Generalized protection and monitoring method based on internet of things intelligent system

Abstract

The application discloses a general-purpose and standardized protection and monitoring method based on an Internet of Things intelligent system, and relates to the field of construction engineering construction safety protection; the general-purpose and standardized protection and monitoring method embeds a general-purpose and embeddable intelligent control box into various types of standardized protection facilities, collects data by using a matched sensor group, realizes real-time self-diagnosis and risk assessment on facilities such as edge guardrails and tool-type step ladders by means of an edge computing algorithm, then outputs a comprehensive risk weight value by using a dynamic risk assessment model, triggers multi-stage linkage alarm according to the comparison result of the comprehensive risk weight value and different grade threshold values, and finally uploads data to the cloud for collaborative management and block chain storage by means of a communication module; the general-purpose and standardized protection and monitoring method solves the problems of lagging monitoring and passive early warning of traditional protection facilities, and improves the intelligent level and management efficiency of construction safety protection.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of construction safety protection, and particularly relates to a general standardized protection and monitoring method based on an intelligent Internet of Things system. BACKGROUND

[0002] In the construction process, standardized protection facilities such as edge guardrails, tool type step ladders, protective fences, and protective sheds are important barriers for protecting the safety of construction personnel.

[0003] However, the traditional monitoring and management method of standardized protection facilities has many technical problems. The monitoring method is backward and relies on manual inspection, making it difficult to achieve real-time detection of the state of protection facilities, leading to a lack of timely discovery of hidden dangers and the inability to provide early warning of risks. There is a lack of systematic method for matching sensors and protection facilities, different types of facilities need to be separately configured with sensors, and the compatibility is poor. The configuration process is complicated and inefficient. The risk assessment model is fixed and cannot be dynamically adjusted according to changes in the construction environment. The accuracy of the assessment results is not high, making it difficult to effectively guide risk prevention and control. The alarm mechanism is single and can only provide simple audible and visual alarms. It is unable to take differentiated intervention measures according to the risk level, and the early warning effect is not good. The data management is scattered, and there is a lack of effective collaborative management and evidence storage mechanism, making it difficult to trace data and not conducive to the whole process management of construction safety.

[0004] These problems seriously restrict the improvement of the level of construction safety protection, and therefore, an intelligent general standardized protection and monitoring method is urgently needed to solve the above problems. SUMMARY

[0005] The present application aims to at least solve one of the technical problems in the prior art. To this end, the present application proposes a general standardized protection and monitoring method based on an intelligent Internet of Things system to solve the following technical problems:

[0006] The traditional monitoring and management method of standardized protection facilities has many technical problems. The monitoring method is backward and relies on manual inspection, making it difficult to achieve real-time detection of the state of protection facilities, leading to a lack of timely discovery of hidden dangers and the inability to provide early warning of risks. There is a lack of systematic method for matching sensors and protection facilities, different types of facilities need to be separately configured with sensors, and the compatibility is poor. The configuration process is complicated and inefficient. The risk assessment model is fixed and cannot be dynamically adjusted according to changes in the construction environment. The accuracy of the assessment results is not high, making it difficult to effectively guide risk prevention and control. The alarm mechanism is single and can only provide simple audible and visual alarms. It is unable to take differentiated intervention measures according to the risk level, and the early warning effect is not good. The data management is scattered, and there is a lack of effective collaborative management and evidence storage mechanism, making it difficult to trace data and not conducive to the whole process management of construction safety.

[0007] In order to solve the above problems, the application provides a general and standardized protection and monitoring method based on an intelligent system of the Internet of Things, comprising the following steps:

[0008] S1: a general and embeddable intelligent control box is quickly embedded into various types of standardized protection facilities, the intelligent control box comprises a main control module, a communication module, a positioning module and a power supply module; the main control module is built-in with an edge computing algorithm;

[0009] S2: based on the data collected by the sensor group matched with the type of the standardized protection facility, the edge computing algorithm is used to perform real-time self-diagnosis and risk assessment on the state of the protection facility, the real-time self-diagnosis and risk assessment comprises self-diagnosis and risk assessment on the edge guardrail, the tool type step ladder, the protection fence and the protection shed roof;

[0010] S3: a dynamic risk assessment model based on the edge computing algorithm is used to perform risk assessment, and a comprehensive risk weight value is output, according to the comparison result of the comprehensive risk weight value and the corresponding threshold value, a multi-level linkage alarm is triggered;

[0011] S4: the local data and analysis results are uploaded to the cloud management platform through the communication module for collaborative management and data storage.

[0012] Preferably, in the step S1, the intelligent control box comprises a main control module, a communication module, a positioning module and a power supply module, comprising:

[0013] The main control module adopts an industrial-grade ARM chip and is built-in with an edge computing algorithm;

[0014] The communication module adopts at least two communication protocols of 4G / 5G, LoRaWAN and Bluetooth Mesh;

[0015] The positioning module adopts UWB indoor positioning and GPS / Beidou dual-mode positioning;

[0016] The power supply module adopts a combination of a solar cell panel and a super capacitor for power supply, and the solar cell panel is designed in an integrated manner with the guardrail stand;

[0017] Meanwhile, the intelligent control box realizes plug-and-play adaptation with different sensor groups through a magnetic waterproof plug.

[0018] Preferably, in the step S2, based on the data collected by the sensor group matched with the type of the standardized protection facility, comprising the following steps:

[0019] According to the function, the standardized protection facility is divided into four core types, including the edge guardrail, the tool type step ladder, the protection fence and the protection shed, and the physical property identification of each type of facility is defined;

[0020] Establish a special risk feature library for each type of facility, and associate it with the core detection parameters;

[0021] Divide the sensors into standardized modules according to their functions, each module having a unique ID and a function description label;

[0022] Through the built-in RFID card reader and visual recognition unit in the intelligent control box, the facility type is identified through two ways;

[0023] Based on the identified facility type, the corresponding risk feature library and sensor configuration requirements are retrieved from the pre-stored database;

[0024] According to the configuration requirements, initialize and verify the corresponding sensor group.

[0025] Preferably, in step S2, the protective facility state is diagnosed in real time, including the following steps:

[0026] The real-time self-diagnosis includes edge guardrail diagnosis, tool-type step ladder diagnosis, protective fence diagnosis, and protective shed roof diagnosis;

[0027] The edge guardrail diagnosis judges the guardrail anchoring state based on vibration spectrum analysis of the vibration sensor, and monitors the guardrail tilt displacement combined with the tilt angle data of the tilt angle sensor;

[0028] The tool-type step ladder diagnosis monitors the load state based on the pedal pressure sensor, monitors the ladder posture using the attitude gyroscope, and detects the step ladder displacement using UWB ranging differential positioning, and detects the pedal icing condition through the anti-skid monitoring electrode;

[0029] The protective fence diagnosis collects wind speed data in real time through an anemometer, judges the strong wind risk combined with the preset wind speed threshold, identifies the illegal removal behavior through video AI camera image data, and synchronously monitors the state of the sound and light alarm;

[0030] The protective shed roof diagnosis monitors the falling impact load through the pressure sensor, and predicts the fatigue degree of the protective shed steel structure based on the strain data chain collected by the structure strain gauge.

[0031] Preferably, in step S3, the dynamic risk assessment model includes the following steps:

[0032] A multi-source heterogeneous data fusion engine is constructed to convert the tilt angle and vibration data of the edge guardrail, the load, attitude and displacement data of the tool-type step ladder, the wind speed and image data of the protective fence, and the pressure and strain data of the protective shed into standardized feature vectors;

[0033] An improved random forest algorithm is used to construct a dynamic risk assessment model, and a feature vector containing various normal and dangerous states and its corresponding true risk level label is used to train the dynamic risk assessment model;

[0034] During system operation, the standardized feature vector constructed in real time is input into the trained model, and the model outputs a comprehensive risk weight value based on the importance of each feature learned during training;

[0035] The comprehensive risk weight value is obtained by coupling calculation of the following parameters: the roadside guardrail sub-risk value is obtained by fusing the product of the vibration frequency spectrum abnormality and the inclination rate, the tool type step ladder sub-risk value is obtained by introducing the coupling coefficient of the load distribution entropy value and the ladder body inclination angle, the protective fence sub-risk value is the weighted sum of the wind speed threshold value duration normalization value and the image recognition violation confidence, and the protective shed sub-risk value is obtained by calculating the weighted sum of the impact load peak value and the fatigue damage degree;

[0036] According to the interval of the comprehensive risk weight value, different levels of risk response are triggered.

[0037] Preferably, the vibration frequency spectrum abnormality is the ratio of the energy of the 20-500Hz frequency band to the total energy.

[0038] Preferably, the load distribution entropy value is obtained by the information entropy formula, specifically:

[0039] H = -∑(h i *log2(h i ))

[0040] Where H represents the load distribution entropy value, h i represents the proportion of the i th pedal pressure to the total pressure.

[0041] Preferably, according to the interval of the comprehensive risk weight value, different levels of risk response are triggered, including the following steps:

[0042] Collect the comprehensive risk weight value historical data sequence of various types of protective equipment in normal state within a preset learning period, perform distribution fitting test on the historical data sequence, determine the optimal probability distribution model, and calculate the corresponding statistics of the model;

[0043] Based on the statistics, the percentile method is used to preliminarily set the initial threshold values of each level of risk, wherein the first level risk initial threshold value T1 is set as the P1 percentile of the historical data sequence; the second level risk initial threshold value T2 is set as the P2 percentile of the historical data sequence; and the third level risk initial threshold value T3 is set as the P3 percentile of the historical data sequence;

[0044] Real-time calculation of the comprehensive risk weight value is compared with the current effective risk threshold value of each level.

[0045] Preferably, the real-time calculation of the comprehensive risk weight value is compared with the current effective risk threshold value of each level, including the following steps:

[0046] If the comprehensive risk weight value is greater than or equal to T1, a local audible and visual alarm is triggered;

[0047] When the comprehensive risk weight value is greater than or equal to T2, an actuator intervention is triggered; the actuator intervention includes: starting the physical self-locking device of the tool-type stair to lock the ladder foot and activating the electronic fence of the edge protection barrier to generate a virtual laser warning screen;

[0048] When the comprehensive risk weight value is greater than or equal to T3, the alarm information and data are synchronously pushed to the cloud and a digital twin emergency deduction plan is started;

[0049] During the system operation, new comprehensive risk weight value data is continuously collected to update the historical data sequence, the distribution fitting test and the percentile calculation are periodically re-executed, the risk threshold values of each level are dynamically updated to generate new risk level threshold values, and the new risk level threshold values are effective in the next cycle.

[0050] Preferably, in the step S4, the data storage is a data storage method based on a block chain, and the use record, maintenance information and flow track data of the protective facility are uploaded to the block chain network for storage.

[0051] The present application has the following beneficial effects:

[0052] The present application can perform real-time self-diagnosis on various protective facilities such as edge protection barriers and tool-type stairs by using the edge computing algorithm built in the intelligent control box and combining the data collected by the matched sensor group, and can timely find the safety hazards existing in the facilities and perform risk assessment, thereby changing the hysteresis of traditional manual inspection.

[0053] The present application establishes a scientific sensor group and protective facility type matching method, realizes plug-and-play and precise adaptation of sensors through facility type division, risk feature library establishment and joint identification, improves sensor configuration efficiency and compatibility, and reduces construction cost.

[0054] The present application uses an improved random forest algorithm to construct a dynamic risk assessment model, which can dynamically adjust model parameters according to historical data and real-time data, and the output comprehensive risk weight value can better reflect the actual risk situation, thereby providing a reliable basis for risk prevention and control.

[0055] The application triggers different levels of risk response through the comparison result of the comprehensive risk weight value and the threshold value, from local sound and light alarm to actuator intervention, to cloud emergency deduction, realizes differentiated risk prevention and control, and improves the effectiveness and pertinence of early warning. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1 It is a method flowchart of the application. DETAILED DESCRIPTION

[0057] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the application.

[0058] Please refer to Figure 1 The application is a general and standardized protection and monitoring method based on an intelligent system of the Internet of Things, including the following steps:

[0059] S1: quickly embed a general and embeddable intelligent control box into various types of standardized protection facilities, the intelligent control box includes a main control module, a communication module, a positioning module and a power supply module; the main control module is built-in with an edge computing algorithm;

[0060] S2: collect data based on a sensor group matched with the type of the standardized protection facility, and use the edge computing algorithm to perform real-time self-diagnosis and risk assessment on the state of the protection facility, the real-time self-diagnosis and risk assessment includes self-diagnosis and risk assessment on a roadside guardrail, a tool type stairway, a protective fence and a protective canopy roof;

[0061] S3: use a dynamic risk assessment model based on the edge computing algorithm to perform risk assessment, output a comprehensive risk weight value, and trigger multi-level linkage alarm according to the comparison result of the comprehensive risk weight value and the corresponding threshold value;

[0062] S4: upload local data and analysis results to a cloud management platform through the communication module for collaborative management and data storage.

[0063] In one embodiment of the application, in the step S1, the intelligent control box includes a main control module, a communication module, a positioning module and a power supply module, including:

[0064] The main control module uses an industrial-grade ARM chip and is built-in with an edge computing algorithm;

[0065] The communication module uses at least two communication protocols in 4G / 5G, LoRaWAN and Bluetooth Mesh;

[0066] The positioning module adopts UWB indoor positioning and GPS / Beidou dual-mode positioning;

[0067] The energy supply module adopts a combination of a solar cell panel and a super capacitor for power supply, and the solar cell panel is designed in an integrated manner with the upright pole of the guardrail;

[0068] Meanwhile, the intelligent control box realizes plug-and-play adaptation with different sensor groups through a magnetic waterproof plug.

[0069] In one embodiment of the present application, in the step S2, the data collected based on the sensor group matched with the type of the standardized protective facility comprises the following steps:

[0070] The standardized protective facility is divided into four core types according to functions, including a roadside guardrail, a tool type step ladder, a protective fence and a protective shed, and the physical property identification of each type of facility is defined;

[0071] A dedicated risk feature library is established for each type of facility, and the core detection parameters are associated;

[0072] The sensors are divided into pluggable standardized modules according to functions, each module is internally provided with a unique ID and a function description label;

[0073] The facility type is identified through the built-in RFID card reader and visual recognition unit of the intelligent control box through two ways;

[0074] Based on the identified facility type, the corresponding risk feature library and sensor configuration requirement are retrieved from the pre-stored database;

[0075] According to the configuration requirement, the corresponding sensor group is initialized and verified.

[0076] Specifically, for each type of facility, construction scene risk investigation is carried out, and a dedicated risk feature library is established, including edge protection fence risk feature library, tool type stair risk feature library, protective fence risk feature library and protective shed risk feature library. According to the construction safety specification, safety threshold, early warning threshold and alarm threshold are set for each core detection parameter, for example, the safety threshold of the edge protection fence inclination angle is ≤3°, the early warning threshold is 3°-5°, and the alarm threshold is ≥5°, and is stored in the local database; The sensor is divided into 6 types of standardized modules according to the detection function, including mechanical perception module (including inclination sensor (range ± 30°, accuracy ± 0.1°), vibration sensor (range 0-50g, frequency response 0-1kHz), pressure sensor (range 0-500kg, accuracy ± 1% FS), used for detecting inclination, vibration and load type parameters), environmental perception module (including anemometer (range 0-30m / s, accuracy ± 0.5m / s), temperature and humidity sensor (temperature -20-70℃, humidity 0-100% RH), harmful gas sensor (detecting CO / NO2, accuracy ± 5% FS), used for environmental parameter detection), positioning perception module (including UWB positioning tag (positioning accuracy ± 10cm), infrared reflection sensor (detection distance 0-5m), used for position and boundary detection), image perception module (using 1080P resolution video AI camera (frame rate 25fps, supporting wide dynamic), used for appearance integrity identification), state perception module (including anti-skid monitoring electrode (impedance measurement range 10kΩ-1MΩ), structure strain gauge (sensitivity coefficient 2.0±1%), used for surface state and structure stress detection) and alarm execution module (including sound and light alarm (105dB buzzer + red and blue flash lamp), electronic fence module (laser warning, range 5m), used for risk early warning output); Each sensor module has a unique ID chip (UUID format) built-in, which sends function description tag information to the outside through Bluetooth broadcast or wired communication mode, including module type, detection parameter, range, communication protocol, etc.; The module shell is pasted with a two-dimensional code label corresponding to the ID, which is convenient for later maintenance identification; The intelligent control box is built-in RFID card reader (working frequency 13.56MHz) read the electronic tag information preset by the protection facility, and start the built-in camera to shoot the appearance image of the facility. The facility type is identified by a locally pre-trained CNN classification model (accuracy ≥ 98%). The results of the two identification methods are cross-verified. If they are consistent, the facility type is determined to be a curbstone guardrail, a tool type stair, a protective fence or a protective shed. Based on the confirmed facility type, the intelligent control box retrieves the corresponding risk feature library and sensor configuration list from the locally pre-stored database. The list includes the required sensor module type, quantity, installation position and communication parameters, etc. For example, the configuration list for the curbstone guardrail is "one inclination sensor (top of the vertical pole) + one vibration sensor (anchor point) + two infrared transceiver sensors (both ends of the guardrail)". The construction personnel connect the corresponding sensor modules to the expansion interface of the intelligent control box through the magnetic waterproof plug according to the configuration list. The interface adopts a mistaken insertion prevention design (different types of sensor plug shapes are different) to ensure correct physical connection. The control box automatically reads the sensor module ID and function description tag connected, and compares it with the configuration list. If the module type matches, the corresponding driver and communication protocol are automatically loaded, and parameters such as sampling frequency and data format are configured. If there is a missing module or a type error, the specific error information is displayed through the indicator light flashing alarm. After the configuration is completed, the control box starts the sensor calibration program: a standard signal is applied to the mechanical sensor (for example, an inclination sensor inputs a 3° standard inclination), and the environmental sensor is placed in a known environmental parameter (for example, an anemometer is placed under a 2m / s standard wind source). The error between the sensor output value and the standard value is compared. If the error is ≤5%, it is considered to pass the calibration. The sensor module that fails the calibration is marked as faulty and is prompted to be replaced.

[0077] In one embodiment of the present application, the real-time self-diagnosis of the protection facility state in step S2 includes the following steps:

[0078] The real-time self-diagnosis includes curbstone guardrail diagnosis, tool type stair diagnosis, protective fence diagnosis and protective shed roof diagnosis.

[0079] The curbstone guardrail diagnosis judges the guardrail anchoring state based on vibration spectrum analysis of the vibration sensor, and monitors the guardrail inclination displacement in combination with the inclination data of the inclination sensor.

[0080] The tool type stair diagnosis monitors the load state based on the pedal pressure sensor, monitors the ladder posture by the posture gyroscope, and detects the stair displacement by UWB ranging differential positioning, and detects the pedal icing condition by the anti-skid monitoring electrode.

[0081] The protective fence diagnosis collects wind speed data in real time through the anemometer, judges the strong wind risk in combination with the preset wind speed threshold, identifies the illegal removal behavior by collecting image data through the video AI camera, and synchronously links the state monitoring of the sound and light alarm.

[0082] The protective shed roof diagnosis monitors the falling object impact load through a pressure sensor and predicts the fatigue degree of the protective shed steel structure based on strain data chains collected by a structural strain gauge.

[0083] In one embodiment of the present application, the dynamic risk assessment model in step S3 includes the following steps:

[0084] A multi-source heterogeneous data fusion engine is constructed to uniformly convert the inclination angle and vibration data of the roadside guardrail, the load, posture and displacement data of the tool-type step ladder, the wind speed and image data of the protective fence, and the pressure and strain data of the protective shed into standardized feature vectors.

[0085] An improved random forest algorithm is used to construct a dynamic risk assessment model, and the dynamic risk assessment model is trained using feature vectors containing various normal and dangerous states and their corresponding true risk level labels.

[0086] During system operation, the standardized feature vectors constructed in real time are input into the trained model, and the model outputs a comprehensive risk weight value based on the importance of each feature learned during training.

[0087] The comprehensive risk weight value is obtained by coupling calculation of the following parameters: the roadside guardrail sub-risk value is obtained by coupling the product of the vibration frequency spectrum abnormality and the inclination change rate, the tool-type step ladder sub-risk value is obtained by introducing the coupling coefficient of the load distribution entropy value and the ladder body inclination angle, the protective fence sub-risk value is the weighted sum of the wind speed threshold value duration normalization value and the image recognition violation confidence, and the protective shed sub-risk value is obtained by calculating the weighted sum of the impact load peak value and the fatigue damage degree.

[0088] According to the interval of the comprehensive risk weight value, different levels of risk response are triggered.

[0089] Specifically, the lap joint multi-interface data access module supports real-time access of data of different types of sensors: mechanical data, environmental data, and positioning and image data; the original data is subjected to multi-stage cleaning processing, including outlier elimination (adopting a 3σ criterion to identify data (for example, inclination > 15°, wind speed > 30 m / s) that exceeds a reasonable range, and replacing it by interpolation with adjacent valid values), noise reduction processing, and time synchronization (taking the GPS timing module built in the intelligent control box as a reference (time accuracy ± 1 ms), adding a unified timestamp to different sensor data to ensure consistency in the time dimension); the preprocessed data is converted into an 8-dimensional weighted feature vector, wherein the definition of each dimension and the conversion method are as follows: the abnormality degree of the roadside guardrail vibration frequency spectrum is normalized to 0-1 by the ratio of the energy in the 20-500 Hz frequency band to the total energy; the roadside guardrail inclination rate is calculated by the inclination difference between adjacent time points, with an adjacent time interval of 1 s, and is normalized to 0-1; the tool-type step ladder load distribution entropy value is calculated by the information entropy formula and is normalized to 0-1; the tool-type step ladder body inclination angle is read by the attitude sensor and is normalized to 0-1; the duration of the wind speed of the protective enclosure exceeding the threshold value is calculated by the ratio to the safety threshold value and is normalized to 0-1; the confidence level of the image recognition violation of the protective enclosure is normalized to 0-1 by the confidence score output by the video AI model; the impact load peak value of the protective shed is obtained by the ratio to the design load and is normalized to 0-1; the fatigue damage degree of the protective shed is calculated based on the rain flow counting method of the strain gauge data chain and is normalized to 0-1; wherein the improved random forest dynamic risk assessment model is constructed and trained as follows: at least 6 months of construction scene data is collected, including: normal state samples: feature vectors of various facilities under safe working conditions (not less than 10,000); dangerous state samples: feature vectors and corresponding real risk level labels of 12 typical risk scenarios including anchor loosening, overload, and strong wind, divided into a training set and a validation set in a 7:3 ratio; the model structure is optimized based on the traditional random forest algorithm, including 1. Decision tree number optimization: the model performance of 50-200 decision trees is tested by grid search method, taking the accuracy of the validation set as the indicator to determine the optimal number of decision trees (recommended 100); 2. Feature weight dynamic adjustment: the attention mechanism is introduced to assign dynamic weights to feature vectors of different facility types (for example, the feature weight of the roadside guardrail in the high-altitude work area is increased by 20%); 3. Node splitting criterion improvement: the Gini index and information gain rate are weighted and fused to select the splitting feature, improving the identification accuracy of small sample risks; the training process: the training set data is used for model iteration training, 5-fold cross-validation is used to control overfitting, the iteration number is 50 times, and the learning rate is 0.01, the model accuracy rate (≥92%), recall rate (≥90%), F1 score (≥91%) are calculated through the validation set, if not up to standard, the training sample is increased or the model parameter is adjusted, the trained model is converted into ONNX format, and is deployed to the edge computing unit (industrial ARM chip) of the intelligent control box to support local real-time inference; when the system runs, the multi-source heterogeneous data fusion engine generates one real-time standardized feature vector every 1s, is input to the deployed dynamic risk assessment model, and the model calculates the sub-risk values of various facilities according to the following company numbers, including: the sub-risk value R1 of the edge guardrail is f1*r1*r2, wherein f1 is a risk correction coefficient of the edge guardrail, the value is 1.2, r1 and r2 are vibration spectrum abnormality and inclination rate respectively; the sub-risk value R2 of the tool type step ladder is f2*H*r3, wherein f2 is a risk correction coefficient of the tool type step ladder, the value is 1.1, H and r3 are load distribution entropy value and ladder body inclination angle respectively; the sub-risk value R3 of the protective enclosure is f3*(0.6*r4+0.4*r5), wherein f3 is a risk correction coefficient of the protective enclosure, the value is 1.0, r4 and r5 are wind speed threshold value duration normalization value and image recognition violation confidence respectively; the sub-risk value R4 of the protective shed is f4*(0.5*r6+0.5*r7), wherein f4 is a risk correction coefficient of the protective shed, the value is 1.3, r6 and r7 are impact load peak value and fatigue damage degree respectively; the comprehensive risk weight value is output by improving the decision tree voting mechanism of the random forest; wherein all the correction coefficients are constructed by the judgment matrix through the analytic hierarchy process, 10 safety engineering experts are invited to score the risk weight of the four types of facilities, and the initial assignment result is: edge guardrail 1.15, tool type step ladder 1.08, protective enclosure 1.00, protective shed 1.25, 12 months of operation data of three typical construction projects are selected, the risk early warning accuracy of the model under different correction coefficients is calculated, the calibration coefficient is equal to the initial amplitude of the expert multiplied by the ratio value of the actual early warning accuracy and the target early warning accuracy, for example, the target early warning accuracy of the protective shed is 90%, the actual test accuracy is 85%, and after calibration, the coefficient = 1.25*(90% / 85%)≈1.3, the other coefficients are normalized and adjusted based on the protective enclosure coefficient (set as 1.0), to ensure that the relative weight relationship is reasonable.

[0090] In one embodiment of the present application, the vibration spectrum abnormality degree is the ratio of the energy of the 20-500Hz frequency band to the total energy.

[0091] In one embodiment of the present application, the load distribution entropy value is obtained by the information entropy formula, specifically:

[0092] H = -∑(h i *log2(h i ))

[0093] Wherein, H represents the load distribution entropy value, h i represents the proportion value of the i th pedal pressure to the total pressure.

[0094] In one embodiment of the present application, the triggering of different levels of risk response according to the interval in which the comprehensive risk weight value is located includes the following steps:

[0095] In one embodiment of the present application, the triggering of different levels of risk response according to the interval in which the comprehensive risk weight value is located includes the following steps:

[0096] Based on the statistical quantity, the initial threshold values of each level of risk are preliminarily set by using the percentile method, wherein the initial threshold value T1 of the first level of risk is set as the P1 percentile of the historical data sequence; the initial threshold value T2 of the second level of risk is set as the P2 percentile of the historical data sequence; and the initial threshold value T3 of the third level of risk is set as the P3 percentile of the historical data sequence.

[0097] The real-time comprehensive risk weight value is calculated and compared with the currently effective threshold values of each level of risk.

[0098] Specifically, the comprehensive risk weight value data in the first month under normal conditions (≥5000) are collected, the statistical quantity is calculated by normal distribution fitting, and the initial threshold values are set by using the percentile method, for example, T1=P90 (90th percentile), T2=P95 (95th percentile), and T3=P99 (99th percentile). Normal state data are collected every 7 days, the distribution consistency is verified by K-S test, and if there is a significant change (p<0.05), the percentile is recalculated to update the threshold values. After each response trigger, the system automatically records the risk value, trigger time, response measure and processing result to form a closed-loop management log as feedback data for model optimization.

[0099] In one embodiment of the present application, the real-time calculation of the comprehensive risk weight value and the comparison with the currently effective threshold values of each level of risk include the following steps:

[0100] If the comprehensive risk weight value is greater than or equal to T1, a local audible and visual alarm is triggered;

[0101] When the comprehensive risk weight value is greater than or equal to T2, an executive intervention is triggered; the executive intervention includes starting the physical self-locking device of the tool-type step ladder to lock the ladder feet and activating the electronic fence of the edge guardrail to generate a virtual laser warning screen.

[0102] When the comprehensive risk weight value is greater than or equal to T3, the alarm information and data are synchronously pushed to the cloud and a digital twin emergency deduction plan is started.

[0103] During the system operation, new comprehensive risk weight value data is continuously collected to update the historical data sequence, the distribution fitting test and the percentile calculation are periodically re-executed, the risk threshold values are dynamically updated to generate new different risk level threshold values and make them effective in the next cycle.

[0104] In one embodiment of the present application, in the step S4, the data archiving is a blockchain-based data archiving method, and the use record, maintenance information and flow track data of the protection facility are uploaded to the blockchain network for archiving.

[0105] The above embodiments are only used to illustrate the technical method of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present application can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present application.

Claims

1. A general-purpose and standardized protection and monitoring method based on an Internet of Things intelligent system, characterized in that, The method comprises the following steps: S1: embedding a general-purpose embeddable intelligent control box into various types of standardized protective facilities, the intelligent control box comprising a main control module, a communication module, a positioning module and a power supply module; the main control module is built-in with edge computing algorithm; S2: collecting data based on a sensor group matched with the type of the standardized protective facility, and using the edge computing algorithm to perform real-time self-diagnosis and risk assessment on the state of the protective facility, the real-time self-diagnosis and risk assessment comprising self-diagnosis and risk assessment on the edge protection barrier, the tool type stairway, the protective fence and the protective shed roof; S3: performing risk assessment using a dynamic risk assessment model based on the edge computing algorithm, outputting a comprehensive risk weight value, and triggering multi-level linkage alarm according to the comparison result of the comprehensive risk weight value and the corresponding threshold value; S4: uploading local data and analysis results to a cloud management platform through the communication module for collaborative management and data storage; In the step S1, the intelligent control box comprises a main control module, a communication module, a positioning module and a power supply module, which comprise: The main control module adopts an industrial-grade ARM chip and is built-in with edge computing algorithm; The communication module adopts at least two communication protocols of 4G / 5G, LoRaWAN and Bluetooth Mesh; The positioning module adopts UWB indoor positioning and GPS / Beidou dual-mode positioning; The power supply module adopts a combination of solar panels and super capacitors for power supply, and the solar panels are designed in an integrated manner with the upright rods of the guardrail; At the same time, the intelligent control box realizes plug-and-play adaptation with different sensor groups through magnetic waterproof plugs; In the step S2, the real-time self-diagnosis on the state of the protective facility comprises the following steps: The real-time self-diagnosis comprises edge protection barrier diagnosis, tool type stairway diagnosis, protective fence diagnosis and protective shed roof diagnosis; The edge protection barrier diagnosis judges the guardrail anchoring state based on vibration spectrum analysis of the vibration sensor, and monitors the guardrail inclination displacement in combination with the inclination data of the inclination sensor; The tool type stairway diagnosis monitors the load state based on the pedal pressure sensor, monitors the ladder posture by the attitude gyroscope, and detects the stair displacement by UWB ranging differential positioning, and detects the pedal icing condition by the anti-skid monitoring electrode; The protective fence diagnosis judges the strong wind risk by real-time collection of wind speed data by the anemometer in combination with the preset wind speed threshold value, and identifies the illegal removal behavior by collecting image data by the video AI camera, and synchronously links the state monitoring of the sound and light alarm; The protective shed roof diagnosis monitors the impact load of falling objects by the pressure sensor, and predicts the fatigue degree of the protective shed steel structure based on the strain data chain collected by the structure strain gauge; In the step S3, the dynamic risk assessment model comprises the following steps: A multi-source heterogeneous data fusion engine is constructed to uniformly convert the inclination and vibration data of the edge protection barrier, the load, attitude and displacement data of the tool type stairway, the wind speed and image data of the protective fence, and the pressure and strain data of the protective shed into standardized feature vectors; An improved random forest algorithm is used to construct a dynamic risk assessment model, and a feature vector containing various normal and dangerous states and its corresponding true risk level label is used to train the dynamic risk assessment model; During system operation, the standardized feature vector constructed in real time is input into the trained model, and the model outputs a comprehensive risk weight value based on the importance of each feature learned during training; The comprehensive risk weight value is obtained by coupling calculation of the following parameters: the sub-risk value of the edge guardrail is obtained by fusing the product of the vibration frequency spectrum abnormality and the inclination rate, the sub-risk value of the tool type step ladder is obtained by introducing the coupling coefficient of the load distribution entropy value and the ladder body inclination angle, the sub-risk value of the protective fence is the weighted sum of the wind speed threshold duration normalization value and the image recognition violation confidence, and the sub-risk value of the protective shed is obtained by calculating the weighted sum of the impact load peak value and the fatigue damage degree; According to the interval of the comprehensive risk weight value, different levels of risk response are triggered. 2.The general-purpose, standardized protection and monitoring method based on the Internet of Things intelligent system according to claim 1, characterized in that, In step S2, data is collected based on the sensor group matching the type of the standardized protective facility, including the following steps: According to the function, the standardized protective facility is divided into four core types, including edge guardrail, tool type step ladder, protective fence and protective shed, and the physical property identification of each type of facility is defined; An exclusive risk feature library is established for each type of facility, and the core detection parameters are associated; The sensors are divided into standardized modules according to the function, each module is built-in with a unique ID and a function description label; The RFID card reader and visual recognition unit built-in the intelligent control box are used to identify the facility type through two ways; Based on the identified facility type, the corresponding risk feature library and sensor configuration requirements are retrieved from the pre-stored database; According to the configuration requirements, the corresponding sensor group is initialized and verified. 3.The general-purpose and standardized protection and monitoring method based on the Internet of Things intelligent system according to claim 1, characterized in that, The vibration frequency spectrum abnormality is the ratio of the energy in the frequency band of 20-500 Hz to the total energy. 4.The general-purpose and standardized protection and monitoring method based on the Internet of Things intelligent system according to claim 1, wherein, The load distribution entropy value is obtained by the information entropy formula, specifically: wherein, represents a load distribution entropy value, represents a proportion value of the first pedal pressure to the total pressure. 5.The general-purpose, standardized protection and monitoring method based on the Internet of Things intelligent system according to claim 1, characterized in that, According to the interval of the comprehensive risk weight value, different levels of risk response are triggered, including the following steps: Collect the historical data sequence of the comprehensive risk weight value of each type of protective equipment in the normal state within a preset learning period, perform distribution fitting test on the historical data sequence, determine the optimal probability distribution model, and calculate the statistical quantity corresponding to the model; Based on the statistical quantity, the percentile method is used to initially set the initial threshold values of each risk level, wherein the initial threshold value T1 of the first level risk is set as the P1 percentile of the historical data sequence; the initial threshold value T2 of the second level risk is set as the P2 percentile of the historical data sequence; and the initial threshold value T3 of the third level risk is set as the P3 percentile of the historical data sequence; The comprehensive risk weight value is calculated in real time, and compared with the currently effective risk threshold values of each level. 6.The general-purpose, standardized protection and monitoring method based on the Internet of Things intelligent system according to claim 5, characterized in that, The comprehensive risk weight value is calculated in real time, and compared with the currently effective risk threshold values of each level, including the following steps: If the comprehensive risk weight value is greater than or equal to T1, the local sound and light alarm is triggered; When the comprehensive risk weight value is greater than or equal to T2, a trigger is sent to an actuator intervention; the actuator intervention includes: starting a physical self-locking device of a tool type step ladder to lock the ladder foot and activating an electronic fence of a curb guard to generate a virtual laser warning screen; When the comprehensive risk weight value is greater than or equal to T3, alarm information and data are synchronously pushed to the cloud and a digital twin emergency deduction plan is started; During system operation, new comprehensive risk weight value data is continuously collected to update the historical data sequence, the distribution fitting test and the percentile calculation are periodically re-executed, the risk threshold values are dynamically updated, new different risk level threshold values are generated, and the new different risk level threshold values take effect in the next period. 7.The general-purpose, standardized protection and monitoring method based on the Internet of Things intelligent system according to claim 1, wherein, In the step S4, the data storage is a data storage method based on a block chain, and the use record, maintenance information and flow track data of the protection facility are uploaded to a block chain network for storage.

Images