A method and system for determining regional-scale axial load spectrum based on multi-level verification and classification

Abstract

This invention discloses a method and system for measuring regional-scale axle load spectra based on multi-level verification and classification. The method includes: collecting axle load data and traffic survey data from multiple stations; performing data integrity, axle load parameter logic, and cross-station consistency checks on the collected data, and selecting station combinations that meet the conditions for spectrum calculation; based on this, identifying the main vehicle types within the region, extracting effective data, and calculating the direction coefficient, lane coefficient, vehicle type distribution coefficient, and axle load distribution coefficient, and plotting the axle load spectrum; further, based on historical measurement data and traffic operation characteristics, predicting the axle load spectrum change trend for the next period. This invention effectively avoids the shortcomings of existing technologies, such as low classification accuracy and strong subjectivity due to fuzzy classification relying on diagrams, and the problem of deviations caused by single-port data being easily affected by equipment errors and environmental interference, thus improving the reliability and accuracy of axle load spectra. It is applicable to various engineering scenarios such as road structure analysis and load simulation.

Description

Technical Field

[0001] This invention relates to traffic data analysis technology, and more particularly to a method and system for measuring regional-scale axle load spectrum based on multi-level verification classification, belonging to the field of vehicle axle load data measurement and analysis. Background Technology

[0002] With the continuous improvement of the digitalization and intelligence of transportation infrastructure, a comprehensive monitoring system has gradually been formed in the highway transportation sector, consisting of various equipment such as dynamic weighing systems, bridge health monitoring systems, and overload detection stations. The widespread deployment of these systems makes it possible to collect vehicle axle load data (such as the number of axles, wheelbase, and axle weight) at high frequency and with wide coverage, thus laying a data foundation for regional traffic load characteristic research and structural response analysis. Especially against the backdrop of continuously advancing the scientification of highway maintenance, bridge safety, and pavement design, axle load data has become an important input variable supporting the optimization of highway engineering design and the formulation of traffic policies.

[0003] Axle load spectrum, as a key indicator for quantitatively describing traffic load characteristics, can reveal the distribution patterns of vehicles with different axle types across various axle load ranges from a statistical perspective. This provides data support for equivalent design axle load conversion, pavement structure fatigue analysis, and load model optimization. Typical applications include pavement structure design, bridge load-bearing capacity assessment, and road service life prediction. However, the accuracy and representativeness of the axle load spectrum highly depend on the authenticity of the source data, the scientific validity of the classification logic, and the consistency of spatial scale, which places higher demands on the data processing workflow.

[0004] Currently, while domestic technical specifications such as the "Specifications for Design of Asphalt Pavement of Highways" (JTG D50-2017) and the "Dimensions, Axle Loads and Mass Limitations of Motor Vehicles, Trailers and Vehicle Combinations" (GB 1589-2016) provide a basic framework for vehicle type classification and load limits, they largely rely on graphic illustrations and empirical judgments for axle load data classification, lacking clear quantitative rules. Especially in the context of big data at the regional scale, traditional classification methods exhibit the following prominent problems: First, the classification standards are highly subjective and difficult to adapt to the increasingly complex vehicle types and diversified traffic loads; second, the classification accuracy is limited by the fuzzy boundaries of the illustrations, leading to large errors in subsequent axle load spectrum calculations; and third, a single data source is easily affected by external factors such as equipment accuracy errors, traffic flow disturbances, and changes in environmental conditions, thus affecting the reliability of the axle load spectrum.

[0005] Furthermore, the consistency of cross-site data is becoming increasingly prominent in practical applications. Differences in factors such as the deployment location of monitoring equipment, the setting of traffic flow intersections, traffic diversion situations, and data collection times lead to inconsistencies in axle load data across different stations within a region. This lack of spatial scale and varying data source reliability further exacerbates the technical difficulty and uncertainty of axle load spectrum mapping. Therefore, it is urgent to develop a quantitative axle load data classification scheme to provide technical support for improving the accuracy of axle load data processing and the efficiency of subsequent engineering applications. Summary of the Invention

[0006] To address the aforementioned technical issues, this invention provides a method and system for determining regional-scale axle load spectra based on multi-level verification and classification. By establishing a regional-scale axle load data processing technology system that combines quantitative classification capabilities, multi-source data fusion, and verification mechanisms, the reliability, representativeness, and engineering applicability of axle load spectrum data are improved, thereby providing more reliable technical support for the design, evaluation, and management of highway engineering.

[0007] Specifically, the technical solution provided by this invention is as follows: A method for determining regional-scale axial load spectra based on multi-level verification classification includes the following steps: S1. Data Acquisition: Collect axle load data from the dynamic weighing system at each station, including the number of axles, axle spacing, and axle load; collect natural traffic volume statistics for different traffic composition categories at each station, referred to as traffic survey data. S2, Multi-level verification classification: Perform data integrity verification, axle load parameter logical verification, and cross-site consistency verification on the collected parameters to obtain the site combination that conforms to the axle load spectrum calculation; S3. Determining the main vehicle models: Statistically determine the proportion of different vehicle models in the region, and select the models with a proportion greater than a set threshold as the main vehicle models in the region; S4. Parameter Calculation and Axle Load Spectrum Drawing: Based on the effective axle load data and traffic data of the station combination that conforms to the axle load spectrum calculation, as well as the main vehicle types in the area, calculate the main coefficients and draw the corresponding axle load spectrum; the main coefficients include the direction coefficient, lane coefficient, vehicle type distribution coefficient, axle load distribution coefficient, and equivalent design axle load conversion coefficient. The axle load spectrum is used to represent the distribution of each axle type in different axle load ranges.

[0008] Furthermore, the data integrity verification includes: verifying the time period, direction, and lane of axle load data and traffic control data to ensure continuous data collection; ensuring that the online rate of the collection equipment is controlled above a set threshold during the collection period; avoiding stations with dynamic weighing equipment only on some lanes; and avoiding areas with complex surrounding environments or special traffic control measures.

[0009] Furthermore, the axle load parameter logical verification is based on three dimensions: number of axles, wheelbase, and total weight. It establishes a precise correspondence between axle load data and vehicle axle load characteristics at the regional scale, achieving accurate and efficient correlation and matching. This follows a judgment logic of first coarse segmentation and then fine segmentation, including the following process: Axle count coarse classification: Obtain the number of vehicle axles through axle load data, classify them according to the number of axles, and remove abnormal data; Wheelbase subdivision: Establish a wheelbase feature threshold library for key axle types at the regional scale. Key axle types should meet the corresponding wheelbase feature threshold requirements. The key axle types include single axle dual wheels for two-axle vehicles, and dual front axle, dual-axle dual wheels, and triple-axle dual wheels for three-axle and above vehicles. Investigate the main vehicle models in the region and establish a vehicle model overview table at the regional scale, further distinguishing different vehicle models through axle type combinations. Establish a total wheelbase threshold library for vehicle models, where the sum of the wheelbases of all vehicle models does not exceed the corresponding total wheelbase threshold. Gross weight classification: Establish a minimum gross weight threshold library for each vehicle type. The sum of the axle weights of each vehicle type should be lower than the corresponding total axle weight threshold. For two-axle vehicles that meet the conditions, further classify them according to their gross weight and set a total weight threshold for two-axle vehicles to distinguish between large two-axle vehicles and small and medium-sized two-axle vehicles. Set the classification accuracy threshold according to the equipment precision of the dynamic weighing system, and count the amount of data to be removed and retained in the logical verification of axle load parameters. The proportion of retained data should not be less than the classification accuracy threshold; otherwise, reclassification verification should be performed.

[0010] Furthermore, the cross-site consistency verification includes the following process: Station combination classification: Based on the road network distribution, axle load data stations are combined with adjacent traffic control data stations to ensure that there are no large intersections for traffic diversion between stations. The station combinations are divided into three categories: no traffic diversion between national, provincial, county, and township roads; no traffic diversion between national and provincial roads; and no traffic diversion between county and township roads. Consistency check coefficient calculation: Calculate the consistency check coefficient for each site combination. The consistency check coefficient is defined as follows: , T i for i The frequency of vehicle passage is the source of axle load data. T 0 represents traffic volume in the traffic survey data; Threshold setting: Based on the traffic distribution of different site combination types, set the threshold for the consistency verification coefficient for different combination types; Threshold matching: Sites that meet the threshold requirements are classified as eligible for axle load spectrum calculation; site combinations that do not meet the threshold requirements are not included in this axle load spectrum calculation.

[0011] Preferably, the threshold settings in each step are determined based on the equipment accuracy and actual traffic conditions, and the station combinations, thresholds, and main vehicle types at the regional scale are updated annually based on the road network construction and traffic detection platform construction, and the axle load spectrum measurement results are issued quarterly.

[0012] Furthermore, the method also includes a next-cycle axle load spectrum prediction step, which is used to deduce the evolution of the axle load spectrum in the next cycle based on existing historical measurement results and traffic operation characteristics, and to obtain in advance the distribution change trend of each axle type in different axle load ranges.

[0013] Furthermore, the next cycle shaft load spectrum prediction step includes the following steps: S501, Construction of Predicted Input Feature Vector Let the current prediction period be t+1, and its corresponding prediction input feature vector be denoted as . X t Its definition is:

[0014] in, This indicates the current period t-th... i Axial type in the first j The spectral distribution coefficient of each axle load interval, i.e., the axle load spectrum obtained by the current period measurement; Corresponding to each shaft type, This represents the axle load interval number. k This represents the total number of axle load intervals. This is the current cycle traffic flow feature vector, used to reflect the potential impact of the current cycle traffic state on the axle load spectrum, and , This represents the average daily traffic flow for the current period. For the proportion of freight vehicles, This represents the proportion of large trucks among all trucks. The average operating speed of the vehicle; Given the current cycle environment and economic factors vector, and , The Holiday Index is the ratio of the average daily traffic flow during holidays to the average daily traffic flow on non-holiday days, reflecting the amplification factor of holidays on traffic peaks. This is a seasonal factor, equal to the ratio of the average quarterly traffic volume to the average annual traffic volume; For the region's quarterly GDP growth rate, This refers to the logistics and transportation index. S502, Construction of Hierarchical Time Series Prediction Model For each type of shaft iA prediction model is established based on historical axle load data to calculate the axle load spectrum in the next cycle t+1. The prediction formula is as follows:

[0015] in: Indicates the predicted first i Axial type in the first j The distribution coefficient of the axle load interval, i.e., the axle load spectrum value of the next period; Indicates the use of the first i Axle-type axial load spectrum prediction model; It contains the t-th period. i Axial type in the first j Distribution coefficient of axle load range That is, the axle load spectrum value for the current period; It contains the t-1th period i Axial type in the first j Distribution coefficient of axle load range That is, the axial load spectrum value of the previous period; It contains the first t - N +1 cycle i Axial type in the first j Distribution coefficient of axle load range , N To reference the total number of historical cycles, it means that only past data is considered when making predictions. N Historical axial load spectrum measurement data for each cycle.

[0016] Furthermore, the prediction model includes linear basic trend prediction, nonlinear dynamic learning, and dynamic fusion; Linear basic trend prediction: First, a seasonally adjusted autoregressive model is used to predict the long-term trend of historical data. The prediction formula is as follows:

[0017] in, The trend term is predicted based on a linear model, i.e., the... i Axial type in the first j Predicted trends within axle load ranges; , , , … These are autoregressive coefficients used to control for the influence of historical data on the predicted trend; This is a seasonality coefficient used to control the magnitude and direction of seasonal fluctuations; S This represents the total number of seasonal terms, used to indicate the number of periods in the time series; This is a periodic fluctuation term used to simulate the impact of seasonality on traffic flow; Nonlinear dynamic learning: Then, an LSTM model is used to capture the nonlinear dynamic characteristics of traffic load. The prediction formula is: , in, This represents the nonlinear prediction result output by the LSTM network, i.e., the predicted [number]th [unit]. i Axial type in the first j Nonlinear prediction values ​​for axle load range; LSTM ( ) represents an LSTM neural network model that uses a gating mechanism to process long-term dependencies in time-series data and outputs a prediction for the next cycle; Dynamic fusion: Finally, the linear basic trend prediction and the nonlinear prediction results are weighted and fused to obtain the final prediction result. The fusion formula is as follows: , in: λ i For the first i The fusion weighting coefficient for axial-type predictions, with a value range of [0,1], determines the contribution of linear and nonlinear predictions to the final result. λ i When the value approaches 1, the prediction relies more on linear trend prediction; conversely, it relies more on the nonlinear dynamic prediction of the LSTM model. The final fused axial load spectrum prediction value, i.e., the first... i Axial type in the first j Prediction results for axle load range.

[0018] A regional-scale axle load spectrum measurement system based on the above method includes a data acquisition module, a multi-level verification and classification module, a major vehicle type determination module, a parameter calculation and axle load spectrum plotting module, and an axle load spectrum prediction module. The data acquisition module is used to collect and obtain axle load data and cross-modulation data from each station; The multi-level verification and classification module is used to perform data integrity verification, axle load parameter logical verification, and cross-site consistency verification on the collected parameters, and to obtain the site combination that conforms to the axle load spectrum calculation. The main vehicle model determination module is used to count the proportion of different vehicle models in a region and select the vehicle models with a proportion greater than a set threshold as the main vehicle models in the region. The parameter calculation and axle load spectrum drawing module is used to calculate the main coefficients and draw the corresponding axle load spectrum based on the effective axle load data and traffic data of the station combination that conforms to the axle load spectrum calculation, as well as the main vehicle types in the area. The axle load spectrum prediction module is used to predict the evolution of the axle load spectrum in the next cycle based on existing historical measurement results and traffic operation characteristics, and to obtain the distribution trend of each axle type in different axle load ranges.

[0019] An electronic device includes a memory, a processor, and a computer program stored in the memory and running on the memory. When the processor executes the program, it implements the steps of the regional-scale axial load spectrum determination method as described above.

[0020] Compared with the prior art, the beneficial effects of the present invention are: 1. To address the lack of precise criteria for classifying vehicle axle load data at the regional scale, this invention constructs a quantitative standard for axle load data classification. This effectively avoids the shortcomings of existing technologies that rely on fuzzy classification based on diagrams, such as low classification accuracy and strong subjectivity. Furthermore, the threshold settings in this standard are highly compatible with regional traffic characteristics and data acquisition equipment parameters. Technicians can easily obtain and calibrate the thresholds based on the actual working conditions of the region, significantly improving the practicality and operability of the technical solution and ensuring its stable application in axle load data processing scenarios in different regions.

[0021] 2. To address the issue that single-port data in existing traffic data collection processes is susceptible to deviations due to equipment errors and environmental interference, this invention innovatively establishes a data cross-verification process between the vehicle load monitoring port and the traffic flow monitoring port. Through cross-verification of data from the two types of ports, and the identification and correction of abnormal data, the impact of single-channel data deviation on subsequent engineering applications is effectively reduced, thereby improving the reliability and accuracy of the axle load spectrum.

[0022] 3. This invention innovatively integrates the axle load spectrum prediction process, supporting the extrapolation of spectrum change trends for the next cycle (quarterly or annually) based on historical measurement data and traffic operation characteristics. This dynamic prediction mechanism, based on time-series data, traffic modulation characteristics, and the evolution of major vehicle types, breaks the limitation of existing technologies that the results are merely the status quo. It achieves a leap from static measurement to dynamic prediction, providing forward-looking input conditions for bridge design, road service life assessment, traffic load simulation, and maintenance strategy formulation, thereby improving the scientific nature and adaptability of infrastructure decision-making.

[0023] 4. This invention, in collaboration with independently developed software, achieves efficient and rapid processing of massive axle load data. At the same time, it clarifies the control standards and operational boundaries of key parameters for technical personnel, ensuring the accuracy of results while significantly improving data processing efficiency, ultimately achieving the dual optimization goals of efficiency and accuracy. Attached Figure Description

[0024] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0025] Figure 1 This is a schematic diagram of the framework of a regional-scale axial load spectrum determination method provided in an embodiment of the present invention; Figure 2 This is an axle load spectrum diagram of a single axle and single wheel provided in an embodiment of the present invention; Figure 3 This is an axle load spectrum diagram of a single-axle dual-wheel assembly provided in an embodiment of the present invention; Figure 4 This is an axial load spectrum of a twin-axis provided in an embodiment of the present invention; Figure 5 This is an axial load spectrum of a triplex shaft provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of software calculation of equivalent design axle load conversion factor provided in an embodiment of the present invention. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative effort are all within the scope of protection of the present invention.

[0027] Example 1 This embodiment provides a method for determining regional-scale axial load spectra based on multi-level verification and classification, such as... Figure 1 As shown, the method mainly includes the following steps: (a) Data collection Data is collected from dynamic weighing systems at sites such as vehicle load monitoring ports, bridge health monitoring systems, and overload detection stations. This mainly includes axle load-related data such as the number of axles, wheelbase, and axle load (hereinafter referred to as "axle load data"). The parameters of the axle load data should include: time (year, month, day), driving direction, lane, total weight, number of axles, axle type, axle load 1, axle load 2, axle load 3, axle load 4, axle load 5, axle load 6, axle load 7, axle load 8, wheelbase 1, wheelbase 2, wheelbase 3, wheelbase 4, wheelbase 5, wheelbase 6, and wheelbase 7.

[0028] Collect natural number statistics of traffic volume for different traffic composition categories from traffic survey points at the traffic flow operation monitoring port (hereinafter referred to as "traffic survey data"). The parameters of the traffic survey data should include: observation station number, observation station name, time (year, month, day), direction of travel, observation mileage, starting station number, ending station number, traffic volume of small and medium-sized passenger vehicles, traffic volume of large passenger vehicles, traffic volume of small trucks, traffic volume of medium-sized trucks, traffic volume of large trucks, traffic volume of extra-large trucks, and traffic volume of containers.

[0029] (ii) Multi-level verification classification 1. First-level verification: Data integrity verification Verify the time period, direction, and lane of axle load data and traffic control data to ensure continuous data collection. The online rate of equipment should be controlled above 95% during the collection period. Avoid stations with dynamic weighing equipment only on some lanes, and avoid areas with complex surrounding environments (such as construction areas, school entrances, large intersections) or special traffic control measures (such as one-way streets, height and weight restrictions, and time-based traffic restrictions for different vehicle types).

[0030] 2. Second-level verification: Logical verification of axle load parameters Using three dimensions—number of axles, wheelbase, and gross weight—a precise correspondence is established between axle load data and vehicle axle load characteristics at the regional scale, achieving accurate and efficient correlation and matching. Following a logic of first coarsely segmenting and then finely segmenting, the main process includes: (1) Coarse classification of axle count. The number of axles of the vehicle is obtained through axle load data and classified according to the number of axles. Data with 1 axle count is listed as abnormal data and removed.

[0031] (2) Wheelbase subdivision. Establish a wheelbase feature threshold library for key axle types at the regional scale. The main axle types should meet the corresponding wheelbase feature threshold requirements. Key axle types include: single axle dual wheels for two-axle vehicles, dual front axles (single tire on each side) for three-axle and above vehicles, dual axle dual wheels, and triple axle dual wheels.

[0032] The wheelbase feature threshold library for critical axle types includes the items listed in the table below. The thresholds can be set statistically based on the design parameters of common vehicles in the region, or the values ​​recommended in this embodiment can be used.

[0033]

[0034] Based on the main vehicle types within the survey area and in conjunction with the main vehicle types in the "Specifications for Design of Asphalt Pavement of Highways" (JTG D50-2017), a vehicle type overview table at the regional scale was established, further distinguishing different vehicle types by axle type combinations. The axle type combination judgment criteria are shown in the table below (taking Class 9, Type 157 and Class 6, Type 115 as examples):

[0035] Establish a total wheelbase threshold library. The sum of the wheelbases of each vehicle model shall not exceed the corresponding total wheelbase threshold. The total wheelbase threshold is set with reference to the maximum design length of the vehicle. Data that does not meet the conditions shall be removed.

[0036] Taking the axle load spectrum measurement of ordinary national and provincial highways in Yangzhou City, Jiangsu Province as an example, this study investigated common trucks in the Yangzhou area and, with reference to the vehicle types covered in the "Specifications for Design of Asphalt Pavement of Highways" (JTG D50-2017), classified them based on information such as gross weight and wheelbase. The specific classification criteria are as follows: A list of vehicle models at a regional scale

[0037] Total Axis Spacing Threshold Library

[0038] (3) Total weight verification. Establish a total weight threshold library. If the sum of the axle weights of each vehicle model is lower than the corresponding total axle weight threshold, the total axle weight threshold is set with reference to the vehicle's unloaded weight. Data that does not meet the conditions is removed. For two-axle vehicles that meet the conditions, further classify them according to their total weight and set a total weight threshold for two-axle vehicles to distinguish between large two-axle vehicles and small and medium-sized two-axle vehicles.

[0039] Total weight threshold library

[0040] (4) Based on the accuracy of the dynamic weighing system equipment, set a classification accuracy threshold (95%), and calculate the amount of data removed and retained in the second-level verification process. The proportion of retained data should not be less than the classification accuracy threshold; otherwise, return to the axle count coarse classification process for reclassification and verification. For example, the amount of data retained after a certain classification of stations on ordinary national and provincial highways in Yangzhou City is as follows:

[0041] 3. Third-level verification: Cross-site consistency verification The consistency verification coefficient-threshold matching method was used to verify the axle load data and cross-site data collected at the regional scale. Based on the cross-site consistency verification results, the stations that can be used for axle load spectrum calculation were selected.

[0042] Define the consistency test coefficient: , In the formula T i for i The frequency of vehicle passages, which is the source of axle load data, will be used to determine the axle load data from overload stations. i Defined as 1, the axle load data of the bridge health monitoring system is defined as 2; T 0 represents the traffic volume in the traffic survey data.

[0043] The main verification process is as follows: (1) Station combination division. Based on the road network distribution, axle load data stations are combined with adjacent traffic control data stations to ensure that there are no large intersections for traffic diversion between stations.

[0044] (2) Calculation of consistency verification coefficient. Calculate the consistency verification coefficient for each site combination.

[0045] (3) Threshold setting. Based on the equipment accuracy and the traffic distribution of different site combination types, set the threshold for the consistency verification coefficient for different combination types.

[0046] (4) Threshold matching. Sites that meet the threshold requirements are classified as "site combinations that meet the axle load spectrum calculation requirements"; site combinations that do not meet the threshold requirements are not included in the axle load spectrum calculation.

[0047] The stations are divided and a consistency verification coefficient is calculated. The consistency verification coefficient is then graded and evaluated as follows: less than 10% is excellent, 10%~20% is good, 20%~30% is average, and greater than 30% is poor. General stations without large intersections for traffic diversion should be "excellent", and the remaining stations should meet the "good" standard.

[0048] For example, the consistency verification coefficients and evaluations of various sites within Yangzhou are as follows:

[0049] The site combination that meets the requirements for axle load spectrum calculation is:

[0050] In the third-level verification, horizontal verification does not distinguish between time periods when compiling data, and vertical verification does not distinguish between vehicle models when compiling data. j "Partial" mainly refers to the vehicles shown in the table below that can represent the main characteristics of mixed traffic flow and can be distinguished by certain classification methods in both traffic control data and axle load data:

[0051] (III) Determination of Main Vehicle Types in the Region The percentage of different car models was calculated on an annual basis, and models with a percentage greater than 0.1% were selected as the main car models.

[0052] Based on the axle load data analysis, the distribution coefficients of various vehicle types at different stations in Yangzhou City were statistically analyzed. (1) The main vehicle types on the road section are Class 2 / 3 type 12 vehicles, Class 6 type 112 / 115 vehicles, and Class 9 type 157 vehicles, with the sum of these four types of vehicles accounting for more than 92% on average; (2) The vehicle types 17, 117, 1122, 1155, 1157 and 1522 in the Class 5 of the Highway Asphalt Pavement Design Specification (JTG D50-2017), as well as the vehicle types 152 and 11522 that are not in the JTG D50-2017 specification, have a small distribution coefficient and generally account for less than 0.1%.

[0053] The main vehicle types were selected from those accounting for more than 0.1% of the total, and the main vehicle types in the city are shown in the table below: List of Main Vehicle Types on National and Provincial Highways in Yangzhou

[0054] Note: The selection of major vehicle models should be updated annually based on the vehicle model distribution.

[0055] (iv) Calculation of key coefficients and plotting of axle load spectrum Calculate the direction coefficient, lane coefficient, vehicle type distribution coefficient, axle load distribution coefficient, and equivalent design axle load conversion coefficient. Based on the axle load distribution coefficient, plot the axle load spectra for single-axle single-wheel, single-axle dual-wheel, dual-axle, and triple-axle configurations. For example... Figure 6 The figure shows the calculation process of the equivalent design axle load conversion factor using calculation software.

[0056] Taking the G233 train route (downstream, chainage range 1579~1601) as an example, the data table to be calculated is selected, and the results are output in chart form. The main calculation results are as follows: (1) Calculation of traffic volume parameters ① Directional coefficient: According to the observation data of the traffic survey station, the directional distribution coefficients of the G233 traffic section for the uphill and downhill directions are 0.50 and 0.50, respectively.

[0057] ② Lane coefficient: According to the weighing data from the overload detection station, the lane distribution coefficients for the first, second, and third lanes (hard shoulder) in the downhill direction are 0.26, 0.73, and 0.01, respectively.

[0058] (2) Vehicle type classification According to the "Vehicle Type and Classification Criteria" in the "Technical Guidelines for Axle Load Spectrum Measurement of Ordinary National and Provincial Highways in Yangzhou City", the vehicle type distribution coefficient results are shown in the table below. The total number of vehicles with 2 axles or more is 488,507, of which the proportion of abnormal or special vehicle types is 0.07%, which are not considered in the calculation.

[0059] Vehicle type distribution coefficient

[0060] (3) Equivalent axle load conversion factor The equivalent axle load conversion factors are shown in the table below:

[0061] (4) Calculation results of axle load spectrum Axle load spectrum of single axle and single wheel as follows Figure 2 As shown, the axle load spectrum of a single axle with two wheels is as follows: Figure 3 As shown, the axial load spectrum of the twin shafts is as follows: Figure 4 As shown, the axial load spectrum of the triple shaft is as follows: Figure 5 As shown.

[0062] The screening, classification, and calculation steps in the axle load spectrum measurement process described above are automated using self-developed software to process massive amounts of data. Key thresholds are set by technical personnel based on equipment accuracy and actual traffic conditions. Furthermore, based on road network construction and traffic monitoring platform development, the station combinations, thresholds, and main vehicle types at the regional scale are updated annually, with results for step four issued quarterly.

[0063] (5) Prediction of the axial load spectrum in the next cycle The aforementioned axle load spectrum determination method is a static measurement, based on data collection and spectrum plotting at fixed periods (such as quarters or years), which has a certain degree of lag and passivity. With seasonal fluctuations in traffic flow structure, changes in regional economic activity intensity, and continuous adjustments in transportation organization methods, road load characteristics exhibit a clear dynamic evolution trend. In this context, relying solely on historical data for axle load spectrum determination often fails to reflect upcoming load changes in a timely manner, thus limiting the effectiveness and guiding power of the spectrum in traffic decision-making, structural design, and forward-looking planning.

[0064] Therefore, this embodiment introduces an axle load spectrum prediction step, aiming to proactively predict the evolution trend of the axle load spectrum for the next cycle (such as the next quarter or the next year) based on existing historical measurement results and traffic operation characteristics, and to obtain the distribution change trends of each axle type in different axle load ranges in advance. This not only helps to build a forward-looking traffic load evolution model, but also provides more time-valued decision-making basis for road structure adaptive design, traffic organization optimization, and traffic facility maintenance. At the same time, through dynamic comparison and feedback between axle load spectrum prediction and measured data, the classification logic and parameter settings can be corrected in reverse, gradually building a self-learning, self-adaptive, and self-updating intelligent load spectrum measurement and prediction system, promoting the transformation of traffic engineering from passive response to proactive perception and prediction, and achieving a higher level of digital and intelligent upgrading.

[0065] Specifically, the axle load spectrum prediction process mainly includes the following steps: Step 1: Construct the prediction input feature set To achieve predictive modeling of axle load spectra at a regional scale, it is necessary to first construct a structured input feature vector to reflect historical spectrum status, traffic conditions, and environmental and economic context. This feature vector not only determines the fitting accuracy of the prediction model but also directly affects the judgment of the temporal evolution trend of the axle load spectrum.

[0066] Let the current prediction period be t+1, and its corresponding prediction input feature be denoted as . X t Its definition is as follows:

[0067] in, This indicates the current period (period t) and the th period. i Axial type in the first j The spectral distribution coefficient of each axle load interval, i.e., the axle load spectrum obtained by measuring in the current period. These correspond to single-axle single-wheel, single-axle double-wheel, double-axle and triple-axle configurations, respectively. This refers to the axle load range number, such as 0~2.5kN, 2.5~5kN, 5~7.5kN, etc. k A range.

[0068] The current period traffic flow feature vector reflects the potential impact of the current period traffic conditions on the axle load spectrum, and includes the following elements: , This represents the average daily traffic flow for the current period. The proportion of freight vehicles (number of trucks / total traffic flow). This refers to the percentage of large trucks (gross weight greater than 20 tons) among all trucks. These parameters, representing average vehicle speeds, can be obtained from traffic control stations and serve as key inputs reflecting changes in traffic composition.

[0069] This represents the vector of environmental and economic factors in the current cycle. , The Holiday Index is the ratio of the average daily traffic flow during holidays to the average daily traffic flow on non-holiday days, reflecting the amplification factor of holidays on traffic peaks. For seasonal factors (such as the Spring Festival travel rush, heating season, peak season for agricultural product transportation, etc.), it is equal to the ratio of the average quarterly traffic volume to the average annual traffic volume. For the region's quarterly GDP growth rate, This is the logistics and transportation index (which can be the Logistics Industry Prosperity Index (LPI) or provincial / municipal freight volume indicators). This part is used to guide the prediction model to learn the impact of non-traffic volume on the load spectrum evolution, thereby improving the prediction model's sensitivity to economic activities.

[0070] Step 2: Hierarchical time series prediction modeling This step combines linear trend prediction with nonlinear dynamic change prediction through hierarchical modeling, and finally outputs the distribution coefficient of each axis type in each axis weight interval.

[0071] For each type of shaft i (For example, single-axle single-wheel, single-axle double-wheel, double-axle, triple-axle), based on historical axle load data and other influencing factors, a multi-level prediction model is established to calculate the axle load spectrum in the next cycle t+1. The specific prediction formula is as follows:

[0072] in: Indicates the predicted first i Axial type in the first j The distribution coefficient of the axle load interval, i.e., the axle load spectrum value of the next period. Indicates the use of the first i The axial prediction model function combines historical data and feature inputs to perform multi-level prediction calculations. The function consists of three parts: basic trend prediction, nonlinear dynamic learning, and dynamic fusion. It contains the t-th period. i Axial type in the first j Distribution coefficient of axle load range That is, the axle load spectrum value for the current period; It contains the t-1th period i Axial type in the first j Distribution coefficient of axle load range That is, the axial load spectrum value of the previous period; It contains the first t - N +1 cycle i Axial type in the first j Distribution coefficient of axle load range , N For reference to the total number of historical cycles, for example when N When the value is 4, the prediction only refers to the historical axial load spectrum measurement data of the past 4 cycles, i.e. .

[0073] (1) Basic trend prediction (linear modeling) First, a seasonally adjusted autoregressive model (SARIMA) is used to predict the long-term trend of historical data. Autoregressive models can capture seasonal fluctuations and long-term trends in traffic flow, helping to extract the more stable components from the data. The formula is as follows:

[0074] in, The trend term is predicted based on a linear model, i.e., the... i Axial type in the first j Predicted trends within axle load ranges; , , , … These are autoregressive coefficients used to control for the influence of historical data on the predicted trend; This is a seasonality coefficient used to control the magnitude and direction of seasonal fluctuations; S This represents the total number of seasonal terms, used to indicate the number of periods in the time series (12 for months, 4 for quarters). This is a periodic fluctuation term used to simulate the impact of seasonality on traffic flow. This model allows for the extraction of long-term trends and seasonal fluctuations in traffic flow, thus providing a foundation for subsequent nonlinear dynamic predictions.

[0075] (2) Nonlinear dynamic learning (deep model) Then, an LSTM (Long Short-Term Memory) model is used to capture the nonlinear dynamic characteristics of traffic load. LSTM can handle long-term dependencies and nonlinear changes, and is particularly suitable for abrupt changes and periodic fluctuations in traffic flow, such as holiday effects and changes in traffic flow structure.

[0076] The prediction formula is: , in, This represents the nonlinear prediction result output by the LSTM network, i.e., the predicted [number]th [unit]. i Axial type in the first j Nonlinear prediction values ​​for axle load ranges. LSTM ( ) represents an LSTM neural network model that uses a gating mechanism to process long-term dependencies in time series data and outputs a prediction for the next cycle.

[0077] LSTM models can effectively capture complex dynamic features in data, such as abrupt changes, holiday effects, and changes in traffic flow structure, thus having a significant advantage in predicting traffic load data.

[0078] (3) Dynamic fusion prediction results Finally, the basic trend prediction and nonlinear prediction results are weighted and fused to obtain the final prediction result.

[0079] The fusion formula is: , in: λ i For the first i The fusion weighting coefficient for axial-type predictions ranges from [0,1]. This coefficient determines the contribution of linear and nonlinear predictions to the final result. λ i When the value approaches 1, the prediction relies more on linear trend prediction; conversely, it relies more on the nonlinear dynamic prediction of the LSTM model. This is the final fused axle load spectrum prediction value, which is the prediction result of the i-th type of axle in the j-th axle load range.

[0080] This weighted fusion method can balance the linear model's ability to capture trends with the LSTM model's ability to adapt flexibly to abrupt changes and nonlinear features, thus providing more accurate and stable prediction results.

[0081] The table below shows the predicted axle load distribution coefficients for a single-axle, single-wheel type 11, model 1222 in various axle load ranges for a certain period:

[0082] The table below shows the predicted axle load distribution coefficients for Type 157, Category 9, of a twin-axle, twin-wheel configuration in various axle load ranges for a certain period:

[0083] Example 2 Based on the above method, this embodiment provides a regional-scale axle load spectrum measurement system. This system is a comprehensive intelligent measurement platform for large-scale traffic load feature extraction. It constructs a highly modular, logically closed-loop functional structure around the technical main lines of data acquisition, multi-level verification, classification extraction, spectrum plotting, and trend prediction. The entire system includes a data acquisition module, a multi-level verification and classification module, a major vehicle type determination module, a parameter calculation and axle load spectrum plotting module, and an axle load spectrum prediction module, which collaboratively realize the automated measurement and forward-looking extrapolation of axle load spectra at a regional scale.

[0084] At the beginning of system operation, the data acquisition module continuously acquires axle load data and traffic composition information for various vehicles through dynamic weighing systems and traffic monitoring systems deployed at multiple monitoring stations within the region. The information collected by this module covers vehicle travel time, lane, direction, axle type, wheelbase, weight of each axle, and traffic flow composition, ensuring comprehensive, continuous, and representative data. Subsequently, the system inputs the collected data into a multi-level verification and classification module. This module, according to set integrity and logical standards, sequentially performs data continuity verification, logical consistency checks, and cross-site data consistency checks, ensuring that the input data meets the stringent requirements for spectral plotting and selecting station combinations with superior data quality as the basis for subsequent analysis.

[0085] Based on high-quality data, the main vehicle type determination module uses statistical analysis methods to identify the distribution of vehicle types within the region, prioritizing vehicle types with a frequency exceeding a set threshold (e.g., 0.1%) as the primary research objects. Following this, the system proceeds to the parameter calculation and axle load spectrum plotting stage. This module utilizes validated data and the identified main vehicle types to calculate key spectrum parameters such as direction coefficient, lane coefficient, vehicle type distribution coefficient, and axle load distribution coefficient, mapping them to the axle load interval distribution for different axle types (e.g., single-axle single-wheel, single-axle dual-wheel, dual-axle, triple-axle, etc.), thereby generating a complete and accurate axle load spectrum, providing quantitative support for road design and load simulation.

[0086] Furthermore, the system also incorporates an axle load spectrum prediction module, which supports dynamic projection of regional axle load spectrum trends for the next period (such as the next quarter or the next year) based on historical measurement results and traffic operation trends. This module integrates time series modeling, regression analysis, or machine learning methods, and can output distribution evolution trend maps of various typical axle types in different axle load ranges, enabling proactive identification and dynamic perception of traffic load status.

[0087] In summary, through modular integration and process collaboration, this system not only supports high-quality classification and spectral construction of historical data, but also has the ability to predict the evolution of spectra in the future, significantly improving the automation, intelligence and practicality of regional-scale axial load spectral determination.

[0088] The above system can execute the regional scale axial load spectrum determination method based on multi-level verification classification as described in Embodiment 1, and has the corresponding functional modules and beneficial effects of the method. For technical details not described in detail in this embodiment, please refer to the regional scale axial load spectrum determination method based on multi-level verification classification provided in Embodiment 1 of this invention.

[0089] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented using software plus a general-purpose hardware platform, or of course, using hardware. Based on this understanding, the above technical solutions, in essence or the parts that contribute to the related technology, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0090] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them. Under the concept of the present invention, the technical features of the above embodiments or different embodiments can also be combined, and there are many other variations of different aspects of the present invention as described above. For the sake of brevity, they are not provided in detail. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. These modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for determining regional-scale axial load spectra based on multi-level verification and classification, characterized in that, Including the following steps: S1. Data Acquisition: Collect axle load data from the dynamic weighing system at each station, including the number of axles, axle spacing, and axle load; collect natural traffic volume statistics for different traffic composition categories at each station, referred to as traffic survey data. S2, Multi-level verification classification: Perform data integrity verification, axle load parameter logical verification, and cross-site consistency verification on the collected parameters to obtain the site combination that conforms to the axle load spectrum calculation; S3. Determining the main vehicle models: Statistically determine the proportion of different vehicle models in the region, and select the models with a proportion greater than a set threshold as the main vehicle models in the region; S4. Parameter Calculation and Axle Load Spectrum Drawing: Based on the effective axle load data and traffic data of the station combination that conforms to the axle load spectrum calculation, as well as the main vehicle types in the area, calculate the main coefficients and draw the corresponding axle load spectrum; the main coefficients include the direction coefficient, lane coefficient, vehicle type distribution coefficient, axle load distribution coefficient, and equivalent design axle load conversion coefficient; the axle load spectrum is used to represent the distribution of each axle type in different axle load ranges. The data integrity verification includes: verifying the time period, direction, and lane of axle load data and traffic control data to ensure continuous data collection; ensuring that the online rate of the collection equipment is controlled above a set threshold during the collection time; avoiding stations with dynamic weighing equipment only on some lanes; and avoiding areas with complex surrounding environments or special traffic control measures. The axle load parameter logical verification is based on three dimensions: number of axles, wheelbase, and gross weight. It establishes a precise correspondence between axle load data and vehicle axle load characteristics at a regional scale, achieving accurate and efficient correlation and matching. The process follows a coarse-to-fine classification logic, including the following steps: Axle count coarse classification: Obtain the number of vehicle axles through axle load data, classify them according to the number of axles, and remove abnormal data; Wheelbase subdivision: Establish a wheelbase feature threshold library for key axle types at the regional scale. Key axle types should meet the corresponding wheelbase feature threshold requirements. The key axle types include single axle dual wheels for two-axle vehicles, and dual front axle, dual-axle dual wheels, and triple-axle dual wheels for three-axle and above vehicles. Investigate the main vehicle models in the region and establish a vehicle model overview table at the regional scale, further distinguishing different vehicle models through axle type combinations. Establish a total wheelbase threshold library for vehicle models, where the sum of the wheelbases of all vehicle models does not exceed the corresponding total wheelbase threshold. Gross weight classification: Establish a minimum gross weight threshold library for each vehicle type. The sum of the axle weights of each vehicle type should be lower than the corresponding total axle weight threshold. For two-axle vehicles that meet the conditions, further classify them according to their gross weight and set a total weight threshold for two-axle vehicles to distinguish between large two-axle vehicles and small and medium-sized two-axle vehicles. Set the classification accuracy threshold according to the equipment precision of the dynamic weighing system, and count the amount of data to be discarded and retained in the logical verification of axle load parameters. The proportion of retained data should not be less than the classification accuracy threshold; otherwise, reclassification verification should be performed. The cross-site consistency verification includes the following process: Station combination classification: Based on the road network distribution, axle load data stations are combined with adjacent traffic control data stations to ensure that there are no large intersections for traffic diversion between stations. The station combinations are divided into three categories: no traffic diversion between national, provincial, county, and township roads; no traffic diversion between national and provincial roads; and no traffic diversion between county and township roads. Consistency check coefficient calculation: Calculate the consistency check coefficient for each site combination. The consistency check coefficient is defined as follows: , T i for i The frequency of vehicle passage is the source of axle load data. T 0 represents traffic volume in the traffic survey data; Threshold setting: Based on the traffic distribution of different site combination types, set the threshold for the consistency verification coefficient for different combination types; Threshold matching: Sites that meet the threshold requirements are classified as eligible for axle load spectrum calculation; site combinations that do not meet the threshold requirements are not included in this axle load spectrum calculation.

2. The method for determining the regional-scale axial load spectrum as described in claim 1, characterized in that, The threshold settings in each step are determined based on the equipment accuracy and actual traffic conditions. Based on the road network construction and traffic detection platform construction, the station combination, thresholds, and main vehicle types at the regional scale are updated annually, and the axle load spectrum measurement results are issued quarterly.

3. The method for determining the regional-scale axial load spectrum as described in claim 2, characterized in that, It also includes a next-cycle axle load spectrum prediction step, which is used to extrapolate the evolution of the axle load spectrum in the next cycle based on existing historical measurement results and traffic operation characteristics, and to obtain the distribution change trend of each axle type in different axle load ranges in advance.

4. The method for determining the regional-scale axial load spectrum as described in claim 3, characterized in that, The next cycle shaft load spectrum prediction step includes the following steps: S501, Construction of Predicted Input Feature Vector Let the current prediction period be t+1, and its corresponding prediction input feature vector be denoted as . X t Its definition is: in, This indicates the current period t-th... i Axial type in the first j The spectral distribution coefficient of each axle load interval, i.e., the axle load spectrum obtained by the current period measurement; Corresponding to each shaft type, This represents the axle load interval number. k This represents the total number of axle load intervals. This is the current cycle traffic flow feature vector, used to reflect the potential impact of the current cycle traffic state on the axle load spectrum, and , This represents the average daily traffic flow for the current period. For the proportion of freight vehicles, This represents the proportion of large trucks among all trucks. The average operating speed of the vehicle; Given the current cycle environment and economic factors vector, and , The Holiday Index is the ratio of the average daily traffic flow during holidays to the average daily traffic flow on non-holiday days, reflecting the amplification factor of holidays on traffic peaks. This is a seasonal factor, equal to the ratio of the average quarterly traffic volume to the average annual traffic volume; For the region's quarterly GDP growth rate, For logistics and transportation index; S502, Construction of Hierarchical Time Series Prediction Model For each type of shaft i A prediction model is established based on historical axle load data to calculate the axle load spectrum in the next cycle t+1. The prediction formula is as follows: in: Indicates the predicted first i Axial type in the first j The distribution coefficient of the axle load interval, i.e., the axle load spectrum value of the next period; Indicates the use of the first i Axle-type axial load spectrum prediction model; It contains the t-th period. i Axial type in the first j Distribution coefficient of axle load range That is, the axle load spectrum value for the current period; It contains the t-1th period i Axial type in the first j Distribution coefficient of axle load range That is, the axial load spectrum value of the previous period; It contains the first t - N +1 cycle i Axial type in the first j Distribution coefficient of axle load range , N To reference the total number of historical cycles, it means that only past data is considered when making predictions. N Historical axial load spectrum measurement data for each cycle.

5. The method for determining the regional-scale axial load spectrum as described in claim 4, characterized in that, The prediction model includes linear basic trend prediction, nonlinear dynamic learning, and dynamic fusion. Linear basic trend prediction: First, a seasonally adjusted autoregressive model is used to predict the long-term trend of historical data. The prediction formula is as follows: in, The trend term is predicted based on a linear model, i.e., the... i Axial type in the first j Predicted trends within axle load ranges; , , , … These are autoregressive coefficients used to control for the influence of historical data on the predicted trend; This is a seasonality coefficient used to control the magnitude and direction of seasonal fluctuations; S This represents the total number of seasonal terms, used to indicate the number of periods in the time series; This is a periodic fluctuation term used to simulate the impact of seasonality on traffic flow; Nonlinear dynamic learning: Then, an LSTM model is used to capture the nonlinear dynamic characteristics of traffic load. The prediction formula is: , in, This represents the nonlinear prediction result output by the LSTM network, i.e., the predicted [number]th [unit]. i Axial type in the first j Nonlinear prediction values ​​for axle load range; LSTM ( ) represents an LSTM neural network model that uses a gating mechanism to process long-term dependencies in time-series data and outputs a prediction for the next cycle; Dynamic fusion: Finally, the linear basic trend prediction and the nonlinear prediction results are weighted and fused to obtain the final prediction result. The fusion formula is as follows: , in: λ i For the first i The fusion weighting coefficient for axial-type predictions ranges from [0,1]. This coefficient determines the contribution of linear and nonlinear predictions to the final result. λ i When the value approaches 1, the prediction relies more on linear trend prediction; conversely, it relies more on the nonlinear dynamic prediction of the LSTM model. The final fused axle load spectrum prediction value, i.e., the first... i Axial type in the first j Prediction results for axle load range.

6. A regional-scale axial load spectrum determination system based on the method of claim 5, characterized in that, It includes a data acquisition module, a multi-level verification and classification module, a main vehicle model determination module, a parameter calculation and axle load spectrum plotting module, and an axle load spectrum prediction module; The data acquisition module is used to collect and obtain axle load data and cross-modulation data from each station; The multi-level verification and classification module is used to perform data integrity verification, axle load parameter logical verification, and cross-site consistency verification on the collected parameters, and to obtain the site combination that conforms to the axle load spectrum calculation. The main vehicle model determination module is used to count the proportion of different vehicle models in a region and select the vehicle models with a proportion greater than a set threshold as the main vehicle models in the region. The parameter calculation and axle load spectrum drawing module is used to calculate the main coefficients and draw the corresponding axle load spectrum based on the effective axle load data and traffic data of the station combination that conforms to the axle load spectrum calculation, as well as the main vehicle types in the area. The axle load spectrum prediction module is used to predict the evolution of the axle load spectrum in the next cycle based on existing historical measurement results and traffic operation characteristics, and to obtain the distribution trend of each axle type in different axle load ranges.

7. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and running thereon, characterized in that, When the processor executes the program, it implements the steps of the regional-scale axial load spectrum determination method according to any one of claims 1 to 5.

Images