Abnormal sound detection method, device and equipment based on multi-source data fusion

Abstract

The invention discloses an abnormal sound detection method, device and equipment based on multi-source data fusion. According to the embodiment of the invention, accurate reproduction of an abnormal sound path spectrum is realized through multi-source data fusion, sound quality quantitative verification and CAE collaborative analysis; a road spectrum is generated through a series of transformations of collected road surface data and working condition data; the road spectrum is imported into a four-stand-column rack control system, the vehicle is placed on a rack, the current second audio data of the vehicle are collected, and the road spectrum is verified through the first audio data and the second audio data; and the qualified road spectrum is imported into the CAE model for analysis, and the priority troubleshooting area is positioned through modal response difference, so that the road spectrum error rate is reduced, and meanwhile, the troubleshooting efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of automotive noise, vibration, and harshness (NVH) detection technology, and in particular to a method, apparatus, and equipment for detecting abnormal noises based on multi-source data fusion. Background Technology

[0002] In the field of automotive NVH testing technology, traditional abnormal noise detection methods have significant technical bottlenecks. Existing solutions mainly rely on subjective evaluation systems or general road spectrum databases for testing. However, due to the limitations of individual cognitive differences in subjective evaluation and the insufficient adaptability of general road spectrum to different scenarios, the accuracy of road spectrum is insufficient, and simulation data is out of sync with actual road spectrum, resulting in low accuracy of abnormal noise detection. In addition, the lack of a closed-loop verification mechanism for sound quality parameters leads to low audio matching degree.

[0003] Therefore, an abnormal noise detection method is needed to improve the accuracy and precision of abnormal noise detection, as well as improve audio matching. Summary of the Invention

[0004] This invention provides a method, apparatus, and device for detecting abnormal noises based on multi-source data fusion, in order to solve existing problems.

[0005] This invention provides an abnormal noise detection method based on multi-source data fusion, the method comprising: Acquire multi-source data, which includes road surface data, operating condition data, and first audio data. The road surface data carries road surface information, the operating condition data carries vehicle dynamic parameters, and the first audio data carries the vehicle's first sound quality parameters. Based on the road surface data and the working condition data, a road spectrum is generated and imported into the four-column test bench control system. The vehicle is placed on the test bench, and the current second audio data of the vehicle is collected. The second audio data carries the second sound quality parameters of the vehicle. Based on the first audio data and the second audio data, the road spectrum is verified. If the verification result is that the road spectrum is qualified, computer-aided engineering (CAE) simulation is performed based on the road spectrum to obtain simulated abnormal noise data. Based on the simulated abnormal noise data and the actual abnormal noise data, high abnormal noise risk areas are determined.

[0006] Furthermore, the road surface data is road surface point cloud data acquired based on a 3D laser scanner; the operating condition data includes vehicle instantaneous speed, wheelbase, and track width. The process of generating a road spectrum based on the road surface data and the operating condition data includes: Based on a preset coordinate transformation algorithm, the road surface point cloud data is transformed in coordinates; A two-dimensional grid is constructed based on the road surface point cloud data after coordinate transformation to obtain the road surface grid. The height value of each grid in the road surface grid is determined by interpolation to obtain the road surface height data matrix. Based on the instantaneous speed of the vehicle, determine the time sequence of the vehicle's travel on the road surface; Based on the time series and the instantaneous speed of the vehicle, determine the vehicle's position on the road surface at any moment in the time series; For each moment in the time series, based on the wheelbase, the track width, and the vehicle's position on the road surface at that moment, the target grid corresponding to each wheel of the vehicle on the road surface grid is determined, and based on the road surface height value corresponding to the target grid in the road surface height data matrix, the road surface height value corresponding to each wheel of the vehicle is determined. The road spectrum is generated by taking the road surface height value corresponding to each wheel of the vehicle at each moment in the time series.

[0007] Furthermore, the coordinate transformation of the road surface point cloud data based on the preset coordinate transformation algorithm includes: For any point cloud in the road surface point cloud data, the coordinates of the point cloud are multiplied by a preset coordinate transformation matrix to obtain the transformed coordinates of the point cloud.

[0008] Furthermore, the process of determining the coordinate transformation matrix includes: Based on a preset first angle, determine a first sub-rotation matrix that rotates around the z-axis by the first angle. Based on the preset second angle, determine the second sub-rotation matrix that rotates around the x-axis by the second angle; Based on the preset third angle, determine the third sub-rotation matrix that rotates around the y-axis by the third angle; The rotation matrix is ​​determined based on the first sub-rotation matrix, the second sub-rotation matrix, and the third sub-rotation matrix; The coordinate transformation matrix is ​​determined based on the rotation matrix and the preset translation vector.

[0009] Furthermore, if it is determined that the range of the road surface point cloud data does not meet the preset constraints, then after performing coordinate transformation on the road surface point cloud data based on the preset coordinate transformation algorithm, the method further includes: If the first length of the road surface point cloud data in the x-axis direction does not reach a first length threshold, then a first difference between the first length threshold and the first length is determined; based on the first difference and the first length, a first number of replications of the road surface point cloud data is determined; based on the first number of replications, the point cloud data is controlled to be periodically replicated in the x-axis direction; and / or, If the second length of the road surface point cloud data in the y-axis direction does not reach the second length threshold, then a second difference between the second length threshold and the second length is determined; based on the second difference and the second length, a second number of copies of the road surface point cloud data is determined; based on the second number of copies, the point cloud data is controlled to be periodically copied in the y-axis direction.

[0010] Further, determining the second number of copies of the road surface point cloud data based on the second difference and the second quantity includes: Determine the ratio of the second difference to the second quantity; Half of the ratio is taken as the second number of copies; The step of controlling the point cloud data to be periodically copied in the y-axis direction according to the second number of copies includes: Based on the second number of copies, the point cloud data is controlled to be periodically copied in the positive y-axis direction; Based on the second number of copies, the point cloud data is controlled to be periodically copied in the negative y-axis direction.

[0011] Further, determining the road height value corresponding to each wheel of the vehicle based on the road height value corresponding to the target grid in the road height data matrix includes: For each wheel of the target vehicle, if there is only one target grid corresponding to that wheel, the road height value corresponding to that target grid in the road height data matrix is ​​determined as the road height value corresponding to that wheel; if there are multiple target grids corresponding to that wheel, the average of the road height values ​​corresponding to each target grid in the road height data matrix is ​​determined as the road height value corresponding to that wheel.

[0012] Furthermore, the road surface data is a road surface image acquired by an image acquisition device; the operating condition data includes the vehicle's instantaneous speed, wheelbase, and track width. The process of generating a road spectrum based on the road surface data and the operating condition data includes: Feature extraction is performed on the road surface image to determine the element attributes and dimensions of the road surface contained in the road surface image; Based on the element attributes and size, determine the road spectrum with the highest matching degree with the road surface from the pre-configured road spectrum library; The road spectrum is updated based on the operating condition data.

[0013] This application embodiment also provides an abnormal noise detection device based on multi-source data fusion, the device comprising: The data acquisition module is used to acquire multi-source data, including road surface data, operating condition data and first audio data. The road surface data carries road surface information, the operating condition data carries vehicle dynamic parameters, and the first audio data carries the vehicle's first sound quality parameters. The processing module is used to generate a road spectrum based on the road surface data and the working condition data, import the road spectrum into the four-column test bench control system, place the vehicle on the test bench, and collect the current second audio data of the vehicle, the second audio data carrying the second sound quality parameters of the vehicle; The detection module is used to verify the road spectrum based on the first audio data and the second audio data. If the verification result is that the road spectrum is qualified, CAE simulation is performed based on the road spectrum to obtain simulated abnormal noise data. Based on the simulated abnormal noise data and the real abnormal noise data, high abnormal noise risk areas are determined.

[0014] This application also provides an electronic device, including: Memory, used to store computer programs; When a processor executes a computer program stored in the memory, it implements the method steps described in any one of the above statements.

[0015] This application embodiment achieves accurate reproduction of abnormal noise road spectrum through multi-source data fusion, sound quality quantification verification, and CAE collaborative analysis. Road spectrum is generated by a series of transformations using collected road surface data and operating condition data. The road spectrum is then imported into a four-column test bench control system. The vehicle is placed on the test bench, and the current second audio frequency of the vehicle is collected. The road spectrum is verified using the first and second audio data. Qualified road spectra are imported into a CAE model for analysis. Modal response differences are used to locate priority areas for investigation, reducing the road spectrum error rate and improving investigation efficiency. Attached Figure Description

[0016] The accompanying drawings, which form part of this specification, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 A schematic diagram of an abnormal noise detection process based on multi-source data fusion is provided for an embodiment of this application; Figure 2 A flowchart illustrating the method for accurately reproducing abnormal noises on a four-pillar bench of a car based on road surface information collection, provided in this embodiment of the application. Figure 3 This is a schematic diagram of an asphalt pavement image provided in an embodiment of this application; Figure 4 A flowchart illustrating the method for replacing and reproducing abnormal noises on a four-pillar bench of a vehicle based on road surface information collection, provided in this embodiment of the application. Figure 5 A schematic diagram of the structure of an abnormal noise detection device based on multi-source data fusion provided in an embodiment of this application; Figure 6 This is a schematic diagram of the electronic device structure provided in an embodiment of this application. Detailed Implementation

[0017] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0018] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this invention is for describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0019] In the field of automotive NVH (Noise, Vibration, and Harshness) testing technology, traditional methods for detecting abnormal noises and vibrations face significant technical bottlenecks. Existing solutions primarily rely on subjective evaluation systems or general road spectrum databases for testing. However, due to limitations in individual cognitive differences in subjective evaluations and insufficient adaptability of general road spectrum databases to specific scenarios, the accuracy of abnormal noise reproduction has consistently remained below 60%. At the data acquisition level, current technologies generally employ a single accelerometer as the core sensor, lacking the ability to collaboratively acquire multi-dimensional data such as road surface geometry, vehicle dynamic conditions, and sound quality parameters, making it difficult to construct a complete database of abnormal noise characteristics.

[0020] More critically, the existing technical system has not yet established a quantitative evaluation standard based on sound quality parameters (such as eight core indicators including sharpness and roughness), resulting in a lack of objective evaluation criteria during bench reproduction. Problem troubleshooting relies on repeated trial and error based on human experience, leading to overall low efficiency. These technical shortcomings directly result in insufficient accuracy in reproducing abnormal noises on benches in different locations, becoming a core pain point restricting the rapid diagnosis of abnormal noise problems during vehicle development.

[0021] In fact, the matching error rate between general road spectrum and actual operating conditions in traditional solutions typically exceeds 40%. The data acquisition mode of a single accelerometer leads to a missing rate of over 35% for key abnormal noise characteristic parameters, and the lack of a quantitative standard for troubleshooting processes can extend the problem localization time by 2-3 times. Therefore, realizing the high-precision road spectrum generation and quantitative evaluation system through multi-source data fusion has become an inevitable path to overcome the current technological bottlenecks.

[0022] Existing technologies have three significant shortcomings in the field of abnormal sound reproduction: 1. Insufficient accuracy of road spectrum: General road spectrum cannot reproduce the actual abnormal noise conditions. Traditional technologies using general road profiles (such as standard models of smooth asphalt roads and damaged asphalt roads) rely on a single sensor (such as an accelerometer) to collect data and lack the road surface point cloud information (including microscopic geometric features such as the depth of bumps and potholes and the spacing between obstacles) obtained by 3D laser scanning. This results in a matching error rate of over 40% with actual abnormal noise conditions. Taking a damaged asphalt road with a bump height of 5-10mm as an example, the general road profile has a significant deviation in the morphological reconstruction of road obstacles and cannot reproduce the abnormal noise conditions caused by irregular road surface excitation during actual driving.

[0023] II. Low audio matching degree: lack of closed-loop verification mechanism for sound quality parameters The existing solution can only collect basic acoustic parameters such as sound pressure level, and has not established a closed-loop verification system for eight sound quality parameters, including sharpness, roughness, and fluctuation. Specifically, the sharpness parameter deviation between the bench-reproduced audio and the original road test audio can be more than 25%, and the roughness index difference exceeds 30%. Furthermore, it does not introduce a quantitative comparison model based on the Euclidean distance algorithm (such as parameter weighting allocation of 30% for the spectrum and 20% for the order), making it impossible to judge the reproduction effect through sound quality consistency, and relying on manual experience to troubleshoot problems.

[0024] III. Weak correlation between CAE simulation and actual road spectrum: simulation data is disconnected from actual road profiles. Traditional CAE simulations, lacking multi-source fusion inputs such as road point clouds, operating parameters, and sound quality data, result in frequency response errors exceeding 15dB between the simulation model and the actual abnormal noise spectrum within the critical frequency band of 20-2000Hz (e.g., a certain vehicle model has an error of 18dB at 100Hz). Furthermore, modal response deviations exceeding 20% ​​occur in unrefined areas with mesh sizes ≥5mm, making it difficult to accurately locate abnormal noise sources at vehicle body joints. In practical applications, the misalignment rate between the first three priority inspection areas and the actual problem points is as high as 45%, severely impacting the efficiency of abnormal noise troubleshooting.

[0025] Based on this, the present invention aims to overcome the technical difficulties of existing magnetic bearings, such as the lack of intelligent adaptive capabilities and the difficulty in automatically optimizing control strategies according to real-time operating conditions, and to provide a method, device and equipment for abnormal noise detection based on multi-source data fusion.

[0026] Figure 1 This application provides a schematic diagram of an abnormal noise detection process based on multi-source data fusion, as illustrated in the embodiments of this application. Figure 1 As shown, the process includes: S101: Acquire multi-source data, which includes road surface data, operating condition data, and first audio data. The road surface data carries road surface information, the operating condition data carries vehicle dynamic parameters, and the first audio data carries the vehicle's first sound quality parameters.

[0027] This embodiment provides a method for detecting abnormal noises based on multi-source data fusion, which can be applied to electronic devices such as servers and PCs.

[0028] The abnormal noise detection method based on multi-source data fusion provided in this application integrates road surface data, working condition data and first audio data to construct a high-precision road spectrum, and realizes accurate reproduction and verification of abnormal noise on a four-column test bench.

[0029] Specifically, electronic devices acquire multi-source data through various channels, including but not limited to road surface data, operating condition data, and primary audio data. Road surface data carries road information, such as point cloud data collected by a 3D laser scanner, which accurately reflects the microscopic geometric features of the road surface, such as the depth of bumps and potholes, and the distance between obstacles. Operating condition data includes dynamic parameters such as vehicle instantaneous speed (recording error ≤ 0.5 km / h), suspension height, axle load, driving mode (e.g., Comfort / Sport / Off-road / Highland mode, Pure Electric / ICE / Hybrid mode, whether regenerative braking is activated), front and rear axle loads, wheelbase / track width, and wheel center acceleration (sampling frequency ≥ 5 kHz). Primary audio data contains eight primary sound quality parameters collected during the vehicle's actual road driving, including but not limited to spectrum, order, sharpness, impact signal frequency, roughness, fluctuation, jitter, speech intelligibility, and loudness.

[0030] S102: Generate a road spectrum based on the road surface data and the working condition data, import the road spectrum into the four-column test bench control system, place the vehicle on the test bench, and collect the current second audio data of the vehicle. The second audio data carries the second sound quality parameters of the vehicle.

[0031] In this embodiment, after obtaining the aforementioned multi-source data, the electronic device enters the virtual road surface reconstruction stage, which involves fusing road surface data with operating condition data to generate a customized road spectrum. In this embodiment, a customized road spectrum is generated by fusing road surface point cloud data with operating condition parameters. Utilizing precise road surface point cloud data obtained through 3D laser scanning, combined with operating condition parameters, the road surface conditions under actual driving conditions can be more realistically reproduced, generating a customized road spectrum that meets specific needs. Iterative processing is performed using operating condition records from the data acquisition stage to generate a road spectrum signal usable on the test bench.

[0032] Among them, electronic devices can generate road profiles based on road surface data and operating condition data.

[0033] The electronic equipment imports the generated road spectrum into the four-post test bench control system and places the vehicle to be tested on the four-post test bench. If the abnormal noise phenomenon is sensitive to ambient temperature, the temperature of the four-post test bench environmental chamber must be adjusted to the temperature at which the abnormal noise occurs one day in advance, and the vehicle must be immersed in this environment for at least 8 hours to ensure that the material condition is consistent with the actual condition. Subsequently, microphones are placed to collect the sound signals currently emitted by the vehicle during the operation of the four-post test bench, forming second audio data. This second audio data contains second sound quality parameters, which are of the same type as the first sound quality parameters in the above embodiment, only the values ​​of each sound quality parameter are different.

[0034] S103: Verify the road spectrum based on the first audio data and the second audio data. If the verification result is that the road spectrum is qualified, perform CAE simulation based on the road spectrum to obtain simulated abnormal noise data, and determine the high abnormal noise risk area based on the simulated abnormal noise data and the real abnormal noise data.

[0035] In this embodiment of the application, after determining the first audio data and the second audio data, the electronic device can perform closed-loop verification of the first audio data and the second audio data.

[0036] Specifically, electronic devices can use the Euclidean distance algorithm to calculate the similarity score between the first audio data and the second audio data. When calculating similarity, different types of sound quality parameters have different weights, and the weights of each type of sound quality parameter can be configured according to requirements.

[0037] For example, the weights for each type of sound quality parameter can be: 30% for spectrum, 20% for order, 15% for sharpness, 10% for roughness, 10% for fluctuation, 5% for jitter, 5% for speech intelligibility, and 5% for loudness.

[0038] The similarity score between two audio data can be determined based on the following formula:

[0039] Where S represents the similarity score. Let i be the weight of the i-th parameter. This represents the measured value of the four-column support frame corresponding to the i-th parameter. This is the original road test baseline value corresponding to the i-th parameter.

[0040] In this embodiment of the application, when the similarity score between the first audio data and the second audio data reaches the similarity score threshold, the electronic device determines that the road spectrum is qualified; otherwise, the electronic device will reconstruct the road spectrum.

[0041] For example, the similarity score threshold can be 85%, that is, when the similarity score S≥85%, the electronic device determines that the road spectrum is qualified; otherwise, it returns to the virtual road reconstruction step for modification until the standard is met.

[0042] In this embodiment of the application, if the electronic device determines that the road spectrum verification is qualified, the electronic device performs CAE-assisted analysis based on the road spectrum.

[0043] Specifically, the electronic device outputs the road spectrum to the CAE simulation model, which then outputs simulated abnormal noise data. This CAE simulation model can cover the key frequency band of 20-2000Hz, with a grid size of ≤5mm in key areas and localized refinement.

[0044] Electronic equipment compares the differences between simulated and actual abnormal noise data (with a set threshold of ±2dB), identifying the three vehicle body attachment points with the largest deviations as priority areas for investigation. By combining mesh analysis of vehicle body deformation, displacement cloud maps, strain energy maps, and modal responses, high-risk areas for abnormal noise are ultimately determined, significantly improving the efficiency of abnormal noise diagnosis and shortening the investigation cycle.

[0045] The abnormal noise detection method based on multi-source data provided in this application effectively solves the problems in traditional abnormal noise detection, such as large errors in general road spectrum matching, lack of sound quality quantification standards, and weak correlation in CAE simulation. It realizes standardized management of the entire process from data acquisition to problem location, improves the overall abnormal noise reproduction accuracy to an error rate of ≤5%, and increases the troubleshooting efficiency by more than 50%.

[0046] To improve the accuracy and precision of abnormal noise detection, based on the above embodiments, in this embodiment, the road surface data is road surface point cloud data acquired by a 3D laser scanner; the operating condition data includes vehicle instantaneous speed, wheelbase, and track width. The process of generating a road spectrum based on the road surface data and the operating condition data includes: Based on a preset coordinate transformation algorithm, the road surface point cloud data is transformed in coordinates; A two-dimensional grid is constructed based on the road surface point cloud data after coordinate transformation to obtain the road surface grid. The height value of each grid in the road surface grid is determined by interpolation to obtain the road surface height data matrix. Based on the instantaneous speed of the vehicle, determine the time sequence of the vehicle's travel on the road surface; Based on the time series and the instantaneous speed of the vehicle, determine the vehicle's position on the road surface at any moment in the time series; For each moment in the time series, based on the wheelbase, the track width, and the vehicle's position on the road surface at that moment, the target grid corresponding to each wheel of the vehicle on the road surface grid is determined, and based on the road surface height value corresponding to the target grid in the road surface height data matrix, the road surface height value corresponding to each wheel of the vehicle is determined. The road spectrum is generated by taking the road surface height value corresponding to each wheel of the vehicle at each moment in the time series.

[0047] In this embodiment, the road surface data can be road surface point cloud data acquired using a 3D laser scanner. The operating condition data may include, but is not limited to, vehicle instantaneous speed, wheelbase, and track width. Specifically, the road surface data is acquired by a handheld 3D laser scanner with a resolution of at least 0.1 mm and a sampling frequency of at least 100 kHz, enabling efficient capture of the road surface's microscopic geometric features. The 3D laser scanner's high speed, high precision, non-contact measurement, and automatic processing capabilities allow for rapid and accurate acquisition of road surface point cloud data. Its precision advantage allows for the capture of subtle geometric changes in the road surface, providing a reliable data foundation for subsequent analysis.

[0048] During the generation of road spectrum, because the road surface point cloud data collected by the portable 3D laser scanner is directional, the electronic device needs to adjust the collected road surface point cloud data to road surface point cloud data in a standard coordinate system that conforms to the vehicle traveling along the horizontal x-axis direction through a preset coordinate transformation algorithm.

[0049] The electronic device constructs a two-dimensional grid based on the coordinate-transformed road surface point cloud data, obtains the road surface grid, and determines the height value of each grid in the road surface grid by interpolation, thus obtaining the road surface height data matrix.

[0050] Specifically, the electronic device constructs a two-dimensional mesh based on the expanded road surface point cloud data to obtain the road surface network. For example, it can be... The axis direction is the direction of vehicle travel. The axis direction is perpendicular to the driving direction, and the dimensions of the road network are set as follows: For each grid cell, the height value of the grid node is calculated using interpolation. ( express Orientation grid index, express (Grid index of direction). A commonly used interpolation method is inverse distance weighted interpolation, which calculates the height of the point to be determined by weighted averaging the height and distance of known points around it. Let the grid node to be determined be... Surrounding Known points ( Then the height value of the grid node. The calculation formula is:

[0051] in, This represents the distance between the point to be determined and the known points. The weighting index is usually taken as... Using the above method, a gridded road surface height data matrix is ​​obtained. .

[0052] Electronic devices determine the time series of a vehicle's journey on the road based on its instantaneous speed. Specifically, if the initial time is 0 and sampling is performed at a fixed sampling interval Δt, the entire time series can be derived.

[0053] For example, electronic devices adjust based on the vehicle's instantaneous speed. (Speed ​​recording error) ), calculate the time series of vehicle travel on the road. Assuming the vehicle starts moving from the starting point, the initial time... With a fixed sampling interval Sampling was performed. The distance traveled by the vehicle within each sampling interval was recorded. ( For the first (instantaneous velocity at the sampling time), then the... Time of each sampling moment In this way, time series of vehicle travel are generated. ,in This represents the total number of samples taken.

[0054] The distance a vehicle travels at any given moment is determined by its instantaneous speed at that moment. Thus, electronic devices can determine the vehicle's position on the road surface at any given moment in the time series based on the time series and the vehicle's instantaneous speed.

[0055] For each moment in the time series, the electronic device determines the target grid corresponding to each wheel of the vehicle on the road surface grid based on the wheelbase, track width, and the vehicle's position on the road surface at that moment. It then determines the road surface height value corresponding to each wheel of the vehicle based on the road surface height value corresponding to the target grid in the road surface height data matrix.

[0056] This electronic device generates a road spectrum based on the road surface height values ​​corresponding to each wheel of the vehicle at each moment in the time series. Specifically, the electronic device arranges the road surface height values ​​corresponding to the four wheels at each moment in chronological order, generating four "height-time" curves, which are then used as the road spectrum. Each curve records the road surface height data corresponding to the four wheels at each moment, with time as the horizontal axis and wheel height as the vertical axis.

[0057] In addition, in this embodiment of the application, the road surface point cloud data collected by the 3D laser scanner may contain noise and redundant information. Electronic devices can filter the collected road surface point cloud data to remove noise points and redundant information, thereby improving the quality of the road surface point cloud data.

[0058] Specifically, electronic devices can employ statistical filtering methods to calculate the average distance between each point and its neighboring points based on the spatial distribution statistical characteristics of road surface point cloud data. This can be achieved using the following formula:

[0059] in, For point coordinates For point Areas coordinates For the domain point The quantity.

[0060] In this embodiment of the application, through refined point cloud processing and vehicle kinematic modeling, a high degree of consistency between the road spectrum and the actual working conditions is ensured, which significantly improves the accuracy and reliability of abnormal noise reproduction.

[0061] To improve the accuracy and precision of abnormal noise detection, based on the above embodiments, in this embodiment, the coordinate transformation of the road surface point cloud data based on a preset coordinate transformation algorithm includes: For any point cloud in the road surface point cloud data, the coordinates of the point cloud are multiplied by a preset coordinate transformation matrix to obtain the transformed coordinates of the point cloud.

[0062] Road surface point cloud data acquired by a 3D laser scanner is directional, but the direction of this point cloud data may deviate from the x-axis direction. In this embodiment, the electronic device can use a preset coordinate transformation matrix to uniformly transform the coordinates of each point in the road surface point cloud data.

[0063] Specifically, the electronic device represents the original coordinates of any point cloud P in the road surface point cloud data as homogeneous coordinates, and then multiplies them with a preset coordinate transformation matrix T to obtain the transformed coordinates.

[0064] For example, electronic devices can perform coordinate transformations using the following formula:

[0065] in, Let P be the original coordinates of any point cloud, and T be the coordinate transformation matrix. The coordinates of point cloud P after transformation.

[0066] To improve the accuracy and precision of abnormal noise detection, based on the above embodiments, in this embodiment, the process of determining the coordinate transformation matrix includes: Based on a preset first angle, determine a first sub-rotation matrix that rotates around the z-axis by the first angle. Based on the preset second angle, determine the second sub-rotation matrix that rotates around the x-axis by the second angle; Based on the preset third angle, determine the third sub-rotation matrix that rotates around the y-axis by the third angle; The rotation matrix is ​​determined based on the first sub-rotation matrix, the second sub-rotation matrix, and the third sub-rotation matrix; The coordinate transformation matrix is ​​determined based on the rotation matrix and the preset translation vector.

[0067] In this embodiment, the coordinate transformation matrix is ​​implemented based on rotation and translation operations in spatial geometry. Specifically, the electronic device determines three rotation angles based on the posture information of the laser scanner in the actual acquisition scene: a first angle around the z-axis. The second angle around the x-axis The third angle around the y-axis These three angles are used to correct deviations in the point cloud data in the heading, pitch, and roll directions, respectively.

[0068] The electronic device can construct three sub-rotation matrices sequentially based on a first angle, a second angle, and a third angle. These three sub-rotation matrices include: rotating around the z-axis by the first angle. First sub-rotation matrix Rotate around the x-axis by a second angle Second sub-rotation matrix and rotate around the y-axis by the third angle The third sub-rotation matrix .

[0069] The first sub-rotation matrix can be represented as:

[0070] The second sub-rotation matrix can be represented as:

[0071] The second sub-rotation matrix can be represented as:

[0072] The electronic device multiplies the first sub-rotation matrix, the second sub-rotation matrix, and the third sub-rotation matrix in a specific order to obtain the total rotation matrix R.

[0073] For example, an electronic device can determine the rotation matrix using the following formula:

[0074] in, Let be a rotation matrix. This is the second sub-rotation matrix. For the third sub-rotation matrix, This is the first sub-rotation matrix.

[0075] In addition to the example above, the electronic device can also multiply the first sub-rotation matrix, the second sub-rotation matrix, and the third sub-rotation matrix in other orders to obtain the total rotation matrix R; for example, ,or, etc., without restriction.

[0076] After determining the rotation matrix, the electronic device can then determine the coordinate transformation matrix T based on the rotation matrix and a preset translation vector. The preset translation vector includes translation components in the x-axis, y-axis, and z-axis directions, respectively.

[0077] For example, the preset translation vector can be represented as , This represents the translation component along the x-axis. This represents the translation component along the y-axis. This represents the translation component along the z-axis.

[0078] The electronic device can determine the coordinate transformation matrix T based on the following formula:

[0079] Where T is the coordinate transformation matrix and R is the rotation matrix. This is the preset translation vector.

[0080] To improve the accuracy and precision of abnormal noise detection, based on the above embodiments, in this embodiment, if it is determined that the range of the road surface point cloud data does not meet the preset constraints, then after performing coordinate transformation on the road surface point cloud data based on the preset coordinate transformation algorithm, the method further includes: If the first length of the road surface point cloud data in the x-axis direction does not reach a first length threshold, then a first difference between the first length threshold and the first length is determined; based on the first difference and the first length, a first number of replications of the road surface point cloud data is determined; based on the first number of replications, the point cloud data is controlled to be periodically replicated in the x-axis direction; and / or, If the second length of the road surface point cloud data in the y-axis direction does not reach the second length threshold, then a second difference between the second length threshold and the second length is determined; based on the second difference and the second length, a second number of copies of the road surface point cloud data is determined; based on the second number of copies, the point cloud data is controlled to be periodically copied in the y-axis direction.

[0081] Based on the above embodiments, to ensure a more accurate road spectrum, the road surface point cloud data used to determine the road spectrum needs to be sufficiently large; that is, the range of the road surface point cloud data is constrained, and the range of the road surface point cloud data needs to meet preset constraint conditions. Therefore, after performing coordinate transformation on the road surface point cloud data, the electronic device will determine whether the range of the road surface point cloud data meets the preset constraint conditions. If the electronic device determines that the range of the road surface point cloud data meets the preset constraint conditions, then the electronic device will execute the subsequent process of constructing the road surface mesh; if the electronic device determines that the range of the road surface point cloud data does not meet the preset constraint conditions, then the electronic device can expand the data range of the road surface point cloud data.

[0082] The preset constraints include a first length threshold that limits the length of the road surface point cloud data in two directions and a second length threshold that limits the width. Generally, the first length threshold corresponds to the x-axis direction and the second length threshold corresponds to the y-axis direction; alternatively, the first length threshold can correspond to the y-axis direction and the second length threshold can correspond to the x-axis direction.

[0083] The specific values ​​of the first and second length thresholds can be configured by technicians according to actual needs. For example, the first length threshold can be set to 50m, and the second length threshold can be set to the vehicle wheelbase W.

[0084] When expanding the data range of road surface point cloud data, electronic devices can use at least one of vertical expansion and horizontal expansion. Vertical expansion expands the point cloud data in the x-axis direction, while horizontal expansion expands the point cloud data in the y-axis direction.

[0085] Specifically, if the first length of the road surface point cloud data in the x-axis direction does not reach the first length threshold, the electronic device determines the first difference between the first length threshold and the first length; and determines the first number of times the road surface point cloud data is copied based on the first difference and the first length; and controls the point cloud data to be copied periodically in the x-axis direction based on the first number of copies.

[0086] Specifically, the electronic device can determine the ratio of the first difference to the first length and use this ratio as the first number of copies of the road surface point cloud data. If the ratio is not an integer, at least one of the following algorithms—rounding up, rounding down, and rounding to the nearest integer—can be used to round the ratio. Alternatively, the ratio can be left unrounded, and the decimal part can be used as a weight to extract a portion of the road surface point cloud data for copying in the last copy cycle. If the ratio is an integer, no operation is required on the ratio.

[0087] For example, if the range of the road surface point cloud data in the x-axis direction is... Then the electronic device can determine the first length of the road surface point cloud data in the x-axis direction. Based on this, if the first length threshold is... Then the electronic device determines the first difference. The electronic device is based on this first difference. and the first length threshold Determine the first replication count ,in, This indicates rounding down. The electronic device sequentially translates the road surface point cloud data k·L0 (k=1,2,…,n) along the x-axis and merges it with the original road surface point cloud data, thus expanding the road surface point cloud data along the x-axis.

[0088] If the second length of the road surface point cloud data in the y-axis direction does not reach the second length threshold, the electronic device determines the second difference between the second length threshold and the second length; and determines the second number of times the road surface point cloud data is copied based on the second difference and the second length; and controls the point cloud data to be copied periodically in the y-axis direction based on the second number of copies.

[0089] Specifically, the electronic device can determine the ratio of the second difference to the second length and use this ratio as the second replication count of the road surface point cloud data. If the ratio is not an integer, at least one of the following algorithms—rounding up, rounding down, and rounding to the nearest integer—can be used to round the ratio. Alternatively, the ratio can be left unrounded, and the decimal part can be used as the weight for replicating a portion of the road surface point cloud data in the last replication cycle. If the ratio is an integer, no operation is required on the ratio.

[0090] For example, if the range of the road surface point cloud data in the y-axis direction is... Then the electronic device can determine the second length of the road surface point data in the y-axis direction. Based on this, if the second length threshold is... Then the electronic device determines the second difference. The electronic device is based on this second difference. and the second length threshold Determine the second replication number ,in, This indicates rounding down. The electronic device sequentially shifts the road surface point cloud data k·W0 (k=1,2,…,n) along the y-axis and merges it with the original road surface point cloud data, thus expanding the road surface point cloud data along the y-axis.

[0091] In this embodiment, the electronic device ensures that the final road surface point cloud data meets the minimum size requirements of the four-column bench test by extending the longitudinal and lateral dimensions as described above, thereby avoiding road spectrum distortion or discontinuous excitation due to excessively small data range and ensuring the authenticity and stability of abnormal noise reproduction.

[0092] To improve the accuracy and precision of abnormal noise detection, based on the above embodiments, in this embodiment, determining the second number of copies of the road surface point cloud data according to the second difference and the second quantity includes: Determine the ratio of the second difference to the second quantity; Half of the ratio is taken as the second number of copies; The step of controlling the point cloud data to be periodically copied in the y-axis direction according to the second number of copies includes: Based on the second number of copies, the point cloud data is controlled to be periodically copied in the positive y-axis direction; Based on the second number of copies, the point cloud data is controlled to be periodically copied in the negative y-axis direction.

[0093] In addition to determining the second number of copies and controlling the periodic copying of point cloud data in the y-axis direction based on the second number of copies as described in the above embodiments, the electronic device can also copy road point cloud data from the positive y-axis direction and the negative y-axis direction respectively.

[0094] Specifically, the electronic device determines a second length threshold and a second difference between the second length and the second length of the road surface point cloud data in the y-axis direction, and uses half of the ratio of this second difference to the second quantity as the second replication count. The electronic device then replicates the road surface point cloud data in both the positive and negative y-axis directions according to this second replication count.

[0095] The electronic device can control the point cloud data to be periodically copied in the positive y-axis direction according to the second copy number; the electronic device can also control the point cloud data to be periodically copied in the negative y-axis direction according to the second copy number.

[0096] For example, if the range of the road surface point cloud data in the y-axis direction is... Then the electronic device can determine the second length of the road surface point data in the y-axis direction. Based on this, if the second length threshold is... Then the electronic device determines the second difference. The electronic device is based on this second difference. and the second length threshold Determine the second replication number ,in, This indicates rounding down. The electronic device sequentially shifts the road surface point cloud data k·W0 (k=1,2,…,n) along the y-axis and merges it with the original road surface point cloud data, thus expanding the road surface point cloud data along the y-axis.

[0097] Building upon the above example, in another possible implementation, the electronic device may also determine in The widths expanded in the positive y-axis direction and the negative y-axis direction are respectively... (like If it is an odd number, it can be (Extend one more unit in the positive axis direction). Based on this, the electronic device determines the number of copies. , This electronic device will convert the raw road surface point cloud data into... Translate sequentially along the positive axis ( ),exist Translate sequentially along the negative axis ( ), and merged with the original road surface point cloud data, to achieve the integration of road surface point cloud data in... The axial expansion ultimately yields road surface point cloud data that satisfies the constraints.

[0098] The lateral expansion provided in this application ensures the uniform distribution of the expanded road surface point cloud data in the lateral direction, avoiding asymmetric excitation caused by unilateral expansion, thereby more realistically simulating the driving state of vehicles on wide roads and improving the physical consistency of abnormal noise reproduction.

[0099] To improve the accuracy and precision of abnormal noise detection, based on the above embodiments, in this embodiment, determining the road height value corresponding to each wheel of the vehicle according to the road height value corresponding to the target grid in the road height data matrix includes: For each wheel of the target vehicle, if there is only one target grid corresponding to that wheel, the road height value corresponding to that target grid in the road height data matrix is ​​determined as the road height value corresponding to that wheel; if there are multiple target grids corresponding to that wheel, the average of the road height values ​​corresponding to each target grid in the road height data matrix is ​​determined as the road height value corresponding to that wheel.

[0100] In this embodiment of the application, when the electronic device determines the road surface height value corresponding to each wheel of the vehicle, for each wheel, after determining the target grid of the wheel in the road surface grid, if the wheel happens to fall within a single grid, the height value of that grid is directly taken as the wheel height; if it crosses multiple grids (such as being located at the intersection of four adjacent grids), the weighted average height is calculated using bilinear interpolation.

[0101] Specifically, the electronic equipment combines the wheelbase L and track width W to determine the coordinate positions of the four wheels on the road surface grid at each time step. For example, the coordinates of the vehicle's front axle center are... The coordinates of the rear axle center are ,and Then the coordinates of the left front wheel are The right front wheel is The coordinates of the left rear wheel are The coordinates of the right rear wheel are: .

[0102] As the vehicle moves, electronic devices calculate the distance traveled based on time sequence and speed, thereby updating the coordinates of the four wheels on the road surface grid. For the coordinates of each wheel... The electronic device calculates the road height value z corresponding to the wheel by searching the gridded road height data matrix Z and using bilinear interpolation.

[0103] For example, let the wheel coordinates be... The corresponding target grid on the road surface grid is , , , The height values ​​of the four target grids are respectively , , and Then the road surface height value corresponding to that wheel The calculation formula is:

[0104] in, , .

[0105] Based on the above embodiments, in the embodiments of this application, Figure 2 This is a flowchart illustrating the method for accurately reproducing abnormal noises from a four-pillar test bench based on road surface information collection, as provided in this application embodiment. Figure 2 As shown, the process includes: S1: Road surface recording: 3D laser scanning to acquire point cloud data; S2: Operating Condition Record: Collects dynamic parameters; S3: Audio Recording: Collects 8 sound quality parameters; S4: Point cloud preprocessing; coordinate transformation processing; data range expansion; road surface point cloud meshing; road spectrum file generation; S5: Import the road spectrum into the four-column test stand; acquire audio signals; determine the comprehensive score S≥85%; if yes, proceed to S6; otherwise, proceed to S4; S6: Output qualified road spectrum to CAE model; compare actual and simulated modal responses, and determine priority areas for investigation.

[0106] To improve the accuracy and precision of abnormal noise detection, based on the above embodiments, in this embodiment, the road surface data is a road surface image acquired by an image acquisition device; the operating condition data includes the vehicle's instantaneous speed, wheelbase, and track width. The process of generating a road spectrum based on the road surface data and the operating condition data includes: Feature extraction is performed on the road surface image to determine the element attributes and dimensions of the road surface contained in the road surface image; Based on the element attributes and size, determine the road spectrum with the highest matching degree with the road surface from the pre-configured road spectrum library; The road spectrum is updated based on the operating condition data.

[0107] In this embodiment, when a laser scanner cannot be used, a road surface image can be captured using an image acquisition device, and a scale bar can be placed in the road surface image to extract road surface element attributes and dimensions (such as convexity, concave depth, obstacle spacing, etc.). Subsequently, these parameters are matched with a preset road spectrum library containing 15 standard road surface types, and the road spectrum with the highest matching degree is selected as the base template.

[0108] Example, Figure 3 This is a schematic diagram of an asphalt pavement image provided in an embodiment of this application.

[0109] Next, the road spectrum is iteratively optimized by combining the collected working condition data (such as instantaneous speed, wheelbase, wheel track, etc.) to generate a road spectrum suitable for four-column test benches.

[0110] For example, electronic devices acquire road surface information by capturing images of the road surface using image acquisition devices. The road surface images must contain scale-like objects with accurate dimensions to extract specific parameters of road surface elements, including key dimensional data such as the depth of bumps / pits and the distance between obstacles, providing a basis for subsequently determining the frequency of impact signals.

[0111] In addition, the electronic device can also match the existing road spectrum library. Based on the road element attributes and size extracted from the road image, the electronic device can select the road spectrum with the highest matching degree from the existing road spectrum library containing 15 standard road surface types (such as smooth asphalt road, rough asphalt road, grooved cement road, etc.), and then combine the vehicle operating condition information (such as instantaneous speed, suspension height, etc.) recorded during the data acquisition stage for iterative optimization to generate a road spectrum suitable for the four-post test bench.

[0112]

[0113]

[0114] Table 1 Table 1 shows the element attributes and dimensions of the road spectrum provided in this application embodiment. As shown in Table 1, different road surfaces correspond to different obstacle distribution densities, bump / dip depths, obstacle widths, and obstacle spacings. The electronic device can match the road surface name that best matches the road surface included in the road surface image from Table 1 based on key dimension data such as the depth of bumps / dips and obstacle spacing in the road surface image.

[0115] Based on the above embodiments, in the embodiments of this application, Figure 4 This application provides a flowchart illustrating the method for replacing and reproducing abnormal noises on a four-pillar bench based on road surface information collection, as shown in the embodiments of this application. Figure 2 Compared to the process shown, Figure 4 Modify procedure S1 to R1 and procedure S4 to R2. Specifically, as follows... Figure 4 As shown, the process includes: R1: Take photos of the road surface; obtain road surface information; S2: Operating Condition Record: Collects dynamic parameters; S3: Audio Recording: Collects 8 sound quality parameters; R2: Select appropriate road spectrum based on road element attributes and dimensions extracted from road surface photos; iterate and optimize by combining vehicle condition information recorded during the data collection phase; S5: Import the road spectrum into the four-column test stand; acquire audio signals; determine the comprehensive score S≥85%; if yes, proceed to S6; otherwise, proceed to S4; S6: Output qualified road spectrum to CAE model; compare actual and simulated modal responses, and determine priority areas for investigation.

[0116] The core technical solution of this application is to achieve accurate reproduction of abnormal noise road spectrum through multi-source data fusion, sound quality quantification verification, and CAE collaborative analysis. 3D laser scanning and multi-sensor synchronous acquisition of road surface point clouds, vehicle operating conditions, and eight sound quality parameters are utilized. A gridded road surface model is generated through point cloud preprocessing, coordinate transformation, and data expansion algorithms. Combined with vehicle kinematics calculations, a four-wheel road spectrum file is generated. A sound quality quantification standard based on the Euclidean distance algorithm is established, and the reproduction effect is verified through weighted scoring using parameters such as spectrum (30% weight) and order (20%). Qualified road spectra are imported into the CAE model for analysis in the 20-2000Hz frequency band, and priority investigation areas are located through modal response differences. When laser scanning is unavailable, an alternative solution can be achieved by analyzing road surface images and matching them with an existing road spectrum library. The overall technology controls the road spectrum error rate to within 5%, improving investigation efficiency by 50%.

[0117] Compared with the prior art, the embodiments of this application have the following advantages: 1. Significantly Improved Reproduction Accuracy: This application's embodiment utilizes 3D laser scanning (resolution ≤ 0.1 mm) and multi-source data fusion technology to generate abnormal noise road spectra with an error rate of ≤ 5% compared to actual operating conditions, a significant optimization compared to the over 40% matching error of traditional general road spectra. Experimental conditions were as follows: a handheld 3D laser scanner was used to collect road surface point clouds, combined with dynamic parameters such as vehicle instantaneous speed (error ≤ 0.5 km / h) and wheel center acceleration (sampling frequency ≥ 5 kHz). Road spectra were generated through point cloud preprocessing, coordinate transformation, and a meshing algorithm. The error rate was then compared with the original road test data in a four-column test bench verification.

[0118] 2. Significantly Improved Troubleshooting Efficiency: This application's embodiment introduces CAE simulation-assisted analysis. By using a mesh-refined model covering the 20-2000Hz frequency band (critical area size ≤5mm), and comparing the difference between actual and simulated modal responses (threshold ±2dB), the abnormal noise location time is shortened by 50%, solving the problem of a 45% misalignment rate between the first three troubleshooting areas and the actual problem point in traditional solutions. In practical applications, a certain vehicle model has reduced its abnormal noise troubleshooting cycle from the traditional 8 hours to 4 hours using this technology.

[0119] 3. Establish a standardized management system: In the embodiments of this application, a quantitative evaluation system based on 8 sound quality parameters is constructed. With the Euclidean distance algorithm of parameters such as spectrum (30% weight), order (20%), and sharpness (15%) as the core, it is determined that the reproduction is qualified through a comprehensive score S≥85%, forming a full-process standardized management from data acquisition to CAE verification, changing the disorderly state of traditional reliance on subjective evaluation. In bench verification, the compliance rate of sound quality parameter consistency is increased to over 90%.

[0120] 4. Enhance multi-scenario adaptability: In the embodiments of this application, for the scenario without laser scanning, through the iterative matching of road surface image analysis and the existing road spectrum library (including 15 standard road surface types), abnormal noise reproduction can still be achieved, expanding the technical application scenarios. Experiments show that the reproduction accuracy of this alternative solution can reach 1.8 times that of the traditional solution under normal road surface conditions.

[0121] Based on the method provided in the above embodiments, the embodiments of this application also provide an abnormal noise detection device based on multi-source data fusion. Figure 5 FIG. is a schematic structural diagram of an abnormal noise detection device based on multi-source data fusion provided by the embodiments of this application. The device includes: A data acquisition module 501, configured to acquire multi-source data, where the multi-source data includes road surface data, working condition data, and first audio data. The road surface data carries road surface information, the working condition data carries dynamic parameters of the vehicle, and the first audio data carries the first sound quality parameters of the vehicle; A processing module 502, configured to generate a road spectrum according to the road surface data and the working condition data, import the road spectrum into a four-column bench control system, place the vehicle on the bench, and acquire the current second audio data of the vehicle. The second audio data carries the second sound quality parameters of the vehicle; A detection module 503, configured to verify the road spectrum according to the first audio data and the second audio data. If the verification result is that the road spectrum is qualified, perform CAE simulation based on the road spectrum to obtain simulation abnormal noise data, and determine a high abnormal noise risk area according to the simulation abnormal noise data and the real abnormal noise data.

[0122] In a possible implementation manner, the road surface data is road surface point cloud data acquired by a 3D laser scanner; the working condition data includes vehicle instantaneous speed, wheelbase, and wheel track; The processing module 502 is specifically used to perform coordinate transformation on the road surface point cloud data based on a preset coordinate transformation algorithm; construct a two-dimensional grid based on the coordinate-transformed road surface point cloud data to obtain a road surface grid, and determine the height value of each grid in the road surface grid by interpolation to obtain a road surface height data matrix; determine the time series of the vehicle's travel on the road surface according to the instantaneous speed of the vehicle; determine the vehicle's position on the road surface at any moment in the time series according to the time series and the instantaneous speed of the vehicle; for each moment in the time series, determine the target grid corresponding to each wheel of the vehicle on the road surface grid according to the wheelbase, the track width and the vehicle's position on the road surface at that moment, and determine the road surface height value corresponding to each wheel of the vehicle according to the road surface height value corresponding to the target grid in the road surface height data matrix; generate the road spectrum based on the road surface height value corresponding to each wheel of the vehicle at each moment in the time series.

[0123] In one possible implementation, the processing module 502 is specifically used to multiply the coordinates of any point cloud in the road surface point cloud data with a preset coordinate transformation matrix to obtain the transformed coordinates of the point cloud.

[0124] In one possible implementation, the processing module 502 is specifically configured to: determine a first sub-rotation matrix for rotating around the z-axis by the first angle based on a preset first angle; determine a second sub-rotation matrix for rotating around the x-axis by the second angle based on a preset second angle; determine a third sub-rotation matrix for rotating around the y-axis by the third angle based on a preset third angle; determine a rotation matrix based on the first sub-rotation matrix, the second sub-rotation matrix, and the third sub-rotation matrix; and determine the coordinate transformation matrix based on the rotation matrix and a preset translation vector.

[0125] In one possible implementation, the processing module 502 is specifically configured to: if it is determined that the range of the road surface point cloud data does not meet a preset constraint condition, and the first length of the road surface point cloud data in the x-axis direction does not reach a first length threshold, then determine a first difference between the first length threshold and the first length; determine a first number of replications of the road surface point cloud data based on the first difference and the first length; control the point cloud data to be periodically replicated in the x-axis direction based on the first number of replications; and / or, if it is determined that the range of the road surface point cloud data does not meet a preset constraint condition, and the second length of the road surface point cloud data in the y-axis direction does not reach a second length threshold, then determine a second difference between the second length threshold and the second length; determine a second number of replications of the road surface point cloud data based on the second difference and the second length; control the point cloud data to be periodically replicated in the y-axis direction based on the second number of replications.

[0126] In one possible implementation, the processing module 502 is specifically configured to determine the ratio of the second difference to the second quantity; take half of the ratio as the second replication count; control the point cloud data to be periodically replicated in the positive y-axis direction based on the second replication count; and control the point cloud data to be periodically replicated in the negative y-axis direction based on the second replication count.

[0127] In one possible implementation, the processing module 502 is specifically configured to, for each wheel of the target vehicle, if the number of target grids corresponding to the wheel is one, determine the road height value corresponding to the target grid in the road height data matrix as the road height value corresponding to the wheel; if the number of target grids corresponding to the wheel is multiple, determine the average of the road height values ​​corresponding to each target grid in the road height data matrix as the road height value corresponding to the wheel.

[0128] In one possible implementation, the road surface data is a road surface image acquired by an image acquisition device; the operating condition data includes vehicle instantaneous speed, wheelbase, and track width. The processing module 502 is specifically used to extract features from the road surface image, determine the element attributes and size of the road surface contained in the road surface image; determine the road spectrum with the highest matching degree with the road surface from a pre-configured road spectrum library based on the element attributes and size; and update the road spectrum based on the working condition data.

[0129] Based on the same inventive concept, this application also provides an electronic device that can realize the functions of the aforementioned abnormal noise detection method based on multi-source data fusion. (Refer to...) Figure 6 The electronic device includes: The system includes at least one processor 601, a memory 603 connected to the at least one processor 601, and a communication interface 602. In this embodiment, the specific connection medium between the processor 601 and the memory 603 is not limited. Figure 6 The example shown is the connection between processor 601 and memory 603 via bus 604. Bus 604 is... Figure 6 The connections between other components are indicated by thick lines and are for illustrative purposes only, not as limiting information. The 604 bus can be divided into address bus, data bus, control bus, etc., for ease of representation. Figure 6 The term is represented by a single thick line, but this does not imply that there is only one bus or one type of bus. Alternatively, the processor 601 can also be called a controller; there is no restriction on the name.

[0130] In this embodiment, memory 603 stores instructions executable by at least one processor 601. By executing the instructions stored in memory 603, at least one processor 601 can perform the IoT device upgrade method described above. Processor 601 can implement... Figure 6 The system shown illustrates the functions of each module.

[0131] The processor 601 is the control center of the device. It can connect to various parts of the control device through various interfaces and lines. By running or executing instructions stored in memory 603 and calling data stored in memory 603, the processor can perform various functions and process data, thereby monitoring the device as a whole.

[0132] In one possible design, processor 601 may include one or more processing units. Processor 601 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, and the modem processor mainly handles wireless communication. It is understood that the modem processor may also not be integrated into processor 601. In some embodiments, processor 601 and memory 603 may be implemented on the same chip; in some embodiments, they may also be implemented on separate chips.

[0133] The processor 601 can be a general-purpose processor, such as a central processing unit (CPU), digital signal processor, application-specific integrated circuit, field-programmable gate array or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the IoT device upgrade method disclosed in the embodiments of this application can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0134] Memory 603, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory 603 may include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic storage, magnetic disk, optical disk, etc. Memory 603 can be any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto. In the embodiments of this application, memory 603 may also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0135] By designing and programming the processor 601, the code corresponding to the IoT device upgrade method described in the foregoing embodiments can be embedded into the chip, enabling the chip to execute the code during operation. Figure 1 The steps of the IoT device upgrade method shown in the embodiment are described. How to design and program the processor 601 is a technique well-known to those skilled in the art and will not be elaborated here.

[0136] Based on the same inventive concept, embodiments of this application also provide a storage medium storing computer instructions that, when executed on a computer, cause the computer to perform the aforementioned abnormal noise detection method based on multi-source data fusion.

[0137] In some possible implementations, various aspects of the noise detection method based on multi-source data fusion provided in this application can also be implemented in the form of a program product, which includes program code. When the program product is run on a device, the program code is used to cause the control device to perform the steps in the IoT device upgrade method according to the various exemplary embodiments of this application described above.

[0138] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting abnormal sound based on multi-source data fusion, characterized in that, The method comprises: acquiring multi-source data, the multi-source data comprising road surface data, working condition data and first audio data, the road surface data carrying road surface information, the working condition data carrying dynamic parameters of a vehicle, and the first audio data carrying first sound quality parameters of the vehicle; generating a road spectrum according to the road surface data and the working condition data, importing the road spectrum into a four-poster rig control system, placing the vehicle on a rig, collecting current second audio data of the vehicle, the second audio data carrying second sound quality parameters of the vehicle; verifying the road spectrum according to the first audio data and the second audio data, and if the verification result is that the road spectrum is qualified, performing computer aided engineering (CAE) simulation based on the road spectrum to obtain simulation abnormal sound data, and determining a high abnormal sound risk area according to the simulation abnormal sound data and real abnormal sound data.

2. The method of claim 1, wherein, The road surface data is road surface point cloud data acquired based on a 3D laser scanner; the working condition data comprises vehicle instantaneous speed, wheelbase and track; The generating of the road spectrum according to the road surface data and the working condition data comprises: performing coordinate conversion on the road surface point cloud data based on a preset coordinate conversion algorithm; constructing a two-dimensional grid based on the road surface point cloud data after coordinate conversion to obtain a road surface grid, and determining height values of each grid in the road surface grid by interpolation to obtain a road surface height data matrix; determining a time sequence of the vehicle driving on the road surface according to the vehicle instantaneous speed; determining a driving position of the vehicle on the road surface at any time in the time sequence according to the time sequence and the vehicle instantaneous speed; for each time in the time sequence, determining a corresponding target grid of each wheel of the vehicle on the road surface grid according to the wheelbase, the track and the driving position of the vehicle on the road surface at the time, and determining a corresponding road surface height value of each wheel of the vehicle according to a road surface height value corresponding to the target grid in the road surface height data matrix; the road spectrum is generated by the road surface height value corresponding to each wheel of the vehicle at each time in the time sequence.

3. The method of claim 2, wherein, The performing of coordinate conversion on the road surface point cloud data based on a preset coordinate conversion algorithm comprises: for any point cloud in the road surface point cloud data, multiplying the coordinates of the point cloud by a preset coordinate transformation matrix to obtain the transformed coordinates corresponding to the point cloud.

4. The method of claim 3, wherein, The determination process of the coordinate transformation matrix comprises: determining a first sub-rotation matrix rotating around the z-axis by a preset first angle; determining a second sub-rotation matrix rotating around the x-axis by a preset second angle; determining a third sub-rotation matrix rotating around the y-axis by a preset third angle; determining a rotation matrix according to the first sub-rotation matrix, the second sub-rotation matrix and the third sub-rotation matrix; determining the coordinate transformation matrix according to the rotation matrix and a preset translation vector.

5. The method according to any of claims 2-4, characterized by, If it is determined that the range of the road point cloud data does not satisfy the preset constraint condition, the method further comprises: If the first length of the road point cloud data in the x-axis direction does not reach a first length threshold, a first difference between the first length threshold and the first length is determined; a first replication number of the road point cloud data is determined according to the first difference and the first length; and the point cloud data is controlled to be periodically replicated in the x-axis direction according to the first replication number; and / or, If the second length of the road point cloud data in the y-axis direction does not reach a second length threshold, a second difference between the second length threshold and the second length is determined; a second replication number of the road point cloud data is determined according to the second difference and the second length; and the point cloud data is controlled to be periodically replicated in the y-axis direction according to the second replication number.

6. The method of claim 5, wherein, The determination of the second replication number of the road point cloud data according to the second difference and the second number comprises: determining a ratio of the second difference and the second number; taking half of the ratio as the second replication number. The control of the point cloud data to be periodically replicated in the y-axis direction according to the second replication number comprises: controlling the point cloud data to be periodically replicated in the positive direction of the y-axis according to the second replication number; and controlling the point cloud data to be periodically replicated in the negative direction of the y-axis according to the second replication number.

7. The method of claim 2, wherein, The determination of the road height value corresponding to each wheel of the vehicle according to the road height value corresponding to the target grid in the road height data matrix comprises: for each wheel of the target vehicle, if the number of target grids corresponding to the wheel is one, the road height value corresponding to the target grid in the road height data matrix is determined as the road height value corresponding to the wheel; and if the number of target grids corresponding to the wheel is multiple, the average of the road height values corresponding to each target grid in the road height data matrix is determined as the road height value corresponding to the wheel.

8. The method of claim 1, wherein, The road data is a road image acquired based on an image acquisition device; and the working condition data comprises vehicle instantaneous speed, wheelbase and track. The generation of the road spectrum according to the road data and the working condition data comprises: feature extraction is performed on the road image to determine the element attribute and size of the road surface contained in the road image; the road spectrum with the highest matching degree to the road surface is determined from a pre-configured road spectrum library according to the element attribute and size; and the road spectrum is updated according to the working condition data.

9. A device for detecting abnormal sound based on multi-source data fusion, characterized in that, The device comprises: a data acquisition module configured to acquire multi-source data, wherein the multi-source data comprises road data, working condition data and first audio data, the road data carries road surface information, the working condition data carries dynamic parameters of a vehicle, and the first audio data carries first sound quality parameters of the vehicle; A processing module is configured to generate a road spectrum based on the road surface data and the working condition data, import the road spectrum into a four-poster rig control system, place the vehicle on a rig, collect second audio data of the vehicle, and the second audio data carries second sound quality parameters of the vehicle. A detection module is configured to verify the road spectrum based on the first audio data and the second audio data, perform CAE simulation based on the road spectrum if the verification result is that the road spectrum is qualified, obtain simulation abnormal sound data, and determine a high abnormal sound risk area based on the simulation abnormal sound data and real abnormal sound data.

10. An electronic device, comprising: The method comprises the following steps: a memory is configured to store a computer program; a processor is configured to execute the computer program stored in the memory, and implement the method steps in any one of claims 1-8.

Images