Vehicle state monitoring method and device, driverless vehicle, equipment and storage medium

Abstract

This disclosure provides a vehicle condition monitoring method, device, autonomous vehicle, equipment, and storage medium, relating to the fields of autonomous driving, electric vehicles, vehicle condition monitoring, deep learning, and image recognition. The method includes: determining multidimensional force data of a target elastic component based on real-time operating data of the target vehicle, using a state association model constructed based on the physical attributes of the target vehicle and deployed on the vehicle-side computing unit; extracting multidimensional deformation data and apparent feature information from the non-contact sensing data of the target elastic component; determining the dynamic mechanical characteristic parameters of the target elastic component based on the multidimensional force data and the multidimensional deformation data; and determining the health status level of the target elastic component based on the dynamic mechanical characteristic parameters and the apparent feature information. This disclosure enables online condition assessment of vehicle chassis elastic components, improving the timeliness of fault identification.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to electric vehicles, vehicle condition monitoring, deep learning, image recognition and autonomous driving technologies. Background Technology

[0002] In vehicle chassis systems, rubber bushings, as key elastic connecting elements, are widely distributed at connection nodes such as control arms, subframes, stabilizer bars, and trailing arms, undertaking the functions of vibration reduction, buffering, and load transfer.

[0003] In existing technologies, monitoring the condition of vehicle bushings primarily relies on periodic maintenance inspections: the vehicle is lifted at fixed mileage or time intervals, and maintenance personnel visually inspect the bushings for abnormalities such as cracks, wear, and rubber delamination. Another common approach is to establish a fatigue life model for the bushings based on bench test and durability test data, and estimate the remaining life of the bushings based on vehicle mileage and usage time. Furthermore, some solutions indirectly infer the changing trends of the bushing condition by monitoring changes in vibration, noise, and other signals during vehicle operation. Summary of the Invention

[0004] This disclosure provides a vehicle condition monitoring method, apparatus, device, storage medium, and unmanned vehicle.

[0005] According to one aspect of this disclosure, a vehicle condition monitoring method is provided, applied to a vehicle-side computing unit. The method includes: determining multidimensional force data of a target elastic component based on real-time operating data of the target vehicle and a state association model deployed on the vehicle-side computing unit, wherein the state association model is a dynamic model constructed based on the physical properties of the target vehicle. Extract multidimensional deformation data and appearance feature information from the non-contact sensing data of the target elastic component; Based on the multidimensional force data and the multidimensional deformation data, the dynamic mechanical property parameters of the target elastic component are determined. Based on the dynamic mechanical property parameters and the apparent characteristic information, the health status level of the target elastic component is determined.

[0006] According to another aspect of this disclosure, a vehicle condition monitoring device is provided, applied to a vehicle-side computing unit, the device comprising: The force data determination module is used to determine the multidimensional force data of the target elastic component based on the real-time operating data of the target vehicle and through the state association model deployed on the vehicle-side computing unit. The state association model is a dynamic model constructed based on the physical properties of the target vehicle. The feature extraction module is used to extract multidimensional deformation data and appearance feature information from the non-contact sensing data of the target elastic component; The mechanical parameter determination module is used to determine the dynamic mechanical property parameters of the target elastic component based on the multidimensional force data and the multidimensional deformation data. The status determination module is used to determine the health status level of the target elastic component based on the dynamic mechanical characteristic parameters and the apparent feature information.

[0007] According to another aspect of this disclosure, an autonomous vehicle is provided, comprising: The vehicle chassis, the suspension system of the vehicle chassis is equipped with rubber bushings containing a metal skeleton; An image acquisition device is installed near the rubber bushing to acquire local image sequence data of the rubber bushing as non-contact sensing data. The vehicle-side computing unit is communicatively connected to the image acquisition device and is used to execute the vehicle status monitoring method provided in any embodiment of this disclosure.

[0008] According to another aspect of this disclosure, an electronic device is provided, comprising: At least one processor; and The memory is communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform any of the methods described in the present disclosure.

[0009] According to another aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to perform any of the methods according to embodiments of this disclosure.

[0010] According to another aspect of this disclosure, a computer program product is provided, including a computer program that, when executed by a processor, implements any of the methods according to embodiments of this disclosure.

[0011] According to the solution of this disclosure embodiment, by fusing and analyzing the dynamic model with non-contact sensing data, online status assessment of the elastic components of the vehicle chassis is realized, improving the timeliness of fault identification.

[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0013] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1This is a schematic flowchart of a vehicle condition monitoring method provided according to an embodiment of this disclosure; Figure 2 This is a block diagram of the overall architecture of the bushing failure monitoring system provided in the embodiments of this disclosure; Figure 3 This is a schematic diagram of the three-zone bushing condition determination logic provided in the embodiments of this disclosure; Figure 4 This is a timing diagram of data interaction between the vehicle and the cloud according to an embodiment of this disclosure; Figure 5 This is an example diagram comparing bushing failure and normal state according to an embodiment of this disclosure; Figure 6 This is a schematic diagram of the installation position of the bushing monitoring miniature camera according to an embodiment of this disclosure; Figure 7 This is a schematic diagram of the vehicle condition monitoring device provided according to an embodiment of the present disclosure; Figure 8 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present disclosure. Detailed Implementation

[0014] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0015] Current maintenance solutions for vehicle chassis systems often employ periodic lifting and visual inspection, along with predictive replacement strategies based on statistical life models. While these methods can maintain basic vehicle availability in traditional manned driving scenarios, for fully autonomous vehicles, the lack of driver feedback means that chassis bushing performance degradation cannot be detected in a timely manner. Periodic lifting inspections are discrete checks, failing to detect faults in their early stages, and frequent lifting increases non-operational time and maintenance costs. Life model-based predictions are influenced by various factors such as manufacturing process differences, ambient temperature, and load history; the actual lifespan of bushings from the same batch can vary significantly across different vehicles, making individualized and accurate predictions impossible with statistical models alone. Furthermore, indirect assessments based on vehicle dynamics response lack sensitivity and struggle to identify specific chassis components malfunctioning, resulting in low accuracy and a high risk of misdiagnosis and over-repair. Traditional solutions can easily lead to undetected bushing performance degradation or even complete failure, impacting vehicle ride comfort and even driving safety.

[0016] To at least partially address one or more of the aforementioned problems and other potential issues, this disclosure proposes a vehicle condition monitoring scheme. By introducing non-contact sensing and data acquisition methods and deeply integrating image recognition technology with the vehicle dynamics model, a multi-dimensional condition assessment and hierarchical early warning system for chassis elastic components is achieved within a vehicle-side edge computing architecture. This method significantly improves the real-time performance of fault identification and the accuracy of assessment, and enhances operational decision-making capabilities through a hierarchical early warning mechanism, enabling on-demand maintenance based on actual conditions.

[0017] This disclosure provides a vehicle condition monitoring method, which can be applied to a vehicle condition monitoring device. The vehicle condition monitoring device is located in an electronic device. This electronic device includes, but is not limited to, an in-vehicle computing device. For example, the in-vehicle computing device includes, but is not limited to, an in-vehicle domain controller, an in-vehicle edge computing unit, an intelligent driving domain controller, or a separate chassis cooperative control unit. In some possible implementations, the method can also be implemented by a processor calling computer-readable instructions stored in memory.

[0018] The embodiments of this disclosure will be described in detail below with reference to the accompanying drawings and examples.

[0019] Figure 1 This is a flowchart illustrating a vehicle condition monitoring method provided according to embodiments of this disclosure. Figure 1 As shown, the method includes at least the following steps: S110. Based on the real-time operating data of the target vehicle, determine the multi-dimensional force data of the target elastic component through the state association model deployed on the vehicle-side computing unit.

[0020] In this embodiment, the target vehicle refers to a motor vehicle requiring chassis health monitoring, such as an unmanned operating vehicle. Real-time operating data can be understood as dynamic operating status indicators generated by the vehicle during operation, including but not limited to vehicle speed, steering angle, braking torque, driving torque, and suspension displacement signals. The state association model is a computational model used to simulate the physical force laws of a vehicle. Specifically, it can be a multibody dynamics model constructed based on the target vehicle's overall mass, geometric dimensions, hard point coordinates, and elastic element characteristics. The target elastic component typically refers to a flexible mechanical component in the chassis suspension system that plays a connecting and buffering role, such as a rubber bushing containing a metal skeleton. Multidimensional force data refers to the combination of mechanical values ​​that the component experiences in different directions in three-dimensional space, such as radial force and axial force components in different directions.

[0021] For example, when a vehicle drives over a speed bump at high speed, the state association model can calculate in real time the transient force values ​​borne by a certain subframe bushing in the X and Y axes of the chassis based on the current deceleration and bump conditions.

[0022] As an alternative implementation, the state association model can also be a data-driven proxy model, which is trained by a large amount of pre-collected real vehicle calibration data. During the inference phase, it takes real-time working condition data as input and directly outputs the force estimation values ​​of each bushing node, thereby reducing online computing overhead.

[0023] By deploying the vehicle's physical model on the vehicle end for real-time stress calculation, the problem that the stress state of the chassis's elastic components under actual driving conditions cannot be directly measured is solved, thus providing a reliable multi-dimensional mechanical benchmark for subsequent stiffness assessment.

[0024] S120. Extract multidimensional deformation data and appearance feature information from the non-contact sensing data of the target elastic component.

[0025] In this embodiment of the disclosure, non-contact sensing data can be understood as observation data acquired by a data acquisition device that does not have direct physical contact with the target being measured. Specifically, it can be a sequence of local images of the chassis acquired by a vehicle-mounted miniature camera. Multidimensional deformation data represents the three-dimensional spatial morphological changes of an elastic component under stress. Appearance feature information refers to the physical texture or structural integrity state of the target surface, including but not limited to rubber crack features, skeleton integrity features, and adhesion features between the rubber and the skeleton.

[0026] like Figure 2 As shown in the AI ​​recognition layer, in specific implementation, the vehicle-side computing unit uses a large model image recognition module to parse multi-dimensional deformation data and surface feature information such as cracks and delamination from the data after image preprocessing.

[0027] As a specific implementation example, the continuous image frames captured by the industrial camera are non-contact sensing data. The offset of the bushing center point in the image obtained from the data belongs to multi-dimensional deformation data, while the identified surface rubber aging marks belong to appearance feature information.

[0028] As an alternative, non-contact sensing data can also be acquired using laser displacement sensors or eddy current sensors. Multidimensional deformation data is obtained by measuring the spatial position changes of various feature points on the bushing surface through a multi-point arrangement. This alternative may offer higher accuracy, but the complexity of sensor arrangement and cost increase accordingly.

[0029] By simultaneously extracting deformation and appearance features through non-contact image sensing, the problem of traditional contact sensors being easily damaged and difficult to deploy in harsh chassis environments is solved, thus realizing non-invasive parallel acquisition of multi-dimensional state information of elastic components.

[0030] S130. Based on multidimensional force data and multidimensional deformation data, determine the dynamic mechanical property parameters of the target elastic component.

[0031] In this embodiment of the disclosure, dynamic mechanical characteristic parameters are used to quantitatively characterize the ability of an elastic component to resist deformation under motion, i.e., stiffness characteristics. Specifically, the theoretical force values ​​obtained at the same time point are mathematically matched with the actual observed deformation values ​​to obtain the current actual stiffness of the component.

[0032] Specifically, by combining the 2000 Newton radial force obtained from the model with the 2 mm radial displacement extracted visually, the dynamic mechanical characteristic parameter of the component at this time is calculated to be 1000 N / mm.

[0033] As an alternative approach, dynamic mechanical property parameters can also be obtained by fitting the slope of the force-displacement relationship curve, rather than by simply calculating the point-by-point quotient. For example, by collecting multiple sets of force-deformation data pairs within a continuous time window and using the least squares fitting method to calculate the slope of the force-displacement curve, a more accurate equivalent stiffness characterization can be obtained in the nonlinear deformation stage.

[0034] By fusing the force data calculated by the model with the deformation data extracted by vision across sources, the problem that a single sensing mode cannot independently complete stiffness assessment is solved, thereby realizing the online quantification of the dynamic mechanical properties of elastic components.

[0035] S140. Based on dynamic mechanical property parameters and apparent characteristic information, determine the health status level of the target elastic component.

[0036] In this embodiment, the health status level is a classification index used to assess the remaining lifespan and failure risk of a component. Specifically, by setting a comprehensive evaluation logic, dynamic mechanical characteristic parameters representing intrinsic performance degradation are cross-compared with apparent characteristic information representing external damage, and the final health rating result is output.

[0037] Specifically, if the stiffness parameter drops abnormally and the appearance information shows obvious tearing on the rubber surface, the system will identify the health status of the component as high-risk.

[0038] As an alternative implementation, health status levels can also use a continuous degradation index instead of discrete levels. Specifically, mechanical property parameters and apparent feature information are input into a weighted scoring model, and the output is a continuous degradation index value between 0 and 1. The operation and maintenance system can set flexible early warning thresholds based on this value to achieve a more flexible hierarchical management strategy.

[0039] According to the solution of this disclosure embodiment, by combining mechanical degradation indicators with visual damage characteristics for comprehensive judgment, the problem of easy omission or misjudgment in single-dimensional monitoring is solved, thereby improving the comprehensiveness and accuracy of health status assessment of elastic components.

[0040] In one possible implementation, the multidimensional deformation data includes translational displacement data in the first dimension and rotational angle data in the second dimension, whereby the rotational angle data includes yaw angle data and torsional angle data. S120 extracts the multidimensional deformation data from the non-contact sensing data of the target elastic component, including the following steps: S121. Identify the outer contour position information of the target elastic component and the relative deflection information of the inner and outer components in the non-contact sensing data.

[0041] In this embodiment, the outer contour position information is used to mark the coordinate distribution of the entire component in the image, and can be a set of pixel-level boundary coordinates of the outer metal frame of the bushing. The relative deflection information of the inner and outer components is used to characterize the angular torsion between the internal spindle and the outer shell of the component.

[0042] In one example, the visual algorithm identifies the pixel coordinates of the outer metal ring of the bushing as the outer contour position information, and identifies the skew state of the internal bolt holes relative to the outer ring as the relative deflection information.

[0043] S122. Determine translational displacement data based on the outer contour position information.

[0044] In this embodiment of the disclosure, the translational displacement data reflects the absolute movement of the target along the linear coordinate axis. Specifically, the pixel difference between the coordinates of the outer contour center and the reference zero coordinate in the current image is compared and multiplied by a pre-calibrated conversion factor between pixels and physical distance to obtain the physical translational distance.

[0045] Taking a specific application scenario as an example, the center of the outer contour moved vertically by 10 pixels, and after conversion, the translational displacement data was determined to be a downward translation of 3 millimeters.

[0046] S123. Determine the rotation angle data based on the relative deflection information.

[0047] In this embodiment of the disclosure, the rotation angle data is decoupled into rotation amounts around different spatial axes. Specifically, a perspective transformation matrix is ​​used to restore the relative deflection information of the plane to three-dimensional spatial features, separating the yaw angle around the radial axis and the torsional angle around the central axis.

[0048] For example, based on the skewness of the internal spindle, the algorithm can parse the rotation angle data at this time, which is skewed by 2 degrees and twisted by 4 degrees.

[0049] According to the scheme of this disclosure, by decoupling the deformation features into two dimensions, translation and rotation, and extracting them separately, the problem that single displacement extraction cannot fully reflect the complex stress deformation of elastic components is solved, thereby achieving high-precision extraction of multi-degree-of-freedom deformation parameters of elastic components and providing a complete data foundation for subsequent multi-dimensional stiffness calculation.

[0050] In one possible implementation, the multidimensional force data includes linear force values ​​and torsional moment values, and the dynamic mechanical characteristic parameters include linear stiffness parameters and angular stiffness parameters, with the angular stiffness parameters including yaw angle stiffness and torsional angle stiffness. S130, based on the multidimensional force data and multidimensional deformation data, determines the dynamic mechanical characteristic parameters of the target elastic component, including the following steps: S131. Perform time-series alignment processing on the multidimensional force data and multidimensional deformation data according to the collection timestamp.

[0051] In this embodiment, the acquisition timestamp represents the precise time node at which the data was generated. Since the vehicle bus system (CAN bus) and the camera system are heterogeneous data sources, their sampling frequencies and transmission delays differ, necessitating time synchronization of the two data streams. Specifically, a time window matching algorithm is used to pair and bind the model force outputs that occur at the closest times with image deformation frames, forming a homogeneous data group.

[0052] In one example, a linear force value of 1500 Newtons generated at a timestamp of 10 minutes 05 seconds 050 milliseconds is aligned with image displacement data captured at the same timestamp.

[0053] S132. Determine the linear stiffness parameter based on the first quotient of the linear force value and the translational displacement data after time alignment.

[0054] In this embodiment, the first quotient is the result of dividing the force in the linear direction by the displacement, characterizing the elastic component's ability to resist deformation in the linear direction. Specifically, the linear stiffness parameter is obtained by processing the aligned two sets of data through a division operation.

[0055] As a specific implementation example, the linear force is 3000 Newtons and the translational displacement is 6 millimeters. The first quotient of the two is 500, thus determining the linear stiffness parameter to be 500 N / mm.

[0056] S133. The torsional moment value after time alignment is divided by the yaw angle data and the torsion angle data to determine the corresponding multiple angular stiffness quotients, and the yaw angle stiffness and torsion angle stiffness are determined according to the multiple angular stiffness quotients.

[0057] In this embodiment of the disclosure, the angular stiffness quotient reflects the proportional relationship between torque and angle change. Specifically, the yaw torque is divided by the yaw angle data to obtain the yaw angle stiffness, and the torsional torque is divided by the torsional angle data to obtain the torsional angle stiffness.

[0058] In one example, the torsional moment is 60 N·m and the torsional angle is 0.12 radians. Dividing these values ​​gives a quotient of 500 N·m / radian, which is then determined as the torsional stiffness.

[0059] According to the solution of this disclosure embodiment, by adopting timestamp alignment and multi-degree-of-freedom quotient calculation, the problem of calculation distortion caused by asynchronous sampling of heterogeneous data sources is solved, thereby realizing accurate calculation of multi-degree-of-freedom stiffness of elastic components in the vehicle edge computing environment.

[0060] In one possible implementation, after step S132, which determines the linear stiffness parameter based on the first quotient of the time-aligned linear force value and the translational displacement data, the method further includes the following steps: S134. Obtain the ambient temperature data of the target elastic component.

[0061] In this embodiment, the ambient temperature data refers to the thermal value of the component's operating space. Since rubber materials are sensitive to temperature, ambient temperature may affect the stiffness assessment results during cold starts or when the temperature has not reached a steady state. Therefore, temperature factors can be considered in the initial stage of stiffness assessment to improve accuracy. Specifically, the current temperature value is read using an ambient temperature sensor integrated into the vehicle chassis.

[0062] Taking a specific application scenario as an example, when driving in the early morning of winter, the ambient temperature data obtained is minus 15 degrees Celsius.

[0063] S135. Find the temperature compensation factor corresponding to the ambient temperature data in the mapping table. The mapping table is a pre-generated correspondence table between temperature values ​​and compensation factor values.

[0064] In this embodiment, the temperature compensation factor is a correction coefficient used to offset the influence of material thermal expansion and contraction and modulus changes on stiffness measurement. The mapping table can be generated through pre-calibrated bench tests. The standard stiffness of the bushing is measured under different temperature conditions, and the compensation factor corresponding to each temperature node is calculated and stored in the vehicle's memory. In specific implementation, the current temperature value is used as the search key to look up the corresponding compensation factor in the mapping table.

[0065] In one example, a lookup of the mapping table reveals that at -15 degrees Celsius, the temperature compensation factor is 0.82 because the elastomer hardens due to the cold.

[0066] S136. Based on the product of the first quotient and the temperature compensation factor, determine the corrected linear stiffness parameters.

[0067] In this embodiment of the disclosure, the product operation is intended to adaptively correct the foundation stiffness under temperature conditions. Specifically, the first quotient calculated in S132 is multiplied by the queried temperature compensation factor to output the equivalent mechanical parameters at the reference temperature.

[0068] For example, the rough linear stiffness calculated using the first quotient is 1000 N / mm. Multiplying this by 0.82 yields a corrected linear stiffness parameter of 820 N / mm. This value excludes the contribution of low-temperature hardening and more accurately reflects the degree of degradation of the bushing rubber material itself.

[0069] According to the scheme of the present disclosure, by introducing an ambient temperature compensation mechanism, the interference of temperature sensitivity on stiffness assessment is reduced under cold start or temperature unsteady conditions, thereby improving the reliability of state assessment under different ambient temperature conditions.

[0070] In one possible implementation, the apparent feature information includes structural boundary features. Before S140 determines the health status level of the target elastic component based on dynamic mechanical property parameters and apparent feature information, the method further includes the following steps: If the structural boundary features do not meet the preset safety boundary conditions, the health status level of the target elastic component is determined as the failure level.

[0071] In this embodiment of the disclosure, structural boundary features describe the absolute integrity limits of the physical shape of the component. A failure level signifies that the component has lost its basic cushioning function and must be immediately shut down. Failure to meet preset safety boundary conditions includes any of the following: The structural boundary feature indicates the presence of a skeleton crack in the target elastic component; the structural boundary feature indicates that the depth of the elastomer crack in the target elastic component is greater than or equal to a preset crack depth threshold; the structural boundary feature indicates a through-grafting separation between the elastomer and the skeleton of the target elastic component; or the structural boundary feature indicates direct contact between the inner and outer skeletons of the target elastic component.

[0072] like Figure 5 The example image showing the comparison between bushing failure and normal conditions visually illustrates the differences in appearance between the normal state and various physical damage limits. In the specific implementation, parallel Boolean trigger logic is established: once the image recognition model detects... Figure 5 Any fatal damage shown, such as severe cracking, limiting failure (metal contact between the inner and outer skeletons), or through-hole debonding, will trigger a veto mechanism, directly outputting the failure result and discontinuing the subsequent stiffness scoring process.

[0073] In one example, when a through-type debonding of half a circumference between the rubber elastomer and the outer metal skeleton is detected, the system no longer performs complex stiffness scoring, but directly determines its health status level as failure level and immediately reports the information to the cloud control center.

[0074] According to the solution of this disclosure embodiment, by setting multiple severe structural damages as unilaterally triggered veto boundary conditions, the problem of delayed response or comprehensive score masking risks that may occur when the algorithm deals with extreme fatal damage is solved, thereby realizing high-priority mandatory protection of the absolute safety bottom line of the vehicle chassis.

[0075] In one possible implementation, the apparent feature information includes a damage state identifier, which characterizes whether damage features exist on the surface of the target elastic component. S120 extracts the apparent feature information from the non-contact sensing data of the target elastic component, including the following steps: S124. Input the non-contact sensing data into the pre-trained feature extraction network model to extract the texture features of at least one pixel channel.

[0076] In this embodiment, the feature extraction network model can be a deep vision model built on a Convolutional Neural Network (CNN) architecture, trained on a proprietary dataset containing normal bushing images and various damaged bushing images. Texture features reflect the spatial arrangement of image pixels in grayscale, and can reflect the deep semantic information of subtle damage such as surface peeling, microcracks, and crazing. Specifically, the RGB matrix acquired by the camera is input into the network layer, and high-dimensional deep semantic feature maps of the corresponding rubber regions are extracted using convolutional kernels.

[0077] As a concrete implementation example, given an image containing mud and bushings on the chassis, the network extracts texture features with sharp variations in the edges of the corresponding rubber areas.

[0078] S125. Perform structured constraint matching on texture features according to the anti-illusion determination instruction to determine the damage state identifier. The anti-illusion determination instruction includes the constraint condition of extracting continuous edge substrings and rejecting the generation of unobserved textures.

[0079] In this embodiment, the anti-illusion determination instruction is a set of hard rule constraint logic used to limit the divergent reasoning of large artificial intelligence models. Structured constraint matching forcibly regulates the network output within a physically reasonable domain. In harsh chassis environments, camera images are easily affected by artifacts such as mud splashes and oil stains, and the visual model may misidentify mud spots as cracks. The anti-illusion determination instruction uses code logic to limit the extracted crack texture to form a continuous substring of pixels in space; if it is merely discrete noise, it is rejected as a damage feature.

[0080] Specifically, although the large model extracted textures resembling cracks, the anti-illusion judgment command found that the pixel sequence was discontinuous (possibly splattered mud spots) after verification, thus rejecting the generation of false damage and ultimately determining that the damage state was marked as having no damage features.

[0081] According to the scheme of this disclosure embodiment, by combining the feature extraction network with the anti-illusion judgment instruction, the extracted texture is subjected to structured constraint verification, which solves the problem that the visual model is easily affected by mud and water artifacts in harsh chassis environments and produces false alarms, thereby realizing intelligent extraction of surface micro-damage features with high fidelity and low false judgment.

[0082] In one possible implementation, S140 determines the health status level of the target elastic component based on dynamic mechanical characteristic parameters and apparent feature information, including the following steps: like Figure 3 The diagram shows the logic flow for the three-zone bushing condition determination. The system employs a two-stage evaluation process: first a rejection, then a matrix-based determination. After passing the rejection condition, the following two-dimensional matrix cross-determination is performed: S142. If the decrease in online stiffness parameter or angular stiffness parameter is greater than or equal to the first threshold, and the damage status indicator indicates the presence of the first damage characteristic, the health status level is determined to be the first risk level.

[0083] In this embodiment of the disclosure, the first threshold characterizes the baseline of severe performance degradation, the first damage characteristic may include severe damage such as cracks that are expanding or signs of delamination, and the first risk level corresponds to... Figure 3 The red zone (failure) state is defined in the code. In practice, a logical AND operation is used to simultaneously verify the depth of mechanical degradation and the severity of apparent defects; a double hit results in a high-risk warning.

[0084] Taking a specific application scenario as an example, the linear stiffness parameter dropped significantly by 25%, and the damage status indicator showed that the crack was expanding. The health status level was determined to be the first risk level in the red zone, and immediate replacement was required.

[0085] S143. If the decrease in online stiffness parameter or angular stiffness parameter is less than the first threshold and greater than or equal to the second threshold, and the damage status indicator indicates the presence of a second damage characteristic, the health status level is determined to be the second risk level.

[0086] In this embodiment of the disclosure, the second threshold characterizes the boundary of slight performance degradation, the second damage characteristic may include minor defects such as early surface cracking, and the second risk level corresponds to... Figure 3The yellow zone (medium risk, observable) status is indicated. In practice, an intermediate value range is set, and the judgment is made in conjunction with minor appearance defects, prompting maintenance personnel to keep a close eye on the component.

[0087] In one example, the stiffness decreased by 10% and the appearance showed early surface cracking, which was identified as the second risk level in the yellow zone, and it was recommended to schedule a key inspection in the next maintenance cycle.

[0088] S144. If the online stiffness parameters and angular stiffness parameters are in an increasing state, and the damage status indicator shows no damage characteristics, the health status level is determined to be a safe level.

[0089] In this embodiment, the rising state reflects the post-vulcanization aging and hardening phenomenon of polymer rubber materials in the early stages of long-term use, which is a normal physical property evolution and corresponds to the safety level. Figure 3 The green zone (no risk) state in the code. In specific implementation, when the logical judgment stiffness increases and visual confirmation shows no tearing, the component is allowed to continue normal use.

[0090] For example, if the linear stiffness and angular stiffness parameters both increase by 10% compared to the initial standard, but the surface is intact and shows no signs of damage, the system determines this to be a normal aging process, classifies it as a green zone safety level, and allows it to continue operating.

[0091] According to the solution of this disclosure embodiment, by cross-judging the direction and magnitude of multi-dimensional stiffness changes with different apparent damages using a two-dimensional matrix, the problem of over-repair caused by the easy obscuring of the objective law of rubber material degradation by single threshold monitoring is solved. This achieves three-zone refined health management that conforms to physical characteristics, and maximizes the service life of components while ensuring safety.

[0092] In one possible implementation, the apparent feature information includes a component interference risk identifier, which characterizes whether there is a risk of motion interference between the target elastic component and surrounding components. The method further includes the following steps: S151. Input non-contact sensing data into a pre-trained feature extraction network model to extract component spacing features.

[0093] In this embodiment, the component spacing feature reflects the minimum physical distance between the elastic component and the surrounding metal support or cantilever in spatial projection. Specifically, the network model outputs the bounding box coordinates of different components, and the Euclidean distance algorithm is used to calculate the shortest gap between the edges.

[0094] In one example, the model simultaneously detects the control arm and the subframe edge, calculating their current component spacing feature as 4 pixel units.

[0095] S152. When the component spacing characteristic is less than or equal to the collision interference threshold, determine the component interference risk indicator.

[0096] In this embodiment, the collision interference threshold is a safety clearance lower limit set by the system to avoid physical impacts and abnormal noises. Specifically, once the distance characteristics calculated in real time trigger the lower limit alarm, a corresponding Boolean risk flag is immediately generated.

[0097] As a specific implementation example, the preset collision interference threshold is 5 pixels. Since the current spacing is compressed to 4 pixels, the system immediately determines to generate a component interference risk label.

[0098] S153. When generating component interference risk indicators, determine the health status level as a warning risk level.

[0099] In this embodiment of the disclosure, since component interference can easily cause sharp abnormal noises from the chassis or hard connection impacts, this warning has a high weight. In specific implementation, through an event callback mechanism, a warning notification will be triggered as long as an interference indicator exists, even if the stiffness index has not yet shown significant degradation.

[0100] In one example, after receiving an interference risk flag, the system determines that the component is at a warning risk level where abnormal noise may occur, and recommends that maintenance personnel arrange for a return visit for inspection.

[0101] According to the solution of this disclosure embodiment, by extracting the component spacing features based on visual images and comparing them with the collision interference threshold, the technical problem of being unable to use conventional acoustic sensors to diagnose abnormal noises due to the harsh chassis acoustic environment is solved, thereby innovatively realizing motion interference and abnormal noise risk prediction purely based on visual images.

[0102] In one possible implementation, the target vehicle includes an autonomous vehicle. The target elastic component includes a rubber bushing with a metal frame in the autonomous vehicle's chassis suspension system. Non-contact sensing data includes sequence data of local images of the chassis acquired by an onboard miniature camera.

[0103] Combination Figure 6 The diagram shows the installation location of the bushing monitoring miniature camera. The vehicle-mounted miniature camera is rigidly fixed to the subframe or near the control arm using a dedicated mounting bracket, and its field of view covers the bushing body and surrounding connecting parts. In practice, the miniature camera continuously acquires local image sequences under alternating chassis loads, providing continuous non-contact sensing data input to the vehicle-side computing unit. A protective cover and supplementary lighting can also be installed outside the image acquisition device: the protective cover protects against impacts from flying stones and mud, while the supplementary lighting provides sufficient illumination in the shadowed areas of the chassis, ensuring that the image acquisition quality is not affected by changes in ambient lighting conditions.

[0104] According to the solution of the present disclosure, by clearly defining the specific physical scenario of the unmanned vehicle chassis, metal-rubber composite bushing and the dedicated arrangement of the on-board camera, the problem of chassis failures not being detected in a timely manner due to the lack of driver feedback in high-level autonomous driving is solved.

[0105] Figure 2 This is a block diagram of the overall architecture of a bushing failure monitoring system provided according to another embodiment of this disclosure. Figure 2 As shown, based on the above embodiments, combined with Figure 4 The timing diagram of data interaction between the vehicle and the cloud is provided. The complete technical solution of this disclosure involves the collaborative work of the vehicle and cloud levels.

[0106] At the vehicle level, miniature cameras and the vehicle controller simultaneously acquire images and operating condition data. The vehicle-side edge computing unit performs image preprocessing and lightweight model inference, and combines this with the vehicle dynamics model to calculate stiffness and determine the three-zone status. Subsequently, the vehicle-to-everything (V2X) transmission module uploads the identification results and status determination information to the cloud control center in real time. At the cloud level, the control center platform receives and aggregates bushing status data from all operating vehicles, and uses a multi-vehicle data aggregation and analysis engine to perform cluster analysis on the bushing degradation of the entire fleet. Once an anomaly is detected in the red or yellow zone, an alarm system is immediately triggered to notify maintenance personnel. The cloud platform automatically generates maintenance work order suggestions and vehicle dispatch instructions based on the bushing status data, completing a closed loop of maintenance operations from single-vehicle monitoring to predictive maintenance for the entire fleet.

[0107] According to the solution of this disclosure embodiment, by adopting a vehicle-cloud integrated architecture that combines vehicle-side intelligent perception and reasoning with cloud-based big data aggregation and monitoring, the problem of large-scale unmanned vehicle fleets being unable to conduct cluster health checks and proactive spare parts management is solved. This achieves a leap from single-vehicle physical monitoring to predictive maintenance and refined scheduling of the entire fleet, significantly improving operation and maintenance scheduling efficiency.

[0108] The following combination Figures 2 to 4 This document summarizes and describes the complete implementation process of the vehicle condition monitoring method according to embodiments of this disclosure. The method includes at least the following steps: S210. Based on the real-time operating data of the target vehicle, the multi-dimensional force data of the target elastic component is determined by the state association model deployed on the vehicle-side computing unit. The state association model is a dynamic model constructed based on the physical properties of the target vehicle.

[0109] S220. Extract multidimensional deformation data and appearance feature information from the non-contact sensing data of the target elastic component.

[0110] Specifically, S220 includes: identifying the outer contour position information of the target elastic component and the relative deflection information of the inner and outer components in the non-contact sensing data; determining translational displacement data based on the outer contour position information; determining yaw angle data and torsional angle data based on the relative deflection information; inputting the non-contact sensing data into a pre-trained feature extraction network model to extract texture features of at least one pixel channel, and performing structured constraint matching on the texture features according to the anti-hallucination judgment instructions to determine the damage state identifier.

[0111] S230. Based on multidimensional force data and multidimensional deformation data, determine the dynamic mechanical property parameters of the target elastic component.

[0112] Specifically, S230 includes: performing time-series alignment processing on the multidimensional force data and multidimensional deformation data according to the acquisition timestamp; determining the linear stiffness parameter based on the first quotient of the time-series aligned linear force value and translational displacement data; determining the yaw angle stiffness and torsional angle stiffness by dividing the time-series aligned torsional torque value by the yaw angle data and torsional angle data, respectively; acquiring the ambient temperature data of the target elastic component, looking up the corresponding temperature compensation factor in the mapping table, and determining the corrected linear stiffness parameter based on the product of the first quotient and the temperature compensation factor.

[0113] S240. Based on dynamic mechanical property parameters and apparent characteristic information, determine the health status level of the target elastic component.

[0114] Specifically, S240 includes: First, a rejection condition check is performed—if the structural boundary characteristics do not meet the preset safety boundary conditions (including any of the following: skeleton cracks, elastomer crack depth exceeding a threshold, through-hole debonding, or direct contact between inner and outer liner tubes), the health status level is directly determined as a failure level. After all rejection conditions are passed, a two-dimensional matrix cross-judgment is performed: if the decrease in online stiffness parameter or angular stiffness parameter is greater than or equal to a first threshold and the damage status indicator indicates the presence of a first damage characteristic, it is determined as a first risk level. If the decrease in online stiffness parameter or angular stiffness parameter is less than the first threshold but greater than or equal to a second threshold and the damage status indicator indicates the presence of a second damage characteristic, it is determined as a second risk level. If the online stiffness parameter and angular stiffness parameter are in an increasing state and the damage status indicator indicates the absence of damage characteristics, it is determined as a safe level. In addition, non-contact sensor data is input into the feature extraction network model to extract component spacing features. When the component spacing features are less than or equal to a collision interference threshold, a component interference risk indicator is determined, and the health status level is determined as a warning risk level.

[0115] Through the complete process from S210 to S240, the vehicle-side computing unit realizes the full-link online processing of the chassis elastic components, from stress calculation, deformation extraction, stiffness assessment to state determination, and transmits the determination results to the cloud via the vehicle network to support the operation control center in making decisions such as maintenance scheduling, operation planning and life data analysis.

[0116] Figure 7 This is a structural schematic diagram of a vehicle condition monitoring device 700 provided according to an embodiment of this disclosure. (See attached diagram.) Figure 7 As shown, the device includes: The force data determination module 701 is used to determine the multidimensional force data of the target elastic component based on the real-time operating condition data of the target vehicle and through the state association model deployed on the vehicle-side computing unit. The state association model is a dynamic model constructed based on the physical properties of the target vehicle.

[0117] The feature extraction module 702 is used to extract multidimensional deformation data and appearance feature information from the non-contact sensing data of the target elastic component.

[0118] The mechanical parameter determination module 703 is used to determine the dynamic mechanical characteristic parameters of the target elastic component based on the multidimensional force data and the multidimensional deformation data.

[0119] The status determination module 704 is used to determine the health status level of the target elastic component based on the dynamic mechanical characteristic parameters and the apparent characteristic information.

[0120] In one possible implementation, the multidimensional deformation data includes translational displacement data in the first dimension and rotational angle data in the second dimension, whereby the rotational angle data includes yaw angle data and torsional angle data. The feature extraction module 702 is specifically used to: identify the outer contour position information of the target elastic component and the relative deflection information of the inner and outer components in the non-contact sensing data; determine the translational displacement data based on the outer contour position information; and determine the rotational angle data based on the relative deflection information.

[0121] In one possible implementation, the multidimensional force data includes linear force values ​​and torsional moment values, and the dynamic mechanical characteristic parameters include linear stiffness parameters and angular stiffness parameters. The angular stiffness parameters include yaw angle stiffness and torsional angular stiffness. The mechanical parameter determination module 703 is specifically used to: perform time-series alignment processing on the multidimensional force data and multidimensional deformation data according to the acquisition timestamp; determine the linear stiffness parameter based on the first quotient of the time-series aligned linear force value and translational displacement data; and determine multiple corresponding angular stiffness quotients by dividing the time-series aligned torsional moment value by the yaw angle data and torsional angle data, and determine the yaw angle stiffness and torsional angular stiffness according to the multiple angular stiffness quotients.

[0122] In one possible implementation, the mechanical parameter determination module 703 is further configured to: acquire ambient temperature data of the target elastic component; look up the temperature compensation factor corresponding to the ambient temperature data in a mapping table; and determine the corrected linear stiffness parameter based on the product of the first quotient and the temperature compensation factor.

[0123] In one possible implementation, the state determination module 704 is further configured to: determine the health status level of the target elastic component as a failure level when the structural boundary features do not meet preset safety boundary conditions. The failure to meet the preset safety boundary conditions includes any of the following: the structural boundary features indicate the presence of a skeleton crack in the target elastic component; the structural boundary features indicate that the crack depth of the elastic body in the target elastic component is greater than or equal to a crack depth threshold; the structural boundary features indicate that there is a through-bonding separation between the elastic body and the skeleton of the target elastic component; or the structural boundary features indicate that the inner skeleton and outer skeleton of the target elastic component are in direct contact.

[0124] In one possible implementation, the feature extraction module 702 is further configured to: input non-contact sensing data into a pre-trained feature extraction network model to extract texture features from at least one pixel channel; and perform structured constraint matching on the texture features according to anti-hallucination determination instructions to determine the damage state identifier.

[0125] In one possible implementation, if the decrease in the online stiffness parameter or angular stiffness parameter is greater than or equal to a first threshold, and the damage status indicator indicates the presence of a first damage characteristic, the status determination module 704 determines the health status level as a first risk level. If the decrease in the online stiffness parameter or angular stiffness parameter is less than the first threshold but greater than or equal to a second threshold, and the damage status indicator indicates the presence of a second damage characteristic, the status determination module 704 determines the health status level as a second risk level. If both the online stiffness parameter and the angular stiffness parameter are in an increasing state, and the damage status indicator indicates the absence of a damage characteristic, the status determination module 704 determines the health status level as a safe level.

[0126] In one possible implementation, the feature extraction module 702 is further configured to: input non-contact sensing data into a pre-trained feature extraction network model to extract component spacing features; and determine a component interference risk indicator if the component spacing features are less than or equal to a collision interference threshold. Upon generating the component interference risk indicator, the state determination module 704 determines the health status level as a warning risk level.

[0127] In one possible implementation, the target vehicle includes an autonomous vehicle. The target elastic component includes a rubber bushing with a metal frame in the autonomous vehicle's chassis suspension system. Non-contact sensing data includes sequence data of local images of the chassis acquired by an onboard miniature camera.

[0128] The specific functions and examples of each module and submodule of the apparatus in this disclosure can be found in the relevant descriptions of the corresponding steps in the above method embodiments, and will not be repeated here.

[0129] This disclosure also provides an autonomous vehicle, including a vehicle chassis, an image acquisition device, and a vehicle-mounted computing unit. The vehicle chassis suspension system includes a rubber bushing with a metal frame. The image acquisition device is installed near the rubber bushing and is used to acquire local image sequence data of the rubber bushing as non-contact sensing data. The vehicle-mounted computing unit is communicatively connected to the image acquisition device and is used to execute any of the vehicle state monitoring methods in this disclosure.

[0130] In one possible implementation, the image acquisition device is equipped with a protective cover and a supplementary light. The protective cover can be made of a high-strength transparent material to resist impacts from flying stones and mud and water, and the supplementary light can use an LED light source to provide sufficient illumination in the dark areas under the chassis.

[0131] The acquisition, storage, and application of any type of information, such as user personal information, involved in the technical solutions disclosed herein comply with relevant laws and regulations and do not violate public order and good morals.

[0132] According to embodiments of this disclosure, this disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0133] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0134] like Figure 8 As shown, device 800 includes a computing unit 801, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 802 or a computer program loaded from storage unit 808 into random access memory (RAM) 803. RAM 803 may also store various programs and data required for the operation of device 800. The computing unit 801, ROM 802, and RAM 803 are interconnected via bus 804. Input / output (I / O) interface 805 is also connected to bus 804.

[0135] Multiple components in device 800 are connected to I / O interface 805, including: input unit 806, such as keyboard, mouse, etc.; output unit 807, such as various types of monitors, speakers, etc.; storage unit 808, such as disk, optical disk, etc.; and communication unit 809, such as network card, modem, wireless transceiver, etc. Communication unit 809 allows device 800 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0136] The computing unit 801 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as vehicle condition monitoring methods. For example, in some embodiments, the vehicle condition monitoring method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on device 800 via ROM 802 and / or communication unit 809. When the computer program is loaded into RAM 803 and executed by the computing unit 801, one or more steps of the vehicle condition monitoring method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the vehicle condition monitoring method by any other suitable means (e.g., by means of firmware).

[0137] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0138] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0139] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0140] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0141] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0142] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0143] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0144] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A vehicle status monitoring method, applied to a vehicle-side computing unit, the method comprising: Based on the real-time operating data of the target vehicle, the multi-dimensional force data of the target elastic component is determined by the state association model deployed on the vehicle-side computing unit; wherein, the state association model is a dynamic model constructed based on the physical properties of the target vehicle. Extract multidimensional deformation data and appearance feature information from the non-contact sensing data of the target elastic component; Based on the multidimensional force data and the multidimensional deformation data, the dynamic mechanical property parameters of the target elastic component are determined; Based on the dynamic mechanical property parameters and the apparent feature information, the health status level of the target elastic component is determined.

2. The method according to claim 1, wherein, The multidimensional deformation data includes translational displacement data in the first dimension and rotational angle data in the second dimension, wherein the rotational angle data includes yaw angle data and torsional angle data. The extraction of multidimensional deformation data from the non-contact sensing data of the target elastic component includes: Identify the outer contour position information of the target elastic component and the relative deflection information of the inner and outer components in the non-contact sensing data; Based on the outer contour position information, the translational displacement data is determined; Based on the relative deflection information, the rotation angle data is determined.

3. The method according to claim 2, wherein, The multidimensional force data includes linear force values ​​and torsional moment values; the dynamic mechanical characteristic parameters include linear stiffness parameters and angular stiffness parameters; the angular stiffness parameters include yaw angle stiffness and torsional angle stiffness. The determination of the dynamic mechanical property parameters of the target elastic component based on the multidimensional force data and the multidimensional deformation data includes: The multidimensional force data and the multidimensional deformation data are time-series aligned according to the collection timestamp; The linear stiffness parameter is determined based on the first quotient of the linear force value after time alignment and the translational displacement data. The torsional moment value after time alignment is divided by the yaw angle data and the torsion angle data to determine multiple corresponding angular stiffness quotients, and the yaw angle stiffness and the torsion angle stiffness are determined according to the multiple angular stiffness quotients.

4. The method according to claim 3, characterized in that, After the step of determining the linear stiffness parameter based on the first quotient of the time-aligned linear force value and the translational displacement data, the method further includes: Obtain the ambient temperature data of the target elastic component; The temperature compensation factor corresponding to the ambient temperature data is found in the mapping table, which is a pre-generated table of correspondence between temperature values ​​and compensation factor values. The corrected linear stiffness parameter is determined based on the product of the first quotient and the temperature compensation factor.

5. The method according to claim 1, wherein, The apparent feature information includes structural boundary features; Before determining the health status level of the target elastic component based on the dynamic mechanical characteristic parameters and the apparent feature information, the method further includes: If the structural boundary features do not meet the preset safety boundary conditions, the health status level of the target elastic component is determined to be a failure level. The failure to meet the preset security boundary conditions includes any of the following situations: The structural boundary features indicate that the target elastic component has a skeleton crack; The structural boundary features indicate that the elastomeric crack depth of the target elastic component is greater than or equal to the crack depth threshold. The structural boundary features indicate that there is a through-bonding between the elastomer and the skeleton of the target elastic component; The structural boundary features indicate that the inner skeleton and outer skeleton of the target elastic component are in direct contact.

6. The method according to claim 3, characterized in that, The apparent feature information includes a damage state identifier, which is used to characterize whether there are damage features on the surface of the target elastic component. The extraction of apparent feature information from the non-contact sensing data of the target elastic component includes: The non-contact sensing data is input into a pre-trained feature extraction network model to extract texture features of at least one pixel channel; The texture features are subjected to structured constraint matching according to the anti-illusion determination instruction to determine the damage state identifier. The anti-illusion determination instruction includes the constraint condition of extracting continuous edge substrings and rejecting the generation of unobserved textures.

7. The method according to claim 6, characterized in that, Determining the health status level of the target elastic component based on the dynamic mechanical characteristic parameters and the apparent feature information includes: If the decrease in the linear stiffness parameter or the angular stiffness parameter is greater than or equal to a first threshold, and the damage status indicator indicates the presence of a first damage characteristic, the health status level is determined to be a first risk level. If the decrease in the linear stiffness parameter or the angular stiffness parameter is less than the first threshold and greater than or equal to the second threshold, and the damage status indicator indicates the presence of a second damage characteristic, the health status level is determined to be a second risk level. If the linear stiffness parameter and the angular stiffness parameter are both increasing, and the damage status indicator shows no damage characteristics, then the health status level is determined to be a safe level.

8. The method according to claim 3, characterized in that, The apparent feature information includes a component interference risk identifier, which is used to characterize whether there is a risk of motion interference between the target elastic component and surrounding components; the method further includes: The non-contact sensing data is input into a pre-trained feature extraction network model to extract component spacing features. If the component spacing features are less than or equal to the collision interference threshold, the component interference risk identifier is determined. Determining the health status level of the target elastic component based on the dynamic mechanical characteristic parameters and the apparent feature information includes: When the component interference risk identifier is generated, the health status level is determined to be a warning risk level.

9. The method according to any one of claims 1 to 8, wherein, The target vehicles include driverless vehicles; The target elastic component includes a rubber bushing with a metal frame in the chassis suspension system of the unmanned vehicle; The non-contact sensing data includes a sequence of local chassis images collected by an onboard miniature camera.

10. A vehicle condition monitoring device, wherein, The device, applied to an on-vehicle computing unit, includes: The force data determination module is used to determine the multidimensional force data of the target elastic component based on the real-time operating condition data of the target vehicle and through the state association model deployed on the vehicle-side computing unit. The state association model is a dynamic model constructed based on the physical properties of the target vehicle. The feature extraction module is used to extract multidimensional deformation data and appearance feature information from the non-contact sensing data of the target elastic component; The mechanical parameter determination module is used to determine the dynamic mechanical characteristic parameters of the target elastic component based on the multidimensional force data and the multidimensional deformation data. The status determination module is used to determine the health status level of the target elastic component based on the dynamic mechanical characteristic parameters and the apparent feature information.

11. An unmanned vehicle, wherein, include: The vehicle chassis, wherein the suspension system of the vehicle chassis is provided with rubber bushings containing a metal skeleton; An image acquisition device is installed near the rubber bushing to acquire local image sequence data of the rubber bushing as non-contact sensing data. The vehicle-side computing unit is communicatively connected to the image acquisition device and is used to execute the vehicle status monitoring method as described in any one of claims 1 to 9.

12. The driverless vehicle according to claim 11, wherein, The image acquisition device is equipped with a protective cover and a supplementary light.

13. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-9.

14. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-9.

15. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-9.

Images