Automobile active noise reduction method and system based on exhaust noise prediction

Abstract

The invention discloses an automobile active noise reduction method and system based on exhaust noise prediction, and relates to the technical field of automobile noise reduction. An automobile active noise reduction system based on exhaust noise prediction comprises a data acquisition module, a driving prediction module, a noise mapping module, a noise correction module, an active noise reduction module and an associated noise reduction module. According to the method, the vehicle exhaust noise influence information and the driving behavior prediction information are fused, a multi-source state sensing and prediction mechanism is constructed, and the perspectiveness and accuracy of exhaust noise modeling and control are remarkably improved; by constructing the vehicle motion prediction model based on the driving style, high-precision deduction of the dynamic state of the vehicle in the future is realized, and an advanced control basis can be provided for active noise reduction, so that the response speed and the noise reduction precision are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile noise reduction, and in particular to an automobile active noise reduction method and system based on exhaust noise prediction. Background Art

[0002] With the rapid development of the automotive industry and increasing consumer demand for enhanced driving comfort, vehicle noise control technology has become a key research area in automotive design. Exhaust noise, as one of the primary noise sources generated during engine operation, directly impacts the acoustic environment quality inside and outside the vehicle, as well as the driving experience. Traditional automotive noise reduction methods rely on passive sound insulation structures or statically tuned active noise reduction systems, making it difficult to precisely control exhaust noise under complex dynamic operating conditions.

[0003] In recent years, active noise cancellation technology has been gradually adopted in the automotive sector, offsetting noise sources through phase-reversed sound waves. However, existing technologies primarily focus on controlling static noise within the vehicle, and their treatment of exhaust noise remains limited. On the one hand, exhaust noise is dynamically influenced by multiple factors, including engine operating conditions, vehicle speed, road conditions, and driving behavior. Its variations exhibit significant temporal characteristics and are difficult to predict. On the other hand, most current active noise cancellation systems lack the ability to predict future noise trends, making it difficult to respond promptly to rapidly changing acoustic environments, which can lead to delayed noise cancellation and unstable results.

[0004] In addition, some studies have attempted to obtain current status information through vehicle sensors to drive noise reduction control strategies, but lack the integrated analysis of predictive factors such as driving behavior characteristics and vehicle network information, making it difficult to form an intelligent noise reduction system with forward-looking, personalized and collaborative optimization capabilities. Summary of the Invention

[0005] This paper proposes an active noise reduction method and system for automobiles based on exhaust noise prediction. By integrating vehicle state perception, driving behavior prediction and dynamic feedback correction, it can achieve accurate modeling, real-time prediction and efficient control of exhaust noise, thereby improving the acoustic performance of the entire vehicle and user experience.

[0006] An automobile active noise reduction method based on exhaust noise prediction, comprising:

[0007] Obtain the vehicle's current exhaust noise impact information, including engine operating parameters, vehicle motion state parameters, and environmental and road parameters; obtain current state prediction information, including driving behavior parameters and vehicle network auxiliary parameters;

[0008] Obtain the current driver's driving style sample and combine it with the state prediction information, calculate it through the vehicle motion prediction model, and obtain the predicted vehicle motion state parameters;

[0009] The noise impact information is input into the noise mapping model corresponding to the vehicle model for calculation, and the exhaust noise parameters at the current moment are output; the vehicle motion state parameters in the exhaust noise impact information are replaced with the predicted vehicle motion state parameters to obtain the predicted exhaust noise impact information, and the information is input into the noise mapping model to obtain the predicted exhaust noise parameters at the next moment;

[0010] Acoustic sensors placed at the exhaust pipe and inside the vehicle are used to obtain exhaust monitoring noise parameters and in-vehicle noise monitoring parameters, respectively. Vehicle state correction samples are optimized in real time by comparing the exhaust monitoring noise parameters with the exhaust noise parameters. The predicted exhaust noise parameters are then weightedly corrected using the vehicle state correction samples to obtain corrected exhaust noise parameters.

[0011] The operating parameters of the noise reduction device at the exhaust pipe are matched according to the modified exhaust noise parameters and controlled.

[0012] As a preferred technical solution of the present invention, a method for active noise reduction of an automobile based on exhaust noise prediction further includes:

[0013] The in-vehicle noise monitoring parameters include first-level noise monitoring parameters and second-level noise monitoring parameters; according to the first-level noise monitoring parameters after the noise reduction device at the exhaust pipe is working, the working parameters of the in-vehicle noise reduction device are matched and controlled; according to the second-level noise monitoring parameters after the noise reduction device at the exhaust pipe and the in-vehicle noise reduction device work together, the two devices are feedback-adjusted respectively; and the exhaust monitoring noise parameters, first-level noise monitoring parameters and second-level noise monitoring parameters when the second-level noise monitoring parameters are lower than the set threshold, and the working parameters of the noise reduction device at the exhaust pipe and the in-vehicle noise reduction device are recorded and organized into a collaborative noise reduction working set, and an active noise reduction collaborative sample corresponding to the vehicle is constructed for auxiliary adjustment of the active noise reduction according to the working parameters of the acoustic sensor and the noise reduction device.

[0014] As a preferred technical solution of the present invention, a method for active noise reduction of an automobile based on exhaust noise prediction further includes:

[0015] Provides a variety of user noise reduction mode options, including quiet mode, sports sound retention mode, and custom noise reduction mode, to achieve personalized adjustment of noise reduction intensity;

[0016] Through the noise mapping models corresponding to different vehicle models, noise data under simulated driving conditions is obtained as the sound simulation parameters of new energy vehicles.

[0017] As a preferred technical solution of the present invention, the vehicle motion prediction model is a deep learning model based on driving style adaptation. The initial weights of the model are determined by the current driver's driving style sample. The driving style sample is generated based on historical driving behavior data and summarized into a driving behavior vector template through cluster analysis or feature vector encoding. The driving behavior data includes acceleration habits, braking mode, steering characteristics, following distance and overtaking frequency.

[0018] As a preferred technical solution of the present invention, the vehicle motion prediction model takes state prediction information as its main input, including driving behavior parameters, vehicle network auxiliary parameters and environmental parameters, and processes them in sequence through the following structure to output predicted vehicle motion state parameters:

[0019] Style guidance layer: adaptively adjusts the initial model parameters according to the driving style samples to achieve personalized modeling;

[0020] Feature fusion layer: normalizes the input state prediction information and extracts static and dynamic mixed features;

[0021] Time series modeling layer: uses a bidirectional long short-term memory network to model the time series dependencies of state changes;

[0022] State regression layer: The output predicted vehicle motion state parameters include acceleration, vehicle speed, gear shift status, engine speed and throttle opening.

[0023] As a preferred technical solution of the present invention, the noise prediction model is a standard vehicle noise mapping model generated by static training, which is used to predict the exhaust noise parameters at the current moment based on the exhaust noise impact information. The model includes:

[0024] Parameter preprocessing layer: Normalizes and aligns the input exhaust noise impact information to obtain corresponding multi-source features;

[0025] Feature encoding layer: Using statistical regression, it compresses and reconstructs multi-source features to extract high-dimensional expressions related to exhaust noise changes;

[0026] Mapping regression layer: Constructs nonlinear functional relationships and outputs target exhaust noise parameters, including exhaust sound pressure level, spectral characteristics, and main frequency band distribution;

[0027] The noise prediction model is a pre-trained model with fixed structure and parameters. It does not undergo self-update or retraining during online operation to ensure inference response speed and system stability.

[0028] As a preferred technical solution of the present invention, the noise prediction model also includes a dynamic correction layer, which operates independently of the noise prediction model. The dynamic correction layer is used to construct a vehicle state correction sample after obtaining the exhaust monitoring noise parameters by comparing and analyzing the deviation between the parameters and the exhaust noise parameters output by the noise prediction model, and perform weighted correction on the predicted exhaust noise parameters at the next moment to obtain the corrected exhaust noise parameters.

[0029] As a preferred technical solution of the present invention, the vehicle state correction sample includes vehicle operating condition offset, sensor feedback deviation, and prediction error trend in a short period;

[0030] The weighted correction adopts a multi-factor fusion strategy, taking into account the historical deviation accumulation value and the current instantaneous difference value to optimize the accuracy and stability of subsequent control instructions;

[0031] The dynamic correction layer provides accuracy compensation for prediction results without interfering with the standard model structure and weights.

[0032] An automobile active noise reduction system based on exhaust noise prediction, comprising:

[0033] Data acquisition module: obtains the vehicle's current exhaust noise impact information and status prediction information;

[0034] Driving prediction module: obtains the predicted vehicle motion state parameters through the vehicle motion prediction model;

[0035] Noise mapping module: obtains exhaust noise parameters and predicts exhaust noise parameters through the noise mapping model;

[0036] Noise correction module: performs weighted correction on the predicted exhaust noise parameters through vehicle state correction samples to obtain corrected exhaust noise parameters;

[0037] Active noise reduction module: Active noise reduction is achieved through noise reduction equipment at the exhaust pipe and noise reduction equipment inside the vehicle;

[0038] Correlation noise reduction module: Auxiliary adjustment of active noise reduction based on active noise reduction collaborative samples.

[0039] The present invention has the following advantages:

[0040] The present invention integrates vehicle exhaust noise impact information with driving behavior prediction information to construct a multi-source state perception and prediction mechanism, significantly improving the foresight and accuracy of exhaust noise modeling and control; by constructing a vehicle motion prediction model based on driving style, it achieves high-precision deduction of future vehicle dynamic states, which can provide an advance control basis for active noise reduction, thereby improving response speed and noise reduction accuracy.

[0041] The present invention introduces a linkage mechanism between the noise mapping model and the predicted exhaust noise parameters, and on this basis uses real-time exhaust noise monitoring data for dynamic correction, establishing a vehicle state correction sample library that is adaptively optimized according to working conditions, effectively improving the adaptability and robustness of the model in complex environments.

[0042] The present invention further realizes the coordinated operation of multiple acoustic wave sensors arranged at the exhaust pipe and inside the vehicle, and combines the hierarchical feedback mechanism of the first-level and second-level noise monitoring parameters to perform closed-loop control and coordinated adjustment of the noise reduction equipment at the exhaust pipe and inside the vehicle, thereby realizing cross-space linkage and intelligent collaborative optimization between active noise reduction devices, and enhancing the consistency and stability of the noise reduction system of the entire vehicle.

[0043] The present invention supports the historical recording and sample construction of collaborative noise reduction data, forming vehicle-specific active noise reduction collaborative samples, providing auxiliary adjustment references for subsequent similar working conditions, with continuous learning and optimization capabilities, and improving the system's intelligence level and personalized adaptation capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only schematic diagrams of the present invention. Those skilled in the art can also derive other drawings based on the provided drawings without inventive effort.

[0045] Figure 1 This is a schematic structural diagram of an automobile active noise reduction system based on exhaust noise prediction adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] To make the objectives, technical solutions, and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0047] Example 1, a method for active noise reduction of an automobile based on exhaust noise prediction, comprising the following steps:

[0048] Step S1: obtaining the vehicle's current exhaust noise impact information, including engine operating parameters, vehicle motion state parameters, and environmental and road parameters; obtaining current state prediction information, including driving behavior parameters and vehicle networking auxiliary parameters;

[0049] Among them, the engine operating parameters include current engine speed, torque output, intake pressure, exhaust temperature and fuel injection amount;

[0050] Vehicle motion state parameters include vehicle speed, acceleration, braking state, gear shifting state, throttle opening, and tire angle;

[0051] Environmental and road parameters include the current road type (such as urban roads, highways, mountain roads), road slope, surface roughness, external temperature, humidity, air pressure, and wind speed and direction;

[0052] The exhaust noise impact information is collected in real time through the vehicle-mounted sensor group, GPS system, IMU (inertial measurement unit), OBD interface and vehicle networking platform to ensure the timeliness and accuracy of the data.

[0053] At the same time, obtain the current state prediction information, including driving behavior parameters and vehicle network auxiliary parameters;

[0054] Among them, driving behavior parameters include at least the driver's acceleration and deceleration patterns, steering rhythm, following behavior, overtaking frequency, braking intensity and fatigue level, all of which are data used to characterize individual driving style;

[0055] The auxiliary parameters of the Internet of Vehicles include at least traffic flow density information synchronized with the cloud, road condition warnings ahead, traffic light timing predictions, navigation planning path information, and vehicle-road collaborative communication data.

[0056] Step S2: Obtaining a driving style sample of the current driver and combining it with the state prediction information, performing calculations using a vehicle motion prediction model to obtain predicted vehicle motion state parameters;

[0057] The driving style sample is constructed based on the current driver's historical driving behavior data. This sample is generated by time windowing and feature extraction of the above data, combined with principal component analysis (PCA) using the K-means clustering algorithm, to generate a driving behavior vector template, forming a driving style feature vector for loading into the neural network.

[0058] The vehicle motion prediction model is a deep learning model based on driving style adaptation. The model has style perception capabilities, and its initial weights are dynamically adjusted based on the current driver's driving style samples, making the model more consistent with individual driving habits during the prediction process;

[0059] The model takes state prediction information as its main input, including driving behavior parameters, vehicle network auxiliary parameters and environmental parameters, and outputs the predicted vehicle motion state parameters after processing them in sequence through the following structure:

[0060] Style guidance layer: By introducing style vectors, the activation function or weight matrix of the neural network is adaptively adjusted to achieve personalized customization of the model structure and parameters;

[0061] Feature fusion layer: This layer normalizes the input multi-dimensional state prediction information to a unified scale, extracts static features (such as road condition level and vehicle type configuration) and dynamic features (such as speed change rate and traffic density fluctuation), and constructs a mixed input vector.

[0062] Time series modeling layer: uses a bidirectional long short-term memory (Bi-LSTM) network to model the temporal dependencies in the input sequence, improving the ability to capture continuous behavior patterns and short-term behavior transitions.

[0063] State regression layer: Based on the above processing results, it outputs predicted vehicle motion state parameters, including key dynamic indicators such as acceleration, vehicle speed, gear shift status, engine speed and throttle opening, providing a data basis for subsequent noise prediction and noise reduction control.

[0064] In addition, to improve the model's generalization ability and adaptability to new driver scenarios, the vehicle motion prediction model can integrate a transfer learning mechanism and introduce an individual sample fine-tuning module based on the general model to achieve rapid personalized deployment.

[0065] Step S3: Inputting the noise impact information into a noise mapping model corresponding to the vehicle model for calculation, and outputting the exhaust noise parameters at the current moment; replacing the vehicle motion state parameters in the exhaust noise impact information with the predicted vehicle motion state parameters to obtain the predicted exhaust noise impact information, and inputting the information into the noise mapping model to obtain the predicted exhaust noise parameters at the next moment; the noise mapping model corresponding to the vehicle model is trained based on the exhaust noise parameters corresponding to the same model standard vehicle under different exhaust noise impact information;

[0066] The noise mapping model corresponding to the vehicle model is constructed through offline training based on exhaust noise parameters collected from standard vehicles of the same model under different exhaust noise influence information, ensuring high matching and generalization capabilities of the model under the vehicle model configuration; sample enhancement and data alignment techniques are introduced during the training process to adapt to the distribution of multiple working conditions, and noise main feature extraction and label smoothing strategies are used to improve training stability.

[0067] The noise prediction model is a standard vehicle noise mapping model generated through static training. It has a fixed structure and parameter configuration and does not perform self-updates or retraining during online operation to ensure computational efficiency and system stability during the inference phase. This model is used to predict exhaust noise parameters in real time, given exhaust noise influencing information. It includes the following structural modules:

[0068] Parameter preprocessing layer: This layer normalizes, fills in missing values, and calibrates the dimensions of the input raw exhaust noise impact information to generate a unified input tensor. This layer supports the fusion and redundancy detection of data from different sources (such as CAN bus parameters and sensor measurements).

[0069] Feature encoding layer: Statistical regression and principal component compression methods are used to reconstruct preprocessed multi-source features and extract potential high-dimensional expression variables that are highly correlated with exhaust noise changes. The encoding method can be ridge regression, PLSR (partial least squares regression), or stacked sparse autoencoder.

[0070] Mapping and regression layer: Constructs a nonlinear function mapping relationship to map the high-dimensional expression to the target exhaust noise parameter space; the output parameters include at least the exhaust sound pressure level, spectrum distribution characteristics (such as the main frequency point and energy concentration bandwidth), and noise fluctuation frequency;

[0071] Dynamic Correction Layer: This correction module runs in parallel with the noise mapping model. When actual exhaust monitoring data is available, it constructs state residual samples based on the deviation between the current actual monitoring value and the model output value. It then uses a time-weighted recursive mechanism to update the error trend model, thereby performing a weighted correction on the predicted exhaust noise parameters at the next moment.

[0072] The weighted correction method uses an exponential moving average (EMA) to make the prediction result closer to the actual exhaust acoustic state of the vehicle.

[0073] Through the above structure and process, high-precision, low-latency prediction of exhaust noise parameters can be achieved during vehicle driving, and real-time optimization of noise estimation results can be performed based on dynamic feedback, thus providing a reliable input basis for subsequent active noise reduction control strategies.

[0074] Step S4: Acquiring exhaust monitoring noise parameters and interior noise monitoring parameters using acoustic wave sensors installed at the exhaust pipe and inside the vehicle, respectively; optimizing vehicle state correction samples in real time by comparing the exhaust monitoring noise parameters with the exhaust noise parameters; and performing weighted correction on the predicted exhaust noise parameters using the vehicle state correction samples to obtain corrected exhaust noise parameters;

[0075] The acoustic wave sensor includes a high-dynamic-response microphone array installed near the exhaust pipe outlet and a low-frequency, high-sensitivity microphone module deployed at a reference point inside the vehicle, which is used to collect the external exhaust noise characteristics and the internal passenger-perceived noise characteristics of the vehicle under different driving conditions;

[0076] The sampling frequency range covers 20Hz to 20kHz, supporting FFT spectrum analysis, time-varying envelope tracking, and segmented sound pressure level assessment to meet the multi-dimensional perception requirements of exhaust noise variations. High-speed CAN or Ethernet communication is used between the acoustic wave sensor and the central control unit to ensure low-latency data synchronization.

[0077] The vehicle state correction sample is used to construct a prediction error compensation mechanism, which comprises:

[0078] Vehicle operating condition offset: indicates the numerical difference between the current engine operating condition, gear status, or acceleration and the preset operating condition of the noise model;

[0079] Sensor feedback deviation: including observation deviation caused by sensor zero drift, ambient temperature change or signal interference;

[0080] Forecast error trend within a short period: Based on the residual sequence between the predicted value and the measured value within a sliding time window (such as the last 3 to 5 periods), its mean drift and change rate are evaluated.

[0081] The weighted correction mechanism adopts a multi-factor fusion strategy:

[0082] Taking into account factors such as historical cumulative deviation (residual integral), current instantaneous error, state trend disturbance, etc., the predicted value is dynamically adjusted by constructing a weighted function. Its calculation form is: in To predict exhaust noise parameters, Δ t is the current residual, is the sliding window average residual, is the residual change rate, λ1, λ2, λ3 are adaptive weight coefficients.

[0083] These corrections are independently implemented by the dynamic correction layer, ensuring that the predictions are accurately compensated without interfering with the main model structure and its parameters. The dynamic correction layer operates outside the inference path and relies solely on measured value feedback and error modeling.

[0084] After the correction is completed, the control strategy of the variable structure noise reduction equipment at the exhaust pipe (such as active noise control ANC module, adjustable exhaust valve, acoustic resistance / acoustic capacity module, etc.) is matched according to the corrected exhaust noise parameters, and its working parameters are automatically adjusted, including the sound wave cancellation frequency band, noise reduction command delay timing, sound source inversion level, etc., to improve the adaptive noise suppression capability of the exhaust system under different working conditions and the listening comfort in the car.

[0085] Step S5: According to the first-level noise monitoring parameters after the noise reduction device at the exhaust pipe works, the working parameters of the noise reduction device in the vehicle are matched and controlled; according to the second-level noise monitoring parameters after the noise reduction device at the exhaust pipe and the noise reduction device in the vehicle work together, the two devices are feedback-adjusted respectively; and the exhaust monitoring noise parameters, the first-level noise monitoring parameters and the second-level noise monitoring parameters when the second-level noise monitoring parameters are lower than the set threshold, and the working parameters of the noise reduction device at the exhaust pipe and the noise reduction device in the vehicle are recorded and organized into a collaborative noise reduction working set, and an active noise reduction collaborative sample corresponding to the vehicle is constructed for auxiliary adjustment of the active noise reduction according to the working parameters of the acoustic sensor and the noise reduction device.

[0086] The in-vehicle noise monitoring parameters are divided into two levels:

[0087] Level 1 noise monitoring parameters: These are acquired through an acoustic wave sensor installed on the rear chassis of the vehicle. They primarily sense the characteristics of residual exhaust noise transmitted into the vehicle after passing through the noise reduction equipment in the exhaust pipe, reflecting the initial effectiveness of external structural noise reduction.

[0088] Secondary noise monitoring parameters: These parameters are obtained through an acoustic wave sensor installed at the head of the driver's seat on the front side of the vehicle. They directly reflect the final noise environment level within the driver's subjective perception area and reflect the overall noise suppression performance of the entire collaborative noise reduction system.

[0089] The noise reduction equipment includes noise reduction equipment at the exhaust pipe: such as electronically controlled exhaust valves, variable structure mufflers, and active noise reduction speaker arrays, which focus on suppressing transient high sound pressure impacts at the exhaust port; and in-vehicle noise reduction equipment: such as headrest speakers in the cabin, active sound insulation modules (ANC systems), door panel standing wave control units, etc., which focus on optimizing the subjective listening experience in the cabin.

[0090] The matching process includes:

[0091] Based on the residual spectrum characteristics reflected by the first-level noise monitoring parameters, the response library of the in-vehicle noise reduction device is called to match the optimal inverted frequency band and sound pressure response amplitude;

[0092] Monitor the secondary noise monitoring parameters. When it is identified that the collaborative noise reduction effect does not meet expectations (such as the sound pressure does not reach the set threshold), feedback is sent to the control interfaces of the two devices respectively to adjust their output parameters, such as activation status, operating frequency band, sound source phase, etc.

[0093] The recording mechanism of the collaborative noise reduction working set is as follows:

[0094] Recording trigger condition: The secondary noise monitoring parameter is lower than the system preset comfort threshold (such as 65dB(A));

[0095] The record fields include: exhaust monitoring noise parameters at the current moment (external sound source intensity and spectrum); primary and secondary noise monitoring parameters (intermediate and final cabin sound pressure); all operating parameters of the two noise reduction devices at the corresponding moment (such as response frequency, phase compensation amount, loudness control level); the above information is structured into a collaborative noise reduction sample data and stored in the noise reduction optimization database.

[0096] These ANC collaborative samples will serve as key input for subsequent model optimization and control decision-making. They are used to: construct a noise reduction device response recommendation model based on historical collaborative effects; quickly retrieve optimal control strategies for similar scenarios from an existing collaborative sample library based on current noise monitoring parameters; and improve the noise reduction system's adaptability and response efficiency under frequently changing operating conditions. Scenario context labels, such as driving speed ranges, engine load ranges, and road surface types, are incorporated into the collaborative samples to establish a multi-label indexing mechanism, enabling more precise collaborative strategy matching and ANC control.

[0097] Example 2, an active noise reduction method for automobiles based on exhaust noise prediction, provides multiple user noise reduction mode options, including quiet mode, sports sound retention mode, and custom noise reduction mode, to achieve personalized adjustment of noise reduction intensity;

[0098] Quiet mode prioritizes eliminating low- and mid-frequency exhaust sound waves and actively adjusts the sound pressure in the driver's seat based on head-mounted sound pressure feedback. It is suitable for scenarios requiring quietness, such as long-distance cruising and nighttime driving.

[0099] The Sport Sound Preservation mode retains some high-frequency sound information and enhances the driving experience by amplifying specific spectrum intervals, adapting to sporty driving needs.

[0100] The custom noise reduction mode allows users to adjust the noise reduction frequency band range, target sound pressure level and response delay based on the vehicle's central control interface or smart terminal, thereby generating a personalized noise reduction configuration and saving it as a user noise reduction file for a long time.

[0101] Through the noise mapping models corresponding to different vehicle models, noise data under simulated driving conditions is obtained as the sound simulation parameters of new energy vehicles.

[0102] For new pure electric vehicles that lack traditional exhaust systems, this method simulates the exhaust sound characteristics of traditional vehicles through a noise mapping model, and combines driving conditions (such as acceleration / shifting simulation signals) to generate corresponding spectrum envelopes to achieve sound reconstruction and personalized feedback.

[0103] The sound wave simulation parameters include: starting excitation sound, rapid acceleration explosion feeling, low-frequency resonance enhancement, etc.

[0104] Users select the target vehicle type (such as muscle cars, German sedans, Japanese performance cars, etc.) in the vehicle personality settings, and the system calls the corresponding sound library and model for real-time generation and playback, enhancing the driving immersion and brand tone of pure electric vehicles.

[0105] Example 3, an automobile active noise reduction system based on exhaust noise prediction, see Figure 1 As shown, it includes the following modules:

[0106] Data acquisition module: obtains the vehicle's current exhaust noise impact information and status prediction information;

[0107] Driving prediction module: obtains the predicted vehicle motion state parameters through the vehicle motion prediction model;

[0108] Noise mapping module: obtains exhaust noise parameters and predicts exhaust noise parameters through the noise mapping model;

[0109] Noise correction module: performs weighted correction on the predicted exhaust noise parameters through vehicle state correction samples to obtain corrected exhaust noise parameters;

[0110] Active noise reduction module: Active noise reduction is achieved through noise reduction equipment at the exhaust pipe and noise reduction equipment inside the vehicle;

[0111] Correlation noise reduction module: Auxiliary adjustment of active noise reduction based on active noise reduction collaborative samples.

[0112] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for active noise reduction of automobile based on exhaust noise prediction, characterized in that: include: Obtain the vehicle's current exhaust noise impact information, including engine operating parameters, vehicle motion state parameters, and environmental and road parameters; obtain current state prediction information, including driving behavior parameters and vehicle network auxiliary parameters; Obtain the current driver's driving style sample and combine it with the state prediction information, calculate it through the vehicle motion prediction model, and obtain the predicted vehicle motion state parameters; The noise impact information is input into the noise mapping model corresponding to the vehicle model for calculation, and the exhaust noise parameters at the current moment are output; the vehicle motion state parameters in the exhaust noise impact information are replaced with the predicted vehicle motion state parameters to obtain the predicted exhaust noise impact information, and the information is input into the noise mapping model to obtain the predicted exhaust noise parameters at the next moment; Acoustic sensors placed at the exhaust pipe and inside the vehicle are used to obtain exhaust monitoring noise parameters and in-vehicle noise monitoring parameters, respectively. Vehicle state correction samples are optimized in real time by comparing the exhaust monitoring noise parameters with the exhaust noise parameters. The predicted exhaust noise parameters are then weightedly corrected using the vehicle state correction samples to obtain corrected exhaust noise parameters. The operating parameters of the noise reduction device at the exhaust pipe are matched according to the modified exhaust noise parameters and controlled.

2. The method for automobile active noise reduction based on exhaust noise prediction according to claim 1, characterized in that: Also includes: The vehicle interior noise monitoring parameters include primary noise monitoring parameters and secondary noise monitoring parameters; According to the first-level noise monitoring parameters after the noise reduction equipment at the exhaust pipe is working, the working parameters of the noise reduction equipment in the vehicle are matched and controlled; according to the second-level noise monitoring parameters after the noise reduction equipment at the exhaust pipe and the noise reduction equipment in the vehicle work together, the two devices are feedback-adjusted separately; and the exhaust monitoring noise parameters, first-level noise monitoring parameters and second-level noise monitoring parameters when the second-level noise monitoring parameters are lower than the set threshold, and the working parameters of the noise reduction equipment at the exhaust pipe and the noise reduction equipment in the vehicle are recorded and organized into a collaborative noise reduction working set, and the active noise reduction collaborative sample corresponding to the vehicle is constructed for auxiliary adjustment of the active noise reduction according to the working parameters of the acoustic sensor and the noise reduction equipment.

3. The method for automobile active noise reduction based on exhaust noise prediction according to claim 1, characterized in that: Also includes: Provides a variety of user noise reduction mode options, including quiet mode, sports sound retention mode, and custom noise reduction mode, to achieve personalized adjustment of noise reduction intensity; Through the noise mapping models corresponding to different vehicle models, noise data under simulated driving conditions is obtained as the sound simulation parameters of new energy vehicles.

4. The method for automobile active noise reduction based on exhaust noise prediction according to claim 1, characterized in that: The vehicle motion prediction model is a deep learning model based on driving style adaptation. The initial weights of the model are determined by the current driver's driving style sample. The driving style sample is generated based on historical driving behavior data and summarized into a driving behavior vector template through cluster analysis or feature vector encoding. The driving behavior data includes acceleration habits, braking patterns, steering characteristics, following distance, and overtaking frequency.

5. The method for active noise reduction of an automobile based on exhaust noise prediction according to claim 4, characterized in that: The vehicle motion prediction model takes state prediction information as its main input, including driving behavior parameters, vehicle network auxiliary parameters and environmental parameters, and processes them in sequence through the following structure to output predicted vehicle motion state parameters: Style guidance layer: adaptively adjusts the initial model parameters according to the driving style samples to achieve personalized modeling; Feature fusion layer: normalizes the input state prediction information and extracts static and dynamic mixed features; Time series modeling layer: uses a bidirectional long short-term memory network to model the time series dependencies of state changes; State regression layer: The output predicted vehicle motion state parameters include acceleration, vehicle speed, gear shift status, engine speed and throttle opening.

6. The method for automobile active noise reduction based on exhaust noise prediction according to claim 1, characterized in that: The noise prediction model is a standard vehicle noise mapping model generated by static training, which is used to predict the exhaust noise parameters at the current moment based on exhaust noise impact information. The model includes: Parameter preprocessing layer: Normalizes and aligns the input exhaust noise impact information to obtain corresponding multi-source features; Feature encoding layer: Using statistical regression, it compresses and reconstructs multi-source features to extract high-dimensional expressions related to exhaust noise changes; Mapping regression layer: Constructs nonlinear functional relationships and outputs target exhaust noise parameters, including exhaust sound pressure level, spectral characteristics, and main frequency band distribution; The noise prediction model is a pre-trained model with fixed structure and parameters. It does not undergo self-update or retraining during online operation to ensure inference response speed and system stability.

7. The method for active noise reduction of automobile based on exhaust noise prediction according to claim 6, characterized in that: The noise prediction model also includes a dynamic correction layer, which operates independently of the noise prediction model. After obtaining the exhaust monitoring noise parameters, the dynamic correction layer is used to construct a vehicle state correction sample by comparing and analyzing the deviation between the parameters and the exhaust noise parameters output by the noise prediction model, and to perform weighted correction on the predicted exhaust noise parameters at the next moment to obtain the corrected exhaust noise parameters.

8. The method for automobile active noise reduction based on exhaust noise prediction according to claim 7, characterized in that: The vehicle state correction samples include vehicle operating condition offset, sensor feedback deviation, and prediction error trend within a short period; The weighted correction adopts a multi-factor fusion strategy, taking into account the historical deviation accumulation value and the current instantaneous difference value to optimize the accuracy and stability of subsequent control instructions; The dynamic correction layer provides accuracy compensation for prediction results without interfering with the standard model structure and weights.

9. An automotive active noise reduction system based on exhaust noise prediction, characterized in that: The system applies any one of claims 1 to 8 of the above-mentioned method for active noise reduction of automobiles based on exhaust noise prediction, comprising: Data acquisition module: obtains the vehicle's current exhaust noise impact information and status prediction information; Driving prediction module: obtains the predicted vehicle motion state parameters through the vehicle motion prediction model; Noise mapping module: obtains exhaust noise parameters and predicts exhaust noise parameters through the noise mapping model; Noise correction module: performs weighted correction on the predicted exhaust noise parameters through vehicle state correction samples to obtain corrected exhaust noise parameters; Active noise reduction module: Active noise reduction is achieved through noise reduction equipment at the exhaust pipe and noise reduction equipment inside the vehicle; Correlation noise reduction module: Auxiliary adjustment of active noise reduction based on active noise reduction collaborative samples.

Images