and the noise prediction signal, and perform fuzzy reasoning using the fuzzy rule base according to the real-time feature vector to generate a noise control strategy, so as to suppress the noise through the corresponding reverse acoustic wave of the noise;

[0128] Wherein, the fuzzy rule base updates the membership function parameters and the control strategy online through self-organizing learning, and dynamically learns the mapping relationship between the features and the noise control by using the fuzzy clustering method; the reverse acoustic wave of the noise is jointly generated by the full-cabin speaker and the headrest speaker, the phase of the reverse acoustic wave of the noise is opposite to the noise prediction signal, and the amplitude of the reverse acoustic wave of the noise is dynamically adjusted according to the noise control strategy.

[0129] It should be noted here that the above comparison module and noise suppression module correspond to the steps in Embodiment 1. The examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1 above. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0130] Embodiment 3

[0131] Embodiment 3 of the present invention provides an electronic device, including a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the steps of the above-mentioned noise suppression method for the rear main reducer assembly of the automobile are completed.

[0132] Embodiment 4

[0133] Embodiment 4 of the present invention provides a computer-readable storage medium for storing computer instructions. When the computer instructions are executed by a processor, the steps of the above-mentioned noise suppression method for the rear main reducer assembly of the automobile are completed.

[0134] Embodiment 5

[0135] Embodiment 5 of the present invention provides a computer program product, including computer programs / instructions, which, when executed by a processor, implement the steps of the above-mentioned method for suppressing noise in the rear main reducer assembly of an automobile.

[0136] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0137] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0138] These computer program instructions can also be loaded onto a computer or other programmable data processing device, and a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0139] In the above embodiments, the descriptions of each embodiment have their own focuses. For parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0140] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. For those skilled in the art, the present invention can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for suppressing noise of a rear main reducer assembly of an automobile, characterized in that: include: Obtain the noise data in the cabin and the vibration data of the rear main reduction assembly, and calculate the difference between the noise data and the vibration data and the vehicle calibration data; If the difference is greater than the preset threshold, the noise data and vibration data are processed through the trained recurrent neural network to obtain the noise prediction signal at the next moment; The recurrent neural network is deployed on an FPGA chip to accelerate the convolution operation of the weight matrix; Acquire the real-time working condition data of the automobile, construct a real-time feature vector representing the noise and working condition based on the real-time working condition data and the noise prediction signal, perform fuzzy reasoning based on the real-time feature vector using the fuzzy rule base, and generate a noise control strategy to suppress noise through the corresponding noise reverse sound wave; Wherein, the fuzzy rule base is constructed and dynamically updated using an improved fuzzy C-means clustering method; The noise reverse sound wave is generated by a sound system including no less than two sound field generating units, the phase of the noise reverse sound wave is opposite to the noise prediction signal, and the amplitude of the noise reverse sound wave is dynamically adjusted according to the noise control strategy.

2. The method for suppressing noise of a rear main reducer assembly of an automobile according to claim 1, characterized in that: Before the noise data and vibration data are processed by the trained recurrent neural network, it also includes: The noise data and vibration data are processed by the FPGA chip to generate corresponding noise frequency domain signals.

3. The method for suppressing noise of a rear main reducer assembly of an automobile according to claim 1, characterized in that: The recurrent neural network is trained with the goal of minimizing the mean square error of the time-frequency characteristics between the noise-free reference signal and the actual working condition vibration signal.

4. The method for suppressing noise of a rear main reducer assembly of an automobile according to claim 1, characterized in that: According to the real-time working condition data and noise prediction signal, the real-time feature vector characterizing the noise and working condition is constructed, including: Perform time domain feature extraction on the noise prediction signal to obtain the corresponding root mean square value; perform bandpass filtering and time-frequency conversion on the noise prediction signal in sequence to obtain the noise prediction frequency domain signal and perform feature extraction to determine the meshing frequency and sideband amplitude; perform normalization on the real-time operating condition data to obtain the standard speed and standard torque; The RMS value, meshing frequency, sideband amplitude, standard speed and standard torque are combined to generate a real-time feature vector.

5. The method for suppressing noise of a rear main reducer assembly of an automobile according to claim 1, characterized in that: The method of generating a noise control strategy by performing fuzzy reasoning based on the real-time feature vector and utilizing a fuzzy rule base is as follows: according to the membership degree of each rule in the fuzzy rule base corresponding to the real-time feature vector, weighted calculation of real-time control parameters is performed and the corresponding noise control strategy is output.

6. The method for suppressing noise of a rear main reducer assembly of an automobile according to claim 1, characterized in that: Using the improved fuzzy C-means clustering method to construct and dynamically update the fuzzy rule base includes: Obtain vibration signals, noise signals and operating parameters and perform preprocessing and feature extraction to obtain feature vectors; The feature vectors are clustered by fuzzy C-means clustering method, fuzzy rule nodes are divided, and the nodes and corresponding fuzzy rules are stored in the fuzzy rule base; The cluster center and width of the membership function are dynamically adjusted according to the real-time feature vector.

7. A noise suppression system for a rear main reducer assembly of an automobile, characterized in that: include: The comparison module is configured to: obtain the noise data in the cabin and the vibration data of the rear main reduction assembly, and calculate the difference between the noise data and the vibration data and the vehicle calibration data; The noise suppression module is configured to: if the difference is greater than a preset threshold, process the noise data and the vibration data through a trained recurrent neural network to obtain a noise prediction signal at the next moment; The recurrent neural network is deployed on an FPGA chip to accelerate the convolution operation of the weight matrix; Acquire the real-time working condition data of the automobile, construct a real-time feature vector representing the noise and working condition based on the real-time working condition data and the noise prediction signal, perform fuzzy reasoning based on the real-time feature vector using the fuzzy rule base, and generate a noise control strategy to suppress noise through the corresponding noise reverse sound wave; Wherein, the fuzzy rule base is constructed and dynamically updated using an improved fuzzy C-means clustering method; The noise reverse sound wave is generated by a sound system including no less than two sound field generating units, the phase of the noise reverse sound wave is opposite to the noise prediction signal, and the amplitude of the noise reverse sound wave is dynamically adjusted according to the noise control strategy.

8. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method for suppressing noise of a rear main reducer assembly of an automobile according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instruction is executed by a processor, the steps of the method for suppressing noise of a rear main reducer assembly of an automobile as described in any one of claims 1 to 6 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction is executed by a processor, the steps of the method for suppressing noise of a rear main reducer assembly of an automobile as described in any one of claims 1 to 6 are implemented.

Images