Hydraulic engineering construction potential safety hazard monitoring method based on Internet of Things

Abstract

The invention relates to the technical field of safety management, and discloses a hydraulic engineering construction potential safety hazard monitoring method based on Internet of Things, which comprises the following steps: S1, real-time data acquisition: arranging a plurality of sensors at a construction site to acquire data and uploading the data in a wireless communication mode, the method comprises the steps of S1, data acquisition and analysis, S2, data pre-processing and analysis: pre-processing acquired data at a cloud or an edge computing platform, S3, adaptive adjustment: automatically adjusting sensor layout and monitoring frequency according to construction stages and environment changes, S4, multi-source data fusion, S5, alarm and emergency: triggering alarm when monitoring data is abnormal, providing emergency measures by means of an intelligent decision support system, and carrying out emergency management. S6, visualization and decision support; and S7, intelligent feedback and optimization. By adopting a dynamic threshold adjustment mechanism and a multi-modal data fusion technology, a more accurate hidden danger identification effect is achieved, the phenomena of missing report and false report are remarkably reduced, and the effectiveness of early warning is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of safety management, and particularly to a method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things. Background Technique

[0002] The construction of water conservancy projects is highly complex and dangerous. If potential safety hazards during the construction process are not discovered in time, serious accidents may occur. Therefore, how to monitor the construction environment, structural status, and construction progress in real time and accurately, and promptly discover potential safety hazards, has become an important issue in construction management.

[0003] Currently, manual hazard detection during the construction of water conservancy projects is not only inefficient but also unable to achieve real-time monitoring and immediate response, which poses a major challenge to construction safety. For the solutions of the prior art corresponding to this technical solution, most existing monitoring systems rely on regular manual inspections, supplemented by video monitoring of fixed-installed cameras. The defects of the prior art are that manual inspections take a long time and it is difficult to cover all risk points, while video monitoring is limited by the perspective and recognition accuracy and often cannot detect hidden safety hazards in time. Summary of the Invention

[0004] In view of the deficiencies of the prior art, the present invention provides a method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things, which solves the problems that real-time and comprehensive safety monitoring of the construction site cannot be achieved, resulting in a lag in hazard discovery, lack of intelligent analysis tools, difficulty in quickly and accurately assessing the risk level and taking countermeasures, low integration of existing monitoring systems, difficult deployment, and high maintenance costs.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things, including the following steps: S1. Real-time data collection: Based on the Internet of Things technology, various sensors are connected to the cloud or edge computing devices through a standardized edge computing gateway, and multiple sensors are deployed at the construction site to collect data and upload it through wireless communication. S2. Data preprocessing and analysis: The collected data is preprocessed on the cloud or edge computing platform, and big data analysis and AI algorithms are used to extract patterns and make intelligent predictions, and a dynamic threshold adjustment mechanism is adopted to identify potential hazards. S3. Adaptive adjustment: Automatically adjust the sensor layout and monitoring frequency according to the construction stage and environmental changes. S4. Multi-source data fusion: Introduce the data of the construction progress management system and fuse it with the sensor data to improve the ability to identify potential hazards. S5. Alarm and emergency: Trigger an alarm when the monitoring data is abnormal, notify relevant personnel through various methods, and provide emergency measures with the help of an intelligent decision support system. S6, Visualization and Decision Support: Use the GIS platform to visually display the safety status, combine deep learning algorithms and expert knowledge graphs to assist in decision-making, and optimize the monitoring model based on historical data; S7, Intelligent Feedback and Optimization: Archive and analyze data after construction, compare differences, and optimize future construction safety management strategies.

[0006] Preferably, in S1, the various types of sensors include temperature and humidity sensors, meteorological sensors, displacement sensors, vibration sensors, and water level sensors. The wireless communication includes NB-IoT, LoRa, and 5G. The cloud devices include Alibaba Cloud ECS instances or Tencent Cloud CVMs. The edge computing devices include Advantech WISE-4010 series or Huawei 5G industrial modules. The data upload method uses encrypted transmission. For sensors with NB-IoT communication, data is transmitted through the operator network using the MQTT protocol, and data encryption uses the AES-128 algorithm; for sensors with LoRa communication, data is transmitted through the LoRaWAN protocol and encrypted using the elliptic curve encryption algorithm. For sensors with 5G communication, data is uploaded through the 5G network using the HTTP / HTTPS protocol.

[0007] Preferably, in S2, the 3σ criterion is used for data cleaning to remove noise and outliers, and principal component analysis is used to reduce the data dimension. The principal components with a contribution rate of more than 95% are extracted. The dynamic threshold adjustment mechanism uses the sliding window algorithm, and the window size is set between 20 - 50 data points according to the monitoring indicators.

[0008] Preferably, in S3, an association model is established by analyzing historical construction and environmental data. During heavy rain, the monitoring frequency of the water level sensor is increased from once every 5 minutes to once every 1 minute based on historical data. When the wind speed is greater than 15 m / s during strong wind weather, the monitoring frequency of the wind speed sensor is increased from once every 2 minutes to once every 30 seconds. During the foundation construction stage, a displacement sensor is added every 10 meters, and the monitoring frequency is increased from once every 10 minutes to once every 3 minutes. During the main structure construction stage, a vibration sensor is added every 15 meters, and the monitoring frequency is increased from once every 8 minutes to once every 5 minutes.

[0009] Preferably, in S4, the data of the construction progress management system is structured data, and the geological exploration data can be semi-structured or unstructured data. When using the feature-level fusion algorithm, wavelet transform is used to extract features from the sensor perception data, and the Apriori algorithm is used to mine frequent itemsets to extract key features from the construction progress management system data. Then, the extracted features are spliced and fused according to the feature dimension ratio of 1:1. When using the decision-level fusion algorithm, the decision results are first independently processed for each data source, and then the decision results are fused by the weighted voting method. Among them, the weight of the decision result of the sensor data is 0.6, the weight of the decision result of the construction progress management system data is 0.2, and the weight of the decision result of the geological exploration data is 0.2.

[0010] Preferably, in S5, the intelligent decision support system is used to formulate emergency measures based on different types of hidden dangers in combination with historical cases. Among them, for the soil humidity anomaly higher or lower than the normal range by 20%, drainage or irrigation suggestions are provided. If the humidity is higher than the normal range, it is recommended to turn on the drainage pump, and the flow rate of the drainage pump is adjusted between 5-10 cubic meters per hour according to the degree of humidity exceeding the standard. When the structural displacement is too large and exceeds 10% of the allowable displacement, measures such as suspending construction and strengthening the structure are given. The strengthening methods include adding supports and pasting carbon fiber cloth. The alarm methods include text messages, APP push, and voice broadcast.

[0011] Preferably, in S6, the GIS platform has a spatial analysis function for spatial statistics and trend analysis of the distribution of safety hazards. The trend analysis includes calculating the density of hazard points and analyzing the distribution trend of hazards in different regions. The deep learning algorithm uses long short-term memory networks or convolutional neural networks and their variants to learn the spatio-temporal characteristics of the data, and constructs a risk assessment model in combination with an expert knowledge graph.

[0012] Preferably, in S7, the archived analysis data is used to deeply mine historical and actual construction data using deep learning algorithms. The mining algorithms use deep autoencoders and generative adversarial networks. By mining, potential safety hazard patterns and rules are discovered, and the safety management strategy and monitoring model are optimized at a frequency of once per quarter. Each optimization adjusts at least 1-3 key parameters. The key parameters include the sensor monitoring frequency, the hazard identification threshold, and the weight of the risk assessment model.

[0013] The present invention provides a method for monitoring construction safety hazards of water conservancy projects based on the Internet of Things. It has the following beneficial effects: 1. By adopting a dynamic threshold adjustment mechanism and multi-modal data fusion technology, the present invention achieves a more accurate hidden danger identification effect, significantly reduces the phenomena of missed reports and false alarms, and improves the effectiveness of early warnings.

[0014] 2. The present invention utilizes a cloud-edge collaboration architecture and a standardized edge computing gateway to achieve a low-cost and easily deployable system design, greatly reducing the initial investment and subsequent operation burden of enterprises.

[0015] 3. By introducing deep learning algorithms and expert knowledge graphs, the automated risk assessment process of the present invention becomes more intelligent, assisting managers in making more scientific, reasonable, and rapid response decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a method flow chart of a method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things according to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0018] Please refer to the attached Figure 1 , an embodiment of the present invention provides a method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things, including the following steps: S1. Real-time data collection: Based on Internet of Things technology, connect various sensors with the cloud or edge computing devices through a standardized edge computing gateway, deploy a variety of sensors at the construction site to collect data and upload it through wireless communication. S2. Data preprocessing and analysis: Preprocess the collected data on the cloud or edge computing platform, extract patterns and make intelligent predictions using big data analysis and AI algorithms, and identify potential hazards using a dynamic threshold adjustment mechanism. S3. Adaptive adjustment: Automatically adjust the sensor layout and monitoring frequency according to the construction stage and environmental changes. S4. Multi-source data fusion: Introduce the data of the construction progress management system and fuse it with the sensor data to improve the ability to identify potential hazards. S5. Alarm and emergency: Trigger an alarm when the monitored data is abnormal, notify relevant personnel through various means, and provide emergency measures with the help of an intelligent decision support system. S6. Visualization and decision support: Use a GIS platform to visually display the safety status, assist in decision-making in combination with deep learning algorithms and expert knowledge graphs, and optimize the monitoring model based on historical data. S7. Intelligent feedback and optimization: Archive and analyze the data after the construction is completed, compare the differences, and optimize the future construction safety management strategy.

[0019] In S1, various sensors include temperature and humidity sensors, meteorological sensors, displacement sensors, vibration sensors, and water level sensors. Wireless communications include NB-IoT, LoRa, and 5G. Cloud devices include Alibaba Cloud ECS instances or Tencent Cloud CVMs. Edge computing devices include Advantech WISE-4010 series or Huawei 5G industrial modules. The data upload method uses encrypted transmission. For sensors communicating via NB-IoT, data is transmitted through the operator network using the MQTT protocol, and data encryption uses the AES-128 algorithm. For sensors communicating via LoRa, data is transmitted through the LoRaWAN protocol and encrypted using the elliptic curve encryption algorithm. For sensors communicating via 5G, data is uploaded through the 5G network using the HTTP / HTTPS protocol.

[0020] Specifically, temperature and humidity sensors, meteorological sensors, displacement sensors, vibration sensors, and water level sensors are deployed. The temperature and humidity sensors monitor the environmental humidity at the construction site. The meteorological sensors collect meteorological data such as wind speed and direction. The displacement sensors detect the displacement of the construction structure. The vibration sensors monitor the vibrations caused by equipment during construction. The water level sensors monitor the changes in the water level of the water body. These sensors upload data to the cloud or edge computing devices through wireless communication technologies such as NB-IoT, LoRa, and 5G. The sensor data is transmitted through encrypted transmission. NB-IoT sensors use the MQTT protocol, and data encryption uses the AES-128 algorithm. LoRa sensors use the LoRaWAN protocol and the elliptic curve encryption algorithm (ECC). 5G sensors use the HTTP / HTTPS protocol for data transmission to ensure data security. The data is uploaded to an Alibaba Cloud ECS instance or a Tencent Cloud CVM instance for processing. The cloud platform provides powerful computing and storage capabilities. Edge computing devices, such as the Advantech WISE-4010 series or Huawei 5G industrial modules, can perform on-site data processing, reduce transmission latency, and improve response speed. Through these sensors and data processing technologies, real-time monitoring and intelligent early warning of potential safety hazards at the construction site are achieved, significantly improving the safety of the construction process.

[0021] In S2, 3σ criterion is used for data cleaning to remove noise and outliers. Principal component analysis is used to reduce the data dimension, and the principal components with a contribution rate of over 95% are extracted. The dynamic threshold adjustment mechanism uses the sliding window algorithm, and the window size is set between 20 - 50 data points according to the monitoring indicators.

[0022] Specifically, the big data analysis technology is used to extract meaningful patterns from sensor data, and artificial intelligence (AI) algorithms are combined to make intelligent predictions on the data. At the same time, a dynamic threshold adjustment mechanism is adopted to dynamically adjust the threshold for hazard identification according to the statistical characteristics of real-time data, construction stages, and environmental factors, etc., so as to identify possible danger signals during the construction process, such as potential safety hazards like abnormal soil moisture and excessive structural displacement. Data preprocessing includes data cleaning and dimensionality reduction. In data cleaning, the 3σ criterion is used to remove noise and outliers, and abnormal data is automatically eliminated by detecting whether the data points exceed 3 times the standard deviation of the mean. Then, principal component analysis (PCA) is used to reduce the dimensionality of the data, and the principal components with a contribution rate exceeding 95% are extracted to reduce redundant information and improve the calculation efficiency; The dynamic threshold adjustment mechanism adopts a sliding window algorithm. This algorithm sets a window with a size between 20 - 50 data points and adjusts the threshold of the monitoring index in real time. When the data change within the window reaches the set standard, the threshold will be adjusted according to the real-time data, so as to more accurately identify potential safety hazards. This mechanism enables the system to flexibly respond to the real-time changes on the construction site and environment, and optimize the accuracy of hazard identification and early warning.

[0023] In S3, an association model is established by analyzing historical construction and environmental data. During rainstorms, according to historical data, the monitoring frequency of the water level sensor is increased from once every 5 minutes to once every 1 minute. When the wind speed is greater than 15 m / s in strong wind weather, the monitoring frequency of the wind speed sensor is increased from once every 2 minutes to once every 30 seconds. During the foundation construction stage, a displacement sensor is added every 10 meters, and the monitoring frequency is increased from once every 10 minutes to once every 3 minutes. During the main structure construction stage, a vibration sensor is added every 15 meters, and the monitoring frequency is increased from once every 8 minutes to once every 5 minutes.

[0024] Specifically, an association model is established by analyzing historical construction data and environmental data. When a rainstorm occurs, according to historical data and the on-site environment, the monitoring frequency of the water level sensor is increased from once every 5 minutes to once every 1 minute to ensure that the water level changes can be captured in time. For strong wind weather, when the wind speed is greater than 15 m / s, the monitoring frequency of the wind speed sensor is increased from once every 2 minutes to once every 30 seconds to ensure that the wind speed changes can be tracked in real time and reduce the risk of wind disasters; During the foundation construction stage, the system will add a displacement sensor every 10 meters according to the construction progress, and increase the monitoring frequency from once every 10 minutes to once every 3 minutes to improve the monitoring accuracy of structural displacement. After entering the main structure construction stage, a vibration sensor will be added every 15 meters, and the monitoring frequency will be increased from once every 8 minutes to once every 5 minutes to ensure that the impact of vibrations generated during construction on the structure can be detected in a timely manner. These automated adjustments can flexibly optimize the monitoring frequency according to the actual construction environment and stage, improving the timeliness and accuracy of monitoring.

[0025] In S4, the data of the construction progress management system is structured data, and the geological exploration data can be semi-structured or unstructured data. When using the feature-level fusion algorithm, wavelet transform is used to extract features from the sensor perception data, and the Apriori algorithm is used to mine frequent itemsets from the construction progress management system data to extract key features. Then, the extracted features are spliced and fused according to the feature dimension ratio of 1:1. When using the decision-level fusion algorithm, the decision results are first independently processed for each data source, and then the decision results are fused through the weighted voting method, where the weight of the decision result of the sensor data is 0.6, the weight of the decision result of the construction progress management system data is 0.2, and the weight of the decision result of the geological exploration data is 0.2.

[0026] Specifically, the data of the construction progress management system is stored in a structured manner, while the geological exploration data may be semi-structured or unstructured data. In feature-level fusion, wavelet transform is used to extract features from the sensor perception data, and the Apriori algorithm is used to mine frequent itemsets from the construction progress management system data to extract key features. Subsequently, splicing and fusion are carried out according to the feature dimension ratio of 1:1 to enhance the correlation between data and improve the comprehensive analysis ability; By introducing multi-modal data from other relevant systems, such as the structured data of the construction progress management system, the semi-structured or unstructured data of geological exploration data, etc., and fusing it with the perception data collected by sensors. Through advanced multi-modal data fusion technologies, such as feature-level fusion, decision-level fusion and other algorithms, the advantages of different sources and different types of data are integrated to reflect the construction status of the water conservancy project, further improving the ability to identify potential safety hazards. Combining the construction progress data can determine whether the structural displacement is within a reasonable range under specific construction links, and combining the geological exploration data can better analyze the impact of soil moisture changes on the project stability. In decision-level fusion, the decision results are obtained after independent processing of each data source and fused through the weighted voting method. The weights are set such that the decision result of the sensor data accounts for 0.6, the construction progress management system data accounts for 0.2, and the geological exploration data accounts for 0.2. This fusion strategy can effectively improve the accuracy of the overall decision-making and ensure that the contributions of different data sources to the identification of construction safety hazards are reasonably reflected.

[0027] In S5, the intelligent decision support system is used to formulate emergency measures based on different types of potential hazards, combined with historical cases. For soil humidity anomalies that are more than 20% higher or lower than the normal range, drainage or irrigation suggestions are provided. If the humidity is higher than the normal range, it is recommended to turn on the drainage pump, and the drainage pump flow rate is adjusted between 5 - 10 cubic meters per hour according to the degree of humidity exceeding the standard. When the structural displacement is too large, exceeding 10% of the allowable displacement, measures such as suspending construction and strengthening the structure are given. The strengthening methods include adding supports and pasting carbon fiber sheets. The alarm methods include text messages, APP push, and voice broadcasts.

[0028] Specifically, by real-time monitoring of key indicators such as soil humidity and structural displacement, corresponding emergency measures are automatically triggered to ensure construction safety. For abnormal soil humidity, when the humidity is more than 20% higher or lower than the normal range, the system recommends taking corresponding measures. If the humidity is higher than the normal range, the system recommends turning on the drainage pump, and the drainage flow rate is adjusted between 5 - 10 cubic meters per hour according to the degree of humidity exceeding the standard. If the humidity is too low, the system recommends irrigation to maintain normal humidity. For abnormal structural displacement, when the displacement exceeds 10% of the allowable range, the system recommends suspending construction and proposes a structural strengthening plan, including measures such as adding supports and pasting carbon fiber sheets, to enhance the structural stability; The system uses various alarm methods such as text messages, APP push, and voice broadcasts to ensure that relevant personnel can receive early warning information in a timely manner and take necessary emergency measures. The implementation of this system can improve the safety during the construction process and reduce the impact of sudden risks on the construction progress and personnel safety.

[0029] In S6, the GIS platform has a spatial analysis function for spatial statistics and trend analysis of the distribution of potential hazards. Trend analysis includes calculating the density of potential hazard points and analyzing the distribution trend of potential hazards in different regions. The deep learning algorithm uses long short-term memory networks or convolutional neural networks and their variants to learn the spatio-temporal characteristics of data, and combines with an expert knowledge graph to construct a risk assessment model.

[0030] Specifically, the platform calculates the density of potential hazard points and analyzes the distribution trend of potential hazards in different regions to help users identify potential risk areas and provide data support for subsequent safety decisions. Trend analysis can reveal the spatio-temporal variation law of potential hazards and provide accurate prediction basis for safety management; The application of deep learning algorithms in this invention improves the accuracy of risk assessment. The platform uses long short-term memory networks (LSTM) or convolutional neural networks (CNN) and their variants to perform deep learning on the spatio-temporal features of data, enabling better capture of the dynamic changes of potential hazards. Combining with an expert knowledge graph, the system constructs an accurate risk assessment model, automatically analyzes potential risks, and provides reasonable suggestions for emergency measures to decision-makers. This technical solution not only improves the accuracy of hazard identification but also enhances the intelligent level of safety management, further ensuring the safety of engineering projects.

[0031] In S7, the archived analysis data is used to deeply mine historical and actual construction data using deep learning algorithms. The mining algorithms adopt deep autoencoders and generative adversarial networks. By mining, potential safety hazard patterns and rules are discovered, and the safety management strategy and monitoring model are optimized at a frequency of once per quarter. Each optimization adjusts at least 1 - 3 key parameters. The key parameters include the sensor monitoring frequency, the hazard identification threshold, and the weights of the risk assessment model.

[0032] Specifically, the mining algorithms adopt advanced technologies such as deep autoencoders (DeepAutoencoders) and generative adversarial networks (GAN) to identify potential safety hazard patterns and rules from a large amount of historical construction data. Through in-depth analysis of these patterns and rules, the system can discover the occurrence trend of hazards in real time and take effective preventive measures in advance; To continuously optimize the safety management strategy and monitoring model, the system is optimized once per quarter to ensure that the safety management is consistent with the actual situation. During each optimization process, the system adjusts at least 1 - 3 key parameters. These key parameters include the monitoring frequency of sensors, the threshold for hazard identification, and the weight allocation in the risk assessment model, etc. By adjusting these key parameters, the system can monitor the hazards during the construction process more accurately and optimize the risk assessment and early warning mechanism according to real-time data, thereby enhancing the overall safety management level and reducing the risk of accidents.

[0033] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principle and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things, characterized in that, It includes the following steps: S1. Real-time data collection: Based on the Internet of Things technology, various sensors are connected to the cloud or edge computing devices through a standardized edge computing gateway. A variety of sensors are deployed at the construction site to collect data and upload it via wireless communication. S2. Data preprocessing and analysis: The collected data is preprocessed on the cloud or edge computing platform. Big data analysis and AI algorithms are used to extract patterns and make intelligent predictions. A dynamic threshold adjustment mechanism is adopted to identify potential hazards. S3. Adaptive adjustment: According to the construction stage and environmental changes, the sensor layout and monitoring frequency are automatically adjusted. S4. Multi-source data fusion: The data of the construction progress management system is introduced and fused with the sensor data to improve the ability to identify potential hazards. S5. Alarm and emergency: When the monitored data is abnormal, an alarm is triggered, and relevant personnel are notified through various methods. An intelligent decision support system is used to provide emergency measures. S6. Visualization and decision support: The GIS platform is used to visually display the safety status. Deep learning algorithms and expert knowledge graphs are combined to assist in decision-making, and the monitoring model is optimized based on historical data. S7. Intelligent feedback and optimization: After the construction is completed, the analysis data is archived, and the differences are compared to optimize the future construction safety management strategy.

2. The method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things according to claim 1, wherein: In S1, the various sensors include temperature and humidity sensors, meteorological sensors, displacement sensors, vibration sensors, and water level sensors. The wireless communication includes NB-IoT, LoRa, and 5G. The cloud devices include Alibaba Cloud ECS instances or Tencent Cloud CVMs. The edge computing devices include Advantech WISE-4010 series or Huawei 5G industrial modules. The data upload method uses encrypted transmission. For sensors using NB-IoT communication, data is transmitted through the operator network using the MQTT protocol, and AES-128 algorithm is used for data encryption. For sensors using LoRa communication, data is transmitted through the LoRaWAN protocol and encrypted using the elliptic curve encryption algorithm. For sensors using 5G communication, data is uploaded through the 5G network using the HTTP / HTTPS protocol.

3. A method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things according to claim 1, characterized in that: In S2, the 3σ criterion is used for data cleaning to remove noise and outliers. Principal component analysis is used to reduce the data dimension, and the principal components with a contribution rate of more than 95% are extracted. The dynamic threshold adjustment mechanism uses a sliding window algorithm, and the window size is set between 20 - 50 data points according to the monitoring index.

4. A method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things according to claim 1, characterized in that: In S3, a correlation model is established by analyzing historical construction and environmental data. During heavy rain, according to historical data, the monitoring frequency of the water level sensor is increased from once every 5 minutes to once every 1 minute. When the wind speed is greater than 15 m / s in strong wind weather, the monitoring frequency of the wind speed sensor is increased from once every 2 minutes to once every 30 seconds. During the foundation construction stage, a displacement sensor is added every 10 meters, and the monitoring frequency is increased from once every 10 minutes to once every 3 minutes. During the main structure construction stage, a vibration sensor is added every 15 meters, and the monitoring frequency is increased from once every 8 minutes to once every 5 minutes.

5. A method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things according to claim 1, characterized in that: In S4, the data of the construction progress management system is structured data, and the geological exploration data can be semi-structured or unstructured data. When using the feature-level fusion algorithm, wavelet transform is used to extract features from the sensor perception data, and the Apriori algorithm is used to mine frequent itemsets to extract key features from the construction progress management system data. Then, the extracted features are spliced and fused according to the feature dimension ratio of 1:

1. When using the decision-level fusion algorithm, the decision results are first independently processed for each data source, and then the decision results are fused through the weighted voting method. Among them, the weight of the decision result of the sensor data is 0.6, the weight of the decision result of the construction progress management system data is 0.2, and the weight of the decision result of the geological exploration data is 0.

2.

6. The method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things according to claim 1, wherein: In S5, the intelligent decision support system is used to formulate emergency measures based on different types of hidden dangers in combination with historical cases. Among them, for the soil humidity being abnormally higher or lower than the normal range by 20%, drainage or irrigation suggestions are provided. If the humidity is higher than the normal range, it is recommended to turn on the drainage pump, and the flow rate of the drainage pump is adjusted between 5-10 cubic meters per hour according to the degree of humidity exceeding the standard. When the structural displacement is too large and exceeds 10% of the allowable displacement, measures such as suspending construction and strengthening the structure are given. The reinforcement methods include adding supports and pasting carbon fiber cloth. The alarm methods include text messages, APP push, and voice broadcast.

7. A method for monitoring potential safety hazards in the construction of water conservancy projects based on the Internet of Things according to claim 1, characterized in that: In S6, the GIS platform has a spatial analysis function for spatial statistics and trend analysis of the distribution of safety hazards. The trend analysis includes calculating the density of hazard points and analyzing the distribution trend of hazards in different regions. The deep learning algorithm uses long short-term memory networks or convolutional neural networks and their variants to learn the spatio-temporal features of the data and constructs a risk assessment model in combination with the expert knowledge graph.

8. A method for monitoring potential safety hazards in water conservancy project construction based on the Internet of Things according to claim 1, characterized in that: In S7, the archived analysis data is used to deeply mine the historical and actual construction data using deep learning algorithms. The mining algorithms use deep autoencoders and generative adversarial networks. By mining, potential safety hazard patterns and rules are discovered, and the safety management strategy and monitoring model are optimized at a frequency of once per quarter. Each optimization adjusts at least 1-3 key parameters. The key parameters include the sensor monitoring frequency, the hazard identification threshold, and the weight of the risk assessment model.

Images