A noise active control method and system for a new energy vehicle

Abstract

The application discloses a new energy automobile noise active control method and system, belongs to the automobile noise reduction technical field, and is used for solving the technical problems that the stability of the existing noise active control technology is poor, the influence of road conditions on cavity noise is not considered, the tire noise cannot be predicted and inhibited in advance according to the road conditions in front, and the noise reduction effect is not ideal. The method comprises the following steps: constructing a target data set based on noise signal characteristics and road point cloud data; training a noise signal prediction model through the target data set; inputting real-time road point cloud data in front of a vehicle into the noise signal prediction model to obtain noise signal prediction characteristics; generating a noise pre-regulation signal and preliminarily inhibiting vehicle noise according to the noise signal prediction characteristics; and inputting the real-time noise signal obtained after preliminary inhibition into an adaptive filter to compensate the noise pre-regulation signal, so as to further inhibit the vehicle noise.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automobile noise reduction technology, and in particular to a noise active control method and system for a new energy vehicle. BACKGROUND

[0002] At present, most new energy vehicles are driven by electric motors. Due to the lack of engine masking, tire cavity resonance noise, road noise, and wind noise are more likely to reduce the ride comfort of the vehicle. Tire cavity resonance noise is generated by the resonance of air inside the vehicle tire cavity excited by road unevenness, and this noise belongs to low-frequency noise, mainly distributed in the range of 30-300Hz. This low-frequency noise is easy to make people feel annoyed, anxious and uncomfortable, and is extremely likely to affect the driving experience of the driver. Therefore, in the field of new energy vehicles, how to suppress tire cavity resonance noise has become an important issue.

[0003] With the progress of science and technology, current vehicle noise control methods are mainly divided into active control technology and passive control technology. The principle of active control technology is to "eliminate noise with sound", that is, by installing specific sensors and speaker systems, the vehicle can monitor and analyze the frequency and phase of the noise in real time, and then generate an opposite sound wave to reduce or eliminate the noise. Passive control technology is to use sound barrier materials or sound absorbing materials at key parts of the vehicle during the engineering prototype stage to weaken the noise. Passive control technology is widely used at present, but the noise suppression effect is not ideal. Active control technology has poor stability and low accuracy, and is less used in practice. In addition, vehicle cavity noise is greatly affected by road conditions. The noise generated by different road conditions is very different. The existing active control technology only focuses on suppressing the actual tire noise signal generated at the current time, but does not consider the influence of road conditions on tire noise. It is impossible to predict and generate a matching noise suppression signal in advance according to the road conditions ahead. SUMMARY

[0004] The embodiments of the present application provide a noise active control method and system for a new energy vehicle, which is used to solve the following technical problems: the existing noise active control technology has poor stability, and does not consider the influence of road conditions on cavity noise, which cannot predict and suppress tire noise in advance according to the road conditions ahead, and the noise reduction effect is not ideal.

[0005] The embodiments of the present application adopt the following technical solutions:

[0006] On the one hand, the embodiments of the present application provide a noise active control method for a new energy vehicle, which comprises: performing a driving test on a target vehicle model, and simultaneously collecting noise signals and road point cloud data in front of the vehicle;

[0007] extracting a time-frequency domain feature of the noise signal to obtain a noise signal feature, and constructing a target data set based on the noise signal feature and the road surface point cloud data; and training a noise signal prediction model through the target data set;

[0008] acquiring real-time road surface point cloud data in front of the vehicle during driving of the target vehicle model, and performing quality evaluation on the real-time road surface point cloud data to obtain a road surface quality evaluation value;

[0009] in a case where the road surface quality evaluation value is lower than a first preset threshold, inputting the real-time road surface point cloud data into the noise signal prediction model to obtain a noise signal prediction feature;

[0010] generating a noise pre-adjustment signal and preliminarily suppressing vehicle noise according to the noise signal prediction feature;

[0011] inputting the real-time noise signal obtained after preliminary suppression into an adaptive filter, compensating the noise pre-adjustment signal, and further suppressing vehicle noise.

[0012] In a feasible implementation, a driving test is performed on a target vehicle model, and noise signals and road surface point cloud data in front of the vehicle are collected, specifically including:

[0013] acceleration sensors are installed at a plurality of positions of the target vehicle model, and an error microphone is installed at a target noise reduction point position;

[0014] a vibration excitation test is performed on the target vehicle model in a stationary state, vibration acceleration signals at the installation positions are acquired through the acceleration sensors, and sound signals at the target noise reduction point position are acquired through the error microphone;

[0015] coherent analysis is performed on the vibration acceleration signals collected at each position to determine a preset number of recommended installation positions of the acceleration sensors, and the acceleration sensors installed at non-recommended installation positions are removed;

[0016] vehicle driving tests are performed on different quality road surfaces at different driving speeds, driving vibration acceleration signals are collected through the acceleration sensors installed at the recommended installation positions, driving sound signals at the target noise reduction point position are acquired through the error microphone, and road surface point cloud data within a preset distance in front of the vehicle are acquired through a laser radar installed at the front of the vehicle;

[0017] the driving vibration acceleration signals are taken as reference signals, and the driving sound signals are taken as error signals to constitute the noise signals.

[0018] In an implementable embodiment, the vibration acceleration signals collected at each position are subjected to coherence analysis to determine a preset number of recommended installation positions of the acceleration sensor, specifically comprising:

[0019] The cross-coherence function value of the vibration acceleration signal collected at each position and the corresponding sound signal is calculated, and a target acceleration sensor with a cross-coherence function value greater than a second preset threshold is screened out;

[0020] Based on the preset number, the target acceleration sensor is arranged and combined, and the re-coherence function value of the vibration acceleration signal in each combination is calculated;

[0021] The installation position of the target acceleration sensor corresponding to the vibration acceleration signal in a combination with the largest re-coherence function value is determined as the recommended installation position.

[0022] In an implementable embodiment, the noise signal is subjected to time-frequency domain feature extraction to obtain a noise signal feature; and based on the noise signal feature and the road surface point cloud data, a target data set is constructed, specifically comprising:

[0023] The noise signal is subjected to time-frequency domain feature extraction based on a preset feature extraction algorithm to obtain a corresponding noise signal feature;

[0024] Based on the vehicle speed, the delay time between the noise signal collected at the same road surface position and the road surface point cloud data is calculated;

[0025] Based on the delay time, the noise signal feature collected at the same road surface position and the road surface point cloud data are one-to-one corresponding to obtain a plurality of data groups, which constitute the target data set.

[0026] In an implementable embodiment, the noise signal is subjected to time-frequency domain feature extraction based on a preset feature extraction algorithm to obtain a corresponding noise signal feature, specifically comprising:

[0027] The reference signal and the error signal in the noise signal are subjected to continuous wavelet transform respectively to obtain a time-frequency distribution matrix of each signal;

[0028] The time-frequency distribution matrix is subjected to singular value decomposition to obtain a singular eigenvalue corresponding to each signal;

[0029] Based on the singular eigenvalue, the inter-class dispersion and the intra-class dispersion of each type of signal are calculated, and the separability threshold is determined according to the inter-class dispersion and the intra-class dispersion;

[0030] Based on the separability threshold, the singular eigenvalue of each signal is evaluated and screened to obtain a singular eigenvalue subset, which constitutes the noise signal feature.

[0031] In a feasible implementation, before training the noise signal prediction model through the target data set, the method further comprises:

[0032] A causal convolution layer and a dilated convolution layer are added between the input layer and the output layer of the convolutional neural network, each convolution layer is connected through a ReLU activation function, and the output end of the last convolution layer is connected to the original input through a jump connection to form a residual link, thereby constituting a local feature extraction module; wherein the causal convolution layer is used to extract the causal relationship between input features, and the dilated convolution layer is used to increase the receptive field of the convolution kernel and capture long-distance dependency between features;

[0033] The output end of the local feature extraction module is connected to the input end of a Transformer encoder to build the noise signal prediction model.

[0034] In a feasible implementation, during driving of the target vehicle model, real-time point cloud data of a road surface in front of the vehicle is acquired, and quality evaluation is performed on the real-time point cloud data of the road surface to obtain a road surface quality evaluation value, specifically including:

[0035] During actual driving of the target vehicle model, real-time point cloud data of a road surface within a preset distance in front of the vehicle is acquired through a laser radar installed at the front of the vehicle;

[0036] Based on the real-time point cloud data of the road surface, a quality parameter of the road surface within the preset distance in front of the vehicle is determined; wherein the quality parameter at least includes one or more of the following: road surface material, road surface flatness, road surface roughness, road surface damage condition, and road surface anti-skid performance;

[0037] The quality parameter is assigned a weight value and weighted calculation is performed to obtain the road surface quality evaluation value.

[0038] In a feasible implementation, according to the noise signal prediction feature, a noise pre-adjustment signal is generated and vehicle noise is preliminarily suppressed, specifically including:

[0039] According to the noise signal prediction feature, frequency information and phase information of the predicted noise signal are extracted;

[0040] Based on the frequency information and the phase information, a noise pre-adjustment signal with the same frequency and opposite phase of the predicted noise signal is generated;

[0041] Based on the current driving speed of the vehicle, a delay time required for the vehicle to travel from the current position to the road surface within the preset distance in front of the vehicle is calculated;

[0042] After the delay time, the noise pre-adjustment signal is played through the in-vehicle speaker system to preliminarily suppress the vehicle noise.

[0043] In an implementable embodiment, the real-time noise signal obtained after preliminary suppression is input into an adaptive filter to compensate the noise pre-adjustment signal, so as to further suppress the vehicle noise, specifically including:

[0044] The real-time noise signal after preliminary suppression is obtained through a noise signal acquisition device installed in the target vehicle model;

[0045] The real-time noise signal is input into an LMS adaptive filter to output a control signal with opposite phase, which is superimposed with the noise pre-adjustment signal and then played through the in-vehicle speaker system to further suppress the vehicle noise.

[0046] On the other hand, the embodiment of the present application also provides a noise active control system for a new energy vehicle, which comprises:

[0047] A model training module is configured to perform a driving test on a target vehicle model, collect noise signals and road point cloud data in front of the vehicle, extract time-frequency domain features of the noise signals to obtain noise signal features, construct a target data set based on the noise signal features and the road point cloud data, and train a noise signal prediction model through the target data set;

[0048] A noise pre-adjustment module is configured to obtain real-time road point cloud data in front of the vehicle during driving of the target vehicle model, perform quality assessment on the real-time road point cloud data to obtain a road quality assessment value, input the real-time road point cloud data into the noise signal prediction model to obtain noise signal prediction features when the road quality assessment value is lower than a first preset threshold, and generate a noise pre-adjustment signal based on the noise signal prediction features to preliminarily suppress the vehicle noise;

[0049] A noise real-time suppression module is configured to input the real-time noise signal obtained after preliminary suppression into an adaptive filter to compensate the noise pre-adjustment signal, so as to further suppress the vehicle noise.

[0050] Compared with the prior art, the noise active control method and system for a new energy vehicle provided by the embodiment of the present application have the following beneficial effects:

[0051] The application obtains road condition data and noise signals through driving test on the target vehicle, and learns the correlation between the road condition data and the noise signals through the convolution neural network model designed by itself, so as to realize the prediction of the noise signal characteristics that may be generated after the vehicle drives to the position according to the road condition in front of the vehicle, generate the suppression signal in advance, compensate for the short delay caused by the operation of the existing noise active control algorithm, avoid the signal suppression lag, and different noise suppression according to different road conditions, improve the noise suppression effect, and make the noise suppression feel smoother.

[0052] After the advance suppression, the application further suppresses the real-time noise signal after the preliminary suppression through the adaptive filter, can further improve the noise suppression effect on the basis of the preliminary suppression, can reduce the noise decibel to a smaller range, and according to the experimental verification result, the noise suppression effect of the algorithm proposed in the application is better than that of the existing active control algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description only illustrate some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor. In the drawings:

[0054] Figure 1 A flow chart of a noise active control method of a new energy vehicle provided by the embodiment of the present application;

[0055] Figure 2 A structural schematic diagram of a noise active control system of a new energy vehicle provided by the embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to make the person skilled in the art better understand the technical solutions in the present application, the following will combine the drawings in the embodiment of the present application to clearly and completely describe the technical solutions in the embodiment of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor should belong to the protection scope of the present application.

[0057] The embodiment of the present application provides a noise active control method of a new energy vehicle, as shown in Figure 1 The noise active control method of the new energy vehicle specifically includes steps S101-S106:

[0058] S101, a driving test is performed on the target vehicle model, and noise signals and road surface point cloud data in front of the vehicle are collected.

[0059] Specifically, in the vehicle model development stage, the target vehicle model is first subjected to excitation test in a static state, and the test process is as follows:

[0060] Acceleration sensors are installed at several positions of the target vehicle model, and error microphones are installed at target noise reduction points. The target noise reduction points can be set as the positions of the driver's ears. Then the exciter is arranged at positions where the target vehicle model is prone to generate low-frequency noise, such as the hub, engine, chassis, etc., to perform excitation test.

[0061] Then the vibration acceleration signals of the installation positions are obtained by the acceleration sensors, and the sound signals of the target noise reduction points are obtained by the error microphones. Then the vibration acceleration signals collected at each position are subjected to coherence analysis to determine the preset number of recommended installation positions of the acceleration sensors, and the acceleration sensors installed at non-recommended installation positions are removed.

[0062] As a feasible implementation, the specific process of obtaining the recommended installation position is as follows: first, the cross-coherence function value of the vibration acceleration collected at each position and the corresponding sound signal is calculated, and the target acceleration sensor with a cross-coherence function value greater than a second preset threshold is selected. Then, based on the preset number, the target acceleration sensors are arranged in combination, and the re-coherence function value of the vibration acceleration signal in each combination is calculated. The installation position of the target acceleration sensor corresponding to the vibration acceleration signal in the combination with the maximum re-coherence function value is determined as the recommended installation position.

[0063] In one embodiment, if 20 acceleration sensors are arranged on the target vehicle model, the cross-coherence function values of the vibration acceleration signals collected by the 20 acceleration sensors and the sound signals collected by the microphones are calculated respectively, and 15 installation positions with cross-coherence function values greater than a certain threshold are preselected. Then, according to the actual demand, the number of acceleration sensors to be installed in the final vehicle model (i.e. the preset number) is set to 4, and then the selected 15 installation positions are combined to obtain 1365 combinations. The re-coherence functions of these combinations are calculated, and the four positions in the combination with the maximum value are determined as the recommended installation positions.

[0064] ​Further, after the end of the static test, vehicle driving tests are carried out on different quality pavements at different driving speeds, and at the same time, driving vibration acceleration signals are collected through the acceleration sensor installed at the recommended installation position, driving sound signals of the target noise reduction point position are obtained through the error microphone, and pavement point cloud data within a preset distance in front of the vehicle are obtained through the laser radar installed at the front of the vehicle. The driving vibration acceleration signal is taken as a reference signal, and the driving sound signal is taken as an error signal to form a noise signal.

[0065] S102, time-frequency domain feature extraction is performed on the noise signal to obtain noise signal features; and based on the noise signal features and the pavement point cloud data, a target data set is constructed.

[0066] Specifically, based on a preset feature extraction algorithm, time-frequency domain feature extraction is performed on the noise signal to obtain corresponding noise signal features.

[0067] As a feasible implementation manner, the preset feature extraction algorithm is: continuous wavelet transform is performed on the reference signal and the error signal in the noise signal respectively to obtain the time-frequency distribution matrix of each signal. Then, singular value decomposition is performed on the time-frequency distribution matrix to obtain the singular eigenvalue corresponding to each signal. Further, based on the singular eigenvalue, the inter-class dispersion and the intra-class dispersion of each class of signals are calculated, and the separability threshold is calculated according to the inter-class dispersion and the intra-class dispersion. The singular eigenvalues greater than the separability threshold are screened to obtain a singular eigenvalue subset, which constitutes the noise signal features.

[0068] Further, based on the vehicle driving speed, the delay time between the noise signal collected at the same pavement position and the pavement point cloud data is calculated. Based on the delay time, the noise signal features collected at the same pavement position are corresponded with the pavement point cloud data one by one to obtain a plurality of data groups, which constitute the target data set.

[0069] In one embodiment, if the time required for the vehicle to drive from the current position to the laser radar scanning position is 1 second according to the vehicle driving speed, the noise signal features collected at the current time are corresponded with the pavement point cloud data collected 1 second ago to form a data group, which is stored in the target data set.

[0070] S103, the noise signal prediction model is trained through the target data set.

[0071] Specifically, a causal convolution layer and a dilated convolution layer are added between the input layer and the output layer of the convolutional neural network, each convolution layer is connected through a ReLU activation function, and the output end of the last convolution layer is added to the original input through a jump connection to form a residual link, which constitutes a local feature extraction module. The causal convolution layer is used to extract the causal relationship between the input features, and the dilated convolution layer is used to increase the receptive field of the convolution kernel to capture the long-distance dependency relationship between the features.

[0072] Further, the output end of the local feature extraction module is connected with the input end of the Transformer encoder to construct a noise signal prediction model. Then, the noise signal prediction model is trained through a target data set until the model converges.

[0073] The causal convolution layer and the dilated convolution layer are added to the convolutional neural network, the local features in the input sequence are captured through the convolution layer to generate a feature map, and then a plurality of convolution layers are stacked through a nonlinear activation function between the convolution layers to enhance the complexity of the model. The causal convolution layer is a unidirectional structure, and only the current and past information is considered at each time step to avoid future information leakage and ensure that the model meets the causal relationship of the real data. The dilated convolution layer can increase the receptive field of the convolution kernel and improve the ability of the model to capture long-distance dependency relationships, thereby improving the computational efficiency without deepening the network level.

[0074] S104、In the process of driving the target vehicle, real-time point cloud data of the road surface in front of the vehicle is obtained, and the real-time point cloud data of the road surface is evaluated to obtain a road surface quality evaluation value.

[0075] Specifically, in the actual driving process of the target vehicle, the real-time point cloud data of the road surface within a preset distance in front of the vehicle is obtained through the laser radar installed at the front of the vehicle.

[0076] Based on the real-time point cloud data of the road surface, the quality parameters of the road surface within the preset distance in front of the vehicle are analyzed, wherein the quality parameters at least include one or more of the following: road surface material, road surface flatness, road surface roughness, road surface damage condition and road surface anti-skid performance.

[0077] Further, the quality parameters are assigned with weights and weighted calculation is performed to obtain the road surface quality evaluation value.

[0078] S105, in the case that the road surface quality evaluation value is lower than the first preset threshold, the real-time point cloud data of the road surface is input into the noise signal prediction model to obtain noise signal prediction features, and a noise pre-adjustment signal is generated according to the noise signal prediction features to preliminarily suppress the vehicle noise.

[0079] Specifically, when the road surface quality evaluation value is lower than a certain threshold, it is preliminarily judged that the road surface in front may cause relatively large noise fluctuation when the vehicle passes through, and therefore the road surface real-time point cloud data is input into the noise signal prediction model at this time to obtain noise signal prediction features.

[0080] Further, frequency information and phase information of the predicted noise signal are extracted according to the noise signal prediction features. Based on the frequency information and the phase information, a noise pre-adjustment signal with the same frequency and opposite phase of the predicted noise signal is generated.

[0081] Further, based on the current driving speed of the vehicle, a delay time required for the vehicle to travel from the current position to the road surface within a preset distance in front is calculated; and after the delay time, the noise pre-adjustment signal is played through the in-vehicle speaker system to cancel the current noise signal, so as to preliminarily suppress the vehicle noise.

[0082] S106, the real-time noise signal obtained after preliminary suppression is input into an adaptive filter to compensate the noise pre-adjustment signal, so as to further suppress the vehicle noise.

[0083] Specifically, the real-time noise signal after preliminary suppression is obtained through the noise signal acquisition device installed in the target vehicle model.

[0084] Further, the real-time noise signal is input into an LMS adaptive filter to output a control signal with opposite phase, which is superimposed with the noise pre-adjustment signal and played through the in-vehicle speaker system, so as to further suppress the vehicle noise.

[0085] As a feasible implementation manner, after the current noise signal is suppressed by the noise pre-adjustment signal, due to reasons such as model prediction accuracy, the suppression effect may not achieve the ideal effect. Therefore, the present application combines the existing adaptive filter to further analyze the noise signal after preliminary suppression, and generates a cancellation signal again to be superimposed with the noise pre-adjustment signal being played, so as to further suppress the noise signal on the basis of pre-adjustment, and greatly reduce the noise decibel value near the driver's ear.

[0086] In addition, the embodiment of the present application also provides a noise active control system of a new energy vehicle, as shown in Figure 2 The noise active control system 200 of the new energy vehicle specifically comprises:

[0087] The model training module 210 is configured to perform driving test on the target vehicle model, simultaneously collect noise signals and road surface point cloud data in front of the vehicle, extract time-frequency domain features of the noise signals to obtain noise signal features, construct a target data set based on the noise signal features and the road surface point cloud data, train a noise signal prediction model through the target data set, and

[0088] The noise pre-adjustment module 220 is configured to obtain real-time point cloud data of a road surface in front of the vehicle during driving of the target vehicle model, and perform quality evaluation on the real-time point cloud data of the road surface to obtain a road surface quality evaluation value; in a case where the road surface quality evaluation value is lower than a first preset threshold value, input the real-time point cloud data of the road surface into the noise signal prediction model to obtain noise signal prediction features; and generate a noise pre-adjustment signal according to the noise signal prediction features and preliminarily suppress vehicle noise.

[0089] The noise real-time suppression module 230 is configured to input the real-time noise signal obtained after preliminary suppression into an adaptive filter, compensate for the noise pre-adjustment signal, and further suppress vehicle noise.

[0090] Each of the embodiments of the present application is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each of the embodiments mainly describes the difference from other embodiments. In particular, for the device, equipment and non-volatile computer storage medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the related parts can be referred to the part of the method embodiment.

[0091] The above describes specific embodiments of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0092] The above only describes the embodiments of the present application and is not intended to limit the present application. The embodiments of the present application can be variously changed and modified by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present application shall be included in the protection scope of the present application.

Claims

1. A method for active noise control in new energy vehicles, characterized in that, The method includes: A driving test was conducted on the target vehicle model, and noise signals and road surface point cloud data in front of the vehicle were collected simultaneously, including: Accelerometers are installed at several locations on the target vehicle model, and error microphones are installed at the target noise reduction points. A vibration test was conducted on the target vehicle while it was stationary. The vibration acceleration signal at the installation location was obtained through the acceleration sensor, and the sound signal at the target noise reduction point location was obtained through the error microphone. Coherent analysis was performed on the vibration acceleration signals collected at each location to determine a preset number of recommended installation locations for the accelerometers. Accelerometers installed at non-recommended locations were then removed. Specifically, this included: Calculate the constant coherence function (CFC) values ​​of the vibration acceleration signal and the corresponding sound signal collected at each location, and filter out target acceleration sensors whose CFC values ​​are greater than a second preset threshold; based on the preset number, arrange and combine the target acceleration sensors, and calculate the recoherence function (RCF) value of the vibration acceleration signal in each combination; determine the installation location of the target acceleration sensor corresponding to the vibration acceleration signal in the combination with the largest recoherence function value as the recommended installation location; Vehicle driving tests were conducted at different speeds on roads of different qualities. At the same time, the driving vibration acceleration signal was collected by the acceleration sensor installed at the recommended installation position, the driving sound signal at the target noise reduction point position was obtained by the error microphone, and the road point cloud data within a preset distance in front of the vehicle was obtained by the lidar installed at the front of the vehicle. The noise signal is constructed by using the driving vibration acceleration signal as a reference signal and the driving sound signal as an error signal. The noise signal is subjected to time-frequency domain feature extraction to obtain noise signal features; and a target dataset is constructed based on the noise signal features and the road surface point cloud data. A causal convolutional layer and a dilated convolutional layer are added between the input and output layers of a convolutional neural network. Each convolutional layer is connected by a ReLU activation function, and the output of the last convolutional layer is added to the original input through a skip connection to form a residual link, which constitutes a local feature extraction module. The causal convolutional layer is used to extract the causal relationship between input features, and the dilated convolutional layer is used to increase the receptive field of the convolutional kernel to capture long-distance dependencies between features. Connect the output of the local feature extraction module to the input of the Transformer encoder to construct the noise signal prediction model; A noise signal prediction model is trained using the target dataset. During the driving process of the target vehicle, real-time point cloud data of the road surface in front of the vehicle is acquired, and the quality of the real-time point cloud data of the road surface is evaluated to obtain the road surface quality evaluation value. When the road surface quality assessment value is lower than a first preset threshold, the real-time point cloud data of the road surface is input into the noise signal prediction model to obtain noise signal prediction features; Based on the noise signal prediction characteristics, a noise pre-conditioning signal is generated and vehicle noise is initially suppressed. The real-time noise signal obtained after initial suppression is input into an adaptive filter to compensate the noise pre-adjustment signal, so as to further suppress vehicle noise.

2. The noise active control method for new energy vehicles according to claim 1, characterized in that, The noise signal is subjected to time-frequency domain feature extraction to obtain noise signal features; Based on the noise signal features and the road surface point cloud data, a target dataset is constructed, specifically including: Based on a preset feature extraction algorithm, time-frequency domain features are extracted from the noise signal to obtain the corresponding noise signal features; Based on the vehicle's speed, the time delay between the noise signal collected at the same road surface location and the road surface point cloud data is calculated. Based on the aforementioned delay time, noise signal features collected at the same road surface location are matched one-to-one with road surface point cloud data to obtain several data groups, which constitute the target dataset.

3. The method for active noise control of a new energy vehicle according to claim 2, characterized in that, Based on a preset feature extraction algorithm, time-frequency domain features are extracted from the noise signal to obtain the corresponding noise signal features, specifically including: Continuous wavelet transform is performed on the reference signal and error signal in the noise signal respectively to obtain the time-frequency distribution matrix of each signal; Singular value decomposition is performed on the time-frequency distribution matrix to obtain the singular eigenvalues ​​corresponding to each signal; Based on the singular eigenvalues, the inter-class dispersion and intra-class dispersion of each type of signal are calculated, and the separability threshold is determined according to the inter-class dispersion and intra-class dispersion. Based on the separability threshold, the singular feature values ​​of each signal are evaluated and filtered to obtain a subset of singular features, which constitute the noise signal features.

4. The noise active control method for a new energy vehicle according to claim 1, characterized in that, During the driving process of the target vehicle, real-time point cloud data of the road surface in front of the vehicle is acquired, and the quality of the real-time point cloud data is evaluated to obtain a road surface quality evaluation value, specifically including: During the actual driving process of the target vehicle, real-time point cloud data of the road surface within a preset distance in front of the vehicle is obtained by the lidar installed at the front of the vehicle. Based on the real-time point cloud data of the road surface, the quality parameters of the road surface within a preset distance ahead are determined; wherein, the quality parameters include at least one or more of the following: road surface material, road surface smoothness, road surface roughness, road surface damage condition, and road surface anti-skid performance; The quality parameters are weighted and weighted to obtain the pavement quality assessment value.

5. The method for active noise control of a new energy vehicle according to claim 1, characterized in that, Based on the noise signal prediction characteristics, a noise pre-conditioning signal is generated to initially suppress vehicle noise, specifically including: Based on the noise signal prediction characteristics, extract the frequency and phase information of the predicted noise signal; Based on the frequency and phase information, a noise pre-conditioning signal with the same frequency but opposite phase as the predicted noise signal is generated. Based on the vehicle's current speed, calculate the delay time required for the vehicle to travel from its current position to the road surface within a preset distance ahead; After the aforementioned delay time, the noise pre-adjustment signal is played through the in-vehicle speaker system to initially suppress vehicle noise.

6. The method for active noise control of a new energy vehicle according to claim 1, characterized in that, The real-time noise signal obtained after initial suppression is input into an adaptive filter to compensate the noise pre-adjustment signal, thereby further suppressing vehicle noise. Specifically, this includes: The noise signal is acquired in real time after initial suppression by the noise signal acquisition device installed in the target vehicle. The real-time noise signal is input into the LMS adaptive filter, which outputs a control signal with opposite phase. This signal is then superimposed on the noise pre-adjustment signal and played through the in-vehicle speaker system to further suppress vehicle noise.

7. A noise active control system for a new energy vehicle, employing the noise active control method for a new energy vehicle as described in any one of claims 1-6, characterized in that, The system includes: The model training module is used to conduct driving tests on the target vehicle model, while simultaneously collecting noise signals and road surface point cloud data in front of the vehicle; extracting time-frequency domain features from the noise signals to obtain noise signal features; and constructing a target dataset based on the noise signal features and the road surface point cloud data; and training a noise signal prediction model using the target dataset. The noise pre-conditioning module is used to acquire real-time point cloud data of the road surface in front of the vehicle during the driving process of the target vehicle, and to perform quality assessment on the real-time point cloud data of the road surface to obtain a road surface quality assessment value; if the road surface quality assessment value is lower than a first preset threshold, the real-time point cloud data of the road surface is input into the noise signal prediction model to obtain noise signal prediction features; and based on the noise signal prediction features, a noise pre-conditioning signal is generated to initially suppress vehicle noise. The real-time noise suppression module is used to input the real-time noise signal obtained after preliminary suppression into an adaptive filter to compensate the noise pre-adjustment signal, so as to further suppress vehicle noise.

Images