Big data intelligent analysis method and system

Abstract

The invention discloses a big data intelligent analysis method and system, and relates to the technical field of big data intelligent analysis, and the method comprises the steps: collecting multi-source monitoring data, such as pavement temperature, humidity, strain and axle load, and calculating a daily average humidity value, a daily maximum temperature gradient value and a normalized strain peak value; constructing a continuous weight based on the strain response, and distinguishing a dry state sample from a wet state sample according to a daily average humidity value; and respectively carrying out weighted linear fitting on the two types of samples to obtain a hysteresis humidity loss temperature amplification index for depicting a humidity effect amplification effect, generating a road section average humidity loss risk index in combination with a continuous weight and a temperature gradient, and realizing road section humidity loss risk sorting according to the road section average humidity loss risk index. According to the invention, automatic quantification and comparative analysis of the moisture damage risk can be realized in a large-scale monitoring data scene, and the accuracy of a road maintenance decision is improved.

Description

Technical Field

[0001] This invention relates to the field of big data intelligent analysis technology, and in particular to a big data intelligent analysis method and system. Background Technology

[0002] During the long-term service of road infrastructure, pavement structures are continuously affected by environmental changes and traffic loads. Temperature fluctuations, moisture infiltration, and vehicle action are constantly superimposed at different time scales, causing the internal humidity state, temperature gradient, and mechanical response of pavement materials to exhibit high time-varying characteristics. With the widespread adoption of intelligent monitoring equipment, various monitoring parameters such as pavement surface temperature, deep temperature, internal humidity, axle load, and strain response can be continuously collected over long periods. The volume of monitoring data is growing exponentially, and pavement health monitoring in the big data environment is gradually becoming an important part of road operation and maintenance. In order to identify potential structural damage trends caused by the combined effects of humidity changes and temperature gradients, it is necessary to extract effective information from massive multi-source monitoring data spanning multiple road sections and dates, analyze the correlation between humidity state changes and temperature-driven strain response, and provide quantitative basis for road section maintenance.

[0003] Existing analytical methods typically rely on simple statistical analysis or empirical threshold judgments of monitored values. They fail to fully leverage the inherent correlations between multi-source monitoring data and cannot automatically identify representative time periods, road sections, or anomalies from large-scale data. In practical engineering, this often leads to problems such as delayed risk assessment, incomparability between different road sections, and difficulty in extracting patterns of wet damage. Especially when monitoring data is massive, unevenly distributed, and of varying quality, the lack of a unified method to automatically filter data, establish relational models, and generate ranking results in big data scenarios hinders the precise and quantitative assessment of wet damage risks across different road sections. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies that lack the ability to comprehensively process long-term monitoring data, cross-segment comparison data, and continuous time-series data, making it difficult to accurately reflect the differences in the impact of temperature changes on roads under dry and humid conditions. Therefore, this invention proposes a big data intelligent analysis method and system.

[0005] To address the problems existing in the prior art, the present invention adopts the following technical solution: A big data intelligent analysis method includes: S1. Collect multi-source monitoring data for road surface monitoring, and determine the daily average humidity, daily maximum temperature gradient and normalized strain peak value of the road section based on the multi-source monitoring data. S2. Construct continuous weights based on normalized strain peak values; S3. Determine the humidity threshold value based on the daily average humidity value to obtain dry state samples and humid state samples; S4. Perform a weighted linear fit on the dry and wet samples, and obtain the hysteresis wet loss temperature amplification index based on the slope ratio of the weighted linear relationship between the dry and wet states. S5. Calculate the average wet damage risk of the road segment based on the hysteresis wet damage temperature amplification index, continuous weights and daily maximum temperature gradient; S6. Sort the road segments according to their average wet damage risk and generate a wet damage risk sorting view.

[0006] Preferably, multi-source monitoring data for road surface monitoring is collected, and the daily average humidity value, daily maximum temperature gradient value, and normalized strain peak value of the road section are determined based on the multi-source monitoring data, including: Collect multi-source monitoring data for road surface monitoring, including: road surface temperature, road surface deep temperature, road surface internal humidity, axle load, and strain peak value corresponding to the axle load; Multi-source monitoring data are labeled using unified time and road segment identifiers; Using the natural day as the time unit, all multi-source monitoring data for each road segment on the same date are grouped to form a time group set for that road segment on that date. The daily average humidity value of the road segment is obtained by averaging the internal humidity of the road surface in the time group set. The absolute difference between the surface temperature and the deep surface temperature in the time group set is calculated, and the maximum value of the absolute difference is taken as the daily maximum temperature gradient value. In the time group set, select the axle load within the preset range, calculate the ratio of the corresponding strain peak value to the axle load, and take the maximum value of the ratio as the normalized strain peak value.

[0007] Preferably, a continuous weight is constructed based on the normalized strain peak value, including: Select each combination of date and road segment for which the daily average humidity, daily maximum temperature gradient and normalized strain peak value have been calculated simultaneously, and define the combination of date and road segment as a sample unit to form a sample unit set. Taking the normalized strain peak values ​​in all sample cells as the object, calculate the median of the normalized strain peak values ​​and the median of the absolute deviations relative to the median. Based on the median of the normalized strain peak and the median of the absolute deviation, the normalized strain peak of each sample unit is standardized to obtain the standardized deviation. Based on the standardized deviation, construct continuous weights for each sample unit in the sample unit set.

[0008] Preferably, a humidity threshold is determined based on the daily average humidity value to obtain dry state samples and humid state samples, including: From the sample unit set, read the date, road segment, daily average humidity value, daily maximum temperature gradient value, normalized strain peak value and continuous weight corresponding to each sample unit. Then, remove duplicates and sort all the daily average humidity values ​​that appear in each sample unit to obtain a set of candidate humidity boundary values ​​composed of daily average humidity values. For each candidate humidity threshold value in the candidate humidity threshold value set, the sample unit set is grouped according to whether the daily average humidity value is less than or equal to the candidate humidity threshold value, forming a dry state sample set and a humid state sample set. In the dry and wet sample sets, the daily maximum temperature gradient value is used as the independent variable, the normalized strain peak value is used as the dependent variable, and continuous weights are used as weights to establish two sets of weighted linear relationships, so as to obtain the linear model parameters under the dry state and the linear model parameters under the wet state. Calculate the weighted residual sum corresponding to the candidate humidity boundary value based on the weighted error of the two linear models on their respective sample sets; Among all candidate humidity cutoff values, the candidate humidity cutoff value with the smallest weighted residual sum is selected as the final humidity cutoff value. The set of dry state samples corresponding to the final humidity threshold value is determined as the final dry state sample set. The set of humid state samples corresponding to the final humidity threshold value is determined as the final humid state sample set.

[0009] Preferably, a weighted linear fit is performed on the dry and wet samples, and the hysteresis wet loss temperature amplification index is obtained based on the ratio of the slope of the weighted linear relationship between the dry and wet states, including: In the final dry state sample set, the daily maximum temperature gradient value is used as the independent variable, the normalized strain peak value is used as the dependent variable, and the continuous weight is used as the weight to establish a weighted linear relationship under the dry state. We performed a weighted least squares fit on the weighted linear relationship under the dry state to obtain the slope value of the temperature sensitivity under the dry state and the uncertainty index corresponding to the slope value of the temperature sensitivity under the dry state. In the final humid state sample set, the daily maximum temperature gradient value is used as the independent variable, the normalized strain peak value is used as the dependent variable, and the continuous weight is used as the weight to establish a weighted linear relationship under humid conditions. We performed a weighted least squares fit on the weighted linear relationship under humid conditions to obtain the slope value of the temperature sensitivity under humid conditions and the uncertainty index corresponding to the slope value of the temperature sensitivity under humid conditions. The ratio of the temperature sensitivity slope value under humid conditions to the temperature sensitivity slope value under dry conditions is used as the hysteresis moisture loss temperature amplification index. The strength of evidence index is composed of the difference between the slope values ​​of temperature sensitivity in the humid state and the slope values ​​of temperature sensitivity in the dry state, the uncertainty index corresponding to the slope value of temperature sensitivity in the dry state, and the uncertainty index corresponding to the slope value of temperature sensitivity in the humid state.

[0010] Preferably, the average wet loss risk of the road segment is calculated based on the hysteresis wet loss temperature amplification index, continuous weights, and the daily maximum temperature gradient, including: In the final set of humid conditions, the daily maximum temperature gradient values ​​are weighted and averaged with continuous weights to obtain the global weighted average temperature gradient under humid conditions. Extract all road segment identifiers that have appeared in all sample units to construct a road segment set; In all sample units, for each combination of date and road segment, the day is determined to be either a wet or dry day based on the relationship between the average daily humidity value and the humidity threshold value of the combination. When it is a wet day, the ratio of the daily maximum temperature gradient value to the global weighted average temperature gradient is calculated. Based on the product of the ratio and the hysteresis wet loss temperature amplification index and the continuous weight, the daily scale wet loss risk contribution of the date and road segment combination is obtained. When the day is in a dry state, the daily scale wet damage risk contribution of the date and road segment combination is set to zero; In the set of road segments, for each road segment, the daily scale wet damage risk contribution of that road segment on all dates is summed to obtain the total wet damage risk of that road segment; In the final wet condition sample set, the continuous weights of all samples belonging to the same road segment are accumulated to obtain the weighted observation total of the road segment under wet conditions; Divide the total wet damage risk by the weighted total observation to obtain the average wet damage risk index of the road segment. When the weighted total observation is zero, the average wet damage risk index is set to zero.

[0011] Preferably, the road segments are sorted according to their average wet damage risk to generate a wet damage risk ranking view, including: In the road segment set, all road segment identifiers are arranged in descending order of average wet damage risk index to obtain an ordered road segment list; In the list of road segments, each road segment is assigned a unique sorting number, and the smaller the number, the higher the risk of wet damage. In the list of ordered road segments, a record is generated for each road segment, including a sorting number, road segment identifier, average wet damage risk index, hysteresis wet damage temperature amplification index, evidence strength index, and humidity threshold value. All records are arranged in ascending order of sorting number to form a sorting view of wet damage risk.

[0012] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention sequentially collects multi-source pavement monitoring data, calculates daily average humidity and daily maximum temperature gradient, extracts normalized strain peak value, distinguishes between dry and wet samples, and performs weighted linear fitting on the two types of samples to obtain the hysteresis wet damage temperature amplification index. Then, it combines continuous weights and daily maximum temperature gradient to calculate the average wet damage risk of each road segment and generate a wet damage risk ranking view. This enables the originally scattered temperature, humidity and strain monitoring data to form a complete analysis link in a big data environment, realizing the automatic quantification and comparative evaluation of the long-term wet damage intensity of different road segments. It solves the problem that it is difficult to uniformly characterize the wet damage risk level under the combined effect of humidity and temperature in massive monitoring data across road segments and dates, resulting in a lack of objective quantitative basis for maintenance decisions.

[0013] 2. This invention introduces a continuous weighting mechanism based on normalized strain peak value and a candidate humidity boundary value search and weighted residual optimization strategy based on daily average humidity value. This ensures that the samples used for modeling come from a large-scale dataset formed by long-term monitoring, and can automatically suppress interference caused by strain anomaly deviation samples and subjective setting of humidity boundaries. Robust weighted linear response relationships are established under dry and humid conditions respectively. The ratio of linear model parameters under the two conditions is used to construct a hysteresis wet loss temperature amplification index, and the change in temperature sensitivity and its uncertainty are combined to form an evidence strength index. Thus, in the context of big data, the significance and reliability of the humidity amplification effect can be quantitatively determined.

[0014] 3. This invention constructs a globally weighted average temperature gradient on samples under humid conditions. Using this as a normalization benchmark, it combines the daily maximum temperature gradient, the lagged wet damage temperature amplification index, and continuous weights of each date and road segment under humid conditions to form the daily contribution of wet damage risk. This contribution is then accumulated at the road segment level to obtain the total wet damage risk. This total risk is then divided by the weighted observation total under humid conditions to form an average wet damage risk index. Finally, the road segments are sorted from largest to smallest according to this index, generating a wet damage risk ranking view. This allows for a direct understanding of the cumulative degree of wet damage and the comparative risk level of each road segment during long-term service, thus providing visualized big data decision support for the priority allocation and targeted reinforcement of limited maintenance resources. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0016] Figure 1 This is a flowchart illustrating a big data intelligent analysis method according to an embodiment of the present invention. Figure 2 This is a functional module diagram of a big data intelligent analysis system provided in an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0018] Example: This example provides a big data intelligent analysis method, see [link to example]. Figure 1 Specifically, including: S1. Collect multi-source monitoring data for road surface monitoring, and determine the daily average humidity, daily maximum temperature gradient and normalized strain peak value of the road section based on the multi-source monitoring data. In embodiments of the present invention, multi-source monitoring data for road surface monitoring is collected, and the daily average humidity value, daily maximum temperature gradient value, and normalized strain peak value of the road segment are determined based on the multi-source monitoring data, including: Collect multi-source monitoring data for road surface monitoring, including: road surface temperature, road surface deep temperature, road surface internal humidity, axle load, and strain peak value corresponding to the axle load; Multi-source monitoring data are labeled using unified time and road segment identifiers; Using the natural day as the time unit, all multi-source monitoring data for each road segment on the same date are grouped to form a time group set for that road segment on that date. Specifically, surface temperature represents the instantaneous temperature state of the outermost layer of pavement material, reflecting the heat distribution formed after the pavement is directly exposed to solar radiation and environmental heat exchange; deep pavement temperature represents the temperature level at deeper locations within the pavement, characterizing the slow heat transfer process within the structure and the overall thermal stability of the roadbed; internal pavement humidity represents the change in water vapor content within the pores of the pavement material, reflecting the moisture fluctuations caused by water infiltration and evaporation; axle load represents the periodic pressure level acting on the pavement when vehicles pass over it, revealing the mechanical disturbance of external loads on the pavement structure; and the peak strain corresponding to the axle load represents the pavement material under vehicle load. The maximum deformation response generated is used to reflect the deformation strength of the structural layer under pressure; a unified time identifier is used to ensure the correspondence between the collection times of different monitoring quantities, so that all monitoring quantities are comparable under the same time scale; a road segment identifier is used to distinguish monitoring objects at different locations, so that the monitoring quantities of the same physical area can be classified into a unified data set; furthermore, the natural day is used as a time unit to represent a fixed time range from 0:00 on each day to 0:00 on the next day, so that all monitoring data on the same day are classified into the same time set, so that various monitoring quantities can be grouped and integrated within a stable time period to form the time group set of the road segment on that date.

[0019] Specifically, when collecting multi-source monitoring data for road surface monitoring, temperature sensors, humidity sensors, and strain gauges are deployed at different locations on the target road section. The axle load is recorded in real time as a vehicle passes by through a data acquisition terminal connected to the axle load metering device, and the strain signal output by the strain gauge is collected simultaneously. After amplification, filtering, and analog-to-digital conversion, the strain peak value corresponding to the axle load is obtained. The road surface temperature is directly measured by temperature sensors deployed on the road surface, the deep road surface temperature is measured by temperature sensors buried at different depths inside the road body, and the internal road surface humidity is measured by humidity sensors buried in the structural layer or base layer. The above-mentioned road surface temperature, deep road surface temperature, internal road surface humidity, axle load, and strain peak value are continuously collected and stored as raw multi-source monitoring data at a preset sampling period.

[0020] Specifically, when labeling multi-source monitoring data, a unified time source is configured for the acquisition terminal, and the internal clock of the acquisition terminal is calibrated with the same time reference. The corresponding time scale is obtained during each sampling, and the time scale is converted into a unified time identifier and written into each multi-source monitoring data record. At the same time, a unique road segment number is set for each monitored physical road segment. This road segment number is used as a road segment identifier and associated with all multi-source monitoring data collected on that road segment. This ensures that each multi-source monitoring data carries both a unified time identifier and a road segment identifier, so that subsequent retrieval and classification can be performed according to time and spatial dimensions.

[0021] Specifically, when grouping data using natural days as the time unit, the collection time of all multi-source monitoring data is first converted into date information according to a unified time identifier, and the time range from 00:00 on the same day to 00:00 on the next day is determined as a natural day interval. Then, for each road segment identifier in the road segment set, all data records with unified time identifiers falling within the same natural day interval and consistent with the road segment identifier are selected from all multi-source monitoring data. These data records are sorted from morning to evening according to the unified time identifier to form the time group corresponding to the road segment on that date. The time groups obtained for each road segment on each date are collected to form the time group set for that road segment on that date, which is used for subsequent calculation of daily average humidity, daily maximum temperature gradient, and normalized strain peak value.

[0022] The daily average humidity value of the road segment is obtained by averaging the internal humidity of the road surface in the time group set. The absolute difference between the surface temperature and the deep surface temperature in the time group set is calculated, and the maximum value of the absolute difference is taken as the daily maximum temperature gradient value. In the time group set, select the axle load within the preset range, calculate the ratio of the corresponding strain peak value to the axle load, and take the maximum value of the ratio as the normalized strain peak value. Specifically, the daily average humidity value represents the average level of the humidity sequence recorded within the pavement material of a given road section within a natural day. It reflects the overall change in water vapor content of the road body on that day and is an effective description of the accumulation, evaporation, and infiltration of moisture within the pavement. The daily maximum temperature gradient value represents the maximum absolute difference in the temperature difference sequence between the surface temperature and the deep pavement temperature within that natural day. It reveals the driving intensity of the temperature difference between rapid heating or cooling of the upper part of the pavement and slow conduction within, thus reflecting the degree of uneven heat propagation in the structure. The normalized peak strain value represents the maximum value of the ratio calculated between the peak strain recorded by the strain gauge under vehicle load and the corresponding axle load. It characterizes the maximum relative deformation capacity of the pavement material under vehicle ballast, allowing for comparison of strain responses under different axle load conditions on a uniform scale to reflect the sensitivity of the structural layer under load.

[0023] Specifically, when acquiring the daily average humidity value, firstly, for each road segment and the corresponding time group set for each natural day, the road surface internal humidity monitoring records in the time group set are read one by one. The units of the collected road surface internal humidity readings are standardized and the range is checked. Readings that are missing, saturated, or significantly exceed the sensor's working range are marked as invalid data and discarded. Then, the remaining valid road surface internal humidity readings are summed, and the number of valid readings is counted. The sum is divided by the number of valid readings to obtain the arithmetic mean of the road surface internal humidity of the road segment for that natural day. The arithmetic mean is recorded as the daily average humidity value of the road segment for that natural day, and is stored in the daily-scale data record along with the road segment identifier and the natural day date for subsequent humidity boundary and risk assessment.

[0024] Specifically, when obtaining the daily maximum temperature gradient value, for each road segment and the corresponding time group set for each natural day, the road surface temperature reading and the deep road surface temperature reading in the time group set are read one by one. The road surface temperature at the same sampling time is subtracted from the deep road surface temperature to obtain the instantaneous temperature difference. The absolute value of this temperature difference is taken as the temperature gradient sample value at that time. During the traversal of the time group set, the current temperature gradient sample value is continuously compared with the recorded maximum temperature gradient sample value. The larger value is updated as the new maximum temperature gradient sample value. After traversing the time group set, the final maximum temperature gradient sample value is determined as the daily maximum temperature gradient value of the road segment for that natural day and written into the daily scale data record of the road segment for that natural day. This data is used for subsequent weighted linear fitting and calculation of the hysteresis wet loss temperature amplification index.

[0025] Specifically, when obtaining the normalized strain peak value, for each road segment and the time grouping set corresponding to each natural day, the axle load value range representing the target vehicle type is first preset. The preset axle load value range is determined according to the traffic composition and vehicle operation characteristics of the target road, and its setting process is based on statistical analysis of long-term traffic monitoring records: the axle load acquisition devices deployed on the road record a large number of axle load values ​​when a large number of vehicles pass through under different days and different traffic flow conditions. By classifying and statistically analyzing the above-mentioned historical axle load data, the typical load distribution range of different vehicle types under the road conditions is identified; then, combined with the provisions of the road design specifications on vehicle axle load levels, the axle load distribution center range of the corresponding vehicle type is taken as the effective stable load range for evaluation, so as to eliminate the strain interference caused by abnormally overloaded vehicles and lightly loaded vehicles; finally, the stable load is... The load range is determined as a preset range of axle load values, so that the calculation of the normalized strain peak value depends only on the stability of the vehicle's operating state. Monitoring records in this time group set that fall within the preset range of axle load are selected. The axle load and the strain peak value collected synchronously in these records are read one by one. The strain peak value is divided by the corresponding axle load to obtain the strain-load ratio of the record. This ratio is used as the normalized strain sample value. During the process of traversing all records that meet the preset range conditions, the current normalized strain sample value is continuously compared with the recorded maximum normalized strain sample value. The larger value is updated as the new maximum normalized strain sample value. After the traversal is completed, the final maximum normalized strain sample value is determined as the normalized strain peak value of the road segment for that natural day and written into the daily scale data record of the road segment for that natural day for subsequent construction of continuous weight and weighted linear fitting relationship.

[0026] S2. Construct continuous weights based on normalized strain peak values; In an embodiment of the present invention, a continuous weight is constructed based on the normalized strain peak value, including: Select each combination of date and road segment for which the daily average humidity, daily maximum temperature gradient and normalized strain peak value have been calculated simultaneously, and define the combination of date and road segment as a sample unit to form a sample unit set. Taking the normalized strain peak values ​​in all sample cells as the object, calculate the median of the normalized strain peak values ​​and the median of the absolute deviations relative to the median. Based on the median of the normalized strain peak and the median of the absolute deviation, the normalized strain peak of each sample unit is standardized to obtain the standardized deviation. Construct continuous weights for each sample unit in the sample unit set based on the standardized deviation. Specifically, after obtaining the daily average humidity value, daily maximum temperature gradient value, and normalized strain peak value, for each road segment's daily-scale data record under each natural day, the corresponding date identifier and road segment identifier are read. The combination of the date identifier and the road segment identifier is used as a unique identifier for a sample unit. The daily average humidity value, daily maximum temperature gradient value, and normalized strain peak value corresponding to this identifier are also stored in the sample unit record. All daily-scale data records are traversed one by one to form a sample unit set containing several sample unit records, which is used to organize subsequent statistical operations under a unified time scale and spatial scale.

[0027] Specifically, taking the normalized strain peak values ​​in the entire sample unit set as the statistical object, the normalized strain peak values ​​of all sample units are sorted according to their numerical values. The normalized strain peak value in the middle of the sorted sequence is taken as the median of the normalized strain peak values. Then, the median is subtracted from the normalized strain peak value of each sample unit, and the absolute value is taken to obtain a set of absolute deviation values. This set of absolute deviation values ​​is sorted again, and the absolute deviation value in the middle position is taken as the median of the absolute deviation, so that this pair of medians can represent the overall strain response level and its dispersion. Then, based on the aforementioned median of the normalized strain peak values ​​and the median of the absolute deviation, the standardization deviation processing is performed on each sample unit. The absolute value of the difference between the normalized strain peak value and the median of the sample unit is divided by the median of the absolute deviation to obtain the standardized deviation of the sample unit. This makes the strain deviation of different road sections and different time periods mapped to a unified dimensionless scale. The smaller the deviation, the closer the standardized deviation is to one; the larger the deviation, the larger the standardized deviation shows an increasing trend.

[0028] Specifically, continuous weights are constructed for each sample unit in the sample unit set based on the standardized deviation. A preset monotonically decreasing function converts the standardized deviation into a weight coefficient between zero and one. This ensures that sample units with smaller standardized deviations receive continuous weights close to one, while those with larger standardized deviations receive continuous weights close to zero. This highlights representative sample units with stable normalized strain peaks during subsequent weighted linear fitting, suppressing the influence of anomalously disturbed sample units on the overall temperature gradient and strain relationship modeling results. The monotonically decreasing function for the continuous weights is determined based on the typical strain response law of road structures under vehicle loads. Its setting process is based on the statistical trends of long-term monitoring data and the mechanical robustness principle of road materials. Under normal service conditions, the strain distribution of road structures exhibits a relatively concentrated main peak range with vehicle loads, meaning that the normalized strain peak of most samples fluctuates around the median. However, during sudden temperature changes, localized moisture anomalies, or accidental sensor disturbances, strain readings with larger deviations appear. These readings contribute unstablely to the temperature gradient and strain relationship model. To make the weighted linear fitting process more robust, after analyzing the standardized deviation of all sample units, it was observed that samples with larger standardized deviations are more likely to correspond to external disturbances or atypical mechanical responses. Therefore, based on this statistical law, a monotonically decreasing weight function was constructed with standardized deviation as the independent variable and sample validity decay as the trend. The monotonicity of this weight function directly reflects the negative correlation between the degree of deviation and the reliability, assigning a weight close to one to samples with small deviations to reflect the stability of their mechanical responses; and assigning a weight close to zero to samples with large deviations to reduce unnecessary interference from such samples on the overall model.

[0029] S3. Determine the humidity threshold value based on the daily average humidity value to obtain dry state samples and humid state samples; In an embodiment of the present invention, a humidity threshold value is determined based on the average daily humidity value to obtain dry state samples and humid state samples, including: From the sample unit set, read the date, road segment, daily average humidity value, daily maximum temperature gradient value, normalized strain peak value and continuous weight corresponding to each sample unit. Then, remove duplicates and sort all the daily average humidity values ​​that appear in each sample unit to obtain a set of candidate humidity boundary values ​​composed of daily average humidity values. For each candidate humidity threshold value in the candidate humidity threshold value set, the sample unit set is grouped according to whether the daily average humidity value is less than or equal to the candidate humidity threshold value, forming a dry state sample set and a humid state sample set. Specifically, the dry state sample set represents a group of samples that are determined to have low pavement moisture levels when all sample units are divided according to the humidity threshold. The daily average humidity value of each sample unit in this group is not greater than the humidity threshold. Its function is to provide information on the response of the pavement structure to temperature changes and vehicle loads under low moisture levels, which is used to establish a linear relationship under dry conditions. The wet state sample set represents a group of samples that are determined to have sufficient internal moisture or be under high humidity conditions when all sample units are divided according to the humidity threshold. The daily average humidity value of each sample unit in this group is greater than the humidity threshold. Its function is to provide information on the response of the pavement structure to temperature changes and vehicle loads under high humidity conditions, which is used to establish a linear relationship under wet conditions. By comparing the two types of sample sets, the influence of moisture level changes on the pavement structure response pattern can be revealed.

[0030] Specifically, when constructing the candidate humidity boundary value set, each sample unit record is first read one by one from the sample unit set to obtain the corresponding date, road segment, and pre-calculated daily average humidity value, daily maximum temperature gradient value, normalized strain peak value, and continuous weight. The date and road segment information are associated with subsequent backtracking, while the daily maximum temperature gradient value, normalized strain peak value, and continuous weight are stored in the sample unit record structure as is. During the traversal of all sample units, each read daily average humidity value is written into a temporary humidity sequence. Obviously missing or invalid daily average humidity values ​​are not added, so that the temporary humidity sequence completely records the daily average humidity values ​​of all valid sample units. After the traversal is completed, the temporary humidity sequence is first deduplicated, and the repeated daily average humidity values ​​are merged into a single value. Then, they are sorted according to the value size from smallest to largest to obtain a candidate humidity boundary value set consisting of multiple distinct daily average humidity values ​​arranged in ascending order. This candidate humidity boundary value set is associated with the sample unit set and stored to provide input for subsequent humidity boundary value search.

[0031] Specifically, when forming dry and humid state sample sets using the candidate humidity boundary value set, the candidate humidity boundary values ​​are first read one by one from the candidate humidity boundary value set. The currently read candidate humidity boundary value is used as the dividing criterion. All sample unit records in the sample unit set are then re-traversed. For each sample unit, its daily average humidity value is read, and it is determined whether the daily average humidity value is less than or equal to the current candidate humidity boundary value. If the daily average humidity value is less than or equal to the current candidate humidity boundary value, the sample unit, along with its date, road segment, daily maximum temperature gradient value, normalized strain peak value, and continuous weight, is added to the current sample set. In the dry state sample set corresponding to the candidate value; when the daily average humidity value is greater than the current candidate humidity threshold, the sample unit and all its associated data are added to the humid state sample set corresponding to the current candidate value; after completing the traversal of all sample units, a pair of complementary dry state sample sets and humid state sample sets are formed for the current candidate humidity threshold, and the pair of sample sets is associated with the corresponding candidate humidity threshold. The above process is repeated until all candidate values ​​in the candidate humidity threshold set have been processed, thus laying the sample foundation for the subsequent weighted linear fitting and residual calculation for each candidate humidity threshold.

[0032] In the dry and wet sample sets, the daily maximum temperature gradient value is used as the independent variable, the normalized strain peak value is used as the dependent variable, and continuous weights are used as weights to establish two sets of weighted linear relationships, so as to obtain the linear model parameters under the dry state and the linear model parameters under the wet state. Calculate the weighted residual sum corresponding to the candidate humidity boundary value based on the weighted error of the two linear models on their respective sample sets; Among all candidate humidity cutoff values, the candidate humidity cutoff value with the smallest weighted residual sum is selected as the final humidity cutoff value. The set of dry state samples corresponding to the final humidity threshold value is determined as the final dry state sample set. The set of humid state samples corresponding to the final humidity threshold value is determined as the final humid state sample set. Specifically, the weighted linear model parameters represent the slope and intercept of the model obtained by weighted least squares fitting when constructing the weighted linear relationship. They are used to characterize the linear response relationship between the daily maximum temperature gradient value and the normalized strain peak value. The weighted residual sum represents the total deviation obtained by summing the weighted residuals of all sample points after establishing weighted linear models on the dry and humid sample sets corresponding to each candidate humidity boundary value. It is used to evaluate whether the candidate humidity boundary value can enable the two sets of samples to form a stable and separated linear relationship. The final humidity boundary value represents the humidity value with the smallest weighted residual sum among all candidate humidity boundary values. It is used as the final humidity threshold to distinguish between dry and humid samples.

[0033] Specifically, in the dry and humid sample sets, the daily maximum temperature gradient value corresponding to each sample unit is used as the independent variable, the normalized strain peak value is used as the dependent variable, and continuous weights are used as weighting coefficients. Weighted least squares fitting operations are performed on the dry and humid sample sets respectively. During the fitting process, the linear relationship between the independent and dependent variables is calculated for each sample unit, so that the sample units with large weights and stable data have a more prominent impact on the fitting results. Finally, the linear model parameters under the dry and humid conditions are obtained, which are used to characterize the response law between the daily maximum temperature gradient value and the normalized strain peak value under the two humid conditions.

[0034] Specifically, after obtaining the linear model parameters under dry and humid conditions, for the current candidate humidity threshold, each sample unit in the dry state sample set is substituted into the corresponding dry state linear model to calculate the predicted normalized strain peak value given by the model. The difference between the predicted and actual normalized strain peak values ​​is then calculated to obtain the error. The square of the error is multiplied by the continuous weight of the sample unit and summed to obtain the weighted residual on the dry state sample set. Similarly, each sample unit in the humid state sample set is substituted into the humid state linear model to calculate the predicted value and obtain the corresponding weighted residual. The weighted residuals of the two sample sets are summed to obtain the weighted residual sum corresponding to the current candidate humidity threshold value, which is used to measure the overall fitting deviation of the two linear models to all samples under the candidate humidity threshold value.

[0035] Specifically, after all candidate humidity thresholds have undergone weighted linear fitting and weighted residual calculation, the weighted residual sums corresponding to each candidate humidity threshold are compared sequentially. The candidate humidity threshold with the smallest value is recorded and determined as the final humidity threshold. This ensures that the dry state sample set and the humid state sample set corresponding to the final humidity threshold have the smallest weighted fitting error in an overall sense, thereby selecting the humidity threshold that can most stably distinguish between dry and humid states from all candidate partitioning methods.

[0036] Specifically, after determining the final humidity threshold, this threshold is used as the dividing standard. Returning to the sample unit set, the daily average humidity value is read for each sample unit. When the daily average humidity value is less than or equal to the final humidity threshold, the sample unit, along with its date, road segment, daily maximum temperature gradient value, normalized strain peak value, and continuous weights, are all included in the final dry state sample set. This forms the dry state sample set under optimal humidity conditions, providing a data foundation for subsequent statistical analysis and model evaluation focused solely on the dry state. Similarly, using the final humidity threshold as the dividing standard, the remaining sample units in the sample unit set are screened. When the daily average humidity value is greater than the final humidity threshold, the sample unit and all its associated data are included in the final humid state sample set. This ensures that the final humid state sample set contains only sample units collected under high humidity conditions, thus providing a complete and consistent data source for constructing response models and analyzing hysteresis wet damage effects under humid conditions, while maintaining correspondence with the final dry state sample set.

[0037] S4. Perform a weighted linear fit on the dry and wet samples, and obtain the hysteresis wet loss temperature amplification index based on the slope ratio of the weighted linear relationship between the dry and wet states. In an embodiment of the present invention, a weighted linear fit is performed on dry state samples and wet state samples, and the hysteresis wet loss temperature amplification index is obtained based on the ratio of the slope of the weighted linear relationship between the dry and wet states, including: In the final dry state sample set, the daily maximum temperature gradient value is used as the independent variable, the normalized strain peak value is used as the dependent variable, and the continuous weight is used as the weight to establish a weighted linear relationship under the dry state. We performed a weighted least squares fit on the weighted linear relationship under the dry state to obtain the slope value of the temperature sensitivity under the dry state and the uncertainty index corresponding to the slope value of the temperature sensitivity under the dry state. In the final humid state sample set, the daily maximum temperature gradient value is used as the independent variable, the normalized strain peak value is used as the dependent variable, and the continuous weight is used as the weight to establish a weighted linear relationship under humid conditions. We performed a weighted least squares fit on the weighted linear relationship under humid conditions to obtain the slope value of the temperature sensitivity under humid conditions and the uncertainty index corresponding to the slope value of the temperature sensitivity under humid conditions. Specifically, when establishing a weighted linear relationship under the dry state in the final dry state sample set, the daily maximum temperature gradient value, normalized strain peak value, and continuous weights corresponding to each sample unit are first read from the final dry state sample set. The daily maximum temperature gradient value is used as the independent variable, the normalized strain peak value as the dependent variable, and the continuous weights as the weighting coefficients of the sample unit. Then, all sample units are organized according to a unified linear model form, and the weighted average value of the independent variable and the weighted average value of the dependent variable are calculated. Based on this, the co-variance between the independent and dependent variables and the weighted dispersion of the independent variable itself are obtained. Then, the slope and intercept of the weighted linear relationship under the dry state are obtained by using the weighted least squares method. This set of slopes and intercepts is used as the parameters of the weighted linear relationship under the dry state to describe the influence of the daily maximum temperature gradient value on the normalized strain peak value under the dry state.

[0038] Specifically, after obtaining the weighted linear relationship under dry conditions, when performing weighted least squares fitting evaluation on the weighted linear relationship under dry conditions, the final dry state sample set is traversed again. The daily maximum temperature gradient value of each sample unit is substituted into the aforementioned linear relationship to calculate the predicted normalized strain peak value, and the difference between the predicted and actual normalized strain peak values ​​is obtained to obtain the residual. The square of the residual is multiplied by the corresponding continuous weight to form the weighted residual. The weighted residuals of all sample units are accumulated to obtain the weighted residual sum. On this basis, the variance estimate and standard deviation estimate of the slope parameter are calculated. The standard deviation estimate, which is jointly determined by the weighted residual sum and the sample size, is used as the uncertainty index corresponding to the temperature sensitivity slope value under dry conditions. At the same time, the slope of the aforementioned linear relationship is used as the temperature sensitivity slope value under dry conditions, thereby providing a measure of the temperature response intensity and its reliability under dry conditions for the subsequent construction of the hysteresis wet loss temperature amplification index.

[0039] Specifically, when establishing the weighted linear relationship under humid conditions in the final humid state sample set, the same processing procedure as for the dry state is adopted. First, the daily maximum temperature gradient value, normalized strain peak value, and continuous weight of each sample unit are read one by one from the final humid state sample set. The daily maximum temperature gradient value is set as the independent variable, the normalized strain peak value is set as the dependent variable, and the continuous weight is used as the weighting coefficient. The weighted average value, co-variance, and weighted dispersion of the independent and dependent variables under humid conditions are calculated. Based on this, the slope and intercept of the weighted linear relationship under humid conditions are solved by the weighted least squares method. This set of slope and intercept is used as the weighted linear relationship parameters under humid conditions to characterize the response strength between the daily maximum temperature gradient value and the normalized strain peak value under humid conditions.

[0040] Specifically, after obtaining the weighted linear relationship under humid conditions, when performing weighted least squares fitting evaluation on the weighted linear relationship under humid conditions, the final humid state sample set is traversed. The daily maximum temperature gradient value of each sample unit is substituted into the humid state linear relationship to calculate the predicted normalized strain peak value. The residual is obtained by the difference between the predicted value and the actual normalized strain peak value. The square of the residual is multiplied by the corresponding continuous weight and summed over all sample units to obtain the weighted residual sum under humid conditions. Based on this, the variance estimate and standard deviation estimate of the slope parameter are obtained. The standard deviation estimate is used as the uncertainty index corresponding to the temperature sensitivity slope value under humid conditions. At the same time, the slope of the weighted linear relationship under humid conditions is used as the temperature sensitivity slope value under humid conditions. Thus, the amplification degree and stability of the normalized strain peak value by the temperature gradient under humid conditions are obtained, which is used to compare with the results under dry conditions and further construct the hysteresis wet loss temperature amplification index.

[0041] The ratio of the temperature sensitivity slope value under humid conditions to the temperature sensitivity slope value under dry conditions is used as the hysteresis moisture loss temperature amplification index. The difference between the slope values ​​of temperature sensitivity in the humid state and the slope values ​​of temperature sensitivity in the dry state, the uncertainty index corresponding to the slope value of temperature sensitivity in the dry state, and the uncertainty index corresponding to the slope value of temperature sensitivity in the humid state are used to form the evidence strength index. Specifically, the hysteresis wet-damage temperature amplification index describes the degree to which a temperature gradient amplifies the structural strain response of a road segment under humid conditions. Its physical meaning lies in characterizing the amplifying effect of a humid environment on temperature-induced strain growth. This index is calculated by comparing the slope value of the temperature sensitivity under humid conditions to that under dry conditions. The slope value of the temperature sensitivity under humid conditions reflects the rate of strain change caused by the maximum daily temperature gradient under humid conditions, while the slope value of the temperature sensitivity under dry conditions reflects the rate of strain change caused by the maximum daily temperature gradient under dry conditions for the same road segment. A larger ratio indicates a greater amplification of strain under humid conditions due to the same temperature gradient change, thus meaning that materials or structures are more prone to hysteresis wet-damage under humid conditions. As a dimensionless parameter, the hysteresis wet-damage temperature amplification index can be used to quantitatively analyze the influence of humidity on the temperature-strain transmission law.

[0042] Specifically, after obtaining the temperature sensitivity slope values ​​for both dry and wet conditions, the data processing module first retrieves the slope results for both states. A division operation is performed, using the wet condition temperature sensitivity slope value as the dividend and the dry condition temperature sensitivity slope value as the divisor, to obtain a dimensionless ratio. This ratio is defined as the hysteresis wet damage temperature amplification index, used to characterize the amplification factor of strain change caused by temperature gradient under wet conditions relative to dry conditions for the same road segment. This provides a single numerical value to characterize the additional temperature sensitivity effect introduced by humidity, offering a unified scale for subsequent wet damage risk assessment and comparison of amplification effects between different road segments.

[0043] Specifically, when constructing the evidence strength index, the slope values ​​of temperature sensitivity in humid and dry states are first read, and the difference between them is calculated as the change in temperature sensitivity. Simultaneously, the uncertainty indices corresponding to the slope values ​​of temperature sensitivity in dry and humid states are read. These two uncertainty indices are combined into a comprehensive uncertainty measure by summing or weighted summing. Then, the evidence strength index is defined by the ratio or mapping relationship between the change in temperature sensitivity and the comprehensive uncertainty measure. The above difference and comprehensive uncertainty are written into the evidence strength index calculation unit. A larger evidence strength index is given for cases where the change is much greater than the comprehensive uncertainty, and a smaller evidence strength index is given for cases where the change is close to or less than the comprehensive uncertainty. Thus, a dimensionless value is used to reflect whether the hysteresis wet damage temperature amplification index is supported by stable and reliable data differences, providing a basis for subsequent judgment of the significance of the wet damage effect and selection of highly reliable sample road sections.

[0044] S5. Calculate the average wet damage risk of the road segment based on the hysteresis wet damage temperature amplification index, continuous weights and daily maximum temperature gradient; In embodiments of the present invention, the average wet loss risk of a road segment is calculated based on the hysteresis wet loss temperature amplification index, continuous weight, and daily maximum temperature gradient, including: In the final set of humid conditions, the daily maximum temperature gradient values ​​are weighted and averaged with continuous weights to obtain the global weighted average temperature gradient under humid conditions. Extract all road segment identifiers that have appeared in all sample units to construct a road segment set; In all sample units, for each combination of date and road segment, the day is determined to be either a wet or dry day based on the relationship between the average daily humidity value and the humidity threshold value of the combination. Specifically, in the final wet state sample set, when performing a weighted average of the daily maximum temperature gradient value with continuous weights, the process first iterates through each sample unit in the final wet state sample set, sequentially reading the daily maximum temperature gradient value recorded in that sample unit and its corresponding continuous weight. For the sample units involved in the calculation, the daily maximum temperature gradient value is multiplied by the continuous weight and then summed to form a weighted sum. At the same time, the continuous weights themselves are summed to form a weighted sum. After all sample units have been traversed, the weighted sum is divided by the weighted sum to obtain a unique scalar. This scalar serves as the global weighted average temperature gradient under wet conditions and is written into the global parameter set for wet damage analysis. It is used subsequently for scale normalization and relative intensity comparison of the daily maximum temperature gradient values ​​of various date and road segment combinations when constructing the daily scale wet damage risk contribution, thereby providing a unified temperature gradient benchmark for the comparability of wet damage risks between different road segments.

[0045] Specifically, when extracting all occurrences of road segment identifiers from all sample units and constructing a road segment set, an empty set is first pre-defined in the data structure to store road segment identifiers. Then, each sample unit record in the sample unit set is traversed, and the road segment identifier corresponding to that sample unit is read from it. The road segment identifier is compared one by one with the existing road segment identifiers in the set. If the set does not yet contain the road segment identifier, the road segment identifier is added to the set. If the set already contains the road segment identifier, it is not added again, until all sample unit records have been traversed. The final set contains only unique road segment identifiers. This set is defined as a road segment set, and a reference relationship is established between it and the sample unit set and the subsequent road segment-level wet damage risk indicators, so that the subsequent risk summarization and sorting operations can use the road segment set as the outer loop unit.

[0046] Specifically, when determining whether a combination of dates and road segments belongs to a wet or dry state day based on the relationship between the average daily humidity value and the humidity threshold value in all sample units, the final humidity threshold value determined in the aforementioned humidity threshold determination step is first read and used as a globally unified judgment threshold. Then, each sample unit in the sample unit set is traversed, and its date and road segment identifiers are read to clarify the date and road segment combination it belongs to. At the same time, the average daily humidity value corresponding to the sample unit is read and compared with the final humidity threshold value. When the average daily humidity value is greater than the final humidity threshold value, the sample unit is logically marked as a wet state day sample. When the average daily humidity value is less than or equal to the final humidity threshold value, the sample unit is marked as a dry state day sample, and this status mark is written into the status field of the sample unit, so that each date and road segment combination is clearly classified as a wet or dry state day.

[0047] When it is a wet day, the ratio of the daily maximum temperature gradient value to the global weighted average temperature gradient is calculated. Based on the product of the ratio and the hysteresis wet loss temperature amplification index and the continuous weight, the daily scale wet loss risk contribution of the date and road segment combination is obtained. When the day is in a dry state, the daily scale wet damage risk contribution of the date and road segment combination is set to zero; In the set of road segments, for each road segment, the daily scale wet damage risk contribution of that road segment on all dates is summed to obtain the total wet damage risk of that road segment; Specifically, the total wet damage risk is used to characterize the cumulative degree of structural damage caused by the combined effects of humidity and temperature on a road segment within an evaluation period. It means that the daily-scale wet damage risk contribution of the road segment is accumulated day by day across all dates, resulting in a comprehensive quantitative index that reflects the strength of the long-term wet damage effect. The total wet damage risk originates from the daily-scale wet damage risk contribution, which is obtained by multiplying the ratio of the daily maximum temperature gradient to the globally weighted average temperature gradient on humid days by the hysteresis wet damage temperature amplification index and continuous weights. This value describes the strain amplification level caused by the combined effects of temperature changes and humidity hysteresis on a single day. Accumulating this value over all dates yields the cumulative intensity of the wet damage impact on the road segment throughout the entire evaluation period. The larger the total wet damage risk, the more significant the cumulative wet damage effect experienced by the road segment, and the more likely the structural materials are to experience long-term fatigue damage, performance degradation, or potential durability risks. Therefore, this index can serve as an important basis for ranking the wet damage sensitivity of road segments and making subsequent maintenance decisions.

[0048] Specifically, when a certain date and road segment combination is marked as a wet state day, the data processing module reads the daily maximum temperature gradient value corresponding to that date and road segment combination and the continuous weight of that combination from the sample unit set. At the same time, it reads the global weighted average temperature gradient and the hysteresis wet damage temperature amplification index under wet conditions from the aforementioned global parameters. It compares the daily maximum temperature gradient value of the combination with the global weighted average temperature gradient to obtain the temperature gradient intensity ratio. Then, it multiplies this ratio by the hysteresis wet damage temperature amplification index and the continuous weight in sequence to obtain the daily scale wet damage risk contribution of that date and road segment combination. This contribution is written into the daily scale risk field of the corresponding sample unit to characterize the wet damage risk intensity of that combination on that day due to the combined effect of temperature gradient and humidity amplification.

[0049] Specifically, when a combination of a date and a road segment is marked as a dry state day, the data processing module also reads the date identifier and road segment identifier of the combination, as well as the corresponding status label. After confirming that the status is a dry state day, it no longer performs the calculation of multiplying the temperature gradient intensity ratio and the hysteresis wet loss temperature amplification index. Instead, it directly assigns the daily-scale wet loss risk contribution of the date and road segment combination to zero and writes it into the daily-scale risk field. In this way, it is clearly indicated that the humidity amplification effect of the combination under dry conditions can be ignored and does not make a positive contribution to the long-term accumulation of wet loss. This makes the subsequent risk summary driven only by the wet state day, which is more in line with the wet loss formation mechanism.

[0050] Specifically, after obtaining the daily-scale wet damage risk contribution for all date and road segment combinations, the data processing module sequentially selects each road segment in the road segment set. For the sample unit records of the current road segment on all dates, it reads the daily-scale wet damage risk contribution for each day. These contributions are accumulated in chronological order or directly in the order of records to obtain the total wet damage risk of the road segment within the evaluation period. The total wet damage risk and the road segment identifier are then written into the road segment-level risk result table, thereby forming a road segment-level comprehensive index that can reflect the long-term cumulative effect of wet damage. This provides basic data for subsequent sorting of road segments according to the size of wet damage risk, screening of key maintenance objects, and assessment of the engineering impact of the lagged wet damage temperature amplification index.

[0051] In the final wet condition sample set, the continuous weights of all samples belonging to the same road segment are accumulated to obtain the weighted observation total of the road segment under wet conditions; Divide the total wet damage risk by the weighted total observation to obtain the average wet damage risk index of the road section. When the weighted total observation is zero, the average wet damage risk index is set to zero. Specifically, the weighted total observations is a comprehensive value representing the number of samples effectively observed in a road segment under wet conditions and their importance. Its formation process involves taking the continuous weights of all samples belonging to the same road segment from the final wet condition sample set, and then summing these continuous weights. This ensures that samples with more observations or higher weights contribute more to the weighted total observations. The continuous weights themselves originate from the standardized deviation of the normalized strain peak value, reflecting the wet damage sensitivity represented by the sample. The weighted total observations include both observational information and a reflection of sample reliability. The average wet damage risk index for a road segment is used to describe the intensity of wet damage risk corresponding to a unit of effective observation during a wet period for a certain road segment. This index is obtained by dividing the total wet damage risk by the total weighted observation, thereby normalizing the total wet damage risk according to the effective number of samples and their continuous weights. This ensures that the risk assessment results are not artificially amplified when there are many observations or large sample weights. When the total weighted observation is zero, it means that no wet condition samples have appeared in the road segment. At this time, an effective wet damage risk assessment cannot be formed. Therefore, the average wet damage risk index is set to zero. This index reflects the actual wet damage exposure level of the road segment during the stage dominated by the humidity amplification effect and can serve as an important basis for ranking and priority maintenance decisions.

[0052] Specifically, in the final wet state sample set, to calculate the weighted total observations of each road segment under wet conditions, an accumulation container with the road segment identifier as the key is first established in the data structure. Then, each sample unit in the final wet state sample set is traversed, and the corresponding road segment identifier and continuous weight value are read. For sample units with the same road segment identifier, their continuous weights are accumulated into the corresponding accumulation container. After the traversal is completed, each road segment identifier corresponds to a value obtained by accumulating the continuous weights of all wet state samples. This value is recorded as the weighted total observations of the road segment under wet conditions and stored together with the road segment identifier in the road segment-level statistical results table. The larger the weighted total observations, the more effective observations the road segment has obtained under wet conditions and the higher the sample weight.

[0053] Specifically, after obtaining the total wet damage risk and weighted observation total for each road segment, the data processing module reads each record of each road segment in the road segment-level statistical results table to calculate the average wet damage risk index for the road segment. First, it determines whether the weighted observation total for the road segment is greater than zero. When the weighted observation total is greater than zero, the corresponding total wet damage risk is used as the dividend, and the weighted observation total is used as the divisor to perform a division operation. The quotient obtained is used as the average wet damage risk index for the road segment, and this index is written into the road segment-level risk results table. When the weighted observation total is equal to zero, it means that no effective wet condition sample was formed for the road segment during the evaluation period. In this case, the average wet damage risk index is directly assigned to zero and recorded in the results table. Through the above operations, the wet damage assessment results of different road segments can still be normalized to a unified scale even if the number of observations and sample quality are inconsistent. The average wet damage risk index reflects the wet damage risk intensity corresponding to a unit of effective observation and can be used for subsequent risk ranking and maintenance priority division among road segments.

[0054] S6. Sort the road segments according to their average wet damage risk and generate a wet damage risk ranking view. In an embodiment of the present invention, road segments are sorted according to their average wet damage risk to generate a wet damage risk ranking view, including: In the road segment set, all road segment identifiers are arranged in descending order of average wet damage risk index to obtain an ordered road segment list; In the list of road segments, each road segment is assigned a unique sorting number, and the smaller the number, the higher the risk of wet damage. In the orderly road segment list, a record is generated for each road segment, including sorting number, road segment identifier, average wet damage risk index, hysteretic wet damage temperature amplification index, evidence strength index and humidity threshold value. All records are arranged in ascending order of sorting number to form a wet damage risk sorting view. Specifically, when arranging all road segment identifiers in the road segment set according to the average wet damage risk index from largest to smallest, a road segment-level statistical table is first created in the data structure. The table records the road segment identifier and the corresponding average wet damage risk index for each road segment. Then, the sorting program is called to traverse all records, using the average wet damage risk index as the sorting key, placing records with larger values ​​first and records with smaller values ​​last. After completion, an ordered list of road segments without duplicate road segment identifiers is output. After obtaining the ordered list of road segments, when assigning a unique sorting number to each road segment, the starting position of the ordered road segment list is taken as the first position. The road segment record in the first row is assigned sorting number one, the road segment record in the next row is assigned sorting number two, and so on until the end of the list. This ensures that each road segment identifier corresponds to a unique positive integer sorting number, and this sorting number field is written into the road segment-level statistics table. In this way, a strict one-to-one correspondence is ensured between the sorting number and the average wet damage risk index. The smaller the sorting number, the higher the average wet damage risk index of the road segment is among all road segments, which makes it easier to quickly identify road segments with higher wet damage risk levels by directly using the sorting number during subsequent display and retrieval.

[0055] Specifically, when generating a wet damage risk ranking view from an ordered list of road segments, the process first reads the ranking number, road segment identifier, average wet damage risk index, and corresponding hysteresis wet damage temperature amplification index, evidence strength index, and humidity threshold value for each road segment record. These fields are then combined to form a complete road segment risk record. All road segment risk records are then rearranged in ascending order of ranking number to form a structured wet damage risk ranking view. This view is stored as a result table that can be displayed on the front end and accessed by external systems. This wet damage risk ranking view retains the order of risk levels between road segments and integrates key parameters such as average wet damage risk level, humidity amplification effect strength, and evidence reliability within the same record. This allows maintenance personnel to complete road segment ranking, risk assessment, and mechanism interpretation within a unified interface when browsing the view, significantly improving the readability and decision support value of big data intelligent analysis results.

[0056] like Figure 2 The diagram shown is a functional block diagram of a big data intelligent analysis system provided in an embodiment of the present invention.

[0057] In this embodiment, the functions of each module / unit are as follows: The data acquisition module is used to collect multi-source monitoring data for road surface monitoring, and to determine the daily average humidity, daily maximum temperature gradient and normalized strain peak value of the road section based on the multi-source monitoring data. The weighting construction module is used to construct continuous weights based on the normalized strain peak value. The humidity boundary module is used to determine the humidity boundary value based on the daily average humidity value, and obtain dry state samples and humid state samples. The amplification fitting module is used to perform weighted linear fitting on dry and wet samples, and obtain the hysteresis wet loss temperature amplification index based on the slope ratio of the weighted linear relationship between dry and wet conditions. The risk calculation module is used to calculate the average wet damage risk of road sections based on the hysteresis wet damage temperature amplification index, continuous weights, and daily maximum temperature gradient. The risk ranking module is used to rank road segments according to their average wet damage risk and generate a wet damage risk ranking view.

[0058] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A big data intelligent analysis method, characterized in that, include: S1. Collect multi-source monitoring data for road surface monitoring, and determine the daily average humidity, daily maximum temperature gradient and normalized strain peak value of the road section based on the multi-source monitoring data. S2. Construct continuous weights based on normalized strain peak values; S3. Determine the humidity threshold value based on the daily average humidity value to obtain dry state samples and humid state samples; S4. Perform a weighted linear fit on the dry and wet samples, and obtain the hysteresis wet loss temperature amplification index based on the slope ratio of the weighted linear relationship between the dry and wet states. S5. Calculate the average wet damage risk of the road segment based on the hysteresis wet damage temperature amplification index, continuous weights and daily maximum temperature gradient; S6. Sort the road segments according to their average wet damage risk and generate a wet damage risk sorting view.

2. The big data intelligent analysis method according to claim 1, characterized in that, Multi-source monitoring data is collected for road surface monitoring. Based on the multi-source monitoring data, the daily average humidity, daily maximum temperature gradient, and normalized strain peak value of the road section are determined, including: Collect multi-source monitoring data for road surface monitoring, including: road surface temperature, road surface deep temperature, road surface internal humidity, axle load, and strain peak value corresponding to the axle load; Multi-source monitoring data are labeled using unified time and road segment identifiers; Using the natural day as the time unit, all multi-source monitoring data for each road segment on the same date are grouped to form a time group set for that road segment on that date. The daily average humidity value of the road segment is obtained by averaging the internal humidity of the road surface in the time group set. The absolute difference between the surface temperature and the deep surface temperature in the time group set is calculated, and the maximum value of the absolute difference is taken as the daily maximum temperature gradient value. In the time group set, select the axle load within the preset range, calculate the ratio of the corresponding strain peak value to the axle load, and take the maximum value of the ratio as the normalized strain peak value.

3. The big data intelligent analysis method according to claim 1, characterized in that, Continuous weights are constructed based on normalized strain peak values, including: Select each combination of date and road segment for which the daily average humidity, daily maximum temperature gradient and normalized strain peak value have been calculated simultaneously, and define the combination of date and road segment as a sample unit to form a sample unit set. Taking the normalized strain peak values ​​in all sample cells as the object, calculate the median of the normalized strain peak values ​​and the median of the absolute deviations relative to the median. Based on the median of the normalized strain peak and the median of the absolute deviation, the normalized strain peak of each sample unit is standardized to obtain the standardized deviation. Based on the standardized deviation, construct continuous weights for each sample unit in the sample unit set.

4. The big data intelligent analysis method according to claim 3, characterized in that, Humidity thresholds are determined based on daily average humidity values, resulting in dry and humid condition samples, including: From the sample unit set, read the date, road segment, daily average humidity value, daily maximum temperature gradient value, normalized strain peak value and continuous weight corresponding to each sample unit. Then, remove duplicates and sort all the daily average humidity values ​​that appear in each sample unit to obtain a set of candidate humidity boundary values ​​composed of daily average humidity values. For each candidate humidity threshold value in the candidate humidity threshold value set, the sample unit set is grouped according to whether the daily average humidity value is less than or equal to the candidate humidity threshold value, forming a dry state sample set and a humid state sample set. In the dry and wet sample sets, the daily maximum temperature gradient value is used as the independent variable, the normalized strain peak value is used as the dependent variable, and continuous weights are used as weights to establish two sets of weighted linear relationships, so as to obtain the linear model parameters under the dry state and the linear model parameters under the wet state. Calculate the weighted residual sum corresponding to the candidate humidity boundary value based on the weighted error of the two linear models on their respective sample sets; Among all candidate humidity cutoff values, the candidate humidity cutoff value with the smallest weighted residual sum is selected as the final humidity cutoff value. The set of dry state samples corresponding to the final humidity threshold value is determined as the final dry state sample set. The set of humid state samples corresponding to the final humidity threshold value is determined as the final humid state sample set.

5. The big data intelligent analysis method according to claim 4, characterized in that, A weighted linear fit is performed on dry and humid samples. The hysteresis temperature amplification index is obtained based on the ratio of the slopes of the weighted linear relationships under dry and humid conditions, including: In the final dry state sample set, the daily maximum temperature gradient value is used as the independent variable, the normalized strain peak value is used as the dependent variable, and the continuous weight is used as the weight to establish a weighted linear relationship under the dry state. We performed a weighted least squares fit on the weighted linear relationship under the dry state to obtain the slope value of the temperature sensitivity under the dry state and the uncertainty index corresponding to the slope value of the temperature sensitivity under the dry state. In the final humid state sample set, the daily maximum temperature gradient value is used as the independent variable, the normalized strain peak value is used as the dependent variable, and the continuous weight is used as the weight to establish a weighted linear relationship under humid conditions. We performed a weighted least squares fit on the weighted linear relationship under humid conditions to obtain the slope value of the temperature sensitivity under humid conditions and the uncertainty index corresponding to the slope value of the temperature sensitivity under humid conditions. The ratio of the temperature sensitivity slope value under humid conditions to the temperature sensitivity slope value under dry conditions is used as the hysteresis moisture loss temperature amplification index. The strength of evidence index is composed of the difference between the slope values ​​of temperature sensitivity in the humid state and the slope values ​​of temperature sensitivity in the dry state, the uncertainty index corresponding to the slope value of temperature sensitivity in the dry state, and the uncertainty index corresponding to the slope value of temperature sensitivity in the humid state.

6. The big data intelligent analysis method according to claim 5, characterized in that, The average wet damage risk of road sections is calculated based on the hysteresis wet damage temperature amplification index, continuous weights, and daily maximum temperature gradient, including: In the final set of humid conditions, the daily maximum temperature gradient values ​​are weighted and averaged with continuous weights to obtain the global weighted average temperature gradient under humid conditions. Extract all road segment identifiers that have appeared in all sample units to construct a road segment set; In all sample units, for each combination of date and road segment, the day is determined to be either a wet or dry day based on the relationship between the average daily humidity value and the humidity threshold value of the combination. When it is a wet day, the ratio of the daily maximum temperature gradient value to the global weighted average temperature gradient is calculated. Based on the product of the ratio and the hysteresis wet loss temperature amplification index and the continuous weight, the daily scale wet loss risk contribution of the date and road segment combination is obtained. When the day is in a dry state, the daily scale wet damage risk contribution of the date and road segment combination is set to zero; In the set of road segments, for each road segment, the daily scale wet damage risk contribution of that road segment on all dates is summed to obtain the total wet damage risk of that road segment; In the final wet condition sample set, the continuous weights of all samples belonging to the same road segment are accumulated to obtain the weighted observation total of the road segment under wet conditions; Divide the total wet damage risk by the weighted total observation to obtain the average wet damage risk index of the road segment. When the weighted total observation is zero, the average wet damage risk index is set to zero.

7. The big data intelligent analysis method according to claim 6, characterized in that, The road segments are sorted according to their average wet damage risk, generating a wet damage risk ranking view, including: In the road segment set, all road segment identifiers are arranged in descending order of average wet damage risk index to obtain an ordered road segment list; In the list of road segments, each road segment is assigned a unique sorting number, and the smaller the number, the higher the risk of wet damage. In the list of ordered road segments, a record is generated for each road segment, including a sorting number, road segment identifier, average wet damage risk index, hysteresis wet damage temperature amplification index, evidence strength index, and humidity threshold value. All records are arranged in ascending order of sorting number to form a sorting view of wet damage risk.

8. A big data intelligent analysis system, characterized in that, The system includes: The data acquisition module is used to collect multi-source monitoring data for road surface monitoring, and to determine the daily average humidity, daily maximum temperature gradient and normalized strain peak value of the road section based on the multi-source monitoring data. The weighting construction module is used to construct continuous weights based on the normalized strain peak value. The humidity boundary module is used to determine the humidity boundary value based on the daily average humidity value, and obtain dry state samples and humid state samples. The amplification fitting module is used to perform weighted linear fitting on dry and wet samples, and obtain the hysteresis wet loss temperature amplification index based on the slope ratio of the weighted linear relationship between dry and wet conditions. The risk calculation module is used to calculate the average wet damage risk of road sections based on the hysteresis wet damage temperature amplification index, continuous weights, and daily maximum temperature gradient. The risk ranking module is used to rank road segments according to their average wet damage risk and generate a wet damage risk ranking view.

Images