A method for controlling the spreading amount of de-bonding anti-icing material based on road surface structure depth

Abstract

The application discloses a kind of based on road surface construction depth's debonding ice material spreading quantity control method, belong to data processing and machine learning technical field, including data acquisition and segmented processing, obtain material performance parameter, model module calculates target spreading quantity, control module issues instruction to solve the technical problem that based on construction feedback data carries out adaptive incremental update to spreading quantity calculation model, to improve the precision of spreading quantity prediction and model environmental adaptability, the present application carries out parameter regression update or model retraining to spreading quantity calculation model using post-construction feedback data, can realize the dynamic adaptive optimization of model on the basis of preserving original model knowledge, significantly improve the prediction accuracy and control robustness of spreading quantity under different construction environments and material conditions.

Description

Technical Field

[0001] This invention belongs to the field of data processing and machine learning technology, and relates to a method for controlling the amount of de-adhesive and anti-icing material sprayed based on pavement texture depth. In particular, it relates to a method for using pavement texture depth data, environmental parameters and material performance parameters as model input features, and realizing model parameter updates and performance optimization through incremental learning or batch retraining. Background Technology

[0002] De-icing and de-icing materials and coating technologies are a new type of active de-icing method proposed in the field of road maintenance in recent years. Compared with traditional de-icing technologies, they have the characteristics of active melting of ice and snow, excellent environmental performance, long-lasting and efficient de-icing, and preventive functions. Therefore, they are increasingly widely used in asphalt pavement de-icing projects. When de-icing and de-icing materials are applied to the road surface, it is necessary to determine the application rate per unit area.

[0003] Existing research indicates that the application rate is closely related to factors such as pavement texture depth, material curing time, and the material's de-icing effect. However, current methods for determining the application rate of de-icing and de-adhesive materials mainly rely on the single parameter of pavement texture depth, failing to fully consider the dynamic changes in environmental parameters (temperature, humidity, snowfall frequency, etc.) and material performance parameters. This results in low accuracy in application rate calculation and poor environmental adaptability.

[0004] More importantly, existing spraying volume calculation models are usually established through offline one-time training. After the model is put into use, it is impossible to dynamically update the model based on feedback information such as the actual spraying volume collected during construction, road surface texture depth remeasurement data, anti-icing performance data, and de-icing effect evaluation data.

[0005] When construction conditions, environmental factors, or material batches change, model prediction biases gradually accumulate, significantly reducing the effectiveness of spraying control. Although some methods attempt to periodically retrain the model, complete retraining requires a large amount of historical data and consumes significant computational resources, making it difficult to achieve rapid model iteration and online optimization. Summary of the Invention

[0006] The purpose of this invention is to provide a method for controlling the amount of de-adhesion and anti-icing material applied based on road surface texture depth. This method solves the technical problem of adaptively updating the application amount calculation model based on construction feedback data to improve the accuracy of application amount prediction and the environmental adaptability of the model.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A method for controlling the application rate of de-adhesion and anti-icing material based on pavement texture depth includes the following steps:

[0009] Step 1: The data acquisition module acquires the pavement texture depth data and environmental parameters of the road section to be constructed, and forms a texture depth sequence distributed along the mileage direction; the pavement texture depth data is obtained by continuous non-contact sampling at a preset sampling interval using a laser texture depth testing device, and then uploaded to the data acquisition module in real time via a data cable;

[0010] The environmental parameters include real-time temperature and humidity parameters collected by environmental sensors at the construction site, as well as snowfall frequency parameters, historical temperature change data, and historical humidity change data obtained from an external meteorological data interface.

[0011] The data acquisition module segments the constructed depth sequence according to a preset mileage interval and calculates the statistical characteristic value of each segment.

[0012] Step 2: The parameter acquisition module acquires the performance parameters of the de-icing and anti-icing material, including anti-icing performance parameters and curing time parameters; the anti-icing performance parameters are ice layer breakage rate indicators obtained based on the falling ball impact test; the curing time parameters are obtained through time gradient testing.

[0013] Step 3: Based on the road surface texture depth data, performance parameters, and environmental parameters, the model module constructs a basic spraying quantity calculation model and a spraying quantity calculation model adapted to different environmental conditions;

[0014] The model module calculates the rate of environmental change based on environmental parameters and determines the current environmental state: if the environment is in a state of sudden change, it selects a spraying amount calculation model that is adapted to different environmental conditions and calculates the target spraying amount; if the environment is in a stable state, it corrects the basic spraying amount calculation model based on environmental parameters and outputs the corrected target spraying amount.

[0015] The model module calculates the target spraying amount based on different mileage locations and generates the correspondence between mileage and spraying amount;

[0016] Step 4: The control module generates a spraying control command based on the target spraying amount and sends the spraying control command to the spraying equipment to control the spraying equipment to automatically adjust the spraying amount of the de-adhesion and anti-icing material.

[0017] Preferably, when performing step 1, the sampling interval of the laser construction depth testing device is no greater than 10 mm;

[0018] The data acquisition module is connected and communicates with the laser structure depth testing device via a data cable;

[0019] The laser-guided depth testing equipment uploads the road surface depth data to the data acquisition module in real time.

[0020] Preferably, when performing step 1, the constructed depth sequence D is specifically represented as: D={(x1,TD1,T1,Hx1),(x2,TD2,T2,Hx2),…,(x n ,TD n ,T n Hx n )};

[0021] Where, x i Let be the mileage position of the i-th sampling point; D is the constructed depth sequence; TD i T represents the pavement texture depth value at the i-th sampling point. i Hx represents the ambient temperature at the i-th sampling point. i Let represent the ambient humidity at the i-th sampling point, and n represent the total number of sampling points.

[0022] Preferably, when performing step 1, segmenting the constructed depth sequence according to a preset mileage interval includes dividing it into fixed-length intervals and dividing it into adaptive intervals; dividing it into fixed-length intervals specifically means dividing each segment into a preset fixed-length interval; dividing it into adaptive intervals specifically means dividing the segments according to the construction depth change gradient, including dividing it into new segments when the construction depth change rate of adjacent sampling points exceeds a preset threshold.

[0023] The construction depth statistical features of each segment include mean, standard deviation, coefficient of variation, range, quantiles, and distribution fitting parameters;

[0024] The mean in the statistical features is used as the input feature of the spray volume calculation model; the standard deviation, coefficient of variation, range, quantiles, and distribution fitting parameters are all used to construct feature value charts and display them.

[0025] Preferably, when performing step 3, the basic spraying volume calculation model is as follows:

[0026] S i =-2.3799-5.783×TDavg i -0.0187×h+0.454×B R +7.7386×TDavg i 2 +0.0206×h 2 -0.0138×B R 2 ;

[0027] Where S i TDavg represents the unit area spray amount corresponding to the i-th segment. i Let B be the mean of the statistical feature values ​​of the construction depth within the i-th segment, h be the curing time parameter, and B be the mean of the feature values. RThese are the anti-icing performance parameters.

[0028] Preferably, when performing step 3, the model module calculates the rate of change of the environment in the construction area based on the environmental parameters and determines the current environmental state: if the change of environmental parameters exceeds a preset threshold, it is determined to be an environmental abrupt change state; if it does not exceed the threshold, it is determined to be an environmental stable state.

[0029] When an environmental abrupt change is detected, the model module executes a model selection strategy, selects a target model from multiple preset spraying quantity calculation models, and outputs S1, S2, or S3 as the target spraying quantity, specifically including:

[0030] When the ambient temperature is below the preset low temperature threshold, the spraying amount calculation model under low temperature conditions is adopted:

[0031] S1 = S × (1 + k1 × |T0 - T|);

[0032] Where k1 represents the low temperature sensitivity coefficient, which is used to characterize the influence weight of temperature changes on the amount of spraying; T0 represents the reference temperature;

[0033] When the ambient humidity is higher than the preset humidity threshold, the spraying amount calculation model under high humidity conditions is adopted:

[0034] S2 = S × (1 + k2 × Hx);

[0035] Where k2 represents the humidity influence coefficient, which is used to characterize the degree of influence of humidity on the ice adhesion characteristics;

[0036] When the frequency of snowfall exceeds a preset threshold, the spraying amount calculation model under high snowfall conditions is adopted:

[0037] S3 = S × (1 + k3 × F);

[0038] Wherein, k3 represents the snowfall impact coefficient, which is used to characterize the degree of influence of snowfall frequency on the ice accumulation effect;

[0039] When the environment is determined to be stable, the model module corrects the basic spraying quantity calculation model based on environmental parameters and outputs S' as the target spraying quantity. The correction form is as follows:

[0040] S'=S+αT+βHx+γF;

[0041] Where S' is the corrected spraying amount; S is the value calculated by the basic model; T is the temperature parameter; Hx is the humidity parameter; F is the snowfall frequency parameter; α, β, and γ are environmental correction coefficients, used to characterize the linear influence of environmental parameters on the spraying amount;

[0042] The model module generates the mileage-spraying amount correspondence based on the calculation results of S1, S2, S3 or S'.

[0043] Preferably, when performing step 3, the model module updates or retrains the spraying volume calculation model based on the feedback data after construction, and the update method adopts incremental learning or batch retraining.

[0044] The feedback data includes actual spraying volume, road surface texture depth remeasurement data, anti-icing performance data, and de-icing effect evaluation data.

[0045] Preferably, when performing model parameter updates, model retraining, or model updates, the specific steps include the following:

[0046] Step S1: Collect post-construction feedback data, including actual spraying volume, pavement texture depth re-measurement data, anti-icing performance data, and de-icing effect evaluation data;

[0047] Step S2: Compare the feedback data with the original model input and output data, and calculate the deviation between the model output spraying amount and the actual spraying amount;

[0048] Step S3: When the deviation exceeds the preset error threshold, the model update mechanism is triggered;

[0049] Step S4: Update the parameters of the basic spraying amount calculation model through regression, or retrain the model based on the new data;

[0050] Step S5: When there are multiple spraying amount calculation models corresponding to multiple environmental conditions: classify the feedback data according to the environmental parameters, and perform independent parameter updates or model retraining for each environmental model.

[0051] Step S6: Use the updated spraying volume calculation model for the spraying volume calculation and control in the next construction cycle.

[0052] The present invention discloses a method for controlling the application rate of de-adhesion and anti-icing materials based on road surface texture depth. This method solves the technical problem of adaptive incremental updating of the application rate calculation model based on construction feedback data to improve the accuracy of application rate prediction and the adaptability of the model to the environment. The present invention utilizes post-construction feedback data to perform parameter regression updates or model retraining on the application rate calculation model. It can achieve dynamic adaptive optimization of the model with low computational cost while retaining the original model knowledge, and significantly improve the accuracy of application rate prediction and control robustness under different construction environments and material conditions. Attached Figure Description

[0053] Figure 1 This is the main flowchart of the present invention;

[0054] Figure 2 This is a flowchart of the model module of the present invention calculating the target spraying amount;

[0055] Figure 3 This is a flowchart of the model update process of the present invention. Detailed Implementation

[0056] Depend on Figures 1-3 The method for controlling the application rate of de-adhesion and anti-icing material based on pavement texture depth, as shown, includes the following steps:

[0057] Step 1: The data acquisition module acquires the pavement texture depth data and environmental parameters of the road section to be constructed, and forms a texture depth sequence distributed along the mileage direction; the pavement texture depth data is obtained through continuous non-contact sampling.

[0058] The road surface texture depth data is obtained by a laser texture depth testing device, with a sampling interval of no more than 10 mm.

[0059] The data acquisition module is connected and communicates with the laser structure depth testing device via a data cable;

[0060] The laser-guided depth testing equipment uploads the road surface depth data to the data acquisition module in real time.

[0061] The laser structure depth testing equipment is also equipped with an environmental sensor to collect temperature and humidity parameters of the construction area in real time and upload them to the data acquisition module.

[0062] In this embodiment, the laser structure depth testing equipment is an XYJ-PH type road structure depth laser testing vehicle, and the environmental sensors include a temperature sensor and a humidity sensor, both of which are built into the XYJ-PH type road structure depth laser testing vehicle.

[0063] The data acquisition module connects to the meteorological data interface via the Internet to obtain meteorological data from third-party meteorological platforms, including snowfall frequency parameters, recent temperature and humidity changes.

[0064] In this embodiment, the data acquisition module, parameter acquisition module, model module, and control module can be integrated into the same industrial computer.

[0065] The industrial control computer serves as the core for on-site data processing and control, possessing high-performance computing capabilities, multi-interface communication capabilities, and industrial-grade environmental adaptability.

[0066] The industrial control computer uses a USB interface or RS232 serial port to connect to the XYJ-PH type pavement texture depth laser testing vehicle via a data cable. It is used to receive pavement texture depth data, temperature parameters, and humidity parameters uploaded by the testing vehicle in real time. The communication protocol adopts the serial communication protocol defined by the software accompanying the testing vehicle, and the data sampling frequency is synchronized with the sampling interval (10mm) of the testing vehicle.

[0067] The industrial control computer uses a CAN bus interface or an RS485 interface, and connects to the controller of the spraying equipment via a control line. It is used to issue spraying control commands and receive status feedback information from the spraying equipment. The communication protocol uses Modbus RTU or a custom protocol, supporting real-time adjustment of the segmented spraying volume.

[0068] The industrial control computer uses an Ethernet interface or a 4G / 5G wireless communication module to connect to a third-party meteorological data platform (such as an API interface) via the Internet to automatically acquire snowfall frequency parameters, recent temperature change data, and humidity change data. The data acquisition frequency can be set to once per hour or dynamically adjusted according to construction needs.

[0069] The industrial control computer is also equipped with a human-machine interface (touch screen) for real-time display of collected data, segmented statistical results, target spraying amount and equipment status, and to receive parameter input and command input from operators.

[0070] The data acquisition module segments the constructed depth sequence according to a preset mileage interval and calculates the statistical characteristic values ​​of each segment. The statistical characteristic values ​​include the mean, standard deviation, coefficient of variation, range, quantiles, and parameters of the constructed depth distribution curve.

[0071] The statistical feature values ​​are used to characterize and assist in the analysis of the structural depth distribution of each segment, wherein the feature value used as input to the spraying amount calculation model is the mean of the structural depth data within the segment.

[0072] In this embodiment, the segmented construction depth sequence is used to construct the mileage-spraying amount correspondence.

[0073] In this embodiment, the constructed depth sequence distributed along the mileage direction is specifically as follows:

[0074] D={(x1,TD1,T1,Hx1),(x2,TD2,T2,Hx2),…,(x n ,TD n ,T n Hx n )};

[0075] Where, x i Let be the mileage position of the i-th sampling point; D is the constructed depth sequence; TD i T represents the pavement texture depth value at the i-th sampling point. i Hx represents the ambient temperature at the i-th sampling point. i Let represent the ambient humidity at the i-th sampling point, and n represent the total number of sampling points.

[0076] For example, in a construction section from K10+000 to K10+100, data can be collected at a sampling interval of 10mm to form a structural depth sequence containing 10,000 sampling points. Each sampling point corresponds to a structural depth value and synchronously collected temperature and humidity data.

[0077] The constructed depth sequence is segmented according to a preset mileage interval, including dividing it into fixed-length intervals and dividing it into adaptive intervals;

[0078] In this embodiment, the priority of adaptive intervals is higher than that of fixed-length intervals.

[0079] The division by fixed length interval specifically means: dividing each segment by a preset fixed length interval, wherein the fixed length interval is 10m or 20m;

[0080] The adaptive interval division specifically involves dividing the data into segments based on the gradient of the construction depth change, including dividing the data into new segments when the rate of change of the construction depth of adjacent sampling points exceeds a preset threshold.

[0081] For example, when the construction depth change rate |TD i -TD i-1 When |>δ (δ is a preset threshold), this position is taken as the new segment starting point. The value of δ ranges from 0.1 mm / m to 0.3 mm / m, preferably 0.2 mm / m. This threshold can be determined according to the road surface smoothness requirements or through calibration on test road sections: in road sections with gradual changes in texture depth, a smaller threshold (such as 0.1 mm / m) can improve segmentation sensitivity; in road sections with large fluctuations in texture depth, a larger threshold (such as 0.3 mm / m) can avoid over-segmentation.

[0082] In this embodiment, when calculating the statistical feature value for each segment, only the TD of each segment is calculated. i TD in datai The statistical characteristics include: averaging all construction depth data within a segment using an arithmetic mean algorithm to obtain the mean; using a standard deviation calculation algorithm to characterize the dispersion of construction depth to obtain the standard deviation; calculating the coefficient of variation by the ratio of the standard deviation to the mean; calculating the range by the difference between the maximum and minimum values ​​within a segment; calculating the quantiles by using a sorting statistical method to calculate a preset percentile (such as P90); and obtaining the parameters of the construction depth distribution curve by using a probability distribution fitting algorithm (such as fitting a normal distribution or a log-normal distribution).

[0083] Step 2: The parameter acquisition module acquires the performance parameters of the de-icing and anti-icing material, including anti-icing performance parameters and curing time parameters;

[0084] The anti-icing performance parameter is the breakage rate index, which is obtained through testing or based on historical data. The anti-icing performance parameter is pre-entered by the user into the parameter acquisition module through the interactive interface.

[0085] The parameter acquisition module communicates with the human-computer interaction interface via an internal data bus.

[0086] The human-computer interface can display the construction depth statistical feature values ​​of each segment calculated in step 1. The display format is generally a table, as shown in Table 1 below: Construction Depth Statistical Feature Values ​​Table.

[0087]

[0088] Table 1

[0089] In this embodiment, the distribution fitting parameters include the type of distribution to be fitted (such as normal distribution or log-normal distribution) and its corresponding distribution parameters (such as mean μ and standard deviation σ). Users can select different segment intervals through the interactive interface to view the corresponding statistical characteristic values ​​and distribution curves in real time.

[0090] The human-computer interface is also used to receive user input of the performance parameters of the de-icing and anti-icing material, including anti-icing performance parameters (breakage rate BR) and curing time parameters (h). Users can enter values ​​via keyboard or touch. After receiving the entered data, the parameter acquisition module performs format and range verification. Once confirmed to be correct, it stores the data in the local database for the model module to call.

[0091] The human-machine interface is also used to receive user commands for starting, stopping, and emergency shutdown, and to display the current working status of the spraying equipment in real time.

[0092] The breakage rate index is obtained through an ice layer destruction test, which is a falling ball impact test.

[0093] The segmentation principle will be explained below with reference to a specific construction section. For example, on a certain highway section from K10+000 to K10+100, which is 100 meters long, the laser structure depth testing vehicle continuously collected data at a sampling interval of 10mm, obtaining a total of 10,000 sampling points.

[0094] Divide the area into fixed-length intervals (segment length 10 meters): Divide K10+000 to K10+100 into 10 segments, each segment being 10 meters long (corresponding to 1000 sampling points). Taking the first segment (K10+000-K10+010) as an example, the 1000 structural depth sampling values ​​within this segment are TD1, TD2, ..., TD... 1000 .

[0095] Adaptive interval division: The adaptive division dynamically determines the segment boundaries based on the gradient of changes in the construction depth. The rate of change threshold δ = 0.2 mm / m is set (i.e., when the change exceeds 0.2 mm per meter, a new segment is created).

[0096] Taking the segment from K10+000 to K10+050 as an example: the structural depth of K10+000-K10+020 is stable between 0.80-0.85mm, and the rate of change does not exceed the threshold, so it belongs to the same segment; the structural depth at K10+020 suddenly changes from 0.82mm to 1.15mm, and the rate of change ΔTD / Δx=(1.15-0.82) / 0.01=33mm / m (sampling interval 0.01m), which far exceeds the threshold δ, so a new segment is cut at K10+020; the structural depth of K10+020-K10+035 fluctuates between 1.10-1.20mm, and the rate of change does not exceed the threshold, so it belongs to the same segment; the structural depth at K10+035 decreases from 1.18mm to 0.75mm, and the absolute value of the rate of change exceeds the threshold, so a new segment is cut again.

[0097] In this embodiment, adaptive segmentation has a higher priority than fixed-length segmentation. Specifically, adaptive segmentation is prioritized based on the gradient of changes in pavement texture depth. If the changes in pavement texture depth within a certain road segment are gradual (without exceeding the threshold), it degenerates into fixed-length segmentation (default segment length 10m). Adaptive segmentation can more accurately capture local abrupt changes in pavement texture depth and avoid mixing areas with significant differences in texture depth into the same segment, thereby improving the precision of the spraying amount calculation.

[0098] In this embodiment, the anti-icing performance parameters of the de-adhesive and anti-icing material were determined by a falling ball impact test, and the ice layer breakage rate was used as the evaluation index. The specific test steps are as follows:

[0099] Step A1: Prepare asphalt mixture specimens according to the mix proportions of the actual asphalt mixture in the project, or obtain them from the actual pavement by core drilling. The specimens can be Marshall specimens or rut slab specimens.

[0100] Step A2: Select two groups of specimens with similar surface morphology, remove surface dust, one group as the test group, and spray the surface of the de-adhesion and anti-icing material to be tested evenly; the other group as the control group, without any spraying treatment, both groups of specimens are left to stand at room temperature until the material is fully cured.

[0101] Step A3: Use appropriate sealing material to create a sealing layer around the specimen to prevent water loss after water injection. Inject a fixed amount of water into the surface of each specimen, with a water volume of 265 mL, which is equivalent to the amount of water accumulated on the surface area of ​​a single rainfall of 15 mm (moderate rain level).

[0102] Step A4: After water injection, the specimen is placed in a low-temperature test chamber at -30℃ and frozen for a sufficient time to ensure that the water layer on the surface of the specimen is completely frozen into ice, in order to simulate the icing environment of asphalt pavement in cold winter.

[0103] Step A5: Remove the frozen specimen and place it on a rigid base. Use a steel ball with a mass of 500g ± 1g to drop freely from a vertical height of 0.5m to strike the surface of the specimen. For Marshall specimens, repeat the impact 10 times; for rutted specimens, repeat the impact 50 times. The impact of the steel ball simulates the breaking effect of vehicle tire load on the road surface ice layer.

[0104] Step A6: Record the total area of ​​ice breakage and cracking after the ice layer breaks, and the length of each extended single crack. Calculate the breakage rate using the following formula as an evaluation index for anti-icing performance:

[0105] B R ={(A+λ×L) / A 总}×100%;

[0106] Among them, B R The breakdown rate is expressed as a percentage (%); A represents the total area of ​​broken and cracked material, in square centimeters (cm²); L represents the length of a single extended crack, in centimeters (cm); λ is the influence coefficient for converting a single crack into area, typically taken as 0.3; A 总 The total area of ​​the test specimen is expressed in square centimeters (cm²). The higher the breakage rate, the better the anti-icing performance of the de-adhesive and anti-icing material.

[0107] The curing time parameter is obtained through time gradient testing or by fitting historical data. The curing time parameter is pre-entered by the user into the parameter acquisition module through the interactive interface.

[0108] In this embodiment, the specific method for determining the curing time parameter through time gradient testing is as follows:

[0109] Step B1: The tentative spraying rate is 0.5 kg / m² (this value is an empirical benchmark value used to standardize test conditions).

[0110] Step B2: Spray the de-adhesion and anti-icing material onto the surface of the Marshall specimen, and prepare multiple specimens in the same batch;

[0111] Step B3: Place the sprayed specimens under different curing times, with time gradient intervals of 10 minutes (e.g., 10 min, 20 min, 30 min...), until after a certain time point, continuing to extend the curing time no longer causes changes in the ice layer breakage rate;

[0112] Step B4: Test the ice breakage rate of each specimen according to the aforementioned falling ball impact test method;

[0113] Step B5: When the ice layer breakage rate no longer changes significantly with the extension of the curing time, the corresponding shortest curing time is the curing time parameter h.

[0114] This curing time parameter reflects the shortest curing time required for the material to fully exert its anti-icing properties and is one of the important input features of the spraying amount calculation model.

[0115] Step 3: The model module constructs a spraying amount calculation model based on the road surface construction depth data, the performance parameters, and the environmental parameters, and calculates the target spraying amount corresponding to different mileage locations according to the spraying amount calculation model;

[0116] The target spraying amount is output in the form of a mileage-spraying amount correspondence, which is used to guide segmented differentiated construction;

[0117] The environmental parameters include temperature, humidity, and snowfall frequency;

[0118] The spraying amount calculation model was constructed by performing multiple regression analysis on historical pavement texture depth data, material performance parameters and corresponding spraying amount data;

[0119] The corresponding spraying volume data is the actual spraying volume used in the historical construction process. The construction effect evaluation data includes anti-icing effect, re-icing situation and service life. In the process of model construction, the construction effect evaluation data is used as the optimization target to fit and solve the parameters of the spraying volume calculation model.

[0120] The specific calculation model for the basic spraying volume is as follows:

[0121] S i =-2.3799-5.783×TDavg i -0.0187×h+0.454×B R +7.7386×TDavg i 2 +0.0206×h 2 -0.0138×B R 2 ;

[0122] Where S i TDavg represents the unit area spray amount corresponding to the i-th segment. i Let B be the statistical characteristic value of the construction depth within the i-th segment, h be the curing time parameter, and B be the value of the construction depth within the i-th segment. R These are the anti-icing performance parameters.

[0123] The model module determines the environmental state based on the time-varying characteristics of environmental parameters, including statistically analyzing the change in snowfall frequency per unit time, calculating the temperature change rate Ts, and calculating the humidity change rate Hxs.

[0124] Calculate the rate of temperature change Ts:

[0125] Ts = ΔT / Δt;

[0126] ΔT = T(t) - T(t - Δt);

[0127] Calculate the humidity change rate Hxs:

[0128] Hxs = ΔHx / Δt;

[0129] ΔHx = H(t) - H(t - Δt);

[0130] Where t represents the current time; Δt is the sampling time interval for environmental parameters, which in this embodiment is between 5 and 30 minutes; ΔT represents the temperature change difference within the sampling time interval Δt; and ΔHx represents the humidity change difference within the sampling time interval Δt.

[0131] When statistically analyzing the changes in snowfall frequency per unit of time, the unit of time is between 1 hour and 24 hours.

[0132] In real-time control scenarios, Δt is preferably 10 minutes, while in construction decision-making scenarios, the unit time is preferably 6 hours or 12 hours.

[0133] An environmental abrupt change is determined when any of the following conditions are met: the rate of temperature change exceeds a preset temperature threshold; the rate of humidity change exceeds a preset humidity threshold; or the frequency of snowfall per unit time exceeds a preset frequency threshold.

[0134] In this embodiment, the temperature change rate threshold ranges from 3℃ / 10min to 10℃ / 10min, preferably 5℃ / 10min; the humidity change rate threshold ranges from 15% / 10min to 30% / 10min, preferably 20% / 10min; and the snowfall frequency threshold ranges from 2 times / hour to 5 times / hour, preferably 3 times / hour. Users can manually adjust these thresholds through the interactive interface.

[0135] If none of these conditions are met, the environment is considered to be in a stable state.

[0136] When an environmental abrupt change is detected, the model module executes a model selection strategy, selects a target model from multiple preset spraying quantity calculation models, and outputs S1, S2, or S3 as the target spraying quantity, specifically including:

[0137] When the ambient temperature is below the preset low temperature threshold, the spraying amount calculation model under low temperature conditions is adopted:

[0138] S1 = S × (1 + k1 × |T0 - T|);

[0139] Wherein, k1 represents the low temperature sensitivity coefficient, which is used to characterize the influence weight of temperature change on the amount of spraying; T0 represents the reference temperature, which in this embodiment is preferably the standard temperature during material performance testing or the regional historical average temperature.

[0140] When the ambient humidity is higher than the preset humidity threshold, the spraying amount calculation model under high humidity conditions is adopted:

[0141] S2 = S × (1 + k2 × Hx);

[0142] Where k2 represents the humidity influence coefficient, which is used to characterize the degree of influence of humidity on the ice adhesion characteristics;

[0143] When the frequency of snowfall exceeds a preset threshold, the spraying amount calculation model under high snowfall conditions is adopted:

[0144] S3 = S × (1 + k3 × F);

[0145] Wherein, k3 represents the snowfall impact coefficient, which is used to characterize the degree of influence of snowfall frequency on the ice accumulation effect;

[0146] The values ​​of k1, k2, and k3 are determined through historical data regression analysis or experimental calibration. In practical applications, users can adjust k1, k2, and k3 according to the climate characteristics of the construction area through the operation interface.

[0147] In this embodiment, the spraying amount calculation model under different environmental conditions is constructed based on historical data under the corresponding environmental conditions.

[0148] When the environment is determined to be stable, the model module corrects the basic spraying quantity calculation model based on environmental parameters and outputs S' as the target spraying quantity. The correction form is as follows:

[0149] S'=S+αT+βHx+γF;

[0150] Where S' is the corrected spraying amount; S is the value calculated by the basic model; T is the temperature parameter; Hx is the humidity parameter; F is the snowfall frequency parameter; α, β, and γ are environmental correction coefficients, used to characterize the linear influence of environmental parameters on the spraying amount;

[0151] In this embodiment, the values ​​of the environmental correction coefficients α, β, and γ are determined in the following ways: by constructing a multivariate regression model based on historical construction data to solve the parameters, by calibrating the parameters using comparative test data under different environmental conditions, or by optimizing the solution using the minimum error fitting method with the construction effect evaluation index as the objective function, and by dynamically updating the solution based on new data during subsequent construction.

[0152] In this embodiment, the spraying amount calculation models under different environmental conditions are all based on the basic spraying amount calculation model, and are obtained by parameter recalibration based on the sample data of the corresponding environmental conditions.

[0153] The model module updates the parameters or retrains the spraying volume calculation model based on post-construction feedback data.

[0154] In this embodiment, the model module updates the parameters or retrains the spray volume calculation model based on post-construction feedback data. The model update is implemented using incremental learning or batch retraining, specifically including:

[0155] Step S1: Collect post-construction feedback data, including actual spraying volume, pavement texture depth re-measurement data, anti-icing performance data, and de-icing effect evaluation data;

[0156] Step S2: Compare the feedback data with the original model input and output data, and calculate the deviation between the model output spraying amount and the actual spraying amount;

[0157] Step S3: When the deviation exceeds the preset error threshold, the model update mechanism is triggered;

[0158] Step S4: Update the parameters of the basic spraying amount calculation model through regression, or retrain the model based on the new data;

[0159] Step S5: When there are multiple spraying amount calculation models corresponding to multiple environmental conditions: classify the feedback data according to the environmental parameters, and perform independent parameter updates or model retraining for each environmental model.

[0160] Step S6: Use the updated spraying volume calculation model for the spraying volume calculation and control in the next construction cycle.

[0161] In this embodiment, the incremental learning method uses existing regression update algorithms such as recursive least squares (RLS) or stochastic gradient descent (SGD) to recursively correct the model parameters using the newly added feedback data, without having to retrain all historical data; the batch retraining method, on the other hand, merges the newly added feedback data with the historical data and then refits the model parameters.

[0162] The distribution fitting parameters are calculated using the maximum likelihood estimation method: when fitting the normal distribution, the mean and standard deviation of the construction depth data within the segment are directly calculated; when fitting the log-normal distribution, the natural logarithm of the data is taken first, and then its mean and standard deviation are calculated.

[0163] Step 4: The control module generates a spraying control command based on the target spraying amount and sends the spraying control command to the spraying equipment to control the spraying equipment to automatically adjust the spraying amount of the de-adhesion and anti-icing material.

[0164] The spraying control command is used to control the spraying equipment to perform segmented or continuous adjustment of the spraying amount.

[0165] In this embodiment, the control module connects to the controller of the spraying device via a CAN bus interface or an RS485 interface, and uses the Modbus-RTU communication protocol for data exchange. Communication parameters are typically set to a baud rate of 9600 bps, 8 data bits, 1 stop bit, and no parity check.

[0166] The control module receives the "mileage-spraying amount correspondence" from the model module. This correspondence is stored in the form of a segmented array, for example: [(K10+000,0.55), (K10+010,0.62), (K10+020,0.48), ...].

[0167] The spraying equipment is equipped with an odometer that provides real-time feedback of the current construction distance to the industrial control computer. The control module then queries the corresponding target spraying volume based on the current location.

[0168] The control module converts the target spray volume into control parameters for the spraying equipment (such as nozzle opening percentage or motor speed) and generates spraying control commands that conform to the Modbus-RTU protocol.

[0169] For example, when the spraying vehicle reaches point K10+010, the control module detects a change in mileage and updates the target spraying rate to 0.48 kg / m² (corresponding to a nozzle opening of 48%). The control module then issues a new command: {01 06 00 10 00 30 4827} (where 0x30 is the hexadecimal representation of 48%). The spraying equipment continuously receives commands during its journey, enabling segmented, continuous, and automatic adjustments along the mileage direction without manual intervention.

[0170] The present invention discloses a method for controlling the application rate of de-adhesion and anti-icing materials based on road surface texture depth. This method solves the technical problem of adaptive incremental updating of the application rate calculation model based on construction feedback data to improve the accuracy of application rate prediction and the adaptability of the model to the environment. The present invention utilizes post-construction feedback data to perform parameter regression updates or model retraining on the application rate calculation model. It can achieve dynamic adaptive optimization of the model with low computational cost while retaining the original model knowledge, and significantly improve the accuracy of application rate prediction and control robustness under different construction environments and material conditions.

Claims

1. A method for controlling the application rate of de-adhesion and anti-icing material based on road surface texture depth, characterized in that: Includes the following steps: Step 1: The data acquisition module acquires the pavement texture depth data and environmental parameters of the road section to be constructed, and forms a texture depth sequence distributed along the mileage direction; the pavement texture depth data is obtained by continuous non-contact sampling at a preset sampling interval using a laser texture depth testing device, and then uploaded to the data acquisition module in real time via a data cable; The environmental parameters include real-time temperature and humidity parameters collected by environmental sensors at the construction site, as well as snowfall frequency parameters, historical temperature change data, and historical humidity change data obtained from an external meteorological data interface. The data acquisition module segments the constructed depth sequence according to a preset mileage interval and calculates the statistical characteristic value of each segment. Step 2: The parameter acquisition module acquires the performance parameters of the de-icing and anti-icing material, including anti-icing performance parameters and curing time parameters; the anti-icing performance parameters are ice layer breakage rate indicators obtained based on the falling ball impact test; the curing time parameters are obtained through time gradient testing. Step 3: Based on the road surface texture depth data, performance parameters, and environmental parameters, the model module constructs a basic spraying quantity calculation model and a spraying quantity calculation model adapted to different environmental conditions; The model module calculates the rate of environmental change based on environmental parameters and determines the current environmental state: if the environment is in a state of sudden change, it selects a spraying amount calculation model that is adapted to different environmental conditions and calculates the target spraying amount. If the environment is stable, the basic spraying amount calculation model is corrected based on environmental parameters, and the corrected target spraying amount is output. The model module calculates the target spraying amount based on different mileage locations and generates the correspondence between mileage and spraying amount; Step 4: The control module generates a spraying control command based on the target spraying amount and sends the spraying control command to the spraying equipment to control the spraying equipment to automatically adjust the spraying amount of the de-adhesion and anti-icing material.

2. The method for controlling the application rate of de-adhesion and anti-icing material based on road surface texture depth as described in claim 1, characterized in that: When performing step 1, the sampling interval of the laser construction depth testing device is no greater than 10 mm; The data acquisition module is connected and communicates with the laser structure depth testing device via a data cable; The laser-guided depth testing equipment uploads the road surface depth data to the data acquisition module in real time.

3. The method for controlling the application rate of de-adhesion and anti-icing material based on road surface texture depth as described in claim 1, characterized in that: When performing step 1, the constructed depth sequence D is specifically represented as: D={(x1,TD1,T1,Hx1),(x2,TD2,T2,Hx2),…,(x n ,TD n ,T n Hx n )}; Where, x i Let be the mileage position of the i-th sampling point; D is the constructed depth sequence; TD i T represents the pavement texture depth value at the i-th sampling point. i Hx represents the ambient temperature at the i-th sampling point. i Let represent the ambient humidity at the i-th sampling point, and n represent the total number of sampling points.

4. The method for controlling the application amount of de-adhesion and anti-icing material based on road surface texture depth as described in claim 1, characterized in that: When performing step 1, the constructed depth sequence is segmented according to a preset mileage interval, including dividing it into fixed-length intervals and dividing it into adaptive intervals. Dividing it into fixed-length intervals specifically means dividing each segment into a preset fixed-length interval. Dividing it into adaptive intervals specifically means dividing the segment according to the gradient of the constructed depth change, including dividing it into new segments when the rate of change of the constructed depth of adjacent sampling points exceeds a preset threshold. The construction depth statistical features of each segment include mean, standard deviation, coefficient of variation, range, quantiles, and distribution fitting parameters; The mean value in the statistical features is used as the input feature of the spray volume calculation model; Standard deviation, coefficient of variation, range, quantiles, and distribution fitting parameters are all used to construct and display feature value charts.

5. The method for controlling the application rate of de-adhesion and anti-icing material based on road surface texture depth as described in claim 1, characterized in that: When performing step 3, the basic spraying volume calculation model is as follows: S i =-2.3799-5.783×TDavg i -0.0187×h+0.454×B R +7.7386×TDavg i 2 +0.0206×h 2 -0.0138×B R 2 ; Where S i TDavg represents the unit area spray amount corresponding to the i-th segment. i Let B be the mean of the statistical feature values ​​of the construction depth within the i-th segment, h be the curing time parameter, and B be the mean of the feature values. R These are the anti-icing performance parameters.

6. The method for controlling the application rate of de-adhesion and anti-icing material based on road surface texture depth as described in claim 1, characterized in that: When performing step 3, the model module calculates the rate of change of the environment in the construction area based on the environmental parameters and determines the current environmental state: if the change of environmental parameters exceeds a preset threshold, it is determined to be an environmental abrupt change state; if it does not exceed the threshold, it is determined to be an environmental stable state. When an environmental abrupt change is detected, the model module executes a model selection strategy, selects a target model from multiple preset spraying quantity calculation models, and outputs S1, S2, or S3 as the target spraying quantity, specifically including: When the ambient temperature is below the preset low temperature threshold, the spraying amount calculation model under low temperature conditions is adopted: S1 = S × (1 + k1 × |T0 - T|); Where k1 represents the low temperature sensitivity coefficient, which is used to characterize the influence weight of temperature changes on the amount of spraying; T0 represents the reference temperature; When the ambient humidity is higher than the preset humidity threshold, the spraying amount calculation model under high humidity conditions is adopted: S2 = S × (1 + k2 × Hx); Where k2 represents the humidity influence coefficient, which is used to characterize the degree of influence of humidity on the ice adhesion characteristics; When the frequency of snowfall exceeds a preset threshold, the spraying amount calculation model under high snowfall conditions is adopted: S3 = S × (1 + k3 × F); Wherein, k3 represents the snowfall impact coefficient, which is used to characterize the degree of influence of snowfall frequency on the ice accumulation effect; When the environment is determined to be stable, the model module corrects the basic spraying quantity calculation model based on environmental parameters and outputs S' as the target spraying quantity. The correction form is as follows: S'=S+αT+βHx+γF; Where S' is the corrected spraying amount; S is the value calculated by the basic model; T is the temperature parameter; Hx is the humidity parameter; F is the snowfall frequency parameter; α, β, and γ are environmental correction coefficients, used to characterize the linear influence of environmental parameters on the spraying amount; The model module generates the mileage-spraying amount correspondence based on the calculation results of S1, S2, S3 or S'.

7. The method for controlling the application amount of de-adhesion and anti-icing material based on road surface texture depth as described in claim 3, characterized in that: When performing step 3, the model module updates or retrains the spraying volume calculation model based on the feedback data after construction. The update method adopts incremental learning or batch retraining. The feedback data includes actual spraying volume, road surface texture depth remeasurement data, anti-icing performance data, and de-icing effect evaluation data.

8. The method for controlling the application amount of de-adhesion and anti-icing material based on road surface texture depth as described in claim 7, characterized in that: When performing model parameter updates, model retraining, or model updates, the specific steps include the following: Step S1: Collect post-construction feedback data, including actual spraying volume, pavement texture depth re-measurement data, anti-icing performance data, and de-icing effect evaluation data; Step S2: Compare the feedback data with the original model input and output data, and calculate the deviation between the model output spraying amount and the actual spraying amount; Step S3: When the deviation exceeds the preset error threshold, the model update mechanism is triggered; Step S4: Update the parameters of the basic spraying amount calculation model through regression, or retrain the model based on the new data; Step S5: When there are multiple spraying amount calculation models corresponding to multiple environmental conditions: classify the feedback data according to the environmental parameters, and perform independent parameter updates or model retraining for each environmental model. Step S6: Use the updated spraying volume calculation model for the spraying volume calculation and control in the next construction cycle.

Images