A tower crane safety management method based on a full-space information system
By using a full-space information system and data fusion model, the problems of insufficient data processing and real-time response of the tower crane safety monitoring platform have been solved, enabling accurate judgment of the tower crane's operating status and comprehensive safety management, thereby improving the safety and efficiency of the construction site.
Patent Information
- Application Number
- CN202410959396.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-17
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2044-07-17
AI Technical Summary
Existing tower crane safety monitoring platforms have limitations in big data processing and real-time response. Their data processing capabilities are insufficient, making it impossible to effectively identify the working status of tower cranes, which prevents the safety monitoring platforms from responding to potential risks in a timely manner.
A tower crane safety management method based on a full-space information system is adopted. By capturing and managing tower crane micro-space elements through the full-space information system, a tower crane micro-space data model is constructed. Data fusion and predictive analysis are performed using a fuzzy grey relational dynamic weight optimization model and a Bayesian hyperparameter optimized CNN-LSTM network model. Safety monitoring sensors and data transmission networks are deployed to establish a tower crane safety monitoring platform.
It improved construction efficiency, reduced construction risks, enabled accurate judgment of tower crane operating status and comprehensive safety management, and enhanced the safety management capabilities of the construction site.
Smart Images

Figure CN118861983B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the technical field of tower crane safety management methods based on a full-space information system, specifically relating to a tower crane safety management method based on a full-space information system. Background Technology
[0002] Construction management is characterized by diverse management elements, complex operational processes, and dynamic, ever-changing timeframes, with safety management being a prominent issue. Traditional construction management typically employs a point-to-point, localized management model, independently managing construction equipment, personnel, and materials. This approach struggles to adapt to the increasingly complex construction management needs arising from urbanization and the expansion of construction project scale. The frequent tower crane accidents in recent years, in particular, highlight the urgent need for digital and intelligent management of construction equipment safety. Currently, there are problems such as incomplete and untimely data collection on tower crane operation and peripheral equipment, and significant susceptibility to human factors, leading to the inability of safety monitoring platforms to accurately identify the working status of tower cranes.
[0003] Currently, research on tower crane safety monitoring platforms has made some progress, mainly focusing on computer-aided systems, real-time monitoring technologies, and intelligent monitoring modes. These technologies have played an important role in improving operators' cognitive abilities and reducing accident risks. However, the data processing capabilities of tower crane safety monitoring platforms are limited, particularly in big data processing and real-time response, exhibiting certain limitations and passively receiving IoT sensing data with insufficient self-analysis capabilities. Summary of the Invention
[0004] The purpose of this invention is to address the limitations of existing tower crane safety monitoring platforms in terms of data processing capabilities, particularly in big data processing and real-time response. These platforms passively receive IoT sensing data but lack sufficient self-analysis capabilities. Therefore, this invention proposes a tower crane safety management method based on a full-space information system. This method effectively improves construction efficiency and reduces construction risks.
[0005] To achieve the above objectives, the present invention adopts the following technical solution:
[0006] Design a tower crane safety management method based on a full-space information system, including the following steps:
[0007] Step 1: Capture and manage tower crane micro-space elements through a full-space information system, construct a tower crane micro-space data model, and collect real-time data streams based on Internet of Things sensing to comprehensively perceive the tower crane and its surrounding environment.
[0008] Step 2: Establish a fuzzy grey relational dynamic weight optimization model to calculate the dynamic relational degree between tower crane safety monitoring variables and tilt angle, providing key monitoring points for the safe operation of tower cranes, and set a time weight optimization model to identify the variables that have the greatest impact on the tilt angle of tower cranes at specific time points during operation.
[0009] Step 3: Use a Bayesian hyperparameter-optimized CNN-LSTM network model to fuse tower crane safety monitoring data and perform predictive analysis on tilt angle data;
[0010] Step 4: Deploy safety monitoring sensors, data transmission networks, servers, and platform data sharing mechanisms. Combine actual construction projects to build a tower crane safety monitoring platform based on a full-space information system to comprehensively manage all elements of the construction site.
[0011] Furthermore, the tower crane micro-space elements include tower crane equipment structural space characteristics, construction personnel flow space characteristics, and tower crane construction environment space characteristics. The tower crane equipment structural space characteristics include the hoisting, slewing, and luffing space characteristics of the tower crane. The construction personnel flow space characteristics include the dynamic space characteristics of personnel flow within the construction work area covered by the tower crane. The tower crane construction environment space characteristics include the space characteristics of dust and environmental meteorological factors.
[0012] Furthermore, in step 1, before capturing the micro-space elements of the tower crane, a hierarchical analysis of the tower crane's safety factors is performed, classifying the tower crane's safety factors into layers, including personnel factors, equipment factors, environmental factors, and process factors.
[0013] Furthermore, a multi-source 3D model is used to construct the tower crane micro-space data model and form a tower crane virtual space model. The tower crane virtual space model includes a building information model, an oblique photogrammetry model, and a GLB model. The source data formats of the building information model, oblique photogrammetry model, and GLB model are converted into 3DTiles file format using a data format conversion tool to form a multi-source model fusion visualization.
[0014] Furthermore, the spatial characteristics of the construction personnel's movement are collected through intelligent equipment, including the configuration of smart safety helmets, deployment of Bluetooth beacons, collection of personnel spatial coordinates and trajectory information, and acquisition of network transmission layer data, for the real-time acquisition of personnel coordinates and trajectory information.
[0015] Furthermore, the structural spatial feature elements of the tower crane equipment include the network transmission layer data acquisition of the sensor equipment perception and signal transmission management system, data visualization and alarm system of the multi-sensor cluster. The sensor equipment perception collects the weight, height, amplitude, slewing direction, wind speed and tilt angle of the tower crane through the main control unit, display, weight sensor, amplitude sensor, height sensor, slewing sensor, wind speed sensor, tilt angle sensor, biometric identifier and control output unit.
[0016] The network transport layer data acquisition is processed by a stepwise filtering and fusion algorithm. During the processing and detection process, if the corresponding value Xn does not arrive at the computing platform synchronously, the value Xn-1 of the previous array is taken. If the value does not arrive at the computing platform within three acquisition cycles, a warning is issued and an error is reported.
[0017] Furthermore, when the spatial characteristics of the tower crane construction environment communicate with the dust monitoring system via the 212 protocol, data from sensors such as wind speed, dust, and noise are obtained.
[0018] Furthermore, in step 2, the method for calculating the dynamic correlation between the tower crane safety monitoring variables and the tilt angle includes the following steps:
[0019] Step 21: First, define the fuzzy weight coefficient to dynamically adjust the weight ratio between the exponentially weighted moving average and the fuzzy average;
[0020] Step 22: For the data at each time point, perform a weighted average of the exponentially weighted moving average weight and the fuzzy average weight to obtain the final weighting coefficient, and use the weighting coefficient to weight the original data.
[0021] Step 23: Calculate the correlation degree using the fused output matrix O to obtain the correlation coefficient, thereby obtaining the correlation degree between each monitored variable and the tilt angle;
[0022] The matrix O is obtained as follows:
[0023] First, the input matrix X of the original data has dimensions j×i, with j attributes and i time points. Information entropy is calculated for each column of attributes in the input matrix, resulting in an information entropy matrix E. The information entropy matrix has the same dimensions as the input matrix, and each element e... ij The information entropy of attribute j at time point i is calculated using the following formula:
[0024]
[0025] In the formula, P is the proportion of the number of samples with value i in index j to the total number of samples, k is a constant, usually k = 1 / ln(n), and n is the number of possible values of factor X;
[0026] The difference coefficient matrix D is calculated using the information entropy matrix E. The difference coefficient represents the degree of difference in attribute values, and the calculation formula is as follows:
[0027] D ij =1―e ij ;
[0028] The attribute weight matrix W is calculated based on the difference coefficient matrix. This matrix measures the contribution of different attributes to the overall correlation. The calculation formula is as follows:
[0029]
[0030] We perform a weighted average of the attributes in each column of the input matrix X to obtain an average matrix M. The indices of the average matrix are the same as those of the input matrix. For each element m... ij This represents the weighted average of attribute j at time point i; finally, the attribute weights W and the average matrix M are weighted and fused to obtain the fused output matrix O, where each element O... ij The value of attribute j after fusion at time point i is represented by the following formula:
[0031]
[0032] Furthermore, in step 3, the method for predictive analysis of the dip angle data includes the following steps:
[0033] Step 31: Real-time data on load, tilt angle, slewing angle, torque, height, wind speed, and tilt angle collected by sensors are input into the standardized data. A sliding window function is defined with a window size of 30. The first 30 data points are used to predict and update the next data point. The first 80% of the dataset is the training set, and the next 20% is the test set.
[0034] Step 32: Convert the data into tensors and use the GPU to perform tensor operations to improve the training speed of the deep learning model;
[0035] Step 33: Split the data into mini-batches. Define the amount of data used in each training iteration as 50. Shuffle the data at the beginning of each epoch. If the sample data of the mini-batch is insufficient, discard it to ensure that all mini-batches are of the same size.
[0036] Step 34: Construct the LSTM framework, which consists of two LSTM hidden layers. The most effective activation function for each hidden layer is the tanh function.
[0037] Step 35: Define an evaluation function, using uniform distribution and random sampling to define the learning rate lr, the hidden layer size hidden_size, the step size step_size, the search range of Gamma and p, optimize the defined objective function using a Bayesian optimization model, set the number of iterations to 100, fine-tune hidden_size and step_size after obtaining the optimal parameters, and then rerun the hyperparameter optimization.
[0038] Step 36: Configure the obtained parameters in the LSTM model, perform iterative training of the model, use the Adam optimizer to update the parameters and calculate the loss function for backpropagation;
[0039] Step 37: Calculate the root mean square error and mean absolute percentage error between the model prediction results and the true labels.
[0040] Furthermore, in step 4, the tower crane safety monitoring platform architecture includes an edge device perception layer, a network transmission layer, a management and control layer, and a decision analysis layer.
[0041] Compared with existing technologies, the tower crane safety management method based on a full-space information system proposed in this invention has the following advantages: This invention effectively improves construction efficiency and reduces construction risks. First, by analyzing the components of the tower crane's entire spatial structure, it extracts sensor data from the tower crane's operation process, including data on lifting weight, wind speed, tilt angle, height, rotation angle, amplitude, and torque. A fuzzy grey relational dynamic weight optimization model is established to analyze the correlation of tower crane monitoring variables at different time points. Next, a BO-CNN-LSTM model is used to fuse tower crane safety monitoring data and predict tilt angle variables, thereby improving the ability to judge the tower crane's operating status. Finally, a tower crane safety monitoring platform based on a full-space information system is built, enabling comprehensive management of various elements at the construction site, such as "people, machinery, materials, methods, and environment," allowing for dynamic identification and control of construction machinery operation and key construction processes, leading to more rational management decisions. Specifically:
[0042] (1) This invention calculates the correlation between key safety monitoring variables during tower crane operation and uses the time-weighted dynamic optimization grey relational analysis method to deeply explore the intrinsic correlation between sensor data at different times and under different environments, providing a scientific basis and effective information for operational decision-making and safety management during tower crane operation.
[0043] (2) This invention addresses the problems of accidental sensor failures, data distortion, and data loss that may occur during tower crane operation by proposing a CNN-LSTM hybrid neural network based on Bayesian optimization algorithm. This network can extract the spatial features of the tower crane and capture the long-term dependencies between sensor data, thereby achieving accurate prediction of key tower crane features. Verification through a series of error indices and fitting relationship graphs confirms a good relationship between the predicted and actual values even when tower crane feature data is missing.
[0044] 3. This invention is a tower crane safety monitoring platform based on PSIS. Combining the requirements for tower crane structural safety monitoring and on-site business needs, it deploys hardware facilities such as sensors, servers, monitoring systems, and networks on-site to achieve real-time data perception, network transmission, and data processing. By establishing a two-dimensional and three-dimensional integrated tower crane safety monitoring platform based on PSIS, it provides users with a comprehensive and multi-angle tower crane safety monitoring perspective, enhancing the safety management capabilities of the construction site. At the same time, through iterative optimization of the platform's functions, it ensures that it can meet the different needs of the construction site. Attached Figure Description
[0045] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:
[0046] Figure 1 This is a flowchart of the present invention;
[0047] Figure 2 This is a flowchart of the present invention;
[0048] Figure 3 This is a schematic diagram of the spatial elements of a tower crane in this invention;
[0049] Figure 4 This is a schematic diagram illustrating the influence of tower crane features in this invention;
[0050] Figure 5 This is a safety factor analysis diagram from the present invention;
[0051] Figure 6 This is a roadmap of model acquisition and visualization technologies in this invention;
[0052] Figure 7 This invention relates to the safety monitoring process for tower cranes;
[0053] Figure 8 This is a flowchart of the empirical process for the hybrid model in this invention;
[0054] Figure 9 This is an architecture diagram of the tower crane safety monitoring platform in this invention. Detailed Implementation
[0055] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0056] The structural features of the present invention will now be described in detail with reference to the accompanying drawings.
[0057] See Figures 1-2 A tower crane safety management method based on a full-space information system includes the following steps:
[0058] S1. Capture and manage tower crane micro-space elements through a full-space information system, construct a tower crane micro-space data model, and collect real-time data streams based on Internet of Things sensing to comprehensively perceive the tower crane and its surrounding environment.
[0059] By utilizing full-space information system technology, the micro-space elements of tower cranes can be effectively captured and managed, constructing a spatial data model of the tower crane. Based on IoT sensing, real-time data streams are collected to comprehensively perceive the tower crane and its surrounding environment. This full-space information system technology not only enables real-time monitoring and analysis of the tower crane's operational status but also provides crucial support for tower crane safety management, thereby improving the reliability and efficiency of monitoring and managing safe tower crane construction.
[0060] See Figure 3 Among them, the micro-space elements of tower cranes include the structural spatial characteristics of tower crane equipment, the spatial characteristics of construction personnel movement, and the spatial characteristics of the tower crane construction environment.
[0061] See Figure 4 The structural spatial characteristics of tower cranes include their hoisting, slewing, and luffing spatial features. Each operation—from hoisting and slewing to luffing—affects the tower crane's spatial characteristics. In the hoisting state, the crane hook rises or falls, causing a change in its spatial position; in the slewing state, the tower crane rotates around its vertical axis. α This causes the crane's working range to expand or contract; while the luffing state involves the extension and retraction (R) of the tower crane's boom, further altering its working range and spatial layout. Simultaneously, factors such as wind speed and load also affect the safe operation of the tower crane itself and directly impact its spatial characteristics. Wind speed directly affects the stability of the tower crane's operation, causing changes in θ and β, while increased load affects its lifting capacity. Furthermore, these factors not only individually affect tower crane operations but also intertwine and mutually constrain each other.
[0062] The spatial characteristics of construction personnel movement include the dynamic spatial features of personnel movement within the construction work area covered by the tower crane. Personnel movement around the tower crane on the construction site is not an isolated or singular behavioral pattern, but rather a complex system involving multi-role collaboration and multi-element interaction. The continuous flow and interaction of personnel within the construction area not only directly affects the progress control of tower crane construction but is also a core factor influencing the effectiveness of the overall safety management system and the improvement of construction efficiency. This phenomenon reveals that in modern construction management, a refined understanding and effective control of personnel movement is crucial for achieving project objectives.
[0063] The spatial characteristics of the tower crane construction environment include dust and environmental meteorological factors. Dust pollution and environmental meteorological factors (humidity, wind speed) at the construction site pose multiple challenges to the safety of tower crane operations. They not only directly affect the operator's vision and judgment but also potentially threaten the stability of the tower crane by altering its frictional properties and inducing lateral dynamic responses, increasing the probability of accidents. Therefore, it is necessary to take targeted dust control measures and fully consider the impact of environmental factors on tower crane operations to ensure both safety and efficiency at the construction site.
[0064] See Figure 5 Before capturing the micro-space elements of the tower crane, a layered analysis of tower crane safety factors is performed, classifying these factors into hierarchical layers: personnel factors, equipment factors, environmental factors, and process factors. Specifically:
[0065] ① Personnel factors involve the operator's skill level, experience, behavior patterns, and compliance with safety regulations and operating procedures.
[0066] ② Equipment factors include the tower crane's structural design, technical condition, operating status, and degree of aging.
[0067] ③ Environmental factors involve the operating environment and natural environment of the tower crane. Operating environment factors mainly include tower crane collisions and collisions with obstacles; natural environment factors during operation mainly include low visibility, high temperature, and high wind speed.
[0068] ④ Process factors refer to factors related to the work process, including work procedures, safe operating procedures, etc.
[0069] Within the technical framework of a tower crane full-space information system, the analysis of safety factors is of paramount importance. By systematically analyzing various safety factors during tower crane operation, a comprehensive safety indicator system can be established, encompassing factors such as personnel operation, machine status, and the working environment. This lays the foundation for subsequent data collection and analysis, making data collection more comprehensive and systematic. Secondly, the analysis of safety factors helps identify and assess potential safety risks, promptly detect potential hazards that could lead to accidents, and take preventative measures before accidents occur. Most importantly, the analysis of safety factors provides a deep understanding of the safety hazards and weaknesses in tower crane operation, providing a basis for developing more effective safety management strategies. Therefore, conducting thorough safety factor analysis can improve the safety and stability of tower crane operations, while ensuring the safety of personnel and equipment at the construction site, thereby improving construction efficiency and quality.
[0070] See Figure 6 A multi-source 3D modeling approach was used to construct a micro-space data model of the tower crane, forming a virtual space model. This virtual space model includes a Building Information Model (BIM), an oblique photogrammetry model, and a GLB model. Data format conversion tools were used to convert the source data formats of the BIM, oblique photogrammetry model, and GLB model into 3DTiles file format, resulting in a multi-source model fusion visualization. The BIM was built using Autodesk Revit software, the oblique photogrammetry model was built using UAV aerial photography and the oblique photogrammetry modeling software ContextCapture, and the GLB model was built using Autodesk Maya software. The space for tower crane construction includes not only the main building but also supporting construction facilities, the construction area, and the surrounding environment. Building the oblique photogrammetry model allows for the acquisition of detailed information on the surrounding terrain features, enabling the verification of on-site feature attributes.
[0071] By using the GLB model, the tower crane model is decomposed into multiple components such as the tower body, boom, pulleys, and hooks, realizing the modular design of the tower crane. This makes the entire model easier to manage and maintain, while also helping users to understand the relationships and characteristics between the various components of the tower crane in a simple and intuitive way.
[0072] The spatial characteristics of construction personnel movement are collected through intelligent equipment. This is achieved by configuring smart safety helmets, deploying Bluetooth beacons, collecting personnel spatial coordinates and trajectory information, and acquiring data at the network transmission layer. This data is used to collect the real-time coordinates and trajectory information of personnel.
[0073] The structural spatial characteristics of tower crane equipment include data acquisition through a network transmission layer of multi-sensor clusters, sensor equipment sensing and signal transmission management systems, data visualization, and alarm systems. Sensor equipment sensing is achieved through a main control unit, display, weight sensors, amplitude sensors, height sensors, slewing sensors, wind speed sensors, tilt sensors, biometric sensors, and control output units. This type of sensor integration features real-time performance, comprehensiveness, accuracy, and automation, collecting data on the tower crane's weight, height, amplitude, slewing orientation, wind speed, and tilt angle. This provides multi-dimensional coverage of the tower crane's operating space, including height, length, and width, while also fully considering the impact of factors such as the hook, luffing trolley, boom's local operating areas, wind speed, and tilt angle on the complex three-dimensional space of the tower crane.
[0074] See Figure 7 The network transport layer data acquisition is processed using a progressive filtering and fusion algorithm. During the processing and detection, if the corresponding value Xn does not arrive at the computing platform synchronously, the value Xn-1 from the previous array is used. If the value does not arrive at the computing platform within three acquisition cycles, an alarm is triggered and an error is reported. This workflow can be roughly divided into four subsystems: multi-sensor cluster, signal transmission and data management system, data visualization, and alarm system. Data from different sensors is combined with the output signal and stored in a database for easy sharing throughout the system. The key lies in the stability and timeliness of wireless transmission and the optimization of data processing. The data center accumulates a large amount of data, and big data analytics is used to input these technologies into the model for analysis and prediction.
[0075] The spatial characteristics of the tower crane construction environment are acquired through communication with the dust monitoring system via the 212 protocol, obtaining data from sensors such as wind speed, dust, and noise. During construction, dust and noise pollution caused by construction vehicles carrying mud, scattering building materials, and construction machinery are extremely serious, becoming one of the main causes affecting urban air quality. Dust and noise pollution generated during construction has consistently been a major source of disputes between construction sites and nearby residents, attracting significant attention from environmental regulatory departments. Real-time monitoring of weather, air quality, and noise in the construction site area is conducted, including environmental parameters such as noise, particulate matter concentration (PM2.5, PM10), wind direction, wind speed, humidity, temperature, and atmospheric pressure. The collected data is analyzed to monitor the implementation of green construction practices and ensure that construction workers operate in a safe external environment.
[0076] S2. Establish a fuzzy grey relational dynamic weight optimization model to calculate the dynamic correlation between tower crane safety monitoring variables and tilt angle, providing key monitoring points for the safe operation of tower cranes. Also, set up a time weight optimization model to identify the variables that have the greatest impact on the tilt angle of tower cranes at specific time points during operation.
[0077] Tower crane safety monitoring variables include the crane's lifting weight, height, radius and slewing angle, wind speed, etc., among which:
[0078] ① Lifting weight and tilt angle: Increasing the lifting weight will cause the tower crane's tilt angle to increase, as this increases the risk of overturning. Therefore, monitoring the tilt angle can help operators determine if there is an overload and take timely measures to adjust the lifting weight or implement support measures to maintain the tower crane's stability.
[0079] ② Inclination angle and height: Changes in inclination angle affect the stability of the tower crane, especially when operating at a large inclination angle, which increases the risk of overturning. Changes in height also affect the stability of the tower crane, especially when there are significant changes in height. Therefore, both inclination angle and height must be considered comprehensively to ensure that the tower crane operates within a safe working range.
[0080] ③ Amplitude and Slewing Angle: Changes in amplitude and slewing angle affect the working range and direction of the tower crane. Appropriate amplitude and slewing angle ensure that the tower crane performs lifting operations in the appropriate position, avoiding safety hazards caused by collisions or operational errors.
[0081] ④ Wind speed, tilt angle, and moment: Increased wind speed increases the risk of tower crane tilting, especially during operations at higher wind speeds. Simultaneously, wind force also affects the tower crane's torque, potentially leading to instability in its operating state. Therefore, in high wind speeds, special attention must be paid to changes in tilt angle and moment to ensure the safe operation of the tower crane. A comprehensive assessment of the safety and stability of the tower crane can be achieved by considering the interrelationships between these variables.
[0082] The tilt angle is a measure of the tower's inclination and a safety indicator of the tower crane. The tilt angle is the main indicator of most tower crane overturning accidents. The tilt angle information of the tower crane directly reflects its degree of tilt during operation. In this embodiment, the tilt angle is used as one of the important features of the system analysis to explore the influence of other variables such as lifting weight, slewing angle, height, and wind speed on the tilt angle of the tower crane.
[0083] The parameters and variables sensed by sensors installed on tower cranes reflect key parameters and operating states during operation, including load, structural stability, operating range, operating height, rotation state, torque, and external environmental factors. Due to the characteristics of tower crane operation, these data often exhibit complex nonlinear relationships. Furthermore, due to uncertainties in environmental factors and noise in sensor data, data incompleteness and uncertainty are common. Since grey relational analysis has strong tolerance for data incompleteness and uncertainty and can effectively handle nonlinear data, it is chosen to comprehensively and accurately assess the relationships between various detected variables during tower crane operation, providing a scientific basis and technical support for the safe operation of tower cranes.
[0084] In continuous tower crane operations, constantly changing environments and conditions, including weather, project progress, and load variations, affect the correlation between various attributes, necessitating consideration of dynamic weights and the fuzzy characteristics of these changes. A fuzzy dynamic weight optimization model comprehensively considers time weights, fuzzy characteristics, and real-time data to more accurately capture the correlations between attributes during tower crane operations. For long-term tower crane operation procedures, time weights are introduced to dynamically adjust the weights of real-time data, optimizing tower crane operational decisions in a timely manner and ensuring both safety and efficiency. The exponentially weighted moving average (EWMA) method is used to determine the time weights.
[0085] The specific definition of the EWMA method is as follows:
[0086] S t =αx t +(1―α)S t―1
[0087] Where S t x represents the exponentially weighted moving average at time point t. t The value is at time point t, and α is a smoothing factor that determines the weight of the most recent observation, typically ranging from 0 to 1. A larger α will give more weight to newer observations, making the average more sensitive, but also more volatile; conversely, a smaller α will make the average more stable.
[0088] The method for calculating the dynamic correlation between tower crane safety monitoring variables and tilt angle includes the following steps:
[0089] S21. First, define the fuzzy weight coefficient to dynamically adjust the weight ratio between the exponentially weighted moving average and the fuzzy average.
[0090] S22. For the data at each time point, the weights of the exponentially weighted moving average and the fuzzy average are weighted and averaged to obtain the final weighting coefficients, and the original data are weighted using the weighting coefficients.
[0091] S23. The correlation degree was calculated using the fused output matrix O, and the correlation coefficient was obtained, thus obtaining the correlation degree between each monitoring variable and the tilt angle.
[0092] The matrix O is obtained as follows:
[0093] First, the input matrix X of the original data has dimensions j×i, with j attributes and i time points. Information entropy is calculated for each column of attributes in the input matrix, resulting in an information entropy matrix E. The information entropy matrix has the same dimensions as the input matrix, and each element e... ij The information entropy of attribute j at time point i is calculated using the following formula:
[0094]
[0095] In the formula, P is the proportion of the number of samples with value i in index j to the total number of samples, k is a constant, usually k = 1 / ln(n), and n is the number of possible values of factor X;
[0096] The difference coefficient matrix D is calculated using the information entropy matrix E. The difference coefficient represents the degree of difference in attribute values, and the calculation formula is as follows:
[0097] D ij =1―e ij ;
[0098] The attribute weight matrix W is calculated based on the difference coefficient matrix. This matrix measures the contribution of different attributes to the overall correlation. The calculation formula is as follows:
[0099]
[0100] We perform a weighted average of the attributes in each column of the input matrix X to obtain an average matrix M. The indices of the average matrix are the same as those of the input matrix. For each element m... ij This represents the weighted average of attribute j at time point i; finally, the attribute weights W and the average matrix M are weighted and fused to obtain the fused output matrix O, where each element O... ij The value of attribute j after fusion at time point i is represented by the following formula:
[0101]
[0102] By calculating the correlation between various monitored variables and the tilt angle, key monitoring points are provided for the safe operation of tower cranes. A time-weighted optimization model is introduced to identify the variables with the greatest impact on the tower crane's tilt angle at specific operational time points, thereby understanding the main risks faced by tower cranes at different construction stages. The analysis results help tower crane operators and project managers take different risk management measures at different times, such as increasing monitoring frequency or implementing special safety precautions during certain periods. Based on time weighting, resource allocation can be optimized, and more detailed dynamic operating procedures can be developed for tower cranes to adapt to the influence of different time periods and environmental conditions, thereby reducing the probability of accidents.
[0103] S3. The CNN-LSTM network model optimized by Bayesian hyperparameters is used to fuse tower crane safety monitoring data and predict and analyze the tilt angle data.
[0104] Reference Figure 8Using n different time steps, each with different data, sensor measurements are applied to a Long Short-Term Time-Tracking Model (LSTM) network based on Bayesian parameter optimization to predict tilt angle values. This quantitatively verifies the feasibility and accuracy of the proposed structure relative to a considered reference. For N samples considered, if x is the true value, These are predicted values. The accuracy of the model can be evaluated by comparing its predicted structure with the actual data using regression indicators. In this embodiment, the indicators are Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Squared Error (MSE), with the specific formulas shown below:
[0105]
[0106] Where n represents the total number of observations used to train the model.
[0107] This embodiment experiments with the construction of one LSTM hidden layer and two LSTM hidden layers. The optimal activation function selected in each hidden layer is the tanh function, while the activation function in the output layer is a linear function. A dropout layer with a dropout rate of 0.1 is inserted between the two LSTM hidden layers. Specifically:
[0108] The method for predictive analysis of dip angle data includes the following steps:
[0109] S31. Real-time data on load, tilt angle, slewing angle, torque, height, wind speed, and tilt angle are collected by sensors. After standardization, a sliding window function is defined with a window size of 30. The first 30 data points are used to predict and update the next data point. The first 80% of the dataset is the training set, and the next 20% is the test set.
[0110] S32. Convert the data into tensors and use the GPU to perform tensor operations to improve the training speed of deep learning models.
[0111] S33. Mini-batch data splitting: Define the amount of data used in each training iteration as 50. Shuffle the data at the beginning of each epoch. If the sample data of the mini-batch is insufficient, discard it to ensure that all mini-batch sizes are consistent.
[0112] S34. Construct an LSTM framework, consisting of two LSTM hidden layers. For each hidden layer, the most effective activation function is the tanh function.
[0113] S35. Define an evaluation function, using uniform distribution and random sampling to define the learning rate lr, the hidden layer size hidden_size, the step size step_size, the search range of Gamma and p, optimize the defined objective function using a Bayesian optimization model, set the number of iterations to 100, fine-tune hidden_size and step_size after obtaining the optimal parameters, and then rerun the hyperparameter optimization.
[0114] S36. Configure the obtained parameters in the LSTM model, perform iterative training of the model, use the Adam optimizer to update the parameters and calculate the loss function for backpropagation;
[0115] Step 37: Calculate the root mean square error and mean absolute percentage error between the model prediction results and the true labels.
[0116] The movement trajectory of a tower crane at different heights and angles, the extension and retraction of the boom, and the positional changes of the hook can all be represented in time-series data. These attributes change over time, and the patterns and trends in the time-series data reflect the spatial movement trajectory and characteristics of the tower crane. Therefore, Convolutional Neural Networks (CNNs) are used to improve the model's sensitivity to features. The outputs of CNN and LSTM models are fused to form a fused feature representation. The introduced attention mechanism can alleviate sensitivity to noise and outliers, making the model more robust and helping to cope with complex environments and uncertainties.
[0117] In this embodiment, the number of channels in the convolutional neural network is set to the dimension of the input data to control the number of local features learned by the model. The convolutional kernel is set to a one-dimensional filter with a width of 3, and a window of size 3 is considered once along the input sequence. The data transpose operation of the convolutional kernel is input into the LSTM, and the local features of the input sequence are automatically extracted through the one-dimensional convolutional layer. Then, the LSTM model captures the temporal information in the sequence.
[0118] S4. Deploy safety monitoring sensors, data transmission networks, servers, and platform data sharing mechanisms to build a tower crane safety monitoring platform based on a full-space information system, integrating actual construction projects to comprehensively manage all elements of the construction site.
[0119] See Figure 9 The tower crane safety monitoring platform comprises an edge device sensing layer, a network transmission layer, a management and control layer, and a decision analysis layer. Among them:
[0120] As the front-end interface of the tower crane construction site monitoring system, the edge device perception layer undertakes the functions of real-time data acquisition and preliminary processing, serving as the data source for the entire monitoring system. By integrating diverse sensing devices, wireless communication components, and intelligent control units, the edge device perception layer achieves real-time perception of the safety status at the tower crane structure and operational process levels, seamlessly transmitting the acquired monitoring data to the network fusion layer for further integration and analysis. Specifically, the sensing devices in the perception layer are scientifically arranged according to the tower crane monitoring targets, monitoring indicator system, and required parameter characteristics, ensuring full coverage monitoring of the tower crane's key load-bearing structures and functional components. These sensors capture key data such as the mechanical state, environmental conditions, and operational behavior during the tower crane's operation in real time. Simultaneously, the perception layer also has the ability to receive alarm information transmitted from the network fusion layer. When receiving an emergency alarm from higher authorities, it can immediately trigger corresponding early warning devices at the construction site, issuing audible and visual warnings or other forms of alarm signals to quickly respond to potential risks and enhance on-site safety warning effects and emergency response capabilities. In summary, the perception layer, with its integrated characteristics of multi-dimensional perception, real-time communication, and intelligent control, builds a solid safety defense line for the tower crane construction site.
[0121] The network transmission layer, serving as the central link in the system architecture, and the network convergence layer undertake crucial functions such as wireless data transmission, localized data processing, on-site early warning, and local data retention. It possesses localized computing, network, and storage resources, enabling both the transmission and aggregation of multi-source heterogeneous data from the sensing layer, and localized data processing to assess the safety status of tower crane structures and operational procedures in real time. Key components include: 1) On-site communication network: Connecting wirelessly to the safety sensing layer and aggregating multi-source heterogeneous monitoring data. Using IoT transmission methods such as Wi-Fi and RFID solves the problems of cumbersome and difficult-to-maintain wiring in traditional monitoring. 2) Data computing nodes: Deployed on the construction site, consisting of hardware with computing capabilities. Receives data collected by the on-site communication network, processes and stores the data, and enables localized safety decision-making. 3) External public network: Primarily built using Ethernet technology, connecting local data computing nodes to an external cloud computing center. Enables data collaboration between the construction site and the cloud computing center.
[0122] The management and decision analysis layer is the core of the tower crane safety monitoring platform. Relying on the computing power of on-site data nodes, it integrates data flows and issues control commands to the network fusion layer, enabling localized safety decision-making and multi-layered management of tower crane safety monitoring: 1) Managing the computing, network, and storage resources of the safety perception layer and the network fusion layer to ensure system stability and reliability. 2) Performing structured processing and storage of collected monitoring data to ensure data integrity and availability. 3) Analyzing various monitoring data, conducting tower crane structural safety assessments, and outputting safety alert information at each level. 4) Achieving integrated safety monitoring and control of tower crane operations to ensure operational safety and effectiveness. 5) Displaying monitoring data and safety alert status in real time, sending safety instructions to relevant personnel, and ensuring the monitoring system's immediate response capability.
[0123] The platform deploys computing resources at the construction site to reduce reliance on cloud data centers. By making safety decisions locally, it can quickly provide early warnings of structural and operational risks to tower cranes, thereby improving the real-time nature of safety management. Furthermore, it performs preliminary preprocessing on the collected data to reduce unnecessary data transmission, effectively lowering latency and storage costs.
[0124] The tower crane safety monitoring platform eliminates information bottlenecks, ensuring real-time updates and seamless sharing of spatial information, providing on-site managers with timely and accurate spatial decision-making support. By integrating full-space information system technology, sensor monitoring networks, on-site monitoring systems, and high-performance localized computing, network, and storage resources, it aims to build a comprehensive, real-time, and intelligent tower crane operation space perception and management system. This will improve construction safety management efficiency, prevent potential accident risks, and promote the efficient and orderly progress of engineering construction.
[0125] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A tower crane safety management method based on a full-space information system, characterized in that, Includes the following steps: Step 1: Capture and manage tower crane micro-space elements through a full-space information system, construct a tower crane micro-space data model, and collect real-time data streams based on Internet of Things sensing to comprehensively perceive the tower crane and its surrounding environment. The tower crane micro-space elements include tower crane equipment structural space characteristics, construction personnel flow space characteristics, and tower crane construction environment space characteristics. The tower crane equipment structural space characteristics include the hoisting, slewing, and luffing space characteristics of the tower crane. The construction personnel flow space characteristics include the dynamic space characteristics of personnel flow within the construction work area covered by the tower crane. The tower crane construction environment space characteristics include the space characteristics of dust and environmental meteorological factors. Step 2: Establish a fuzzy grey relational dynamic weight optimization model to calculate the dynamic relational degree between tower crane safety monitoring variables and tilt angle, providing key monitoring points for the safe operation of tower cranes, and set a time weight optimization model to identify the variables that have the greatest impact on the tilt angle of tower cranes at specific time points during operation. The method for calculating the dynamic correlation between tower crane safety monitoring variables and tilt angle includes the following steps: Step 21: First, define the fuzzy weight coefficient to dynamically adjust the weight ratio between the exponentially weighted moving average and the fuzzy average. Step 22: For the data at each time point, perform a weighted average of the exponentially weighted moving average weight and the fuzzy average weight to obtain the final weighting coefficient, and use the weighting coefficient to weight the original data. Step 23: Calculate the correlation degree using the fused output matrix O to obtain the correlation coefficient, thereby obtaining the correlation degree between each monitored variable and the tilt angle; The matrix O is obtained as follows: First, the input matrix X of the original data has dimensions... , Each attribute and For each time point of data, information entropy is calculated for each column of the input matrix, resulting in an information entropy matrix E. The dimension of the information entropy matrix is the same as that of the input matrix, and each element... Represents attributes At the point of time Information entropy, calculation formula: ; In the formula, p is the median value of index j. The proportion of the sample size to the total number of scripts, where k is a constant, usually... , where n is the number of possible values for factor X; The difference coefficient matrix D is calculated using the information entropy matrix E. The difference coefficient represents the degree of difference in attribute values, and the calculation formula is as follows: ; The attribute weight matrix W is calculated based on the difference coefficient matrix. This matrix measures the contribution of different attributes to the overall correlation. The formula for calculating the weight matrix is as follows: ; A weighted average is calculated for each column of the input matrix X to obtain an average matrix m. The indices of the average matrix are the same as those of the input matrix. For each element... Representing attributes At the point of time The weighted average is calculated; finally, the attribute weights W and the average matrix m are weighted and fused to obtain the fused output matrix O, where each element... Representing attributes At the point of time The fused value is calculated using the following formula: ; Step 3: Use a Bayesian hyperparameter-optimized CNN-LSTM network model to fuse tower crane safety monitoring data and perform predictive analysis on tilt angle data; Step 4: Deploy safety monitoring sensors, data transmission networks, servers, and platform data sharing mechanisms. Combine actual construction projects to build a tower crane safety monitoring platform based on a full-space information system to comprehensively manage all elements of the construction site.
2. The tower crane safety management method based on a full-space information system according to claim 1, characterized in that, In step 1, before capturing the micro-space elements of the tower crane, a hierarchical analysis of the tower crane's safety factors is performed, which stratifies the tower crane's safety factors into layers, including personnel factors, equipment factors, environmental factors, and process factors.
3. The tower crane safety management method based on a full-space information system according to claim 1, characterized in that, The tower crane micro-space data model is constructed using a multi-source 3D model approach, forming a tower crane virtual space model. The tower crane virtual space model includes a building information model, an oblique photogrammetry model, and a GLB model. The source data formats of the building information model, oblique photogrammetry model, and GLB model are converted into 3DTiles file format using a data format conversion tool, forming a multi-source model fusion visualization.
4. The tower crane safety management method based on a full-space information system according to claim 1, characterized in that, The spatial characteristics of the construction personnel's movement are collected by intelligent equipment. This is achieved through the configuration of smart safety helmets, deployment of Bluetooth beacons, collection of personnel spatial coordinates and trajectory information, and acquisition of network transmission layer data. This data is used for the real-time acquisition of personnel coordinates and trajectory information.
5. The tower crane safety management method based on a full-space information system according to claim 1, characterized in that, The structural spatial features of the tower crane equipment include data acquisition from the network transmission layer of a multi-sensor cluster sensor equipment perception and signal transmission management system, data visualization, and alarm system. The sensor equipment perception collects the tower crane's weight, height, amplitude, slewing orientation, wind speed, and tilt angle through the main control unit, display, weight sensor, amplitude sensor, height sensor, slewing sensor, wind speed sensor, tilt angle sensor, biometric identifier, and control output unit. The network transport layer data acquisition is processed by a stepwise filtering and fusion algorithm. During the processing and detection process, if the corresponding value Xn does not arrive at the computing platform synchronously, the value Xn-1 of the previous array is taken. If the value does not arrive at the computing platform within three acquisition cycles, a warning is issued and an error is reported.
6. The tower crane safety management method based on a full-space information system according to claim 1, characterized in that, The spatial characteristics of the tower crane construction environment are obtained by communicating with the dust monitoring system via the 212 protocol to acquire data from wind speed, dust, and noise sensors.
7. A tower crane safety management method based on a full-space information system according to claim 1, characterized in that, In step 3, the method for predictive analysis of the dip angle data includes the following steps: Step 31: Real-time data on load, tilt angle, slewing angle, torque, height, wind speed, and tilt angle collected by sensors are input into the standardized data. A sliding window function is defined with a window size of 30. The first 30 data points are used to predict and update the next data point. The first 80% of the dataset is the training set, and the last 20% is the test set. Step 32: Convert the data into tensors and use the GPU to perform tensor operations to improve the training speed of the deep learning model; Step 33: Split the data into mini-batches. Define the amount of data used in each training iteration as 50. Shuffle the data at the beginning of each epoch. If the sample data of the mini-batch is insufficient, discard it to ensure that all mini-batches are of the same size. Step 34: Construct the LSTM framework, consisting of two LSTM hidden layers. For each hidden layer, select the most effective activation function. function; Step 35: Define an evaluation function, using a uniform distribution and random sampling to define the learning rate. Hidden layer size Step length Gamma and The search range is defined, and the objective function is optimized using a Bayesian optimization model. The number of iterations is set to 100. After obtaining the optimal parameters, the search is then... and Make fine adjustments, and then rerun the hyperparameter optimization. Step 36: Configure the obtained parameters in the LSTM model, perform iterative training of the model, use the Adam optimizer to update the parameters and calculate the loss function for backpropagation; Step 37: Calculate the root mean square error and mean absolute percentage error between the model prediction results and the true labels.
8. A tower crane safety management method based on a full-space information system according to claim 1, characterized in that, In step 4, the tower crane safety monitoring platform includes an edge device perception layer, a network transmission layer, a management and control layer, and a decision analysis layer.
Citation Information
Patent Citations
Highway construction optimization method and system based on big data
CN117710156A
Building tower crane operation monitoring system and method
CN118209154A
