Predictive Maintenance Method and System Based on Road Infrastructure Monitoring

Abstract

The present invention discloses a predictive maintenance method and system based on road infrastructure monitoring, which relates to the technical field of road operation and maintenance. The method includes: collecting weekly traffic flow time series information and weekly pedestrian flow time series information; collecting annual temperature time series information, annual humidity time series information, and annual precipitation time series information; whenever the first timing satisfies the first period, activating the first road infrastructure status prediction model to obtain the first road infrastructure status prediction result; whenever the second timing satisfies the second period, activating the second road infrastructure status prediction model to obtain the second road infrastructure status prediction result; obtaining a road infrastructure status fusion result through a road infrastructure status fusion model; extracting abnormal status types and abnormal trigger time information. It solves the technical problems of long detection cycles and lagging maintenance response times in the prior art, and achieves the technical effects of improving the accuracy of road infrastructure status prediction and the timeliness of maintenance.

Description

Technical Field

[0001] This application relates to the technical field of road operation and maintenance, and specifically relates to a predictive maintenance method and system based on road infrastructure monitoring. Background Art

[0002] The maintenance and management of road infrastructure are crucial to ensure the efficient and safe operation of the transportation system. With the increasing traffic flow and environmental changes, traditional road maintenance methods are difficult to meet the increasingly complex requirements. Currently, the operation and maintenance management of road infrastructure mainly rely on regular inspections and manual patrols. However, the coverage of manual patrols is limited, and the changes in road conditions cannot be monitored in real time; although regular inspections can detect some problems in a timely manner, due to the long detection cycle and lag in response time, the best repair opportunity is often missed, and it is difficult to comprehensively consider environmental factors such as temperature, humidity, and precipitation of road facilities, which all affect the service life and safety of road infrastructure, thereby affecting the safety and reliability of the road.

[0003] Therefore, in the related technologies at the present stage, there are technical problems such as a long detection cycle and a lag in maintenance response time. Summary of the Invention

[0004] This application provides a predictive maintenance method and system based on road infrastructure monitoring, solves the technical problems of a long detection cycle and a lag in maintenance response time existing in the prior art, and achieves the technical effect of improving the accuracy of road infrastructure status prediction and the timeliness of maintenance.

[0005] This application provides a predictive maintenance method based on road infrastructure monitoring. The method includes: collecting weekly vehicle flow time series information and weekly pedestrian flow time series information; collecting annual temperature time series information, annual humidity time series information, and annual precipitation time series information; whenever a first timer satisfies a first period, activating a first road infrastructure status prediction model to process the weekly vehicle flow time series information and the weekly pedestrian flow time series information to obtain a first road infrastructure status prediction result, and restarting the count of the first timer from 0, where the first period is one week; whenever a second timer satisfies a second period, activating a second road infrastructure status prediction model to process the annual temperature time series information, the annual humidity time series information, and the annual precipitation time series information to obtain a second road infrastructure status prediction result, and starting to count the second period from 0, where the second period is one year; through a road infrastructure status fusion model, obtaining a road infrastructure status fusion result for the first road infrastructure status prediction result and the second road infrastructure status prediction result; extracting the abnormal status type and abnormal trigger time information of the road infrastructure status fusion result, and performing pre-maintenance configuration.

[0006] In a possible implementation, activate the first road infrastructure status prediction model, process the weekly traffic flow time series information and the weekly pedestrian flow time series information to obtain a first road infrastructure status prediction result, and also perform the following processing: obtain an initial road infrastructure status matrix; perform time-domain aggregation on the weekly traffic flow time series information and the weekly pedestrian flow time series information to obtain a first time-domain binary flow, a second time-domain binary flow, up to a Qth time-domain binary flow, where any one time domain has a traffic flow identifier and a pedestrian flow identifier; input the initial road infrastructure status matrix, the first time-domain binary flow, the second time-domain binary flow, up to the Qth time-domain binary flow into the first road infrastructure status prediction model, and output the first road infrastructure status prediction result.

[0007] In a possible implementation, perform time-domain aggregation on the weekly traffic flow time series information and the weekly pedestrian flow time series information to obtain a first time-domain binary flow, a second time-domain binary flow, up to a Qth time-domain binary flow, and also perform the following processing: perform time-domain aggregation on the weekly traffic flow time series information to obtain a first time-domain traffic flow, a second time-domain traffic flow, up to an Nth time-domain traffic flow; perform time-domain aggregation on the weekly pedestrian flow time series information to obtain a first time-domain pedestrian flow, a second time-domain pedestrian flow, up to an Mth time-domain pedestrian flow; fuse the first time-domain traffic flow, the second time-domain traffic flow, up to the Nth time-domain traffic flow with the first time-domain pedestrian flow, the second time-domain pedestrian flow, up to the Mth time-domain pedestrian flow to obtain the first time-domain binary flow, the second time-domain binary flow, up to the Qth time-domain binary flow.

[0008] In a possible implementation, input the initial road infrastructure status matrix, the first time-domain binary flow, the second time-domain binary flow, up to the Qth time-domain binary flow into the first road infrastructure status prediction model, and output the first road infrastructure status prediction result, and also perform the following processing: input the initial road infrastructure status matrix, the first time-domain binary flow, and a first time-domain step size into the first road infrastructure status prediction model to obtain a first time-domain road infrastructure status prediction result; input the first time-domain road infrastructure status prediction result, the second time-domain binary flow, and a second time-domain step size into the first road infrastructure status prediction model to obtain a second time-domain road infrastructure status prediction result; until inputting the (Q - 1)th time-domain road infrastructure status prediction result, the Qth time-domain binary flow, and a Qth time-domain step size into the first road infrastructure status prediction model to obtain the first road infrastructure status prediction result.

[0009] In a possible implementation, when constructing the first road infrastructure status prediction model, the following processing is further performed: Based on the type of road infrastructure, collect weekly vehicle flow time series record information, weekly pedestrian flow time series record information, and first road infrastructure status time series record information; perform time-domain aggregation on the weekly vehicle flow time series record information and the weekly pedestrian flow time series record information to obtain first time-domain binary flow record information, second time-domain binary flow record information up to Yth time-domain binary flow record information; according to the first time domain, the second time domain up to the Yth time domain, extract the road infrastructure status at the start time of the first time domain, the road infrastructure status at the start time of the second time domain up to the road infrastructure status at the start time of the Yth time domain, and the road infrastructure status at the end time of the Yth time domain from the first road infrastructure status time series record information; use the road infrastructure status at the start time of the first time domain and the first time-domain binary flow record information as the first time-domain input data of the long short-term memory neural network, use the road infrastructure status at the start time of the first time domain as the first time-domain output supervision data of the long short-term memory neural network, use the road infrastructure status at the start time of the second time domain and the second time-domain binary flow record information as the second time-domain input data of the long short-term memory neural network, use the road infrastructure status at the start time of the second time domain as the second time-domain output supervision data of the long short-term memory neural network, until using the road infrastructure status at the start time of the Yth time domain and the Yth time-domain binary flow record information as the Yth time-domain input data of the long short-term memory neural network, and use the road infrastructure status at the end time of the Yth time domain as the Yth time-domain output supervision data of the long short-term memory neural network to train the first road infrastructure status prediction model.

[0010] In a possible implementation, when constructing the second road infrastructure status prediction model, the following processing is further performed: Based on the type of road infrastructure, collect first sample data, where the first sample data includes annual temperature time series record information, annual humidity time series record information, annual precipitation time series record information, and second road infrastructure status time series record information; where the average annual vehicle flow of the first sample data is less than or equal to the vehicle flow threshold, and the average annual pedestrian flow of the first sample data is less than or equal to the pedestrian flow threshold; train the second road infrastructure status prediction model according to the annual temperature time series record information, the annual humidity time series record information, the annual precipitation time series record information, and the second road infrastructure status time series record information.

[0011] In a possible implementation, when constructing the road infrastructure status fusion model, the following processing is also performed: Based on the type of road infrastructure, second sample data is collected, where the second sample data includes second sample weekly traffic flow time series record information, second sample weekly pedestrian flow time series record information, second sample annual temperature time series record information, second sample annual humidity time series record information, second sample annual precipitation time series record information, and second sample road infrastructure status information; The first road infrastructure status prediction model is used to process the second sample weekly traffic flow time series record information and the second sample weekly pedestrian flow time series record information to obtain first road infrastructure status identification data; The second road infrastructure status prediction model is used to process the second sample weekly pedestrian flow time series record information, the second sample annual temperature time series record information, the second sample annual humidity time series record information, and the second sample annual precipitation time series record information to obtain second road infrastructure status identification data; Using the second sample road infrastructure status data as supervision and the first road infrastructure status identification data and the second road infrastructure status identification data as inputs, the road infrastructure status fusion model is trained.

[0012] This application also provides a predictive maintenance system based on road infrastructure monitoring, including: a traffic flow and pedestrian flow information collection module for collecting weekly traffic flow time series information and weekly pedestrian flow time series information; a meteorological data collection module for collecting annual temperature time series information, annual humidity time series information, and annual precipitation time series information; a first status prediction result obtaining module for activating the first road infrastructure status prediction model whenever a first timing satisfies a first period, processing the weekly traffic flow time series information and the weekly pedestrian flow time series information to obtain a first road infrastructure status prediction result, and restarting the count of the first timing from 0, where the first period is one week; a second status prediction result obtaining module for activating the second road infrastructure status prediction model whenever a second timing satisfies a second period, processing the annual temperature time series information, the annual humidity time series information, and the annual precipitation time series information to obtain a second road infrastructure status prediction result, and starting to count the second period from 0, where the second period is one year; a status prediction result fusion module for fusing the first road infrastructure status prediction result and the second road infrastructure status prediction result through a road infrastructure status fusion model to obtain a road infrastructure status fusion result; a pre-maintenance configuration execution module for extracting the abnormal status type and abnormal trigger time information of the road infrastructure status fusion result and executing the pre-maintenance configuration.

[0013] The predictive maintenance method and system based on road infrastructure monitoring proposed in this application are intended to collect weekly traffic flow time series information and weekly pedestrian flow time series information; collect annual temperature time series information, annual humidity time series information, and annual precipitation time series information; whenever the first timing meets the first period, activate the first road infrastructure status prediction model to obtain the first road infrastructure status prediction result; whenever the second timing meets the second period, activate the second road infrastructure status prediction model to obtain the second road infrastructure status prediction result; through the road infrastructure status fusion model, obtain the road infrastructure status fusion result; extract the abnormal status type and abnormal trigger time information, and execute the pre-maintenance configuration. This solves the technical problems of long detection cycles and lagging maintenance response times in the prior art, and achieves the technical effects of improving the accuracy of road infrastructure status prediction and the timeliness of maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments of the present disclosure will be briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the operations described above or below do not necessarily need to be executed precisely in sequence. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more steps can be removed from these processes.

[0015] Figure 1 Schematic flowchart of the predictive maintenance method based on road infrastructure monitoring provided by the embodiment of the present application;

[0016] Figure 2 Schematic structural diagram of the predictive maintenance system based on road infrastructure monitoring provided by the embodiment of the present application.

[0017] Description of reference numerals: Traffic and pedestrian flow information collection module 10, meteorological data collection module 20, first status prediction result acquisition module 30, second status prediction result acquisition module 40, status prediction result fusion module 50, pre-maintenance configuration execution module 60. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The above description is only an overview of the technical solutions of this application. In order to be able to understand the technical means of this application more clearly, it can be implemented in accordance with the content of the specification. And in order to make the above and other purposes, features, and advantages of this application more obvious and understandable, the following specifically illustrates the specific embodiments of this application.

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the following will further describe this application in detail with reference to the accompanying drawings. The described embodiments should not be construed as limitations on this application. All other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this application.

[0020] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments. However, it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict. The terms "first / second" merely distinguish similar objects and do not represent a specific order for the objects. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application.

[0021] Embodiments of this application provide a predictive maintenance method based on road infrastructure monitoring, as Figure 1 shown. The method includes:

[0022] Step S100, collecting weekly vehicle flow time series information and weekly pedestrian flow time series information.

[0023] Preferably, the vehicle flow and pedestrian flow data within a week are collected and recorded periodically and organized and processed in a time series manner to obtain the weekly vehicle flow time series information and the weekly pedestrian flow time series information. Specifically, the weekly vehicle flow time series information refers to recording the traffic flow on the road (usually measured by the number of vehicles passing through a specific road) within a week at an hourly or finer time granularity, such as the number of vehicles passing through a certain section of the road per hour, vehicle types (such as cars, large trucks, etc.), vehicle speeds, etc., to reflect information such as the traffic load, traffic peak and trough periods of the road, which helps to predict the road usage status, evaluate the road load, calculate road wear and damage, etc.; similarly, the weekly pedestrian flow time series information refers to recording the pedestrian flow on the road or in a specific area within a week at an hourly or finer time granularity (usually measured by the number of pedestrians passing through a certain location), such as the number of pedestrians passing through a certain intersection or sidewalk per hour, and can also be further refined into the activity types of pedestrians (such as commuting peak, tourist flow, etc.), which helps to evaluate the load on the road surface and infrastructure (such as sidewalks, traffic lights, etc.), understand the pedestrian activity patterns in different time periods, and thus assist in predicting the road usage situation.

[0024] Step S200: Collect annual temperature time-series information, annual humidity time-series information, and annual precipitation time-series information.

[0025] Preferably, meteorological data at different times within a year are collected, including temperature, humidity, and precipitation, and then annual temperature time-series information, annual humidity time-series information, and annual precipitation time-series information are formed. Specifically, the annual temperature time-series information refers to the temperature change conditions recorded at different times (such as daily, hourly, etc.) within a year. For example, data such as the daily maximum temperature, minimum temperature, and average daily temperature in a certain area are collected. The annual temperature information can reflect the seasonal climate changes and provide long-term data support for the impact of temperature on roads and infrastructure (such as road surfaces, bridges, tunnels, etc.). Temperature changes can affect the expansion and contraction of road materials and the freezing and thawing processes, thereby affecting the structural stability and durability of roads; the annual humidity time-series information refers to the air humidity data recorded at different times (such as daily, hourly, etc.) within a year. For example, data such as the hourly air humidity and the daily average humidity are recorded. The annual humidity information plays an important role in evaluating the impact of climate change, precipitation, evaporation, and other phenomena on road facilities in different seasons. A high-humidity environment may cause road materials (such as asphalt and concrete) to be corroded by moisture, affecting their structural stability and also accelerating the aging of the road surface; the annual precipitation time-series information refers to the precipitation data recorded at different time periods (such as daily, hourly, etc.) within a year. For example, the daily precipitation or the hourly precipitation is recorded, reflecting the precipitation changes in different seasons. The annual precipitation information is crucial for evaluating the long-term impact of precipitation on road infrastructure. Long-term precipitation (such as rain, snow, heavy rain, etc.) will affect the durability of roads. Especially in rainy and snowy weather, water accumulation may cause corrosion, settlement, or structural damage to road facilities.

[0026] Step S300: Whenever the first timing meets the first period, activate the first road infrastructure status prediction model, process the weekly vehicle flow time-series information and the weekly pedestrian flow time-series information, obtain the first road infrastructure status prediction result, and restart the statistics of the first timing from 0, where the first period is one week.

[0027] Preferably, the first timing refers to recording time points during the operation of the system as time progresses. It is usually a sign of a certain time-triggered mechanism. For example, status updates are performed at regular intervals. When the first timing meets the first cycle, the first road infrastructure status prediction model is activated to process the weekly vehicle flow time series information and the weekly pedestrian flow time series information. Here, the first cycle is one week, that is, whenever the time reaches the duration of one week, the first road infrastructure status prediction model is activated to process the weekly vehicle flow time series information and the weekly pedestrian flow time series information. Specifically, the first road infrastructure status prediction model is based on the collected weekly vehicle flow time series information and the weekly pedestrian flow time series information to predict the status of the road infrastructure, including the usage status of the road, possible damages or situations requiring maintenance. For example, by analyzing the vehicle and pedestrian flow data during peak traffic hours, it can be predicted which roads may experience wear or damage due to overuse, and this is used as the first road infrastructure status prediction result. When a complete cycle (i.e., one week) ends, the first timing is restarted from 0, that is, the statistical data of the previous cycle is cleared and new cycle data recording starts again to ensure accurate tracking and prediction of the data within each week.

[0028] Furthermore, step S300 further includes step S310 of obtaining the initial state matrix of the road infrastructure; step S320 of performing time-domain aggregation on the weekly vehicle flow time series information and the weekly pedestrian flow time series information to obtain the first time-domain binary flow, the second time-domain binary flow up to the Qth time-domain binary flow, where any one time domain has a vehicle flow identifier and a pedestrian flow identifier; step S330 of inputting the initial state matrix of the road infrastructure, the first time-domain binary flow, the second time-domain binary flow up to the Qth time-domain binary flow into the first road infrastructure status prediction model and outputting the first road infrastructure status prediction result.

[0029] Preferably, the existing condition data of the road infrastructure is obtained, including road surface damage conditions, traffic flow, environmental impacts, historical fault records, etc. These data are represented as a numerical matrix of the current health state of the road facilities, that is, the initial state matrix of the road infrastructure is obtained. The initial state matrix of the road infrastructure can contain different types of values, such as whether there are cracks, potholes, etc. in the road surface condition, the current traffic flow of the road, and information on the health status of the road infrastructure such as road surface wear and damage to traffic facilities.

[0030] Preferably, the time-domain aggregation of the weekly vehicle flow time-series information and the weekly pedestrian flow time-series information means summarizing and transforming the weekly vehicle flow time-series information and the weekly pedestrian flow time-series information at a certain time interval (such as hours, days, etc.) into binary flows, that is, summarizing and correlating the vehicle flow and the pedestrian flow at the same time to obtain two identifiers, the vehicle flow identifier and the pedestrian flow identifier. Among them, the vehicle flow identifier is the traffic flow within a certain time period, usually the number of vehicles or the traffic density, and the pedestrian flow identifier is the pedestrian flow within the same time period, usually the number of pedestrians passing through a specific section. Then, multiple time-domain binary flows are obtained, including the first time-domain binary flow, the second time-domain binary flow up to the Qth time-domain binary flow. The first time-domain binary flow includes the vehicle flow and the pedestrian flow in a certain time period (such as the first hour), the second time-domain binary flow includes the vehicle flow and the pedestrian flow data in the second time period, covering Q time periods, representing the changes in traffic load in different time periods. Q is a positive integer greater than 1.

[0031] Preferably, the initial state matrix of the road infrastructure and the binary flow data obtained through time-domain aggregation (including the binary flows in the first time domain, the second time domain, up to the Qth time domain) are input into the first road infrastructure state prediction model. Based on the input data (such as traffic flow, pedestrian flow, initial state matrix of the infrastructure, etc.), the model predicts the future health state of the road infrastructure (such as possible damage, faults, etc.), which may include the prediction of the road health condition, such as whether new damage may occur and whether maintenance is required; the risk assessment within a future period of time, predicting which sections may have problems due to excessive traffic load, aging or environmental factors, etc. Then, it is output as the first road infrastructure state prediction result to help more accurately evaluate the health condition of the road infrastructure and future maintenance requirements.

[0032] Furthermore, step S320 further includes step S321, performing time-domain aggregation on the weekly vehicle flow time-series information to obtain the first time-domain vehicle flow, the second time-domain vehicle flow up to the Nth time-domain vehicle flow; step S322, performing time-domain aggregation on the weekly pedestrian flow time-series information to obtain the first time-domain pedestrian flow, the second time-domain pedestrian flow up to the Mth time-domain pedestrian flow; step S323, fusing the first time-domain vehicle flow, the second time-domain vehicle flow up to the Nth time-domain vehicle flow with the first time-domain pedestrian flow, the second time-domain pedestrian flow up to the Mth time-domain pedestrian flow to obtain the first time-domain binary flow, the second time-domain binary flow up to the Qth time-domain binary flow.

[0033] Preferably, the time-domain aggregation of the weekly vehicle flow time-series information means summarizing these vehicle flow data by specific time periods. The vehicle flow within each time period (time domain) will be processed into an aggregated value, so as to obtain more simplified and operable flow information. For example, based on the data of a week, it is divided into several time periods (such as every hour), and then the vehicle flow within each time period is summarized. Eventually, the first time-domain vehicle flow, the second time-domain vehicle flow up to the Nth time-domain vehicle flow are obtained. Each time domain represents the total vehicle flow within a period of time. Among them, the first time-domain vehicle flow refers to the total vehicle flow within the first time period (for example: the first hour or the first day), the second time-domain vehicle flow refers to the vehicle flow in the second time period, and so on. N is a positive integer greater than 1.

[0034] Preferably, similarly, the time-domain aggregation of the weekly pedestrian flow time-series information means summarizing or processing these pedestrian flow data by specific time periods. For example, the pedestrian flow data are divided into multiple time periods such as hours and days, and the flow data within each time period are summarized. Eventually, the first time-domain pedestrian flow, the second time-domain pedestrian flow up to the Mth time-domain pedestrian flow are obtained. Each time domain represents the total pedestrian flow within that time period. Among them, the first time-domain pedestrian flow refers to the pedestrian flow within the first time period (for example: the first hour or the first day), the second time-domain pedestrian flow refers to the pedestrian flow in the second time period, and so on. M is a positive integer greater than 1.

[0035] Preferably, after completing the time-domain aggregation of the vehicle flow and the pedestrian flow, the first time-domain vehicle flow, the second time-domain vehicle flow up to the Nth time-domain vehicle flow and the first time-domain pedestrian flow, the second time-domain pedestrian flow up to the Mth time-domain pedestrian flow are fused, that is, the values of the vehicle flow and the pedestrian flow within each time period are combined to form a binary flow. This includes combining the first time-domain vehicle flow and the first time-domain pedestrian flow to obtain the first time-domain binary flow representing the comprehensive traffic condition within that time period. Among them, the binary flow represents the comprehensive data considering both the vehicle flow and the pedestrian flow within a time period. Through time-domain fusion, the impacts of both the vehicle flow and the pedestrian flow on the road infrastructure can be evaluated simultaneously, so as to obtain a more comprehensive prediction of the road state and provide a more comprehensive road health assessment.

[0036] Further, step S330 further includes step S331, inputting the initial state matrix of the road infrastructure, the first time-domain binary traffic flow, and the first time-domain step length into the first road infrastructure state prediction model to obtain the first time-domain road infrastructure state prediction result; step S332, inputting the first time-domain road infrastructure state prediction result, the second time-domain binary traffic flow, and the second time-domain step length into the first road infrastructure state prediction model to obtain the second time-domain road infrastructure state prediction result; step S333, until the (Q - 1)th time-domain road infrastructure state prediction result, the Qth time-domain binary traffic flow, and the Qth time-domain step length are input into the first road infrastructure state prediction model to obtain the first road infrastructure state prediction result.

[0037] Preferably, the first time-domain step length, the second time-domain step length to the Qth time-domain step length refer to the time span (i.e., step length) of each time domain, which is used to control the prediction period of the model. It may be a specific time period (such as hours, days, etc.). Using the initial state matrix of the road infrastructure, the first time-domain binary traffic flow, and the first time-domain step length as input data and providing them to the first road infrastructure state prediction model to evaluate the road state in the first period, the output result is the first time-domain road infrastructure state prediction result, including states such as wear and cracks; then using the first time-domain road infrastructure state prediction result as input, together with the second time-domain binary traffic flow and the second time-domain step length, input them into the first road infrastructure state prediction model to evaluate the road state in the second period, and the output result is the second time-domain road infrastructure state prediction result; following the same logic, using the prediction result of the previous time domain as one of the inputs for the current time domain, and then combining the binary traffic flow and step length of the current time domain to predict the state of the current time domain, and finally processing up to the Qth time domain to output the first road infrastructure state prediction result, thereby dynamically predicting the health state of the road infrastructure.

[0038] Further, step S300 further includes step S340 of collecting weekly vehicle flow time series record information, weekly pedestrian flow time series record information, and first road infrastructure status time series record information based on the type of road infrastructure; step S350 of performing time-domain aggregation on the weekly vehicle flow time series record information and the weekly pedestrian flow time series record information to obtain first time-domain binary flow record information, second time-domain binary flow record information, up to Yth time-domain binary flow record information; step S360 of extracting the road infrastructure status at the start time of the first time domain, the road infrastructure status at the start time of the second time domain, up to the road infrastructure status at the start time of the Yth time domain, and the road infrastructure status at the end time of the Yth time domain from the first road infrastructure status time series record information according to the first time domain, the second time domain, up to the Yth time domain; step S370 of using the road infrastructure status at the start time of the first time domain and the first time-domain binary flow record information as the first time-domain input data of the long short-term memory neural network, using the road infrastructure status at the start time of the first time domain as the first time-domain output supervision data of the long short-term memory neural network, using the road infrastructure status at the start time of the second time domain and the second time-domain binary flow record information as the second time-domain input data of the long short-term memory neural network, using the road infrastructure status at the start time of the second time domain as the second time-domain output supervision data of the long short-term memory neural network, until using the road infrastructure status at the start time of the Yth time domain and the Yth time-domain binary flow record information as the Yth time-domain input data of the long short-term memory neural network, and using the road infrastructure status at the end time of the Yth time domain as the Yth time-domain output supervision data of the long short-term memory neural network to train the first road infrastructure status prediction model.

[0039] Preferably, according to the type of road infrastructure, collect weekly vehicle flow time series record information (recording the change of vehicle flow over time within a week), weekly pedestrian flow time series record information (recording the change of pedestrian flow over time within a week), and first road infrastructure status time series record information (recording the status of the road at different time points, including wear degree, crack conditions, etc.), and then perform time-domain aggregation on the weekly vehicle flow time series record information and the weekly pedestrian flow time series record information, that is, aggregate the weekly vehicle flow and weekly pedestrian flow time series record information by time period into first time-domain binary flow record information, second time-domain binary flow record information, up to Yth time-domain binary flow record information, and each binary flow record information contains the comprehensive data of vehicle flow and pedestrian flow within the corresponding time domain.

[0040] Preferably, according to the first time domain, the second time domain up to the Y-th time domain, extract the road infrastructure state at the starting moment of each time domain from the first road infrastructure state time series record information, including the road infrastructure state at the starting moment of the first time domain, the road infrastructure state at the starting moment of the second time domain up to the road infrastructure state at the starting moment of the Y-th time domain, and the road infrastructure state at the end moment of the Y-th time domain, for supervising model training and guiding the model to learn to predict changes in road conditions within each time period; then construct a prediction model based on the Long Short-Term Memory Neural Network (LSTM), using the road infrastructure state at the starting moment of the first time domain and the first time domain binary traffic record information as the input data of the first time domain of the Long Short-Term Memory Neural Network, and using the road infrastructure state at the starting moment of the second time domain as the output supervision data to guide the Long Short-Term Memory Neural Network to learn how to predict the road state in the next time domain based on the initial road state and traffic data.

[0041] Preferably, use the road infrastructure state at the starting moment of the second time domain and the second time domain binary traffic record information as the input data of the second time domain of the Long Short-Term Memory Neural Network, and use the road infrastructure state at the starting moment of the second time domain as the output supervision data of the second time domain of the Long Short-Term Memory Neural Network, until using the road infrastructure state at the starting moment of the Y-th time domain and the Y-th time domain binary traffic record information as the input data of the Y-th time domain of the Long Short-Term Memory Neural Network, and use the road infrastructure state at the end moment of the Y-th time domain as the output supervision data of the Y-th time domain of the Long Short-Term Memory Neural Network, to capture the time dynamic characteristics of the road state. By using the starting and ending states of each time domain as supervision data to guide training, ensure that the predicted results output by the model are consistent with the actually recorded states, and learn to predict the state in the next time period derived from the current state and traffic conditions. Among them, the Long Short-Term Memory Neural Network is suitable for processing long time series data, can remember remote historical information (such as the initial state) and combine the current input for prediction, is suitable for predicting road states with strong time dependence. The road infrastructure state prediction model can dynamically predict the future state of the road infrastructure, combine short-term traffic load and long-term state changes well, and provide strong data support and decision-making basis for subsequent prediction and maintenance.

[0042] Step S400, whenever the second timing meets the second cycle, activate the second road infrastructure state prediction model, process the annual temperature time series information, the annual humidity time series information, and the annual precipitation time series information to obtain the second road infrastructure state prediction result, and start counting from 0 for the second cycle, where the second cycle is one year.

[0043] Preferably, the second timing refers to the time mark when a certain time point reaches a one-year cycle, that is, the time reaches a complete annual cycle, which is a periodically triggered moment. When the second timing meets the second cycle, the second road infrastructure status prediction model is activated to process the annual temperature time series information, annual humidity time series information, and annual precipitation time series information. Among them, the second cycle refers to a one-year time cycle, that is, every year, the annual meteorological data is restarted for statistics and processing to predict the status of the road infrastructure. Specifically, the second road infrastructure status prediction model is used to process the annual meteorological data, and the long-term trend of the road infrastructure status is predicted through the annual meteorological data (such as annual temperature, humidity, precipitation, etc.). At the end of each annual cycle, this model is activated, and the temperature, humidity, and precipitation data of the past year are used as inputs to predict the long-term health status of the road. For example, it predicts problems such as aging, crack expansion, and structural damage that may occur on certain roads due to temperature fluctuations, humidity changes, or precipitation, and then obtains the second road infrastructure status prediction result. When the second cycle (i.e., one year) ends, the second cycle is counted from 0. That is, at the end of the one-year cycle, all statistical information related to the annual meteorological data is reset, and new data collection and analysis for the new year are started, so as to continuously and dynamically evaluate the impact of annual meteorological changes on the road infrastructure and adjust the maintenance plan in a timely manner.

[0044] Preferably, the first sample data is collected according to the type of road infrastructure, including annual temperature time series record information, annual humidity time series record information, annual precipitation time series record information, and second road infrastructure status time series record information. The annual temperature time series record information records the temperature fluctuations in each time period within a year (such as the average temperature per month, per day, etc.). The annual humidity time series record information records the air humidity conditions in different time periods within a year (such as daily average humidity, humidity fluctuations, etc.). The annual precipitation time series record information records the time series data of precipitation within a year, such as monthly or daily precipitation, which helps to analyze the long-term impact of meteorological conditions on the road. The second road infrastructure status time series record information records the long-term status of the road facilities, including information such as road wear, cracks, and settlements, and is used to analyze the health changes of the road facilities under different meteorological conditions.

[0045] Preferably, the average vehicle flow and the average pedestrian flow in the first sample data should meet the set threshold conditions, that is, the annual average value of the vehicle flow cannot exceed a certain specific value, which means that the area corresponding to the data set does not encounter excessive traffic load, and the annual average value of the pedestrian flow should also be lower than a preset threshold, restricting the range of the sample data, so that the samples for model training focus on road infrastructure with lower traffic load. The vehicle flow threshold and the pedestrian flow threshold avoid road conditions under high traffic load conditions, so as to analyze the response and aging mode of the road under relatively normal or lower load conditions; then the time series record information of the annual temperature, humidity, precipitation and the second road infrastructure status is used to train the second road infrastructure status prediction model.

[0046] Preferably, a prediction model is constructed based on a long short-term memory neural network. The time series record information of the annual temperature, the time series record information of the annual humidity, and the time series record information of the annual precipitation are used as input data, and the time series record information of the second road infrastructure status is used as output supervision data to train the prediction model, and the second road infrastructure status prediction model is obtained, which can predict the status of road infrastructure according to meteorological conditions (such as temperature, humidity, precipitation), and help predict possible changes in future road facilities. For example, higher humidity or precipitation may cause the road facilities to corrode or age faster, while high temperature may cause the expansion and cracking of the road surface material, so as to formulate a more efficient pre-maintenance plan.

[0047] Step S500, through the road infrastructure status fusion model, the first road infrastructure status prediction result and the second road infrastructure status prediction result are used to obtain the road infrastructure status fusion result.

[0048] Preferably, the road infrastructure status fusion model is used to fuse the first road infrastructure status prediction result and the second road infrastructure status prediction result obtained based on different data sources (such as vehicle flow, pedestrian flow and meteorological data such as temperature, humidity and precipitation). Usually, some mathematical or statistical methods (such as weighted average, data fusion algorithm, machine learning, etc.) are used to combine the two prediction results (the first prediction result and the second prediction result) to obtain the road infrastructure status fusion result. Among them, the road infrastructure status fusion model not only considers the impact of short-term traffic load, but also considers the role of long-term environmental factors to more accurately and comprehensively reflect the true health status of road facilities.

[0049] Further, step S500 further includes step S510 of collecting second sample data based on the type of road infrastructure, where the second sample data includes second sample weekly vehicle flow time series record information, second sample weekly pedestrian flow time series record information, second sample annual temperature time series record information, second sample annual humidity time series record information, second sample annual precipitation time series record information, and second sample road infrastructure status information; step S520 of processing the second sample weekly vehicle flow time series record information and the second sample weekly pedestrian flow time series record information according to the first road infrastructure status prediction model to obtain first road infrastructure status identification data; step S530 of processing the second sample weekly pedestrian flow time series record information, the second sample annual temperature time series record information, the second sample annual humidity time series record information, and the second sample annual precipitation time series record information according to the second road infrastructure status prediction model to obtain second road infrastructure status identification data; and step S540 of training the road infrastructure status fusion model with the second sample road infrastructure status data as the supervision and the first road infrastructure status identification data and the second road infrastructure status identification data as the inputs.

[0050] Preferably, collecting second sample data according to the type of road infrastructure includes second sample weekly vehicle flow time series record information (vehicle flow data within a week in the second sample), second sample weekly pedestrian flow time series record information (pedestrian flow data within a week in the second sample), second sample annual temperature time series record information (recording the temperature change situation in a year in the second sample), second sample annual humidity time series record information (humidity change situation in a year in the second sample), second sample annual precipitation time series record information (precipitation data in a year in the second sample), and second sample road infrastructure status information (road facility status data at different time points in the second sample, including information such as the health condition, damage degree, wear, and cracks of the road). Processing the second sample weekly vehicle flow time series record information and the second sample weekly pedestrian flow time series record information with the first road infrastructure status prediction model outputs first road infrastructure status identification data, that is, the prediction result of the traffic flow data based on the first model, which is used to represent the health status of the road during this period.

[0051] Preferably, the second sample weekly pedestrian flow time series record information, the second sample annual temperature time series record information, the second sample annual humidity time series record information, and the second sample annual precipitation time series record information are processed using the second road infrastructure status prediction model to output the second road infrastructure status identification data, that is, the prediction result based on the second model, to evaluate the health status of the road under these meteorological conditions; finally, the first road infrastructure status identification data and the second road infrastructure status identification data are used as input data, and the second sample road infrastructure status data is used as the supervision data for fusion training to obtain a road infrastructure status fusion model, which is used to combine the prediction results from two prediction models (traffic flow - based and meteorological condition - based), combining short - term and long - term impacts, thereby improving the prediction accuracy of the future status of road facilities, obtaining a more comprehensive and accurate road status prediction, and further generating a more accurate road maintenance prediction.

[0052] Step S600, extract the abnormal status type and abnormal trigger time information of the road infrastructure status fusion result, and execute pre - maintenance configuration.

[0053] Preferably, by analyzing the fusion result of the road infrastructure status, abnormal conditions are identified, and appropriate preventive maintenance measures are taken based on this abnormal information. Specifically, through the predicted result of the road infrastructure status after fusion, it is found which specific road problems or failure types may occur. The abnormal status can be problems with the road structure, surface damage, excessive wear, cracks, settlement, corrosion, etc. For example, it is detected that a certain section of the road has pavement cracks due to excessive traffic flow, or abnormal situations such as water accumulation caused by excessive precipitation or bridge corrosion; the abnormal trigger time information refers to the specific time point when problems or failures are found in the prediction of the road infrastructure status. According to the data cycle (such as one week or one year), it is determined when the abnormality occurs, usually corresponding to specific traffic loads or meteorological conditions. For example, it is found that a certain section of the road has an abnormality during a specific time period (such as peak hours or bad weather), which may be because the traffic flow is too large, the precipitation is too much, or the temperature is too low at this time, resulting in road problems; pre - maintenance configuration means that after the abnormal status and its trigger time are found, maintenance measures are taken in advance according to the prediction result. Among them, pre - maintenance configuration is a prediction - based maintenance strategy, which can avoid safety accidents caused by infrastructure damage by identifying potential problems in advance, thereby reducing maintenance costs, increasing the service life and safety of the road.

[0054] In the above text, reference is made to Figure 1 has been described in detail the prediction maintenance method based on road infrastructure monitoring according to the embodiments of the present invention. Next, reference will be made to Figure 2 describe a prediction maintenance system based on road infrastructure monitoring according to the embodiments of the present invention.

[0055] The predictive maintenance system based on road infrastructure monitoring according to an embodiment of the present invention is used to solve the technical problems of long detection cycles and lagging maintenance response times existing in the prior art, and achieves the technical effects of improving the accuracy of road infrastructure status prediction and the timeliness of maintenance. The predictive maintenance system based on road infrastructure monitoring includes: a traffic flow and pedestrian flow information collection module 10, a meteorological data collection module 20, a first status prediction result obtaining module 30, a second status prediction result obtaining module 40, a status prediction result fusion module 50, and a pre-maintenance configuration execution module 60.

[0056] The traffic flow and pedestrian flow information collection module 10 is used to collect weekly traffic flow time series information and weekly pedestrian flow time series information; the meteorological data collection module 20 is used to collect annual temperature time series information, annual humidity time series information, and annual precipitation time series information; the first status prediction result obtaining module 30 is used to activate the first road infrastructure status prediction model whenever the first timing meets the first period, process the weekly traffic flow time series information and the weekly pedestrian flow time series information, obtain the first road infrastructure status prediction result, and restart counting the first timing from 0, where the first period is one week; the second status prediction result obtaining module 40 is used to activate the second road infrastructure status prediction model whenever the second timing meets the second period, process the annual temperature time series information, the annual humidity time series information, and the annual precipitation time series information, obtain the second road infrastructure status prediction result, and start counting the second period from 0, where the second period is one year; the status prediction result fusion module 50 is used to fuse the first road infrastructure status prediction result and the second road infrastructure status prediction result through a road infrastructure status fusion model to obtain a road infrastructure status fusion result; the pre-maintenance configuration execution module 60 is used to extract the abnormal status type and abnormal trigger time information of the road infrastructure status fusion result and execute the pre-maintenance configuration.

[0057] Next, the specific configuration of the first status prediction result obtaining module 30 will be described in detail. The first status prediction result obtaining module 30 further includes: obtaining an initial road infrastructure status matrix; performing time-domain aggregation on the weekly traffic flow time series information and the weekly pedestrian flow time series information to obtain a first time-domain binary flow, a second time-domain binary flow, up to a Qth time-domain binary flow, where any one time domain has a traffic flow identifier and a pedestrian flow identifier; inputting the initial road infrastructure status matrix, the first time-domain binary flow, the second time-domain binary flow, up to the Qth time-domain binary flow into the first road infrastructure status prediction model, and outputting the first road infrastructure status prediction result.

[0058] Next, the specific configuration of the first state prediction result obtaining module 30 will be further described in detail. The first state prediction result obtaining module 30 further includes: performing time-domain aggregation on the weekly vehicle flow time series information to obtain the first time-domain vehicle flow, the second time-domain vehicle flow, up to the Nth time-domain vehicle flow; performing time-domain aggregation on the weekly pedestrian flow time series information to obtain the first time-domain pedestrian flow, the second time-domain pedestrian flow, up to the Mth time-domain pedestrian flow; fusing the first time-domain vehicle flow, the second time-domain vehicle flow, up to the Nth time-domain vehicle flow with the first time-domain pedestrian flow, the second time-domain pedestrian flow, up to the Mth time-domain pedestrian flow to obtain the first time-domain binary flow, the second time-domain binary flow, up to the Qth time-domain binary flow.

[0059] Next, the specific configuration of the first state prediction result obtaining module 30 will be further described in detail. The first state prediction result obtaining module 30 further includes: inputting the initial state matrix of the road infrastructure, the first time-domain binary flow, and the first time-domain step length into the first road infrastructure state prediction model to obtain the first time-domain road infrastructure state prediction result; inputting the first time-domain road infrastructure state prediction result, the second time-domain binary flow, and the second time-domain step length into the first road infrastructure state prediction model to obtain the second time-domain road infrastructure state prediction result; until inputting the (Q - 1)th time-domain road infrastructure state prediction result, the Qth time-domain binary flow, and the Qth time-domain step length into the first road infrastructure state prediction model to obtain the first road infrastructure state prediction result.

[0060] Next, the specific configuration of the first state prediction result obtaining module 30 will be further described in detail. The first state prediction result obtaining module 30 further includes: collecting weekly traffic flow time series record information, weekly pedestrian flow time series record information, and first road infrastructure state time series record information based on the type of road infrastructure; performing time domain aggregation on the weekly traffic flow time series record information and the weekly pedestrian flow time series record information to obtain first time domain binary traffic flow record information, second time domain binary traffic flow record information up to the Y-th time domain binary traffic flow record information; according to the first time domain, the second time domain up to the Y-th time domain, extracting the road infrastructure state at the start time of the first time domain, the road infrastructure state at the start time of the second time domain up to the road infrastructure state at the start time of the Y-th time domain, and the road infrastructure state at the end time of the Y-th time domain from the first road infrastructure state time series record information; using the road infrastructure state at the start time of the first time domain and the first time domain binary traffic flow record information as the first time domain input data of the long short-term memory neural network, using the road infrastructure state at the start time of the first time domain as the first time domain output supervision data of the long short-term memory neural network, using the road infrastructure state at the start time of the second time domain and the second time domain binary traffic flow record information as the second time domain input data of the long short-term memory neural network, using the road infrastructure state at the start time of the second time domain as the second time domain output supervision data of the long short-term memory neural network, until using the road infrastructure state at the start time of the Y-th time domain and the Y-th time domain binary traffic flow record information as the Y-th time domain input data of the long short-term memory neural network, and using the road infrastructure state at the end time of the Y-th time domain as the Y-th time domain output supervision data of the long short-term memory neural network to train the first road infrastructure state prediction model.

[0061] Next, the specific configuration of the second state prediction result obtaining module 40 will be described in detail. The second state prediction result obtaining module 40 further includes: collecting first sample data based on the type of road infrastructure, where the first sample data includes annual temperature time series record information, annual humidity time series record information, annual precipitation time series record information, and second road infrastructure state time series record information; where the average annual traffic flow of the first sample data is less than or equal to the vehicle flow threshold, and the average annual pedestrian flow of the first sample data is less than or equal to the pedestrian flow threshold; training the second road infrastructure state prediction model according to the annual temperature time series record information, the annual humidity time series record information, the annual precipitation time series record information, and the second road infrastructure state time series record information.

[0062] Next, the specific configuration of the status prediction result fusion module 50 will be described in detail. The status prediction result fusion module 50 further includes: collecting second sample data based on the type of road infrastructure, where the second sample data includes second sample weekly traffic flow time series record information, second sample weekly pedestrian flow time series record information, second sample annual temperature time series record information, second sample annual humidity time series record information, second sample annual precipitation time series record information, and second sample road infrastructure status information; processing the second sample weekly traffic flow time series record information and the second sample weekly pedestrian flow time series record information according to the first road infrastructure status prediction model to obtain first road infrastructure status identification data; processing the second sample weekly pedestrian flow time series record information, the second sample annual temperature time series record information, the second sample annual humidity time series record information, and the second sample annual precipitation time series record information according to the second road infrastructure status prediction model to obtain second road infrastructure status identification data; using the second sample road infrastructure status data as supervision and using the first road infrastructure status identification data and the second road infrastructure status identification data as inputs to train the road infrastructure status fusion model.

[0063] The prediction and maintenance system based on road infrastructure monitoring provided by the embodiments of the present invention can execute the prediction and maintenance method based on road infrastructure monitoring provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing the method.

[0064] Although this application makes various references to certain modules in the system according to the embodiments of this application, however, any number of different modules can be used and run on the user terminal and / or the server. The included units and modules are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for easy distinction from each other and do not limit the protection scope of the present invention.

[0065] The above specific embodiments do not constitute a limitation on the protection scope of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims

1. A predictive maintenance method based on road infrastructure monitoring, characterized in that Including: Collecting weekly vehicle flow time series information and weekly pedestrian flow time series information; Collecting annual temperature time series information, annual humidity time series information and annual precipitation time series information; Whenever the first timer meets the first period, activate the first road infrastructure status prediction model, process the weekly vehicle flow time series information and the weekly pedestrian flow time series information to obtain the first road infrastructure status prediction result, and restart counting the first timer from 0, where the first period is one week; Whenever the second timer meets the second period, activate the second road infrastructure status prediction model, process the annual temperature time series information, the annual humidity time series information and the annual precipitation time series information to obtain the second road infrastructure status prediction result, and start counting the second period from 0, where the second period is one year; Fusing the first road infrastructure status prediction result and the second road infrastructure status prediction result through a road infrastructure status fusion model to obtain a road infrastructure status fusion result; Extracting the abnormal status type and abnormal trigger time information of the road infrastructure status fusion result and executing pre-maintenance configuration; The first road infrastructure status prediction result is used to reflect the status of road infrastructure affected by traffic flow in the short term, including the usage status of the road, damage or the need for repair; The second road infrastructure status prediction result reflects the status of road infrastructure affected by meteorological factors in the long term, including problems such as aging, crack expansion, and structural damage of the road due to temperature fluctuations, humidity changes or precipitation meteorological factors; The road infrastructure status fusion result is used to reflect the true health status of road infrastructure, including problems with the road structure, surface damage, excessive wear, cracks, settlement, corrosion abnormal status.

2. The predictive maintenance method based on road infrastructure monitoring according to claim 1, wherein, Whenever the first timer meets the first period, activate the first road infrastructure status prediction model, process the weekly vehicle flow time series information and the weekly pedestrian flow time series information to obtain the first road infrastructure status prediction result, including: Obtaining the initial state matrix of road infrastructure; Performing time-domain aggregation on the weekly vehicle flow time series information and the weekly pedestrian flow time series information to obtain the first time-domain binary flow, the second time-domain binary flow until the Qth time-domain binary flow, where any time domain has a vehicle flow identifier and a pedestrian flow identifier; Inputting the road infrastructure initial state matrix, the first time-domain binary flow, the second time-domain binary flow until the Qth time-domain binary flow into the first road infrastructure status prediction model and outputting the first road infrastructure status prediction result.

3. The predictive maintenance method based on road infrastructure monitoring according to claim 2, wherein, Performing time-domain aggregation on the weekly vehicle flow time series information and the weekly pedestrian flow time series information to obtain the first time-domain binary flow, the second time-domain binary flow until the Qth time-domain binary flow, including: Performing time-domain aggregation on the weekly vehicle flow time series information to obtain the first time-domain vehicle flow, the second time-domain vehicle flow until the Nth time-domain vehicle flow; Performing time-domain aggregation on the weekly pedestrian flow time series information to obtain the first time-domain pedestrian flow, the second time-domain pedestrian flow until the Mth time-domain pedestrian flow; Fuse the first time-domain vehicle flow, the second time-domain vehicle flow up to the Nth time-domain vehicle flow, and the first time-domain pedestrian flow, the second time-domain pedestrian flow up to the Mth time-domain pedestrian flow to obtain the first time-domain binary flow, the second time-domain binary flow up to the Qth time-domain binary flow.

4. The predictive maintenance method based on road infrastructure monitoring according to claim 2, characterized in that, Input the initial state matrix of the road infrastructure, the first time-domain binary flow, the second time-domain binary flow up to the Qth time-domain binary flow into the first road infrastructure state prediction model, and output the first road infrastructure state prediction result, including: Input the initial state matrix of the road infrastructure, the first time-domain binary flow and the first time-domain step size into the first road infrastructure state prediction model to obtain the first time-domain road infrastructure state prediction result; Input the first time-domain road infrastructure state prediction result, the second time-domain binary flow and the second time-domain step size into the first road infrastructure state prediction model to obtain the second time-domain road infrastructure state prediction result; Until input the (Q - 1)th time-domain road infrastructure state prediction result, the Qth time-domain binary flow and the Qth time-domain step size into the first road infrastructure state prediction model to obtain the first road infrastructure state prediction result.

5. The predictive maintenance method based on road infrastructure monitoring according to claim 1, characterized in that The construction steps of the first road infrastructure state prediction model include: Collect weekly vehicle flow time series record information, weekly pedestrian flow time series record information and the first road infrastructure state time series record information based on the type of road infrastructure; Perform time-domain aggregation on the weekly vehicle flow time series record information and the weekly pedestrian flow time series record information to obtain the first time-domain binary flow record information, the second time-domain binary flow record information up to the Yth time-domain binary flow record information; According to the first time-domain, the second time-domain up to the Yth time-domain, extract the road infrastructure state at the start time of the first time-domain, the road infrastructure state at the start time of the second time-domain up to the road infrastructure state at the start time of the Yth time-domain, and the road infrastructure state at the end time of the Yth time-domain from the first road infrastructure state time series record information; Use the road infrastructure state at the start time of the first time-domain and the first time-domain binary flow record information as the first time-domain input data of the long short-term memory neural network, use the road infrastructure state at the start time of the first time-domain as the first time-domain output supervision data of the long short-term memory neural network, use the road infrastructure state at the start time of the second time-domain and the second time-domain binary flow record information as the second time-domain input data of the long short-term memory neural network, use the road infrastructure state at the start time of the second time-domain as the second time-domain output supervision data of the long short-term memory neural network, until use the road infrastructure state at the start time of the Yth time-domain and the Yth time-domain binary flow record information as the Yth time-domain input data of the long short-term memory neural network, use the road infrastructure state at the end time of the Yth time-domain as the Yth time-domain output supervision data of the long short-term memory neural network to train the first road infrastructure state prediction model.

6. The predictive maintenance method based on road infrastructure monitoring according to claim 1, characterized in that, The steps for constructing the second road infrastructure status prediction model are as follows: Based on the type of road infrastructure, collect the first sample data, where the first sample data includes annual temperature time series record information, annual humidity time series record information, annual precipitation time series record information, and second road infrastructure status time series record information; Among them, the average annual traffic flow of the first sample data is less than or equal to the vehicle flow threshold, and the average annual pedestrian flow of the first sample data is less than or equal to the pedestrian flow threshold; According to the annual temperature time series record information, the annual humidity time series record information, the annual precipitation time series record information, and the second road infrastructure status time series record information, train the second road infrastructure status prediction model.

7. The predictive maintenance method based on road infrastructure monitoring according to claim 1, characterized in that The steps for constructing the road infrastructure status fusion model are as follows: Based on the type of road infrastructure, collect the second sample data, where the second sample data includes second sample weekly traffic flow time series record information, second sample weekly pedestrian flow time series record information, second sample annual temperature time series record information, second sample annual humidity time series record information, second sample annual precipitation time series record information, and second sample road infrastructure status information; According to the first road infrastructure status prediction model, process the second sample weekly traffic flow time series record information and the second sample weekly pedestrian flow time series record information to obtain the first road infrastructure status identification data; According to the second road infrastructure status prediction model, process the second sample weekly pedestrian flow time series record information, the second sample annual temperature time series record information, the second sample annual humidity time series record information, and the second sample annual precipitation time series record information to obtain the second road infrastructure status identification data; Using the second sample road infrastructure status information as the supervision and the first road infrastructure status identification data and the second road infrastructure status identification data as the inputs, train the road infrastructure status fusion model.

8. A predictive maintenance system based on road infrastructure monitoring, characterized in that, The system is used to implement the predictive maintenance method based on road infrastructure monitoring according to any one of claims 1 to 7, and the system includes: A traffic flow and pedestrian flow information collection module for collecting weekly traffic flow time series information and weekly pedestrian flow time series information; A meteorological data collection module for collecting annual temperature time series information, annual humidity time series information, and annual precipitation time series information; A first status prediction result obtaining module for activating the first road infrastructure status prediction model whenever the first timing meets the first period, processing the weekly traffic flow time series information and the weekly pedestrian flow time series information to obtain the first road infrastructure status prediction result, and restarting the statistics of the first timing from 0, where the first period is one week; A second status prediction result obtaining module for activating the second road infrastructure status prediction model whenever the second timing meets the second period, processing the annual temperature time series information, the annual humidity time series information, and the annual precipitation time series information to obtain the second road infrastructure status prediction result, and starting to count the second period from 0, where the second period is one year; A state prediction result fusion module, which is used to fuse the first road infrastructure state prediction result and the second road infrastructure state prediction result through a road infrastructure state fusion model to obtain a road infrastructure state fusion result; A pre-maintenance configuration execution module, which is used to extract the abnormal state type and abnormal trigger moment information of the road infrastructure state fusion result and execute the pre-maintenance configuration.

Images