Method and system for simulating noise of expected automobile working conditions based on reference noise matching

Abstract

The present invention provides a method and system for simulating the noise of expected working conditions of an automobile based on reference noise matching, and relates to the technical field of noise simulation. The method provided by the present invention comprises extracting features from the reference noise data to obtain reference noise features of each reference vehicle model under each original working condition; inputting the sound field influencing factors of the vehicle model to be tested into a pre-trained target model to obtain the predicted noise features output by the target model; obtaining a target reference vehicle model matching the vehicle model to be tested by comparing the predicted noise features with each reference noise feature; and replaying the reference noise data of the target reference vehicle model under expected working conditions in a noise playback cavity matching the vehicle model to be tested. The present invention enables the OEM to simulate and evaluate noise features at an early stage, thereby reducing design changes caused by noise problems.

Description

Technical Field

[0001] The present invention relates to the technical field of noise simulation, and in particular to a method and system for simulating noise of expected automobile working conditions based on reference noise matching. Background Art

[0002] With the development of intelligent connected vehicles, cars have changed from a simple means of transportation to a "third living space" that integrates transportation, information, entertainment and other functions. This transformation not only improves the user's driving experience, but also puts higher requirements on various acoustic systems in the cockpit (such as voice interaction system, accident emergency call system, hands-free calling system, audio entertainment system, etc.). The quality and performance of the vehicle acoustic system has become one of the key factors affecting consumers' driving experience and even car purchase decisions. At present, in order to enable good communication between the vehicle terminal and the public safety answering point (PSAP), better use of social resources, and save citizens' lives, higher requirements are put forward for call voice quality. In addition, the performance of the vehicle acoustic system is closely related to driving safety. Research by the U.S. Highway Safety Administration (NHTSA) shows that poor performance of the vehicle acoustic system will increase the probability of driver distraction, thereby increasing the risk of traffic accidents. In stark contrast to the relatively quiet listening environment in a home or professional audio room, the vehicle acoustic system must demonstrate its performance in a variable and noisy noise environment. The intensity and frequency of vehicle driving noise and wind noise are affected by the vehicle speed and the design of the vehicle itself, showing diversity. In addition, the closed space inside the car and the reflective properties of the materials will cause multiple reflections and absorption of sound, resulting in a reverberation effect, which further exacerbates the complexity of the sound field. Therefore, integrating the vehicle acoustic system into the car is full of challenges. The arrangement of microphones and speakers, noise reduction algorithms, etc. will significantly affect the performance of the vehicle acoustic system.

[0003] In order to analyze the performance of the vehicle acoustic system during actual use by users, the impact of the noise inside the car needs to be considered when conducting relevant tests on the vehicle acoustic system. In the early development stage of the vehicle acoustic system, testers often sit in the car and make subjective evaluations of the vehicle acoustic system when the vehicle is driving to the corresponding working conditions. This method has the disadvantages of high cost and long cycle, and it is easy to mislead the tuning of the product due to the problems of difficulty in quantification, poor repeatability and poor objectivity. With the development of testing technology, the vehicle acoustic system test is currently carried out in the whole vehicle quiet test room using standardized acoustic dummies (equipped with artificial ears and artificial mouths for receiving and sending test audio signals). This type of test method uses objective physical quantities to test the vehicle acoustic system, which has high objectivity and repeatability. In order to test the performance of the vehicle acoustic system during actual use by users, a set of vehicle operating noise simulation system is also required to simulate the noise environment inside the car. When the vehicle acoustic system is working, car noise has the greatest impact on users and vehicle microphones. Therefore, the vehicle operating noise simulation system needs to accurately restore the sound field at the user's ear point and the vehicle microphone point.

[0004] The order of automobile development is concept development, detailed design, engineering development, prototype manufacturing, testing and verification, and pre-production. The layout of vehicle speakers and microphones is often in the early stages of concept development and detailed design, but the vehicle operating noise needs to be collected from prototype manufacturing to pre-production. The automotive industry urgently needs a vehicle operating noise simulation system that can simulate noise in the early stages of automobile development to resolve the above contradictions. Summary of the invention

[0005] In view of the above problems, the present invention proposes a method and system for simulating the expected operating noise of an automobile based on reference noise matching, which is used to simulate and evaluate noise characteristics at an early stage, thereby reducing design changes caused by noise problems.

[0006] In a first aspect, an embodiment of the present invention provides a method for simulating noise of an expected automobile operating condition based on reference noise matching, comprising:

[0007] Collect reference noise data of various reference models under various original working conditions;

[0008] Performing feature extraction on the reference noise data to obtain reference noise features of each reference vehicle model under each original working condition;

[0009] Inputting the sound field influencing factors of the vehicle model to be tested into a pre-trained target model to obtain the predicted noise characteristics output by the target model;

[0010] Obtaining a target reference vehicle model that matches the vehicle model to be tested by comparing the predicted noise feature with each reference noise feature;

[0011] In a noise playback chamber that matches the vehicle type to be tested, reference noise data of the target reference vehicle type under expected operating conditions are played back.

[0012] Optionally, collect reference noise data of various reference models under various original working conditions, including:

[0013] Control each reference vehicle model to operate under each original working condition;

[0014] When the reference vehicle model is driving under any original working condition, reference noise data is collected through multi-point microphones in the reference vehicle model.

[0015] Optionally, before inputting the sound field influencing factors of the vehicle model to be tested into the pre-trained target model, the following steps are also included:

[0016] Collecting training samples, wherein the training samples include samples of sound field influencing factors of multiple reference vehicle models and noise feature labels;

[0017] The target model is trained using the training samples.

[0018] Optionally, the target model is a multilayer perceptron.

[0019] Optionally, obtaining a target reference vehicle model matching the vehicle model to be tested by comparing the predicted noise feature with each reference noise feature includes:

[0020] Calculating a difference vector between the predicted noise feature and each reference noise feature;

[0021] Obtaining a matching probability between the predicted noise feature and each reference noise feature according to the difference vector and the normalized exponential function;

[0022] The reference vehicle model from which the reference noise feature with the highest matching probability comes is determined as the target reference vehicle model.

[0023] Optionally, before replaying the reference noise data of the target reference vehicle model under the expected working condition in the noise replay chamber matching the vehicle model to be tested, the method further includes:

[0024] Summarize a reference noise data set under each original working condition, wherein the reference noise data set includes reference noise data of multiple vehicle models, recorded by microphones at multiple locations, and in multiple frequency bands;

[0025] Clustering the reference noise data sets under various original working conditions to obtain various typical working conditions and a typical reference noise data set under each typical working condition;

[0026] The expected operating condition is any typical operating condition.

[0027] Optionally, in a noise playback chamber matching the vehicle type to be tested, playing back reference noise data of the target reference vehicle type under expected operating conditions includes:

[0028] Determine the preset microphone arrangement for the vehicle model to be tested;

[0029] Screening reference noise data of the target reference vehicle model under expected operating conditions according to the preset microphone arrangement;

[0030] The screened reference noise data is played back using a power amplifier and a speaker group that are consistent with the preset microphone arrangement.

[0031] Optionally, the method also includes: when collecting reference noise in a reference vehicle model, and when playing back reference noise data in a noise playback cavity, using a noise equalization front end to perform sound field equalization in the noise playback cavity and in the reference vehicle model.

[0032] In a second aspect, an embodiment of the present invention provides a vehicle expected operating condition noise simulation system based on reference noise matching, comprising:

[0033] Microphone, used to collect reference noise data of various reference vehicle models under various original working conditions;

[0034] A controller is used to extract features from the reference noise data to obtain reference noise features of each reference vehicle model under each original working condition; input the sound field influencing factors of the vehicle model to be tested into a pre-trained target model to obtain predicted noise features output by the target model; and obtain a target reference vehicle model that matches the vehicle model to be tested by comparing the predicted noise features with each reference noise feature;

[0035] The noise playback system is used to play back the reference noise data of the target reference vehicle model under the expected working conditions in a noise playback cavity matched with the vehicle model to be tested.

[0036] Optionally, the noise playback cavity is a completely enclosed space having a resonance frequency that matches the vehicle type to be tested; the noise data in the noise playback cavity is reflected multiple times to form a reverberation effect, and noise data of different frequencies are filtered.

[0037] During the automotive development phase, the layout of onboard speakers and microphones is often in the early stages of concept development and detailed design, but the vehicle operating noise needs to be collected during prototype manufacturing and pre-production. If traditional solutions are used, when the acoustic system of the vehicle prototype cannot achieve the expected performance, the early design needs to be overturned, increasing development costs and extending the development cycle.

[0038] Therefore, the present invention collects noise data under different working conditions of the vehicle during the prototype manufacturing to pre-production stage of vehicle development, uses a noise playback system to play back the noise, and simulates the noise under the working condition of the vehicle, and the beneficial effects are:

[0039] 1. Early noise simulation: Traditional methods usually collect noise only in the pre-production stage of vehicle prototype manufacturing. When problems with the acoustic system are found, the early design may need to be overturned, increasing development costs and cycles. The present invention collects noise data in the prototype manufacturing stage and uses a noise playback system for simulation, so that the OEM can simulate and evaluate noise characteristics at an early stage, thereby reducing design changes caused by noise problems.

[0040] 2. Establishment of reference noise database and scene optimization: By recording real vehicle noise under various working conditions of various vehicle models, a reference noise database consisting of reference noise data was established. Then the working conditions were simplified into a small number of typical working conditions using clustering methods. This not only reduces the number of processing scenes and improves test efficiency, but also covers the main noise characteristics under different vehicle working conditions.

[0041] 3. Multilayer Perceptron (MLP) modeling: Use MLP to establish a relationship model between multiple samples of sound field influencing factors (structure, power, material, size, etc.) and noise characteristics. This machine learning method can accurately capture complex nonlinear relationships, enhance the accuracy of noise characteristic prediction of the tested vehicle model, and make the development process more scientific and data-driven.

[0042] 4. Probability matching and noise playback: The probability value of the target reference model to which the tested model belongs is calculated through the normalized exponential function (Softmax function), and it is matched with the existing reference noise data. For the successfully matched models, the corresponding noise in the reference noise library is played back to simulate the noise environment inside the car. This process not only improves the accuracy of noise evaluation, but also provides OEMs with a more flexible design verification method.

[0043] 5. Reduce development costs and cycles: By simulating and predicting noise characteristics in advance, this project greatly reduces the design changes that may occur later, reduces development costs, and shortens the development cycle. This has significant economic benefits and market competitiveness for automobile development companies.

[0044] 6. By setting up a noise playback cavity, the sound field environment can be simulated before the vehicle model is manufactured; at the same time, the noise playback cavity is a completely closed space with a resonance frequency that matches the vehicle model to be tested; the noise data is reflected multiple times to form a reverberation effect, and the noise data of different frequencies are filtered; the noise equalization front end is used to equalize the sound field in the noise playback cavity, which improves the authenticity of the noise field restoration in the car.

[0045] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above description and other purposes, features and advantages of the present invention more obvious and easy to understand, the preferred embodiments are specifically cited and described in detail as follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0047] Figure 1 A flowchart of a method for simulating noise of expected vehicle operating conditions based on reference noise matching provided by an embodiment of the present invention;

[0048] Figure 2 A schematic diagram of 16 microphone locations and their arrangement in a reference vehicle model provided in an embodiment of the present invention;

[0049] Figure 3 A schematic diagram of the structure of a system for simulating and emulating noise under expected vehicle operating conditions based on reference noise matching provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0050] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0051] In the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", "connected", "fixed" and the like should be understood in a broad sense, for example, it can be connected, detachably connected, or integrated; it can be mechanically connected or electrically connected; it can be directly connected or indirectly connected through an intermediate medium, it can be the internal connection of two elements or the interaction relationship between two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0052] Example 1

[0053] The embodiment of the present invention provides a method for simulating noise of expected vehicle working conditions based on reference noise matching, which is applicable to the situation of simulating the noise sound field in the noise playback cavity simulating the vehicle model to be tested in the early stage when the vehicle model to be tested has not yet been manufactured. Figure 1 The method for simulating the expected working condition noise of an automobile based on reference noise matching provided in this embodiment includes the following steps:

[0054] S110. Collect reference noise data of various reference vehicle models under various original working conditions.

[0055] The reference vehicle model is a vehicle model already available on the market. Specifically, each reference vehicle model is controlled to operate under each original working condition; when the reference vehicle model is driving under any original working condition, the reference noise data is collected through the multi-point microphones in the reference vehicle model. The collected reference noise data is stored in the reference noise library.

[0056] In some examples, 16 original conditions are set to ensure that the reference noise library can fully cover the noise scenarios of various vehicle operating conditions. These original conditions include combinations of vehicle speed (0, 60, 80, 120 km / h) and air conditioning air volume settings (off, low, medium, high).

[0057] Table 1 Information of original working condition

[0058]

[0059] Under all original working conditions, the air conditioner is set to external circulation, with only the face-blowing air duct opened and the other air ducts closed. The direction of the air outlet is adjusted to the middle angle between the horizontal and vertical angles of the air duct, the sunroof is closed, and all side windows are closed.

[0060] Vehicle status of the reference model: There is no abnormal sound during driving, and the vehicle status is normal without any alarm signal. The tire model is the factory model, the pattern has no obvious wear and uneven wear, and the cold inflation pressure of the tire should meet the requirements of the vehicle's technical conditions. The fuel, lubricating oil, brake fluid, ignition timing, and injection time used by the vehicle should meet the requirements of the vehicle's technical conditions or the current national standards. When recording / collecting noise, the engine coolant temperature should be within the normal range. It is recommended to drive at a medium speed for a distance before recording. When recording / collecting noise, there is no other load on the vehicle except the driver, the noise recording personnel (1 person, sitting in the co-pilot seat), and the necessary test equipment, and the passengers on the vehicle shall not make any sound. When the vehicle is driving under any original working condition, the manual transmission uses the highest gear corresponding to the vehicle speed; the automatic transmission should not change the gear manually. There are no wipers, turn signal sounds, automatic sound devices, and other devices that affect the test, and they are in a closed working state.

[0061] 100 vehicles with significant differences were selected, including different body structures, power, sizes, and materials, which can cover the reference models available on the market. The above 16 working conditions of more than 100 different models were recorded at multiple locations. Microphones were arranged at the recording locations. Figure 2The figure shows the locations of 16 microphones and their arrangement in a reference vehicle model. Each microphone is represented by a number from 1 to 16. The specific locations are as follows:

[0062] Microphones No. 1 and 2: Located on the upper left and upper right of the central control screen. The microphones face upward.

[0063] Microphones 3 to 5 are located on the edge of the front ceiling light module, close to the passenger seat. 3 is located on the left side of the edge, 4 is located in the middle of the edge, and 5 is located on the right side of the edge. The microphones face backwards.

[0064] Microphones 6 to 9 are located on the edge of the sun visor near the passenger seat, with 6 located on the left side of the left sun visor edge, 7 located in the middle of the left sun visor edge, 8 located in the middle of the right sun visor edge, and 9 located on the right side of the right sun visor edge. The microphones face backwards.

[0065] Microphones No. 10-11: Located at the top of the left and right B-pillars, respectively. The microphones face upward.

[0066] Microphones No. 12 to 13: located directly above the driver and co-pilot's heads. If the microphones cannot be placed directly above the driver and co-pilot's heads due to the sunroof, place the microphones on the left and right edges of the sunroof, as close as possible to the driver and co-pilot's heads. The microphones face backwards.

[0067] Microphone No. 14: Located directly above the head of the middle passenger in the rear row. If the microphone cannot be placed directly above the head of the middle passenger in the rear row due to the sunroof, place the microphone at the rear edge of the sunroof, as close as possible to the head of the middle passenger in the rear row. The microphone faces backwards.

[0068] Microphones No. 15-16: located directly above the heads of the left and right passengers in the second row. If the microphones cannot be placed directly above the heads of the left and right passengers in the second row due to the sunroof, the microphones should be placed on the left and right edges of the sunroof, as close as possible to the heads of the left and right passengers in the second row. The microphones should face backwards.

[0069] S120, extracting features from the reference noise data to obtain reference noise features for each reference vehicle model under each original working condition.

[0070] For the reference noise data (essentially a piece of audio) collected by each microphone for each vehicle under each working condition, the octave sound pressure level (a total of 10 frequency bands) in the range of 31.5~16000Hz is extracted as the reference noise feature.

[0071] X is a set of reference noise features. Each row represents a reference noise feature row vector under a certain working condition, including m feature data, where m = number of vehicles × number of noise recording points × number of frequency bands. These data are stored in a matrix X, the dimension of which is 16 (number of working conditions) × m (number of features).

[0072] ;

[0073] S130 , inputting the sound field influencing factors of the vehicle model to be tested into a pre-trained target model to obtain the predicted noise characteristics output by the target model.

[0074] The vehicle model to be tested is a vehicle model that has not yet been manufactured, but its sound field influencing factors that have an impact on the in-vehicle sound field are known, including but not limited to: vehicle structure, power, material and size.

[0075] Optionally, the target model is a multilayer perceptron (MLP), which is used to map the input sound field influencing factors to noise characteristics, such as octave band sound pressure level. That is, the internal noise characteristics of the vehicle to be tested are predicted based on its characteristics. Before using the target model, a multilayer perceptron needs to be built and trained.

[0076] The multilayer perceptron includes an input layer, a hidden layer, and an output layer. The factors affecting the sound field are set as input features, including vehicle structure, power, material, size, etc. Assume that the input vector composed of these factors is A. The MLP contains one or more hidden layers. In the hidden layer, the output of the neuron can be represented by weighted summation without using an activation function. l The output of the hidden layer is h l , and its calculation formula is:

[0077] ;

[0078] in: Indicates l The weight matrix of the layer. Indicates l The bias vector of the layer. h0=A represents the input vector.

[0079] The output layer is used to calculate the noise characteristics of the vehicle , and its calculation formula is:

[0080] ;

[0081] in: Represents the weight matrix of the output layer. Represents the bias vector of the output layer. is the output of the last hidden layer. represents the calculated noise characteristics.

[0082] Collect training samples: including samples of sound field influencing factors of various reference models and noise feature labels; use the training samples to train the target model. Through the trained MLP model, for the sound field influencing factor A of a given model under test, test , we can get the corresponding noise characteristics , called the prediction noise feature.

[0083] ;

[0084] S140: Obtain a target reference vehicle model that matches the vehicle model to be tested by comparing the predicted noise feature with each reference noise feature.

[0085] The predicted noise characteristics output by the model are compared with the reference noise characteristics of the reference vehicle model. Through logistic regression, the probability value of the tested vehicle matching the existing vehicle model is obtained, and then the tested vehicle is matched with the reference vehicle model with existing noise data.

[0086] Optionally, a similarity score between the predicted noise feature and each reference noise feature is calculated; based on the similarity score and a normalized exponential function, a matching probability between the predicted noise feature and each reference noise feature is obtained; the reference vehicle model from which the reference noise feature with the highest matching probability comes is determined as the target reference vehicle model; and then the reference noise data of the target reference vehicle model under each original working condition is determined in the reference noise library for playback use.

[0087] Optionally, a similarity score between the predicted noise feature and each reference noise feature is calculated using the following formula:

[0088] ;

[0089] in, is the score of the i-th vehicle type category. m is Number of features. is the kth feature data under the ith reference noise feature.

[0090] The Softmax function is used to calculate the matching probability between the predicted noise feature and each reference noise feature. The Softmax calculation formula is:

[0091]

[0092] Where: C is the number of vehicle type categories 100. Indicates that the tested vehicle belongs to the i-th vehicle type category probability.

[0093] S150: Play back reference noise data of the target reference vehicle model under expected operating conditions in a noise playback chamber that matches the vehicle model to be tested.

[0094] The noise playback cavity is a completely enclosed space (used to simulate the cockpit) with a resonance frequency that matches the vehicle model to be tested. When the frequency of the external sound matches the resonance frequency of the cavity, it will cause a strong resonance of the cavity and amplify the sound. The noise data in the noise playback cavity is reflected many times to form a reverberation effect. The length of the cavity reverberation time is affected by factors such as the size, shape, material of the cavity and the sound absorption inside. The noise playback cavity filters noise data of different frequencies. Sounds of certain frequencies may be absorbed or reflected by the cavity, while sounds of other frequencies can pass smoothly. The resonance frequency, reverberation effect and filtering capability of the noise playback cavity should match the vehicle model to be tested.

[0095] In order to cover many reference models, this embodiment uses up to 16 microphones to collect noise data. In fact, the microphone arrangement in the model to be tested should be less than 16. In order to restore the real noise sound field, the reference noise data needs to be screened according to the microphone arrangement in the model to be tested. Optionally, the preset microphone arrangement of the model to be tested is determined (for example, arranged on both sides of the seat); according to the preset microphone arrangement, the reference noise data of the target reference model under the expected working condition is screened, that is, firstly, the target microphone closest to or overlapping with the preset microphone arrangement position is screened from the 16 microphones, and then the reference noise data collected by the target microphone is screened from the 16 groups of reference noise data of the target reference model under the expected working condition. Then, the power amplifier and speaker group consistent with the preset microphone arrangement are used to play back the screened reference noise data. Preferably, the power amplifier and speaker group use 8 full-range speakers and 1 subwoofer to simulate the background noise in the car. The interference of the sound transmission path between the speaker, the terminal microphone of the vehicle under test and the test system should be reduced, and the speaker should not resonate.

[0096] Furthermore, when collecting reference noise in the reference vehicle model and playing back reference noise data in the noise playback cavity, a noise equalization front end is used to perform sound field equalization in the noise playback cavity and the reference vehicle model. Specifically, sound field equalization can eliminate the influence of errors in speaker frequency response, placement, and directivity. Sound field equalization ensures that each playback sound field and the recorded noise sound field meet the following requirements:

[0097] (a) The level difference should be within ±1dB;

[0098] (b) In the frequency range of 100 Hz to 1000 Hz, the amplitude of the complex coherence (normalized cross-correlation spectrum) should be greater than 0.9 (measured in 1 / 3 octave bands);

[0099] (c) The phase of the complex coherence should be accurate to within ±10 degrees in the frequency range of 100 Hz to 1000 Hz and to within ±30 degrees in the frequency range of 1000 Hz to 1500 Hz (measured in 1 / 3 octave bands);

[0100] (d) The difference between the original reference noise octave band sound pressure level and the simulated noise octave band sound pressure level for each microphone should be within ±3 dB in the frequency range of 50 Hz to 10,000 Hz and within ±6 dB in the frequency range of 10,000 Hz to 16,000 Hz (measured in 1 / 3 octave bands);

[0101] (e) Within the frequency range of 50 Hz to 20,000 Hz, the average spectrum accuracy of all microphones should be within ±3 dB.

[0102] Optionally, the expected operating condition may be any original operating condition, so that after determining the target reference vehicle model, reference noise data of the target reference vehicle model under a certain original operating condition is selected from the reference noise library for playback.

[0103] Example 2

[0104] In some embodiments, the noise environment of the vehicle to be tested can be tested by testing only a small number of representative working conditions in the noise playback chamber, thereby improving the test efficiency. Therefore, before replaying the reference noise data of the target reference vehicle under the expected working conditions in the noise playback chamber matching the vehicle to be tested, a small number of typical working conditions are selected from a plurality of original working conditions.

[0105] Preferably, a reference noise data set under each original working condition is summarized, and the reference noise data set includes reference noise data of multiple vehicle models, recorded by multiple microphones, and in multiple frequency bands; the reference noise data sets under multiple original working conditions are clustered to obtain multiple typical working conditions and a typical reference noise data set under each typical working condition. The number of typical working conditions is less than the number of original working conditions.

[0106] Optionally, this embodiment uses the k-means clustering method to obtain a variety of typical working conditions. First, the elbow method is used to automatically determine k (the number of clusters). The clustering cost (the sum of squared errors within the cluster, abbreviated as SSE) under different k values ​​is calculated. For each possible k value (for example, any number from 1 to 10), the corresponding SSE is calculated:

[0107] ;

[0108] Where: k is the current number of clusters. C i is the ith cluster. x jis assigned to cluster C i The jth characteristic vector in , the characteristic vector is a reference noise data set under a certain original working condition, that is, the row vector in the above X matrix. μ i It is a cluster C i The center of mass.

[0109] Then, plot the change of SSE under different k values ​​and find the elbow point (the point where the SSE decreases significantly slower). The k value corresponding to the elbow point is the optimal number of clusters. The present invention applies the elbow rule to select = 4, that is, the optimal number of clusters obtained by the elbow rule is = 4. Next, randomly select A reference noise dataset is taken as the initial centroid. The initial centroid is For each feature vector x j (j = 1, 2,..., 100), calculate its distance to all centroids, find the nearest centroid in Euclidean order, and assign the feature vector to the corresponding cluster C i The Euclidean distance between the feature vector and the centroid is calculated as:

[0110] ;

[0111] Among them, k is a variable. is the kth characteristic data under the jth working condition.

[0112] According to the following rules, the characteristic vector x j Assign to cluster C i :

[0113] ;

[0114] For each cluster , calculate the new center of mass , which is the centroid after t+1 iterations. The centroid is the mean of all feature vectors in the cluster:

[0115]

[0116] in, Represents cluster C i The number of feature vectors in .

[0117] Check if the centroid has changed. If the centroid change is less than a preset threshold ε (usually set to a very small value, such as ε = 10 -6), or the maximum number of iterations is reached, the algorithm stops, otherwise it continues to iterate. After several iterations, the final Cluster Heart . Assume that four cluster centers are finally calculated: . Calculate each feature vector The Euclidean distance from each cluster center, select the feature vector closest to each cluster center The corresponding working condition is a typical working condition. As a typical reference noise data set.

[0118] This embodiment selects 4 typical working conditions from the aforementioned 16 working conditions through a clustering algorithm, covering typical vehicle driving noise collection scenarios.

[0119] After determining the typical working conditions, the reference noise data of each reference vehicle model under the typical working conditions, recorded by multi-point microphones, and in multiple frequency bands can be stored in the reference noise library. Then, the reference noise data of the target reference vehicle model under any typical working condition can be replayed in the noise playback chamber.

[0120] Example 3

[0121] This embodiment provides a vehicle expected operating condition noise simulation system based on reference noise matching, see Figure 3 , including: a microphone, a controller and a noise playback system, the noise playback system includes a noise equalization front end, a noise playback cavity, a power amplifier and a speaker group.

[0122] The microphone is used to collect reference noise data of various reference vehicle models under various original working conditions to build a reference noise library. The controller is used to extract features from the reference noise data to obtain reference noise features of each reference vehicle model under each original working condition; input the sound field influencing factors of the vehicle model to be tested into a pre-trained target model to obtain the predicted noise features output by the target model; and obtain the target reference vehicle model that matches the vehicle model to be tested by comparing the predicted noise features with each reference noise feature. The noise playback system is used to play back the reference noise data of the target reference vehicle model under the expected working conditions in a noise playback cavity that matches the vehicle model to be tested.

[0123] When collecting reference noise in a reference vehicle model and when playing back reference noise data in a noise playback cavity, a noise equalization front end is used to perform sound field equalization in the noise playback cavity and in the reference vehicle model.

[0124] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for simulating the expected working condition noise of an automobile based on reference noise matching, characterized in that: Collect reference noise data of various reference models under various original working conditions; Performing feature extraction on the reference noise data to obtain reference noise features of each reference vehicle model under each original working condition; Inputting the sound field influencing factors of the vehicle model to be tested into a pre-trained target model to obtain the predicted noise characteristics output by the target model; Obtaining a target reference vehicle model that matches the vehicle model to be tested by comparing the predicted noise feature with each reference noise feature; In a noise playback chamber that matches the vehicle type to be tested, reference noise data of the target reference vehicle type under expected operating conditions are played back.

2. The method according to claim 1, characterized in that Collect reference noise data of various reference models under various original working conditions, including: Control each reference vehicle model to operate under each original working condition; When the reference vehicle model is driving under any original working condition, reference noise data is collected through multi-point microphones in the reference vehicle model.

3. The method according to claim 1, characterized in that: Before inputting the sound field influencing factors of the vehicle model to be tested into the pre-trained target model, it also includes: Collecting training samples, wherein the training samples include samples of sound field influencing factors of multiple reference vehicle models and noise feature labels; The target model is trained using the training samples.

4. The method according to claim 3, characterized in that The target model is a multi-layer perceptron.

5. The method according to claim 1, characterized in that By comparing the predicted noise feature with each reference noise feature, a target reference vehicle model matching the vehicle model to be tested is obtained, including: Calculating a difference vector between the predicted noise feature and each reference noise feature; Obtaining a matching probability between the predicted noise feature and each reference noise feature according to the difference vector and the normalized exponential function; The reference vehicle model from which the reference noise feature with the highest matching probability comes is determined as the target reference vehicle model.

6. The method according to claim 1, characterized in that Before replaying the reference noise data of the target reference vehicle model under the expected working condition in the noise replay cavity matched with the vehicle model to be tested, the method further includes: Summarize a reference noise data set under each original working condition, wherein the reference noise data set includes reference noise data of multiple vehicle models, recorded by microphones at multiple locations, and in multiple frequency bands; Clustering the reference noise data sets under various original working conditions to obtain various typical working conditions and a typical reference noise data set under each typical working condition; The expected operating condition is any typical operating condition.

7. The method according to claim 1, characterized in that In a noise playback chamber that matches the vehicle type to be tested, reference noise data of the target reference vehicle type under expected operating conditions is played back, including: Determine the preset microphone arrangement for the vehicle model to be tested; Screening reference noise data of the target reference vehicle model under expected operating conditions according to the preset microphone arrangement; The screened reference noise data is played back using a power amplifier and a speaker group that are consistent with the preset microphone arrangement.

8. The method according to claim 1, characterized in that The method further comprises: When collecting reference noise in a reference vehicle model and when playing back reference noise data in a noise playback cavity, a noise equalization front end is used to perform sound field equalization in the noise playback cavity and in the reference vehicle model.

9. A vehicle expected operating noise simulation system based on reference noise matching, characterized in that: include: Microphone, used to collect reference noise data of various reference vehicle models under various original working conditions; A controller is used to extract features from the reference noise data to obtain reference noise features of each reference vehicle model under each original working condition; input the sound field influencing factors of the vehicle model to be tested into a pre-trained target model to obtain predicted noise features output by the target model; and obtain a target reference vehicle model that matches the vehicle model to be tested by comparing the predicted noise features with each reference noise feature; The noise playback system is used to play back the reference noise data of the target reference vehicle model under the expected working conditions in a noise playback cavity matched with the vehicle model to be tested.

10. The system according to claim 9, characterized in that The noise playback cavity is a completely enclosed space and has a resonance frequency that matches the vehicle model to be tested; The noise data in the noise playback cavity is reflected multiple times to form a reverberation effect, and the noise data of different frequencies are filtered.

Images