CN113312744B - Vehicle load prediction method and system - Google Patents
Vehicle load prediction method and system Download PDFInfo
- Publication number
- CN113312744B CN113312744B CN202110378341.2A CN202110378341A CN113312744B CN 113312744 B CN113312744 B CN 113312744B CN 202110378341 A CN202110378341 A CN 202110378341A CN 113312744 B CN113312744 B CN 113312744B
- Authority
- CN
- China
- Prior art keywords
- vehicle load
- vehicle
- monte carlo
- load characteristics
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
Abstract
The invention relates to a vehicle load prediction method and a system, wherein the method comprises the following steps: acquiring the determined vehicle load characteristics through a road monitoring system, and loading the vehicle load characteristics into a pre-established and trained multi-level Monte Carlo model, wherein the multi-level Monte Carlo model acquires the prediction results of all vehicle load characteristics according to pre-trained random vectors corresponding to the uncertain vehicle load characteristics and the determined vehicle load characteristics; the multilevel Monte Carlo model comprises a plurality of layers of nodes, each layer of nodes respectively represents a prediction layer of vehicle load characteristics, each node represents a distribution model of vehicle load characteristic vectors on corresponding dimensions under a certain condition, and the vehicle load characteristic vectors are vectors formed by all vehicle load characteristics. Compared with the prior art, the method has higher robustness, and can statistically and effectively solve the problem that the vehicle load cannot be directly acquired in the visual information.
Description
Technical Field
The invention relates to the technical field of vehicle load prediction, in particular to a vehicle load prediction method and system.
Background
With the rapid development of socioeconomic, the road traffic load increases year by year. Passenger vehicle loads are rapidly evolving due to vehicle type restrictions and other modes of transportation, such as: passenger transport loads of high-speed rails, airplanes and the like gradually tend to be smooth, but the annual average load of freight vehicles tends to increase year by year. This means that as highway transportation is further developed, bridges built according to the existing standards will face higher and higher bearing capacity requirements, and the service life of the bridges will be greatly shortened due to inevitable damage to the bridge structure caused by increasing transportation loads. Therefore, the construction of the vehicle load model based on the actually measured traffic load has great significance for bridge design verification, safety maintenance, performance evaluation and operation management.
However, bridge deck vehicles have high randomness and currently lack an effective means for observation. Currently, the mainstream technical method is to construct a vehicle load model through vehicle load data collected by a bridge floor dynamic weighing system (WIM). The bridge deck dynamic weighing system can accurately record the vehicle load key information such as the vehicle speed, the vehicle length, the total weight, the axle type, the axle number, the axle weight and the like of each vehicle passing through the weighing section, and can record the lane through which the vehicle runs according to the layout position of the weighing sensors. However, since the WIM system data only expresses the actual traffic load characteristics of each lane of a certain section of the bridge deck, and the situations such as whether the vehicle changes lanes or the specific positions of wheel loads cannot be known, the vehicle load model established based on the WIM data cannot accurately reflect the actual vehicle load level of each area of the bridge deck for the overall bridge structure.
Disclosure of Invention
The invention aims to overcome the defect that the vehicle load model established based on WIM data cannot accurately reflect the actual vehicle load level of each area of a bridge floor in the prior art, and provides a vehicle load prediction method and a vehicle load prediction system.
The purpose of the invention can be realized by the following technical scheme:
a vehicle load prediction method comprising the steps of:
the vehicle video data is collected by the road monitoring system, and the data is processed to obtain the determined vehicle load characteristics,
loading the determined vehicle load characteristics into a pre-established and trained multilayer Monte Carlo model, and comparing the determined vehicle load characteristics with all preset vehicle load characteristics by the multilayer Monte Carlo model to obtain uncertain vehicle load characteristics; obtaining the prediction results of all vehicle load characteristics according to the pre-trained random vector corresponding to the uncertain vehicle load characteristics and the confirmed vehicle load characteristics;
the multilevel Monte Carlo model comprises a plurality of layers of nodes, each layer of nodes respectively represents a prediction layer of vehicle load characteristics, each node represents a distribution model of vehicle load characteristic vectors on corresponding dimensions under a certain condition, and the vehicle load characteristic vectors are vectors formed by all vehicle load characteristics;
the training of the multi-level Monte Carlo model comprises the following steps: and constructing a sample set of the vehicle load characteristic vector, loading the sample set into the multilevel Monte Carlo model, and training to obtain a training value of a random vector corresponding to each layer of nodes in the multilevel Monte Carlo model.
Further, the expression of the multilevel monte carlo model is as follows:
ξ 0 ＝F 0 (x 0 )
X＝[x 0 ,x 1 ,...,x n }
in the formula, xi 0 Is the first random vector and is the second random vector,
is a joint probability distribution function of the j node on the ith layer in the multi-layer Monte Carlo model i To satisfy (0, 1)]Uniformly distributed random numbers, X is a random vector, X n The nth vehicle load signature.
And further, acquiring dynamic weighing data through a dynamic weighing system, and constructing a vehicle load characteristic vector sample set by combining vehicle load characteristics acquired by a road monitoring system for training a multi-level Monte Carlo model.
Further, the vehicle load characteristics include a vehicle traveling direction, a vehicle category, a vehicle size, a total vehicle weight, a vehicle axle combination pattern, an axle type, and an axle weight.
Further, the vehicle axle combination pattern is represented by a code.
The present invention also provides a vehicle load recognition system, comprising:
the vision acquisition module is used for acquiring vehicle video data through the road monitoring system, processing the data to acquire the determined vehicle load characteristics and the vehicle tracking result,
the vehicle load prediction module is used for loading the determined vehicle load characteristics into a pre-established and trained multi-level Monte Carlo model, and the multi-level Monte Carlo model compares the determined vehicle load characteristics with all preset vehicle load characteristics to obtain uncertain vehicle load characteristics; obtaining the prediction results of all vehicle load characteristics according to the pre-trained random vector corresponding to the uncertain vehicle load characteristics and the confirmed vehicle load characteristics;
the multilevel Monte Carlo model comprises a plurality of layers of nodes, each layer of nodes respectively represents a prediction layer of vehicle load characteristics, each node represents a distribution model of vehicle load characteristic vectors on corresponding dimensions under a certain condition, and the vehicle load characteristic vectors are vectors formed by all vehicle load characteristics;
the vehicle load space-time distribution identification module is used for combining a vehicle tracking result and prediction results of all vehicle load characteristics to construct a vehicle load space-time distribution model;
a model training module for training a multi-level Monte Carlo model, the training process comprising: and constructing a sample set of the vehicle load characteristic vector, loading the sample set into the multilevel Monte Carlo model, and training to obtain a training value of a random vector corresponding to each layer of nodes in the multilevel Monte Carlo model.
Further, the expression of the multilevel monte carlo model is as follows:
ξ 0 ＝F 0 (x 0 )
X＝{x 0 ,x 1 ,...,x n }
in the formula, xi 0 Is the first random vector and is the second random vector,
is a joint probability distribution function of the j node on the ith layer in the multi-layer Monte Carlo model i To satisfy (0, 1)]Uniformly distributed random numbers, X being a random vector, X n The nth vehicle load characteristic.
Further, dynamic weighing data are obtained through a dynamic weighing system, and a vehicle load characteristic vector sample set is constructed by combining vehicle load characteristics obtained by a road monitoring system and is used for training a multi-level Monte Carlo model.
Further, the vehicle load characteristics include a vehicle traveling direction, a vehicle category, a vehicle size, a total vehicle weight, a vehicle axle combination pattern, an axle type, and an axle weight.
Further, the vehicle axle combination pattern is represented by a code.
Compared with the prior art, the invention has the following advantages:
the invention provides a vehicle load inference method based on a multilevel Monte Carlo model, aiming at the problems that visual information can not directly obtain vehicle load and the vehicle load information is difficult to obtain in the construction of a vehicle load distribution model, and having the following advantages:
(1) The multilevel Monte Carlo model constructed based on the Monte Carlo principle can effectively learn the random characteristics of the vehicle load.
(2) The multi-level Monte Carlo model vehicle load inference method for simulating the vehicle load feature vector has high robustness, and can effectively solve the problem that the vehicle load cannot be directly acquired in visual information statistically.
(3) The real bridge application with the river-Yin bridge as the background proves that the multilevel Monte Carlo vehicle load inference method which integrates the computer vision vehicle identification and tracking results is used for constructing various space-time distribution models and refined vehicle load spectrum models. The related research results can be directly applied to management and maintenance of the bridge in the operation period, and have important significance on further bridge performance evaluation and fatigue performance research.
Drawings
FIG. 1 is a block diagram of a vehicle load recognition system incorporating a multi-level Monte Carlo model and computer vision techniques according to an embodiment of the present invention;
FIG. 2 is a multi-level Monte Carlo model suitable for random vehicle load feature vector simulation according to an embodiment of the present invention;
FIG. 3 is a line graph of the gross vehicle weight distribution for vehicles on different axles based on MLMC modeling and WIM data statistics in an embodiment of the present invention;
FIG. 4 is a line graph of gross vehicle weight distribution for different vehicle classes based on MLMC modeling and WIM data statistics in an embodiment of the present invention;
FIG. 5 is a line graph of the gross vehicle weight distribution at three load inferences for five spatial points according to an embodiment of the present invention;
FIG. 6 is a comparison graph of the daily average flow distribution and the annual average flow distribution of vehicles of different grades in the embodiment of the present invention, wherein (6 a) is a line graph of the daily average flow distribution of the passenger 1 vehicle type, and (6 b) is a line graph of the daily average flow distribution of the cargo 5 vehicle type;
FIG. 7 is a bar chart of the time distribution of the flow of each model of the river-yin bridge in the embodiment of the present invention;
FIG. 8 is a line diagram of the average daily flow distribution of each model of the Jiangyin bridge in the embodiment of the invention;
FIG. 9 is a vehicle speed distribution boxline diagram of each grade of vehicle in an embodiment of the invention;
FIG. 10 is a frequency spectrum of a traveling vehicle acting on a bridge deck according to an embodiment of the present invention;
FIG. 11 is a spatial distribution plot of average wheel load for an embodiment of the present invention;
FIG. 12 is a diagram illustrating a result of identifying a high-occupancy area of a bridge deck based on a transverse distribution model according to an embodiment of the present invention;
FIG. 13 is a diagram illustrating the identification result of a vulnerable (fatigue prone) area of a bridge deck based on a transverse distribution model in an embodiment of the present invention;
FIG. 14 is a model diagram of a vehicle load spectrum in accordance with an embodiment of the present invention;
FIG. 15 is a schematic view of a single-axle single tire in an embodiment of the present invention;
FIG. 16 is a schematic view of a single-axle twin tire in an embodiment of the present invention;
FIG. 17 is a schematic view of a dual spindle single tire in an embodiment of the present invention;
FIG. 18 is a schematic view of a dual spindle single and dual tire in accordance with an embodiment of the present invention;
FIG. 19 is a schematic view of a two-axis twin in an embodiment of the invention;
FIG. 20 is a schematic view of a triple-axis twin tire in accordance with an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations.
Thus, the following detailed description of the embodiments of the present invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures.
In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. indicate orientations or positional relationships based on orientations or positional relationships shown in the drawings or orientations or positional relationships that the present product is conventionally placed in use, and are only for convenience of describing the present invention and simplifying the description, but do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus, should not be construed as limiting the present invention.
It should be noted that the terms "first" and "second" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or to implicitly indicate the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of the present application, "a plurality" means two or more unless specifically limited otherwise.
Furthermore, the terms "horizontal", "vertical" and the like do not imply that the components are required to be absolutely horizontal or pendant, but rather may be slightly inclined. For example, "horizontal" merely means that the direction is more horizontal than "vertical" and does not mean that the structure must be perfectly horizontal, but may be slightly inclined.
Example 1
In consideration of the fact that the observation range of vehicle load monitoring can be effectively expanded by using a computer vision technology, but the load information of the vehicle cannot be directly acquired only through the vision information, the embodiment provides a vehicle load identification system integrating a multilevel Monte Carlo model and the computer vision technology.
Firstly, realizing large-view and long-term actual traffic data acquisition based on a vehicle detection and tracking method, comprising the following steps: vehicle type, vehicle speed, vehicle body size, vehicle running track, running direction, time and other information.
Secondly, a vehicle load inference method based on a multilevel Monte Carlo model is provided to solve the problem that the vehicle load cannot be directly acquired in visual information.
And thirdly, the full collection of the actual traffic flow load data of the bridge deck is realized by fusing the vehicle tracking result with a multi-level Monte Carlo model obtained based on WIM data training.
Specifically, referring to fig. 1, a vehicle load prediction system includes:
the vision acquisition module is used for acquiring vehicle video data through the road monitoring system and processing the data to acquire determined vehicle load characteristics,
the vehicle load prediction module is used for executing a preset vehicle load inference method, and the vehicle load inference method comprises the steps of loading the determined vehicle load characteristics into a pre-established and trained multilayer Monte Carlo model, and comparing the determined vehicle load characteristics with all preset vehicle load characteristics by the multilayer Monte Carlo model to obtain uncertain vehicle load characteristics; obtaining the prediction results of all vehicle load characteristics according to the pre-trained random vector corresponding to the uncertain vehicle load characteristics and the confirmed vehicle load characteristics;
the multilevel Monte Carlo model comprises a plurality of layers of nodes, each layer of nodes respectively represents a prediction layer of vehicle load characteristics, each node represents a distribution model of vehicle load characteristic vectors on corresponding dimensions under a certain condition, and the vehicle load characteristic vectors are vectors formed by all the vehicle load characteristics;
a model training module for training a multi-level Monte Carlo model, the training process comprising: and constructing a sample set of the vehicle load characteristic vector, loading the sample set into the multilevel Monte Carlo model, and training to obtain a training value of a random vector corresponding to each layer of nodes in the multilevel Monte Carlo model.
The vehicle load recognition system of the present embodiment is described in detail below.
1. Monte Carlo method
The Monte Carlo method (Monte Carlo method), also called computer random simulation method, is a statistical test method guided by probability statistical theory, which is proposed with the development of electronic computer science and technology. The basic idea is that when the solution is the probability of an event or the expectation of a random variable, they can be used as the solution to the problem by some "trial" method to obtain the probability of the event occurring or the average value of the random variable. The core theoretical basis for sampling samples by adopting the Monte Carlo method is the law of large numbers. The law of large numbers is a law describing the results of a considerable number of repeated tests, according to which it is known that the greater the number of samples, the closer to the true value on average. On the contrary, if the probability distribution function of a certain sample is known, a large number of samples conforming to the distribution can be constructed by means of random sampling, and such sampling method is called monte carlo simulation, and the using method is generally divided into three steps:
1.1 Construct and describe probabilistic processes
For some variable that inherently has random characteristics, for example: vehicle load, vehicle position, etc. The probability problem can be correctly described and modeled by constructing the probability of occurrence of each type of sample by counting certain samples.
1.2 To achieve sampling from a known probability distribution
After a probability model representing a problem is obtained, since the model is composed of various probability distributions, generating random variables (or random vectors) of known probability distributions becomes a basic means for simulation by the monte carlo method. The most basic, and important one of these, is to generate pseudo-random numbers from a uniform distribution (or rectangular distribution) where the probability distribution is (0, 1), while with the aid of pseudo-random numbers, random sampling can be done from a known distribution.
1.3 To establish various estimators
Generally, after random sampling is performed by a structured probabilistic model, a value of a random variable may be determined as a solution to a desired problem, and such a solution is called unbiased estimation.
2. Characteristic vector of vehicle load
The vehicle load has a highly random characteristic, and therefore various characteristics thereof express a random process, and the present embodiment refers to a vector composed of these random processes as a vehicle load characteristic vector. Generally, the vehicle load-related characteristics include: vehicle direction of travel, vehicle type, vehicle size, vehicle gross weight, vehicle axle combination pattern, axle type, axle weight, and the like.
The vehicle axle combination mode is a mode in which axles of a certain vehicle are aligned and combined. Specifically, common axle types include several types as shown in table 1. Therefore, a combination of vehicle axles of a certain vehicle can be expressed in terms of a permutation combination of these common vehicle axle types. For example: the axle combination mode of the common car is I1I1; the axle combination mode for a six-axle truck can be I1I22I32;
TABLE 0 common vehicle axle types
In practical applications, the specific configuration of the vehicle load feature vector may be defined according to a data format collected by the WIM system.
3. Multilevel Monte Carlo model
Referring to fig. 2, according to the monte carlo theory, resampling to the known distribution can obtain the value of a certain random variable. For sampling the vehicle load feature vector, some distribution model with correlation of each dimension needs to be constructed, and for this purpose, this embodiment proposes a method called "Multi-Layer Monte Carlo model" (MLMC).
For a certain vehicle load characteristic vector X = { X = 0 ,x 1 ,...,x n The calculation formula is as follows:
ξ 0 ＝F 0 (x 0 )
in the formula, xi 0 Is the first random vector and is the second random vector,
is a joint probability distribution function of j node on the ith layer in a multilayer Monte Carlo model i To satisfy (0, 1)]Uniformly distributed random numbers, X being a random vector, X n Is the n-thA vehicle load characteristic.
For the vehicle load feature vector, as can be seen from fig. 1, the multi-level monte carlo model is a flexible multi-level network structure. And each node in the network represents a distribution model of the random variable on the corresponding dimension under a certain condition of the vehicle load characteristic vector. These distribution models can be obtained by learning from measured data. After the model is trained, the sampling of the vehicle load characteristics can be realized by introducing random vectors composed of random numbers with the same size. It should be noted here that such a model may also take some certain variable as input, and only sample other variables that are uncertain, which is advantageous for fusing computer vision vehicle detection and tracking results, and providing a basis for achieving the correlation of visual information and load information.
4. Multilevel Monte Carlo model training and validation
In order to verify that the multilevel monte carlo model method proposed in this section fuses the computer vision processing result and the vehicle load information obtained based on the WIM data, the section introduces the field test developed in a certain bridge.
4.1 overhead bridge floor road monitoring
In the experiment, all-weather bridge deck pavement monitoring videos recorded in high-altitude overhead shots in 2017, 10 month, 27 to 2017, 11 month and 1 day are selected as video data. A set of perspective imaging computer vision system with a sampling frequency of 25 frames per second is deployed at the midpoint of the beam in the pylon. In addition, in order to expand the shooting field of vision as much as possible, the camera is vertically installed according to the linear structural characteristics of the bridge, namely, the long edge of the image is parallel to the road surface in the direction along the bridge.
It should be noted that such a monitoring view will inevitably produce oblique projection distortions of the image. In order to facilitate subsequent vehicle detection and algorithm verification, the monitoring images are uniformly corrected by adopting projection transformation according to the bridge design drawing and the road surface marking object.
In the test, 2016 dynamic weighing data from a bridge toll station is adopted as a vehicle load data source for defining vehicle load characteristics and training a multi-level Monte Carlo model. According to the obtained WIM data format, the experimental vehicle load characteristic vector is defined as follows:
vehicle grade, vehicle direction of travel, axle split mode, ith axle weight, \8230
Wherein the vehicle grade is the vehicle grade specified in the national highway toll Standard of China. For passenger vehicles, there are four classes of passenger 1, passenger 2, passenger 3, and passenger 4, based on their rated seating positions. The cargo vehicles are classified into cargo 1, cargo 2, cargo 3, cargo 4, and cargo 5, in total, into five grades according to their load capacities.
TABLE 2 vehicle size and vehicle grade mapping relationship Table
4.2 vehicle load simulation and fusion
In 2016, 2.4 million weighing data samples were recorded at toll stations in a bridge and used to train a Monte Carlo model containing 17 layers and 193509 nodes. Based on the results of vehicle detection and tracking, the model is used to infer load estimates for each vehicle observed.
To verify the accuracy of the method to vehicle load estimation, fig. 3 and 4 compare the gross vehicle weight distribution for different axle count vehicles and different vehicle classes based on MLMC simulation and WIM data statistics. In particular, as can be seen from fig. 3, the gross vehicle weight distributions of three-axle, four-axle and five-axle vehicles are almost uniform, while the average value of the gross vehicle weight distribution of two-axle vehicles based on MLMC is significantly smaller than that of two-axle vehicles based on WIM, mainly because the video recording time is exactly the weekend and the number of cars passing through the deck is significantly higher than the annual average value. All curves in fig. 4 are fitted well, which fully explains that the method proposed in this chapter can well learn various random characteristics of bridge deck vehicle load, and combines computer visual information to realize reliable vehicle load inference.
Furthermore, to verify the robustness of the method, fig. 5 makes statistics on the vehicle load at five different points in the field of view, comparing the difference in the vehicle gross weight distribution in three load inferences. Therefore, the results of the three load inferences are basically consistent, which shows that the method has strong robustness for load characteristic inference. That is to say, the multilevel monte carlo model trained by a large amount of data can be used as a reliable vehicle load characteristic simulator, and the fusion of vehicle load data and visual signals can be well realized statistically under the condition that the vehicle detection and tracking result based on computer vision is used as an input parameter, so that the large-field acquisition of the actual vehicle load is realized. Therefore, a computer vision vehicle detection and tracking technology integrating multilevel Monte Carlo simulation can be used for observing and constructing an actual vehicle load model required by bridge maintenance in an operation period.
5. Space-time distribution characteristic of vehicle load of certain bridge
5.1 bridge floor traffic flow model
Based on actual vehicle loads collected from 10/month 27 in 2017 to 11/month 1 in 2017, time distribution of traffic of each vehicle type of a certain bridge in an observation period is constructed in fig. 7. As can be seen from the figure, the traffic flow of a certain bridge mainly comprises cars and trucks, the average daily total traffic flow is 7.5 million functions, and the peak phenomenon of the passenger flow of friday and sunday is obvious. Fig. 8 counts the distribution of various types of vehicles per hour on a single day. As can be seen, the daily car occurs primarily from 8 am to 19 pm and there are two peak periods of approximately 4000 volume flows per hour at 9 am and 17 pm. In comparison, the average flow of a truck is relatively uniform at each moment, but slightly higher at night than during the day.
Furthermore, by comparing the various levels of vehicle daily distributions established by the WIM data and the video data, it can be seen that the overall regularity of the annual daily distribution is consistent with the days of both the car and the heavy vehicle during the observation period, but the afternoon peak hours of heavy vehicle are not apparent compared to the annual average level because the weekend is included in the observation period. Fig. 6 compares the average daily traffic distribution for passenger 1 and cargo 5 vehicles.
5.2 bridge floor speed model
The statistical analysis of the flow velocity distribution of the bridge deck vehicles is very important for bridge management, and the smooth traffic condition of the bridge deck can be reflected by analyzing the overall flow velocity of the traffic flow. Fig. 9 compares the vehicle speed distribution at each level for passenger and logistics. As can be seen, the larger the vehicle load, the lower its average vehicle speed. On average, the average speed of class 1 passenger vehicles is about 80 km/h, which is related to the speed limit of a certain bridge. The average speed of large passenger vehicles and trucks is centered around 65 km/h, especially the heaviest class 5 vehicles, which are the slowest due to the requirements of transportation safety, 57.5 km/h.
5.3, spatial distribution model
The associated spatial distribution is obtained by converting the vehicle travel path into a wheel path and applying it to the deck. Fig. 10 shows a statistical frequency spectrum of the running vehicles acting on the bridge deck. Fig. 11 is a statistical map of the vehicle average load action space based on the wheel load track. The higher the color temperature, the higher the frequency and vice versa. Statistically, most vehicles travel in the inner lanes (S-N L3 and N-S L3), with the major contribution coming from the cars. In addition, the position where the vehicle load acts on the deck is concentrated near the lane line, and the larger load is distributed more to the outer lane, which indicates that trucks and buses prefer to travel on the middle and outer lanes.
Based on the spatial distribution spectrum, a refined transverse distribution model of any cross section can be easily extracted. Taking the cross section of x =960 as an example, fig. 12 shows the transverse distribution of the wheel action frequency on this section and identifies a high deck footprint, concentrated in the range of the inner 3 to 12U ribs. Figure 13 shows the lateral distribution of the average wheel load strength across the section and identifies the vulnerable areas of the deck and steel box girder roof, concentrated in the outboard range of the 14 to 24U ribs, with the 20 to 24 areas being most prone to fatigue problems.
5.4 vehicle load Spectrum
According to the obtained bridge deck vehicle load course data, the load on any point of the bridge deck can be counted to construct a vehicle load spectrum model. Fig. 14 shows the load spectrum at points 24, 64, 100, 137, 158, 196, 234, 274 across the x =960 cross-section, and it is clear that points closer to the outside are subjected to greater vehicle loads, that is, these locations are more susceptible to fatigue cracking and pavement distress.
6. Summary of the invention
The embodiment provides a vehicle load inference method based on a multilevel Monte Carlo model aiming at the problems that the visual information can not directly obtain the vehicle load and the vehicle load information is difficult to obtain in the construction of a bridge vehicle load distribution model, and researches show that:
6.1 A multilevel Monte Carlo model constructed based on the Monte Carlo principle can effectively learn the random characteristics of the vehicle load.
6.2 The multi-level Monte Carlo model vehicle load inference method for simulating the vehicle load characteristic vector has higher robustness, and can effectively solve the problem that the vehicle load cannot be directly acquired in the visual information statistically.
6.3 The real bridge application with the river-yin bridge as the background proves that the multilevel Monte Carlo vehicle load inference method which integrates the computer vision vehicle identification and tracking results is used for constructing various space-time distribution models and refined vehicle load spectrum models. The related research results can be directly applied to the management and maintenance of the bridge in the operation period, and have important significance for further bridge performance evaluation and fatigue performance research.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations can be devised by those skilled in the art in light of the above teachings. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.
Claims (6)
1. A vehicle load prediction method characterized by comprising the steps of:
the vehicle video data is collected by the road monitoring system, and the data is processed to obtain the determined vehicle load characteristics,
loading the determined vehicle load characteristics into a pre-established and trained multilayer Monte Carlo model, and comparing the determined vehicle load characteristics with all preset vehicle load characteristics by the multilayer Monte Carlo model to obtain uncertain vehicle load characteristics; obtaining the prediction results of all vehicle load characteristics according to the pre-trained random vector corresponding to the uncertain vehicle load characteristics and the confirmed vehicle load characteristics;
the multilevel Monte Carlo model comprises a plurality of layers of nodes, each layer of nodes respectively represents a prediction layer of vehicle load characteristics, each node represents a distribution model of vehicle load characteristic vectors on corresponding dimensions under a certain condition, and the vehicle load characteristic vectors are vectors formed by all the vehicle load characteristics;
the training of the multi-level Monte Carlo model comprises the following steps: constructing a sample set of vehicle load characteristic vectors, loading the sample set into a multilevel Monte Carlo model, and training to obtain a training value of a random vector corresponding to each layer of nodes in the multilevel Monte Carlo model;
the expression of the multilevel Monte Carlo model is as follows:
ξ 0 ＝F 0 (x 0 )
X＝{x 0 ，x 1 ，...，x n }
in the formula, xi 0 Is the first random vector, and the second random vector,
is a joint probability distribution function of j node on the ith layer in a multilayer Monte Carlo model i To satisfy (0, 1)]Uniformly distributed random numbers, X is a random vector, X n Is the nth vehicle load characteristic;
the vehicle load characteristics include vehicle heading, vehicle category, vehicle size, vehicle gross weight, vehicle axle combination pattern, axle type, and axle weight.
2. The vehicle load prediction method according to claim 1, characterized in that dynamic weighing data is obtained through a dynamic weighing system, and the vehicle load feature vector sample set is constructed in combination with vehicle load features obtained by a road monitoring system, and is used for training a multi-level Monte Carlo model.
3. A vehicle load prediction method according to claim 1, characterized in that the vehicle axle combination pattern is represented by coding.
4. A vehicle load recognition system, comprising:
the vision acquisition module is used for acquiring vehicle video data through the road monitoring system and processing the data to acquire the determined vehicle load characteristics and vehicle tracking results,
the vehicle load prediction module is used for loading the determined vehicle load characteristics into a pre-established and trained multi-level Monte Carlo model, and the multi-level Monte Carlo model compares the determined vehicle load characteristics with all preset vehicle load characteristics to obtain uncertain vehicle load characteristics; obtaining the prediction results of all vehicle load characteristics according to the pre-trained random vector corresponding to the uncertain vehicle load characteristics and the confirmed vehicle load characteristics;
the multilevel Monte Carlo model comprises a plurality of layers of nodes, each layer of nodes respectively represents a prediction layer of vehicle load characteristics, each node represents a distribution model of vehicle load characteristic vectors on corresponding dimensions under a certain condition, and the vehicle load characteristic vectors are vectors formed by all the vehicle load characteristics;
the vehicle load space-time distribution identification module is used for combining a vehicle tracking result and prediction results of all vehicle load characteristics to construct a vehicle load space-time distribution model;
the model training module is used for training a multi-level Monte Carlo model, and the training process comprises the following steps: constructing a sample set of vehicle load characteristic vectors, loading the sample set into a multilevel Monte Carlo model, and training to obtain a training value of a random vector corresponding to each layer of nodes in the multilevel Monte Carlo model;
the expression of the multilevel Monte Carlo model is as follows:
ξ 0 ＝F 0 (x 0 )
X＝{x 0 ，x 1 ，...，x n }
in the formula, xi 0 Is the first random vector and is the second random vector,
is a joint probability distribution function of the j node on the ith layer in the multi-layer Monte Carlo model i To satisfy (0,1)]Uniformly distributed random numbers, X being a random vector, X n Is the nth vehicle load characteristic;
the vehicle load characteristics include vehicle heading, vehicle category, vehicle size, vehicle gross weight, vehicle axle combination pattern, axle type, and axle weight.
5. The vehicle load recognition system according to claim 4, wherein dynamic weighing data is obtained through a dynamic weighing system, and the vehicle load feature vector sample set is constructed by combining vehicle load features obtained by a road monitoring system, and is used for training a multi-level Monte Carlo model.
6. A vehicle load identification system according to claim 4 wherein the vehicle axle combination pattern is represented by a code.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110378341.2A CN113312744B (en) | 2021-04-08 | 2021-04-08 | Vehicle load prediction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110378341.2A CN113312744B (en) | 2021-04-08 | 2021-04-08 | Vehicle load prediction method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113312744A CN113312744A (en) | 2021-08-27 |
| CN113312744B true CN113312744B (en) | 2022-11-18 |
Family
ID=77371923
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110378341.2A Active CN113312744B (en) | 2021-04-08 | 2021-04-08 | Vehicle load prediction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113312744B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104732013A (en) * | 2015-02-12 | 2015-06-24 | 长安大学 | Method for recognizing load of single vehicle passing through multi-girder type bridge |
| CN104933284A (en) * | 2015-02-12 | 2015-09-23 | 长安大学 | Random traffic flow simulation method in road and bridge based on measured data |
| CN109002622A (en) * | 2018-07-26 | 2018-12-14 | 广州大学 | A kind of random lower Large Span Bridges totality load response evaluation method of wagon flow effect |
| CN109167956A (en) * | 2018-05-21 | 2019-01-08 | 同济大学 | The full-bridge face traveling load spatial distribution merged based on dynamic weighing and more video informations monitors system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110232216A (en) * | 2019-05-15 | 2019-09-13 | 南京金蓝智慧城市规划设计有限公司 | Bituminous pavement failure analysis method based on multi-level fuzzy judgment |
| CN112347668B (en) * | 2020-09-29 | 2022-04-12 | 华东交通大学 | Steel bridge deck fatigue reliability assessment method based on probabilistic fracture mechanics |
-
2021
- 2021-04-08 CN CN202110378341.2A patent/CN113312744B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104732013A (en) * | 2015-02-12 | 2015-06-24 | 长安大学 | Method for recognizing load of single vehicle passing through multi-girder type bridge |
| CN104933284A (en) * | 2015-02-12 | 2015-09-23 | 长安大学 | Random traffic flow simulation method in road and bridge based on measured data |
| CN109167956A (en) * | 2018-05-21 | 2019-01-08 | 同济大学 | The full-bridge face traveling load spatial distribution merged based on dynamic weighing and more video informations monitors system |
| CN109002622A (en) * | 2018-07-26 | 2018-12-14 | 广州大学 | A kind of random lower Large Span Bridges totality load response evaluation method of wagon flow effect |
Non-Patent Citations (4)
| Title |
|---|
| Identification of moving loads based on the information fusion of weigh-in-motion system and multiple camera machine vision;Dan DH , Ge LF , Yan XF;《Measurement》;20190517;全文 * |
| 基于WIM数据的多车道随机车流模拟及汽车荷载效应;吕硕;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20160215;全文 * |
| 基于多层蒙特卡罗的火箭结构动力学不确定性分析;袁赫, 刘莉, 康杰;《弹箭与制导学报》;20190630;第3章 * |
| 基于实桥调查的交通荷载极限作用效应分析;杨帆, 徐俊, 袁帅;《结构工程师》;20161231;第1-2章 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113312744A (en) | 2021-08-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109448369B (en) | Real-time operation risk calculation method for expressway | |
| Lin et al. | A review of travel-time prediction in transport and logistics | |
| CN105869402B (en) | Express highway section speed modification method based on polymorphic type floating car data | |
| CN104850676B (en) | A kind of random traffic flow simulation analogy method of highway bridge | |
| CN106250613A (en) | A kind of wheel service state security domain is estimated and method for diagnosing faults | |
| CN104933284B (en) | The random wagon flow analogy method of a kind of highway bridge based on measured data | |
| CN105741553A (en) | Method for identifying parking road segment in vehicle track based on dynamic threshold | |
| CN111063204B (en) | Expressway vehicle speed prediction model training method based on toll station flow | |
| CN107957259A (en) | Wheelmark cross direction profiles measuring system and measuring method | |
| Jeng et al. | Real-time vehicle classification using inductive loop signature data | |
| Bridgelall | Characterizing ride quality with a composite roughness index | |
| CN113312744B (en) | Vehicle load prediction method and system | |
| Lee | Freeway travel time forecast using artifical neural networks with cluster method | |
| Zhang et al. | Multi-sensor graph transfer network for health assessment of high-speed rail suspension systems | |
| CN114023065A (en) | Algorithm for intelligently diagnosing intersection service level by utilizing video analysis data | |
| CN116092037A (en) | Vehicle type identification method integrating track space-semantic features | |
| CN116597642A (en) | Traffic jam condition prediction method and system | |
| Ghanim et al. | An artificial intelligence approach to estimate travel time along public transportation bus lines | |
| CN115240111A (en) | Real-time back calculation method for traffic load of medium and small-span bridge | |
| CN114446031A (en) | Multi-device and multi-dimensional data fusion analysis-based inspection station management method and system | |
| CN114663992A (en) | Multi-source data fusion expressway portal positioning method | |
| Wei et al. | Applying data fusion techniques to traveler information services in highway network | |
| CN110956808A (en) | Heavy truck traffic flow prediction method based on non-full-sample positioning data | |
| CN107085074B (en) | A method of classification monitoring motor-vehicle tail-gas | |
| Zhou et al. | A hybrid virtual–real traffic simulation approach to reproducing the spatiotemporal distribution of bridge loads |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |