A road surface flatness detection device and detection method based on vehicle vibration response
By installing an acceleration sensor on the inspection vehicle and using a generative adversarial neural network to optimize road surface roughness samples, the problem of poor recognition results caused by relying on theoretical models in existing technologies is solved. This achieves low-cost and efficient road surface smoothness detection, which is suitable for rapid evaluation and decision support of large-scale road networks.
Patent Information
- Application Number
- CN202411264518.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-10
- Publication Date
- 2025-11-07
- Estimated Expiration
- 2044-09-10
AI Technical Summary
Existing methods for identifying road surface smoothness based on vehicle vibration response are highly dependent on theoretical models, resulting in poor identification performance when measurement errors or parameters are inaccurate, and are also costly or inefficient.
By combining generative adversarial neural networks with vehicle vibration response, and by installing acceleration sensors on the front and rear axles of the detection vehicle, the generative adversarial neural network generates road surface roughness samples. The generator and discriminator are optimized by combining physical constraint equations, which reduces the dependence on theoretical models and improves the accuracy and efficiency of identification.
It achieves low-cost and efficient road surface smoothness detection, reduces dependence on vehicle parameters, improves recognition accuracy and robustness, and is suitable for rapid assessment and decision support of large-scale road networks.
Smart Images
Figure CN118999435B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of road and rail engineering detection, and particularly relates to a road roughness detection device and method based on vehicle vibration response. BACKGROUND
[0002] Roughness is an important indicator of road performance and driving comfort. Identifying road roughness is of great significance for timely maintenance and management of roads. There are currently three main methods for identifying road roughness. The first method is direct measurement using a profilometer, ruler, level, and other devices. Although the accuracy is guaranteed, this method is time-consuming and labor-intensive. The second method is to identify road roughness using a laser profilometer, ground penetrating radar, optical imaging, and point cloud model. Although the accuracy is good, these instruments are very expensive. The third method is to indirectly identify road roughness based on vehicle vibration response. Although the accuracy is relatively low, this method has the advantages of low cost and high measurement efficiency. In recent years, with the development of big data technology, the method of indirectly identifying road roughness based on vehicle vibration response has become increasingly popular. However, existing methods of indirectly identifying road roughness based on vehicle vibration are highly dependent on theoretical models of vehicle vibration. When there are measurement errors or vehicle parameters cannot be measured, the method based on theoretical models often fails to achieve satisfactory results.
[0003] Therefore, in order to reduce the cost of detecting road roughness and avoid the problem of existing methods of indirectly identifying road roughness based on vehicle vibration being highly dependent on theoretical models, it is necessary to combine advanced machine learning models and use big data analysis to indirectly measure road roughness based on vehicle vibration response. SUMMARY
[0004] The purpose of the present application is to provide a road roughness detection device and method based on vehicle vibration response, in order to reduce the cost of detecting road roughness and avoid the problem of existing methods of indirectly identifying road roughness based on vehicle vibration being highly dependent on theoretical models.
[0005] The present application is implemented by the following technical solution: a road roughness detection method based on vehicle vibration response, characterized by the following steps:
[0006] S1. Acceleration sensors are attached to the front and rear axles of a two-axle detection vehicle without a shock absorption system, and the detection vehicle is driven over uneven road surfaces to measure and record the vibration acceleration responses of the front and rear axles. and
[0007] S2. A generative adversarial neural network is established to generate vibration acceleration responses of the front and rear axles of the vehicle and and road roughness samples rg ;
[0008] S3, generating a response by a discriminator of the generative adversarial neural network and with the measured response and performing anastomosis discrimination, constructing a discrimination loss;
[0009] S4, generating an acceleration response and obtaining the corresponding displacement response by numerical integration and and establishing a two-axis vehicle vibration equation and and r g physical constraint equation, constructing a generation loss containing physical constraint loss and adversarial loss;
[0010] S5, first reduce the discrimination loss to optimize the discriminator, and then reduce the generation loss to optimize the generator;
[0011] S6, the road roughness sample r generated by the optimized generative adversarial neural network g is the unevenness of the road to be identified.
[0012] Further, the generative adversarial neural network comprises a generator and a discriminator, and the generator and the discriminator are both composed of five layers of feedforward neural network.
[0013] Further, the physical constraint equation in step S4 is expressed as follows:
[0014]
[0015] In the formula: d1 and d2 are the distances between the vehicle center of gravity and the front and rear axles, respectively, d is the distance between the front and rear axles of the vehicle and has d=d1+d2, k1 is the spring stiffness of the front axle tire of the vehicle, k2 is the spring stiffness of the rear axle tire of the vehicle, M v is the mass of the vehicle, J v is the moment of inertia of the vehicle center of gravity, is the road roughness sample contacted by the front wheel, is the road roughness sample contacted by the rear wheel.
[0016] Further, the discrimination loss in step S3 is expressed as follows:
[0017]
[0018] In the formula: E(.) is a mathematical expectation function, D(.) is a discriminator of the generative adversarial neural network, t represents a time, z is an input variable of the generative adversarial neural network, and is generated through a standard normal distribution.
[0019] Further, the generation loss in the step S4 is expressed as follows:
[0020] Gen loss = Adv loss + beta * Phy loss
[0021]
[0022] In the formula: Adv loss represents an adversarial loss of the generative adversarial neural network, Phy loss represents a physical constraint loss of a physical constraint equation, and beta represents a weight coefficient of the physical constraint loss.
[0023] A road surface flatness detection device based on vehicle vibration response, characterized in that it comprises a front axle, a rear axle, a vehicle frame, a traction frame, an acceleration sensor and a data acquisition system; the vehicle frame is fixedly supported on the front axle and the rear axle, and the traction frame and a weight box are installed on the vehicle frame; the acceleration sensors are respectively installed on the front axle and the rear axle, and the data acquisition system is arranged in the weight box of the vehicle frame and connected with the acceleration sensors.
[0024] Further, the front axle and the rear axle are respectively formed by connecting left and right two wheels and a steel bar, the vehicle frame is fixedly supported on the steel bars of the front axle and the rear axle, and the acceleration sensors are respectively installed at the center positions of the steel bars of the front axle and the rear axle.
[0025] The road surface flatness detection device and method based on vehicle vibration response have the following beneficial effects:
[0026] Significant cost-effectiveness: the present application indirectly evaluates the road surface flatness by using vehicle vibration response, which greatly reduces the detection cost compared with the traditional method of directly measuring the road surface flatness by using a profilometer, a ruler, a level and other direct measurement tools. Meanwhile, compared with high-precision but expensive equipment such as a laser profilometer and a ground penetrating radar, the present application realizes higher economy under the premise of ensuring a certain detection accuracy.
[0027] High efficiency and convenience: the road surface flatness is detected by using vehicle vibration response, which does not require a complex installation and calibration process, can quickly and continuously obtain road surface state data, and significantly improves the detection efficiency. This is particularly important for large-scale road network maintenance management, and can quickly provide decision basis.
[0028] Data-driven and model fusion: combined with advanced generative adversarial neural network (GAN) technology, this method not only utilizes the powerful data processing capability of machine learning, but also integrates the vehicle vibration theoretical model, realizing the joint driving of physical data and mathematical model. This method reduces the excessive dependence on vehicle parameters and improves the accuracy and robustness of identification.
[0029] Strong applicability: the detection device is designed as a simple double-axle detection vehicle, which is compact in structure and easy to install and deploy. By adjusting the weight of the pressure box, the vehicle driving state under different load conditions can be simulated, further enhancing the universality and reliability of the detection results.
[0030] Fast decision support: the road roughness identification results provided by the present application can provide timely and accurate decision basis for road repair and maintenance. By quickly evaluating the roughness of different road sections in the road network, the repair work of severely damaged road sections can be prioritized, thereby improving the traffic quality and driving safety of the entire road network. BRIEF DESCRIPTION OF DRAWINGS
[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. The drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the structures shown in the drawings.
[0032] Figure 1 The flowchart of the detection method of the present application;
[0033] Figure 2 The schematic diagram of the detection device of the present application;
[0034] Figure 3 The schematic diagram of the detection device axle of the present application;
[0035] Figure 4 The measured vehicle front axle vibration acceleration response schematic diagram of the detection vehicle of the embodiment of the present application passing through the uneven road;
[0036] Figure 5 The measured vehicle rear axle vibration acceleration response schematic diagram of the detection vehicle of the embodiment of the present application passing through the uneven road;
[0037] Figure 6 The feedforward neural network structure schematic diagram of the generator and discriminator of the generative adversarial neural network of the embodiment of the present application;
[0038] Figure 7 The physical constraint loss change schematic diagram of the generative adversarial neural network of the embodiment of the present application with the training sample;
[0039] Figure 8 Fig. 1 is a schematic diagram of the change of the adversarial loss of the generative adversarial neural network of the embodiment of the present application with the training sample;
[0040] Figure 9 Fig. 1 is a schematic diagram of the change of the adversarial loss of the generative adversarial neural network of the embodiment of the present application with the training sample;
[0041] In the figure, 1 is a front axle, 2 is a rear axle, 3 is a wheel, 4 is a steel bar, 5 is a frame, 6 is a traction frame, 7 is a weight box, 8 is an acceleration sensor, and 9 is a data acquisition system. DETAILED DESCRIPTION
[0042] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the accompanying drawings can be arranged and designed in various different configurations.
[0043] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative labor are within the scope of protection of the present application.
[0044] As shown in Figures 1-9 A road surface flatness detection device based on vehicle vibration response, which comprises a front axle 1, a rear axle 2, a frame 5, a traction frame 6, an acceleration sensor 8 and a data acquisition system 9. The front axle 1 and the rear axle 2 are respectively connected by left and right two wheels 3 and a steel bar 4 to form. The frame 5 is fixedly supported on the steel bars 4 of the front axle 1 and the rear axle 2, and the traction frame 6 and the weight box 7 are installed on the frame 5. The acceleration sensor 8 is respectively installed at the center position of the steel bars 4 of the front axle 1 and the rear axle 2, and the data acquisition system 9 is arranged in the weight box 7 of the frame 5 and connected with the acceleration sensor 8. The detection device is connected with a vehicle having a power system through the traction frame 6, and is pulled forward by the vehicle.
[0045] A road surface flatness detection method based on vehicle vibration response, comprising the following main steps:
[0046] S1, pasting acceleration sensors on the front axle and the rear axle of the detection vehicle respectively, collecting the vibration acceleration response of the vehicle passing through uneven road surface and As shown in Figure 4 and Figure 5 ;
[0047] S2, establishing a generative adversarial neural network to generate vibration acceleration responses of the front axle and the rear axle of the vehicle and and a road roughness sample r g ;
[0048] S3, performing a consistency discrimination on the generated response and and the measured response and to construct a discrimination loss;
[0049] S4, obtaining corresponding displacement responses and by numerical integration of the generated acceleration responses and and establishing a two-axle vehicle vibration equation and and r g physical constraint equation to construct a generative loss containing a physical constraint loss and an adversarial loss;
[0050] S5, first reducing the discrimination loss to optimize the discriminator, and then reducing the generative loss to optimize the generator;
[0051] S6, the road roughness sample r g generated by the generative adversarial neural network after optimization is the unevenness of the road to be identified.
[0052] The generative adversarial neural network comprises a generator and a discriminator, and both the generator and the discriminator are composed of five layers of feedforward neural networks, as shown in Figure 6 ;
[0053] The physical constraint equation is expressed as follows:
[0054]
[0055] In the formula, d1 and d2 are the distances from the vehicle center of gravity to the front axle and the rear axle respectively, d is the distance between the front axle and the rear axle of the vehicle and has d=d1+d2, k1 is the spring stiffness of the front axle tire of the vehicle, k2 is the spring stiffness of the rear axle tire of the vehicle, M v is the mass of the vehicle, J v is the moment of inertia of the vehicle center of gravity, is the road roughness sample contacted by the front wheel, is the road roughness sample contacted by the rear wheel.
[0056] The discrimination loss of the discriminator is expressed as follows: by reducing the discrimination loss, the discriminator cannot discriminate the similarities and differences between the generated data and the measured data.
[0057]
[0058] In the formula: E(.) is a mathematical expectation function, D(.) is a discriminator of the generative adversarial neural network, t represents a time, z is an input variable of the generative adversarial neural network, and is generated through a standard normal distribution.
[0059] The generation loss of the generator is expressed as follows:
[0060]
[0061] In the formula: Adv loss represents an adversarial loss of the generative adversarial neural network, Phy loss represents a physical constraint loss of a physical constraint equation, and β represents a weight coefficient of the physical constraint loss.
[0062] Changes of the generation loss and the physical constraint loss of the generator in the iteration process are as follows: Figure 7 and Figure 8 ;
[0063] Finally, the identified road flatness profile is as follows: Figure 9 .
[0064] In the embodiment, the vehicle parameter values of the detection vehicle are as shown in Table 1.
[0065] Table 1 Vehicle parameter values of the detection vehicle
[0066]
[0067] In the above embodiments, the basic principles and main features of the present application and the advantages of the present application are described. It should be understood by those skilled in the art that the present application is not limited by the above embodiments, and the above embodiments and descriptions in the specification are only to illustrate the principles of the present application. Any modification and change made by those skilled in the art without departing from the spirit and scope of the present application should be within the protection scope of the appended claims of the present application.
Claims
1. A method for detecting road roughness based on a vehicle vibration response, characterized by, The method comprises the following steps: S1, respectively paste acceleration sensor on the front axle and rear axle of a two-axle detection vehicle, make the detection vehicle pass through uneven road, respectively measure and record the vibration acceleration response of the front axle and rear axle and S2, establish a generative adversarial neural network to generate vibration acceleration responses of the front axle and the rear axle of the vehicle and and a road roughness sample r g ; S3, discriminating the generated response by a discriminator of the generative adversarial neural network and with the measured response and to perform the anastomosis discrimination, and construct a discrimination loss S4, generate acceleration response and The corresponding displacement response is obtained by numerical integration and And through the two-axis vehicle vibration equation and and r g The physical constraint equation is constructed, including the physical constraint loss and the adversarial loss. S5, first reducing the discriminant loss to optimize the discriminator, and then reducing the generation loss to optimize the generator; S6, the roughness sample r generated by the generated adversarial neural network after optimization is completed g is the unevenness of the road surface to be identified; The physical constraint equation in the step S4 is expressed as follows: where d1 and d2 are the distances of the vehicle's center of gravity from the front and rear axles, d is the distance between the front and rear axles and has d = d1 + d2, k1 is the spring rate of the front axle tires of the vehicle, k2 is the spring rate of the rear axle tires of the vehicle, M is the mass of the vehicle, J v is the moment of inertia of the vehicle's center of gravity, v is the moment of inertia of the vehicle's center of gravity, is a sample of the road roughness at the front wheel contact, is a sample of the road roughness at the rear wheel contact; The generation loss in the step S4 is expressed as follows: Gen loss = Adv loss + β · Phy loss In the formula: Adv loss The adversarial loss of the generative adversarial neural network, Phy loss The physical constraint loss of the physical constraint equation, β represents the weight coefficient of the physical constraint loss, E(.) is the mathematical expectation function, D(.) is the discriminator of the generative adversarial neural network, t represents the time, z is the input variable of the generative adversarial neural network, and is generated through a standard normal distribution.
2. The method of claim 1, wherein The generative adversarial neural network comprises a generator and a discriminator, and both the generator and the discriminator are composed of five layers of feedforward neural networks.
3. The method of claim 1, wherein The discriminant loss in the step S3 is expressed as follows: In the formula, E(.) is a mathematical expectation function, D(.) is a discriminator of the generative adversarial neural network, t represents a time point, and z is an input variable of the generative adversarial neural network, which is generated through a standard normal distribution.