Road typhoon disaster real-time monitoring system and method

Abstract

The invention discloses a real-time monitoring system for road typhoon disasters. The real-time monitoring system comprises a composite detection sensor and an equipment terminal, wherein the composite detection sensor consists of a rainfall sensor, a wind speed sensor, a signal transmission antenna and a fixed base; the equipment terminal is composed of a central processing unit, a main display screen, an auxiliary display screen, a prompting lamp, an operator and a signal receiving antenna. The invention further discloses a road typhoon disaster real-time monitoring method. According to the method, typhoon key disaster-inducing factors are screened, disaster-inducing risk grades are divided, accurate prediction and intelligent warning are carried out, and a prevention and treatment scheme is provided in a targeted manner, so that the difficulty of typhoon disaster prevention and treatment of a current road construction level is solved, road safety management during typhoon is improved, and property loss and casualties of a construction unit are avoided.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of road construction disaster prevention, and particularly relates to a real-time monitoring system and method for road typhoon disaster. BACKGROUND

[0002] When roads are constructed in coastal areas, they are disturbed and damaged by typhoons every year. Typhoons have a huge impact on construction progress and cost, have the characteristics of high frequency, large disaster intensity, wide influence range, etc. A large number of typhoons have caused road damage at home and abroad, which not only causes economic losses and delays in construction, but also causes casualties of construction personnel. Therefore, analyzing the disaster-causing factors of typhoons and accurately predicting the damage level of typhoons on roads has important theoretical significance and practical guiding role for road construction safety.

[0003] At present, typhoon disaster prediction mainly focuses on provincial and above ranges, lacks subdivision and real-time dynamic risk assessment for road construction, and the typhoon disaster prediction method also has certain limitations. Therefore, a real-time monitoring system and method for road typhoon disaster are proposed. SUMMARY

[0004] To solve the above problems in the prior art, the application provides a real-time monitoring system and method for road typhoon disaster, which divides, accurately predicts, intelligently warns and proposes a targeted prevention scheme for the typhoon key disaster-causing factors, disaster risk level, solves the difficulty of typhoon disaster prevention at the current road construction level, improves road safety management during typhoons, and avoids property loss and personnel casualties of construction units.

[0005] The technical scheme to achieve the above-mentioned purposes is as follows: One of the application is a real-time monitoring system for road typhoon disaster, comprising: a composite detection sensor and a device terminal; wherein, The composite detection sensor is composed of a rainfall sensor, a wind speed sensor, a signal transmission antenna and a fixed base; The device terminal is composed of a central processor, a main display screen, a secondary display screen, a prompt light, an operator and a signal receiving antenna; The rainfall sensor is used to collect typhoon rainfall data; The wind speed sensor is used to collect typhoon wind speed data; The central processor is used to introduce a CNN-LSTM model to predict the road disaster level, and send color-coded optical signals according to the predicted level; It is also used to judge the disaster level at this moment and the disaster level 5 seconds ago; Also used for dynamic contribution analysis of 8 influence factors, by introducing the coefficient of variation CV to calculate each influence factor separately, analyze the factor that leads to the highest contribution of disaster level, according to the factor with the highest contribution, start the specific preplan; The main display screen is used for displaying the specific preplan; The auxiliary display screen is used for displaying the typhoon parameters near the road at this time in real time; The prompt light is used for displaying the color-coded optical signal sent according to the predicted level; The operator is used for switching the prediction level at different times and the corresponding specific preplan; The signal transmission antenna of the composite detection sensor and the signal receiving antenna of the equipment terminal establish communication, which is used for real-time transmission of sensor data.

[0006] Preferably, if the central processor predicts the level V, the disaster degree is very serious, and the prompt light flashes red; If the central processor predicts the level IV, the disaster degree is serious, and the prompt light is always red; If the central processor predicts the level III, the disaster degree is relatively serious, and the prompt light is always yellow; If the central processor predicts the level II, the disaster degree is relatively light, and the prompt light flashes green; If the central processor predicts the level I, the disaster degree is light, and the prompt light is always green.

[0007] Preferably, the central processor judges the disaster level at this moment and the disaster level 5s ago, if the disaster level increases, the prompt light flashes red-yellow-green alternately, and lasts for 5s.

[0008] The second road typhoon disaster real-time monitoring method of the application comprises: Step S1, using the relative economic loss per kilometer, the average personnel casualty and the relative delay time length per kilometer as the risk level division standard, the specific level of the disaster road is evaluated; Step S2, according to the correlation analysis of typhoon influence factors, 8 disaster factors are introduced as the influence factors of typhoon damage to the road; Step S3, introducing the CNN-LSTM model to the central processor, using 335 groups of road data damaged by typhoon in coastal and Hainan areas in recent 20 years to train the CNN-LSTM model; Step S4, combining the road design drawings, using GIS-BIM to model the road, and through finite element analysis, the road model is discretized into The regular grid can be used, and each network center point can be used as a candidate sensor placement point. A multi-objective optimization model is constructed using the NSGA-III algorithm to screen all candidate points. Step S5: After selecting the optimal detection points for wind speed and rainfall, place a typhoon multi-parameter composite detection sensor at the location. The composite detection sensor collects data on typhoon wind speed and rainfall every 5 seconds. Step S6: Arrange the collected data in sequence to build a dynamic database. Based on the trained CNN-LSTM model and the normalized ps function, predict the current road disaster level and send color-coded optical signals according to the predicted level. Step S7: The secondary display screen displays the typhoon parameters near the road in real time, and the central processing unit judges the disaster level at this moment and the disaster level 5 seconds ago. If the disaster level increases, the indicator light flashes red-yellow-green alternately for 5 seconds. Step S8: The central processing unit performs dynamic contribution analysis on the eight influencing factors. By introducing the coefficient of variation (CV), each influencing factor is calculated separately, and the factor that contributes the most to the disaster severity is analyzed. Step S9: Based on the factor with the greatest contribution, activate the contingency plan accordingly and intelligently output treatment suggestions on the main display screen.

[0009] Preferably, in step S1, the specific levels of the affected roads include: Level I, Level II, Level III, Level IV, and Level V; wherein, If the affected road is classified as Level I, the degree of damage is considered minor, and the relative economic loss per kilometer is... The average number of casualties was The relative delay time per kilometer of construction period ; If the affected road is classified as Level II, the degree of damage is relatively minor, and the relative economic loss per kilometer is [not specified]. The average number of casualties was The relative delay time per kilometer of construction period ; If the affected road is classified as Level III, the damage is considered severe, with a relative economic loss per kilometer. The average number of casualties was The relative delay time per kilometer of construction period ; If the affected road is classified as Level IV, the degree of damage is considered severe, and the relative economic loss per kilometer is [not specified]. The average number of casualties was The relative delay time per kilometer of construction period ; If the affected road is classified as Class V, the degree of damage is considered extremely severe, with a relative economic loss per kilometer. , the average personnel casualty is , the relative delay time per kilometer of the construction period .

[0010] Preferably, in step S2, the 8 disaster-causing factors include: maximum wind speed, cumulative rainfall, maximum rainfall, negative terrain, soil loose coefficient, average building height, road drainage capacity, and total value of on-site equipment.

[0011] Preferably, in step S3, the data is normalized by the MATLAB built-in function normalize and then transmitted to the CNN-LSTM model for training. In the CNN-LSTM model structure, the CNN has 24 1x3 size two-dimensional convolution kernel modules, and the LSTM has a total of 64 memory cells. According to the 8 influence factors and 1 disaster level in the database, the trained model and the normalized mapminmax_ps function are packaged and stored in the central processing unit by the "save" function.

[0012] Preferably, in step S4, the NSGA-III algorithm is used to construct a multi-objective optimization model to screen all candidate points, and the specific operation is as follows: Let the candidate sensor point set be , the road discrete unit set be , and the wind speed sensor deployment decision variable be ; Let represent the distance between the candidate point and the road unit , and define the exponential decay influence function as: ; ; In the formula, is the monitoring influence weight of the wind speed sensor on the road unit at the candidate point , and the larger the value, the stronger the wind speed monitoring ability of the unit, is the influence scale of the wind speed sensor, and the influence weight decays to about 1 / e for every increase in distance, is the monitoring influence weight of the rainfall sensor on the road unit at the candidate point , is the influence scale of the rainfall sensor; The objective function includes: Maximize the weighted coverage rate: ; ; ; In the formula, is the vulnerability weight or importance of the road unit , the larger the value, the more critical the unit, is the weight parameter of the wind speed factor, used to adjust the contribution of wind speed to coverage, is the wind speed coverage accumulation of road unit k, which depends on which candidate points are deployed with wind speed sensors, is the weight parameter of the rainfall factor, used to adjust the contribution of rainfall to coverage, is the rainfall coverage accumulation of road unit k, is whether to deploy wind speed sensors at candidate points , 1 = deploy, 0 = do not deploy, is whether to deploy rainfall sensors at candidate points , 1 = deploy, 0 = do not deploy; Minimize the total deployment cost: ; In the formula, is, .

[0013] Maximize model observation sensitivity: In the formula, is the model prediction sensitivity contribution value of road unit when wind speed sensors are deployed at candidate points , estimated by historical data, model gradient or finite difference, is the model prediction sensitivity contribution value of road unit when rainfall sensors are deployed at candidate points .

[0014] Preferably, in step S9, after each typhoon disaster, the predicted grade is automatically compared with the actual disaster result to generate an actual grade label, and the actual post-disaster data is fed back to the central processor and added to the incremental learning training of the CNN-LSTM model to update the weight parameters of the model and dynamically optimize and self-learn; wherein the weight parameters include the wind speed risk weight at the candidate point and the rainfall risk weight at the candidate point; Assuming the observation value of the th disaster-causing factor at the last 12 sampling time points is , then the mean and standard deviation of the factor in the sliding window are: ， ; introducing the stability constant , the coefficient of variation of this factor is obtained : ; In the formula, is the stability constant ; The CV value of each factor is normalized to obtain the relative fluctuation contribution rate : ; Meanwhile, the central processing unit uses the trained CNN-LSTM model to predict the disaster grade, and calculates the sensitivity of the input factor; The sensitivity is realized by the finite difference method, that is, a small disturbance is added to the first factor, and the change rate of the model output grade is calculated : : ; Among them, indicates the disaster grade score predicted by the model; The sensitivity is also normalized: ; Finally, according to the model prediction accuracy, the weight coefficient is determined as 0.85, the weight of the on-site sensor data fluctuation is 0.15, and the dynamic contribution degree calculation formula of the factor is: ; The central processing unit sorts all factors , if the maximum contribution factor meets , the corresponding special plan of the factor is directly triggered; if the cumulative contribution of the top 3 factors exceeds 0.7, the joint plan is triggered.

[0015] Compared with the prior art, the beneficial effects of the present application are: the present application carries out typhoon detection point layout through GIS-BIM road modeling + finite element discretization road division + NSGA-III algorithm multi-objective optimization triple technology, solves the problem of monitoring blind area; while increasing the accuracy of typhoon detection, optimizing resource allocation, reducing the layout point, and reducing the detection cost; typhoon parameter measurement is carried out every 5s, so that the staff can receive the change of typhoon more timely; The present application proposes a multi-index road disaster classification basis caused by typhoon, including one kilometer relative economic loss, average casualty, and one kilometer relative delay of construction period, so that the construction unit can more comprehensively evaluate the disaster situation of the road when facing the typhoon, and avoid one-sided misjudgment; The application introduces a CNN-LSTM-CV combined decision model with faster emergency response to predict the road disaster level, on the one hand, the CNN extracts spatial features (such as terrain, drainage capacity) and the LSTM captures time sequence rules (such as wind speed / rainfall mutation), both of which improve the prediction accuracy and enhance the dynamic response capability; on the other hand, the CV analyzes the real-time judgment of the dominant factor of typhoon damage, accurately locates the risk source, and then puts forward the coping strategy, thereby improving the emergency disposal ability of the construction unit under the typhoon disaster; The application converts the prediction result of the road disaster level into light display through optical coding, so that the damage of the typhoon to the road can be visualized and intuitive, and the staff can receive information more efficiently. The composite sensor with tower-shaped rainproof design and the terminal double-screen interactive monitoring method have strong industrial practicality. BRIEF DESCRIPTION OF DRAWINGS

[0016] The accompanying drawings are used to provide a further understanding of the application, and constitute a part of the specification, together with the embodiments of the application, to explain the application, and do not constitute a limitation on the application. In the drawings: Figure 1 is a composite detection sensor structure diagram in the road typhoon disaster real-time monitoring system of the application; Figure 2 is an equipment terminal structure diagram in the road typhoon disaster real-time monitoring system of the application; Figure 3 is a flowchart of a road typhoon disaster real-time monitoring method of the application; Figure 4 is a typhoon disaster factor correlation analysis diagram in the application; Figure 5 is a CNN-LSTM model structure diagram in the application; Figure 6 is a model training accuracy line chart in the application; Figure 7 is a grid-based multi-target detection point layout diagram in the application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only a part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0018] As shown in Figure 1 , 2 A road typhoon disaster real-time monitoring system, comprising: a composite detection sensor and an equipment terminal; wherein, The composite detection sensor is composed of a rainfall sensor 1, a wind speed sensor 2, a signal transmission antenna 3 and a fixed base 4, and the tower-shaped top has a certain rainproof effect; The device terminal is composed of a central processing unit 5, a main display screen 6, a secondary display screen 7, a prompt light 8, an operator 9 and a signal receiving antenna 10; The rainfall sensor 1 is used to collect typhoon rainfall data; The wind speed sensor 2 is used to collect typhoon wind speed data; The central processing unit 5 is used to introduce a CNN-LSTM model to predict the road disaster level and send color-coded optical signals according to the predicted level; It is also used to judge the disaster level at this moment and the disaster level 5 seconds ago; It is also used to analyze the dynamic contribution of the eight influence factors, calculate each influence factor by introducing the coefficient of variation CV, analyze the factor with the highest contribution to the disaster level, and start the corresponding contingency plan according to the factor with the highest contribution; The main display screen 6 is used to display the corresponding contingency plan; The secondary display screen 7 is used to display the real-time typhoon parameters near the road at this time; The prompt light 8 is used to display the color-coded optical signals sent according to the predicted level; The operator 9 is used to switch the prediction level at different times and the corresponding contingency plan; The signal transmission antenna 3 of the composite detection sensor and the signal receiving antenna 10 of the device terminal establish communication for real-time transmission of sensor data.

[0019] In the embodiment, if the central processing unit 5 predicts a level of V, the disaster degree is very serious, and the prompt light 8 flashes red; If the central processing unit 5 predicts a level of IV, the disaster degree is serious, and the prompt light 8 is red and always on; If the central processing unit 5 predicts a level of III, the disaster degree is relatively serious, and the prompt light 8 is yellow and always on; If the central processing unit 5 predicts a level of II, the disaster degree is relatively light, and the prompt light 8 flashes green; If the central processing unit 5 predicts a level of I, the disaster degree is light, and the prompt light 8 is green and always on.

[0020] In the embodiment, the central processing unit 5 judges the disaster level at this moment and the disaster level 5 seconds ago, and if the disaster level increases, the prompt light 8 flashes red-yellow-green alternately for 5 seconds.

[0021] As Figure 3As shown, a road typhoon disaster real-time monitoring method comprises: Step S1, using the relative economic loss per kilometer, the average personnel casualty and the relative delay time length per kilometer as the risk grade division standard, the specific grade of the disaster road is evaluated.

[0022] In the embodiment, in combination with the Highway Engineering Construction Safety Risk Identification and Control Implementation Guide and the Railway Tunnel Engineering Risk Management Technical Specification, considering that the investment funds and construction volume of different roads are different, in order to facilitate unified quantitative index, the relative economic loss per kilometer, the average personnel casualty and the relative delay time length per kilometer are used as the risk grade division standard, the specific grade of the disaster road is evaluated, the grade division is as shown in Table 1, and the specific grade of the disaster road includes I, II, III, IV and V grades. If the grade of the disaster road is I, the disaster degree is light, the relative economic loss per kilometer is , the average personnel casualty is , and the relative delay time length per kilometer is . If the grade of the disaster road is II, the disaster degree is relatively light, the relative economic loss per kilometer is , the average personnel casualty is , and the relative delay time length per kilometer is . If the grade of the disaster road is III, the disaster degree is relatively serious, the relative economic loss per kilometer is , the average personnel casualty is , and the relative delay time length per kilometer is . If the grade of the disaster road is IV, the disaster degree is serious, the relative economic loss per kilometer is , the average personnel casualty is , and the relative delay time length per kilometer is . If the grade of the disaster road is V, the disaster degree is very serious, the relative economic loss per kilometer is , the average personnel casualty is , and the relative delay time length per kilometer is . Table 1 Road disaster grade Step S2, according to the typhoon influence factor correlation analysis, eight disaster-causing factors are introduced as the influence factors of the typhoon on the road damage, wherein the typhoon disaster-causing factor correlation analysis diagram is as shown in Figure 4 .

[0023] In the embodiment, the 8 disaster-causing factors include: maximum wind speed, cumulative rainfall, maximum rainfall, negative terrain, loose soil coefficient, average building height, road drainage capacity, and total value of on-site equipment.

[0024] Step S3, introduce the CNN-LSTM model to the central processor 5, use 335 sets of coastal and Hainan area in recent 20 years damaged by typhoon road data to train the CNN-LSTM model, the data is normalized by MATLAB built-in function normalize, and then transmitted to the CNN-LSTM model for training, in the CNN-LSTM model structure, the CNN has 24 1*3 size two-dimensional convolution kernel modules, the LSTM has 64 memory units, which will intelligently allocate weight coefficients according to the 8 influence factors and 1 disaster level in the database, the trained model and the normalized mapminmax_ps function will be packaged and stored in the central processor 5 by the "save" function, wherein the CNN-LSTM model structure diagram is as shown in Figure 5 , and the model training accuracy line chart is as shown in Figure 6 .

[0025] Step S4, combine the road design drawings, use GIS-BIM to model the road, and through finite element analysis, the road model is discretized into regular grid, each network center point can be used as a candidate sensor arrangement point, and a multi-objective optimization model is constructed by NSGA-III algorithm to screen all candidate points, and the grid multi-objective detection point layout diagram is as shown in Figure 7 .

[0026] In the embodiment, the NSGA-III algorithm is used to construct a multi-objective optimization model to screen all candidate points, and the specific operation is as follows: Let the candidate sensor point set be , the road discrete unit set be , and the wind speed sensor deployment decision variable be defined as ; Let represent the distance between the candidate point and the road unit , and define the exponential decay influence function as ; ; In the formula, is the monitoring influence weight of the wind speed sensor on the road unit on the candidate point , and the larger the value is, the stronger the wind speed monitoring ability of the unit is, is the influence scale of the wind speed sensor, and the distance increases by , the impact weight decays to about original 1 / e, is the rainfall sensor on candidate point , the monitoring impact weight of road unit , is the impact scale of rainfall sensor; The objective function includes: (1) maximize the weighted coverage rate: ; ; ; In the formula, is the vulnerability weight or importance of road unit , the larger the value, the more critical the unit, is the weight parameter of wind speed factor, used to adjust the contribution of wind speed to coverage rate, is the wind speed coverage accumulation of road unit k, which depends on which candidate point is deployed with wind speed sensor, is the weight parameter of rainfall factor, used to adjust the contribution of rainfall to coverage rate, is the rainfall coverage accumulation of road unit k, is whether to deploy wind speed sensor at candidate point , 1=deploy, 0=not deploy, is whether to deploy rainfall sensor at candidate point , 1=deploy, 0=not deploy; (2) minimize the total deployment cost: ; In the formula, is, is; (3) maximize the model observation sensitivity: In the formula, is the model prediction sensitivity contribution value of road unit when wind speed sensor is deployed at candidate point , estimated by historical data, model gradient or finite difference, is the model prediction sensitivity contribution value of road unit when rainfall sensor is deployed at candidate point .

[0027] NSGA-III algorithm is used for Pareto optimal search, and according to different construction areas, the layout scheme set meeting the cost, rainfall and wind speed coverage is obtained.

[0028] Step S5, after selecting the best wind speed and rainfall detection point, placing a typhoon multi-parameter composite detection sensor at the site, the composite detection sensor collects typhoon wind speed and rainfall every 5s.

[0029] Step S6, arrange the collected data in order, construct a dynamic database, predict the road disaster level at this moment according to the trained CNN-LSTM model and the normalized ps function, and send color-coded optical signals according to the predicted level.

[0030] Step S7, the secondary display screen 7 displays the typhoon parameters near the road in real time, and the central processor 5 judges the disaster level at this moment and the disaster level 5s ago, if the disaster level increases, the prompt light 8 red-yellow-green alternately flashes for 5s.

[0031] Step S8, the central processor 5 analyzes the dynamic contribution of the 8 influence factors, calculates each influence factor separately by introducing the coefficient of variation CV, and analyzes the factor with the highest contribution to disaster level.

[0032] Step S9, according to the factor with the largest contribution, start the corresponding plan, and output the treatment suggestion on the main display screen 6.

[0033] In the embodiment, it is assumed that the observed value of the first disaster factor in the last 12 sampling times is , then the mean and standard deviation of this factor in the sliding window are: ， ; Introducing a stable constant , the coefficient of variation of the factor is : ; In the formula, is the stable constant ; Normalize the CV value of each factor to get the relative fluctuation contribution rate : ; At the same time, the central processor 5 uses the trained CNN-LSTM model to predict the disaster level, and calculates the sensitivity of the input factor; The sensitivity can be realized by finite difference method, that is, adding a small disturbance to the first factor, calculating the change rate of the model output level: ; wherein, represents the model-predicted disaster grade score; The sensitivity is also normalized: ; Finally, the weight coefficient is determined as 0.85 according to the model prediction accuracy, and the weight of the on-site sensor data fluctuation is 0.15, and the dynamic contribution degree calculation formula of the factor ; The central processor 5 sorts the of all factors, if the maximum contribution degree factor meets , the special plan corresponding to the factor is directly triggered; if the cumulative contribution degree of the top 3 factors exceeds 0.7, the joint plan is triggered. In order to avoid misjudgment caused by short-term fluctuation, the system requires that this condition must be maintained for two consecutive sampling periods (10s) to truly start the corresponding treatment scheme.

[0034] In the embodiment, after each typhoon disaster occurs, the predicted grade is automatically compared with the actual disaster result, the actual grade label is generated, and the actual post-disaster data is fed back to the central processor 5 and added to the incremental learning training of the CNN-LSTM model, and the weight parameters of the model are updated, and dynamic optimization and self-learning are performed; wherein the weight parameters include the wind speed risk weight at the candidate point and the rainfall risk weight at the candidate point.

[0035] Finally, it should be noted that: the above is only the preferred embodiment of the present application, and is not used to limit the present application, although the present application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.​

Claims

1. A real-time monitoring system for road typhoon disasters, characterized in that, include: Composite detection sensors and equipment terminals; among which, The composite detection sensor consists of a rainfall sensor (1), a wind speed sensor (2), a signal transmission antenna (3), and a fixed base (4); The equipment terminal consists of a central processing unit (5), a main display screen (6), a secondary display screen (7), an indicator light (8), an operator (9), and a signal receiving antenna (10); Rainfall sensor (1) is used to collect typhoon rainfall data; Wind speed sensor (2) is used to collect typhoon wind speed data; The central processing unit (5) is used to introduce the CNN-LSTM model, predict the road disaster level, and send color-coded optical signals according to the predicted level. It is also used to compare the disaster level at this moment with the disaster level 5 seconds ago; It is also used to perform dynamic contribution analysis on eight influencing factors. By introducing the coefficient of variation (CV), each influencing factor is calculated separately, and the factor that contributes the most to the disaster level is analyzed. Based on the factor with the largest contribution, the emergency plan is activated in a targeted manner. The main display screen (6) is used to display the targeted activation plan; The secondary display screen (7) is used to display the typhoon parameters near the road in real time. Indicator light (8) is used to display the color-coded optical signal sent according to the predicted level; The operator (9) is used to switch the prediction level at different times and the corresponding targeted activation plan; The signal transmission antenna (3) of the composite detection sensor establishes communication with the signal receiving antenna (10) of the device terminal for real-time transmission of sensor data.

2. The real-time monitoring system for road typhoon disasters according to claim 1, characterized in that, If the central processing unit (5) predicts a level of V, then the degree of disaster is very serious, and the indicator light (8) flashes red. If the central processing unit (5) predicts a level of IV, then the degree of disaster is severe, and the indicator light (8) will be constantly red. If the central processing unit (5) predicts a level III disaster, the disaster severity is relatively severe, and the indicator light (8) will be constantly yellow. If the central processing unit (5) predicts a level II disaster, the degree of disaster is relatively mild, and the indicator light (8) flashes green. If the central processing unit (5) predicts a level I disaster, the degree of disaster is mild, and the indicator light (8) will remain green.

3. The real-time monitoring system for road typhoon disasters according to claim 1, characterized in that, The central processing unit (5) judges the disaster level at this moment and the disaster level 5 seconds ago. If the disaster level increases, the indicator light (8) flashes red-yellow-green alternately for 5 seconds.

4. A method for real-time monitoring of road typhoon disasters, characterized in that, include: Step S1: Using relative economic loss per kilometer, average casualties, and relative delay time per kilometer as risk level classification standards, the specific level of the affected roads is assessed. Step S2: Based on the correlation analysis of typhoon impact factors, eight disaster-causing factors are introduced as factors affecting typhoon damage to roads. Step S3: Introduce the CNN-LSTM model into the central processing unit (5) and train the CNN-LSTM model using 335 sets of road data from coastal and Hainan regions damaged by typhoons over the past 20 years. Step S4: Based on the road design drawings, model the road using GIS-BIM. Through finite element analysis, discretize the road model into... The regular grid can be used, and each network center point can be used as a candidate sensor placement point. A multi-objective optimization model is constructed using the NSGA-III algorithm to screen all candidate points. Step S5: After selecting the optimal detection points for wind speed and rainfall, place a typhoon multi-parameter composite detection sensor at the location. The composite detection sensor collects data on typhoon wind speed and rainfall every 5 seconds. Step S6: Arrange the collected data in sequence to build a dynamic database. Based on the trained CNN-LSTM model and the normalized ps function, predict the current road disaster level and send color-coded optical signals according to the predicted level. Step S7, the secondary display screen (7) displays the typhoon parameters near the road in real time, and the central processing unit (5) judges the disaster level at this moment and the disaster level 5 seconds ago. If the disaster level increases, the indicator light (8) flashes red-yellow-green alternately for 5 seconds. Step S8, the central processing unit (5) performs dynamic contribution analysis on the eight influencing factors, and calculates each influencing factor separately by introducing the coefficient of variation (CV), and analyzes the factor that contributes the most to the disaster level. Step S9: Based on the factor with the greatest contribution, activate the contingency plan accordingly and intelligently output treatment suggestions on the main display screen (6).

5. A method for real-time monitoring of road typhoon disasters according to claim 4, characterized in that, In step S1, the specific levels of the affected roads include: Level I, Level II, Level III, Level IV, and Level V; among them, If the affected road is classified as Level I, the degree of damage is considered minor, and the relative economic loss per kilometer is... The average number of casualties was The relative delay time per kilometer of construction period ; If the affected road is classified as Level II, the degree of damage is relatively minor, and the relative economic loss per kilometer is [not specified]. The average number of casualties was The relative delay time per kilometer of construction period ; If the affected road is classified as Level III, the damage is considered severe, with a relative economic loss per kilometer. The average number of casualties was The relative delay time per kilometer of construction period ; If the affected road is classified as Level IV, the degree of damage is considered severe, and the relative economic loss per kilometer is [not specified]. The average number of casualties was The relative delay time per kilometer of construction period ; If the affected road is classified as Class V, the degree of damage is considered extremely severe, with a relative economic loss per kilometer. The average number of casualties was The relative delay time per kilometer of construction period .

6. The method for real-time monitoring of road typhoon disasters according to claim 4, characterized in that, In step S2, the eight disaster-causing factors include: maximum wind speed, cumulative rainfall, maximum rainfall, negative terrain, soil looseness coefficient, average building height, road drainage capacity, and total value of on-site equipment.

7. A method for real-time monitoring of road typhoon disasters according to claim 4, characterized in that, In step S3, the data is normalized by the built-in MATLAB function normalize and then transmitted to the CNN-LSTM model for training. In the CNN-LSTM model structure, the CNN has 24 two-dimensional convolutional kernel modules of size 1×3 and the LSTM has a total of 64 memory units. The weight coefficients are intelligently allocated according to the 8 influencing factors and 1 disaster level in the database. The trained model and the normalized mapminmax_ps function will be packaged and stored in the central processing unit (5) through the "save" function.

8. A method for real-time monitoring of road typhoon disasters according to claim 4, characterized in that, In step S4, the NSGA-III algorithm is used to construct a multi-objective optimization model, and all candidate points are screened. The specific operation is as follows: Let the set of candidate sensor points be... The set of discrete road units is Define the decision variables for wind speed sensor deployment. ; make Indicate candidate points With road unit The distance between them is defined by the exponentially decaying influence function: ； ； In the formula, Candidate points The wind speed sensor on the road unit The monitoring impact weight is determined by the value; a larger value indicates a stronger wind speed monitoring capability for that unit. For the influence scale of the wind speed sensor, the distance increases by 1 / 2. The impact weight decays to approximately 1 / e of its original value. Candidate points Rainfall sensors on the road unit The monitoring impact weight, The scale of influence of rainfall sensors; The objective function includes: Maximize weighted coverage: ； ； ； In the formula, For road units The vulnerability weight or importance is indicated by a higher value, which means the unit is more critical. This is a weighting parameter for the wind speed factor, used to adjust the contribution of wind speed to coverage. The cumulative wind speed coverage for road unit k depends on which candidate points have wind speed sensors deployed. This is a weighting parameter for the rainfall factor, used to adjust the contribution of rainfall to coverage. Let k be the cumulative rainfall cover of road unit k. Whether at the candidate point Deploy wind speed sensors: 1 = deployed, 0 = not deployed. Whether at the candidate point Deploy rainfall sensors: 1 = deployed, 0 = not deployed; Minimize total deployment cost: ； In the formula, For the candidate point The cost of deploying wind speed sensors includes equipment price, installation costs, and maintenance costs. For candidate points The cost of deploying rainfall sensors; Maximize model observation sensitivity: In the formula, For candidate points When deploying wind speed sensors, for road units The model's predicted sensitivity contribution value is estimated using historical data, model gradients, or finite differences. For candidate points When deploying rainfall sensors, for road units The model predicts the sensitivity contribution value.

9. A method for real-time monitoring of road typhoon disasters according to claim 4, characterized in that, In step S9, after each typhoon disaster, the predicted level is automatically compared with the actual disaster result, the actual level label is generated, and the actual post-disaster data is fed back to the central processor (5) and added to the incremental learning training of the CNN-LSTM model to update the model's weight parameters, dynamically optimize and self-learn; among which, the weight parameters include the wind speed risk weight at the candidate point and the rainfall risk weight at the candidate point. Assume the first The observed values ​​of each disaster-causing factor at the most recent 12 sampling times are The mean of the factor within the sliding window. with standard deviation They are respectively: ， ； Introducing stability constant The coefficient of variation of this factor is obtained. : ； In the formula, Stability constant ; Normalize the CV values ​​of each factor to obtain the relative volatility contribution rate. : ； Meanwhile, the central processing unit (5) uses the trained CNN-LSTM model to predict the disaster level and calculate the sensitivity of the input factors; Sensitivity is achieved through the finite difference method, that is, for the ... Each factor increases a small perturbation. Calculate the rate of change of the output level of the model. : ； in, This represents the disaster severity score predicted by the model; Sensitivity is also normalized: ； Ultimately, based on the model's prediction accuracy, a weighting coefficient of 0.85 was determined, and the weighting for fluctuations in on-site sensor data was 0.

15. The formula for calculating dynamic contribution is: ； The central processing unit (5) processes all factors. Sort the data, if the maximum contribution factor satisfies... If the cumulative contribution of the first three factors exceeds 0.7, the specific contingency plan corresponding to that factor will be triggered directly; if the cumulative contribution of the first three factors exceeds 0.7, the joint contingency plan will be triggered.

Images