A Dynamic Safety Assessment Method for Steel Structure Lifting Structures Based on Multi-Source Monitoring Data

Abstract

This invention discloses a dynamic safety assessment method for steel structure lifting based on multi-source monitoring data. The method includes the following steps: Step 1: Determine the probability distribution model of uncertainties during steel structure lifting construction; Step 2: Determine the specific safety criteria and early warning function for construction; Step 3: Establish a finite element model, perform sensitivity analysis, and determine the main control items and auxiliary control items; Step 4: Use the main control items and their early warning function values ​​as the dataset for model training; Step 5: Split the dataset into training, validation, and test sets; Step 6: Train the model to obtain the evaluation model; Step 7: Install corresponding sensors on the steel structure according to the main and auxiliary control items; Step 8: When the early warning function value reaches the early warning threshold, execute the emergency plan; otherwise, continuously observe the early warning function value and record its real-time changes. This invention relates to the field of large-span steel structure construction projects involving extreme risks and can solve the problem of time-consuming structural safety assessment during steel structure lifting processes in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of construction of large-span steel structure projects with severe risks, and in particular to a dynamic assessment method for the safety of steel structure lifting structures based on multi-source monitoring data. Background Technology

[0002] Currently, steel structure corridors are widely used as transportation passages between buildings. For ease of construction, assembling large steel structure corridors on the ground and then lifting them as a whole to the design elevation has become a common construction method. It is important to note that the construction of such a large-span steel structure involves many uncertainties, such as asynchronous lifting at the two sides, instantaneous wind pressure changes, and component position deviations. These factors significantly impact the safety of the overall lifting process of the large steel structure. Currently, to ensure construction safety, finite element analysis under different working conditions is often performed on the steel structure before lifting to guarantee the safety of the construction process.

[0003] However, current safety assessments during construction are often based solely on the maximum stress and deformation reported by deployed sensors, failing to comprehensively reflect the structural morphology of complex steel structures during construction. Furthermore, uncertainties in actual construction are dynamically changing, and preliminary condition analyses cannot fully cover all potential situations that may arise during the overall lifting of large steel structure corridors. While feeding real-time changes in load conditions or boundary conditions during construction into a finite element model for calculation results is time-consuming, and subsequent evaluation based on these results is also time-consuming, unsuitable for the rapid and efficient construction process required for lifting operations, and unable to effectively guide engineering construction. On the other hand, a large amount of monitoring data obtained during steel structure construction, such as temperature changes, changes in member stress levels, and wind speed, is not effectively utilized. Therefore, a dynamic safety assessment method for steel structure lifting based on multi-source monitoring data is needed to address the problem of time-consuming structural safety assessments during steel structure lifting processes in existing technologies. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic assessment method for the structural safety of steel structure lifting based on multi-source monitoring data, which can solve the problem of long time consumption in the existing technology for assessing the structural safety of steel structure lifting process.

[0005] This invention is implemented as follows:

[0006] A dynamic safety assessment method for steel structure lifting structures based on multi-source monitoring data includes the following steps:

[0007] Step 1: Based on the actual situation of the project, determine the probability distribution model of the structural uncertainties during the steel structure lifting construction;

[0008] Step 2: Determine the specific safety criteria and early warning functions for steel structure lifting construction based on the specifications or project requirements;

[0009] Step 3: Establish a finite element model based on the drawings, conduct sensitivity analysis, and determine the main control items and auxiliary control items;

[0010] Step 4: Use the main control items and their corresponding early warning function values ​​from Step 3 as the dataset for model training, where the early warning function values ​​serve as the output values ​​of the multi-input data;

[0011] Step 5: Split the model's dataset into a training set, a validation set, and a test set;

[0012] Step 6: Train the model using machine learning algorithms to obtain the evaluation model;

[0013] Step 7: Install corresponding sensors on the steel structure according to the main control items and auxiliary control items;

[0014] Step 8: Set the warning threshold and formulate an emergency plan. When the warning function value reaches the warning threshold, execute the emergency plan; otherwise, continue to observe the warning function value and record its real-time changes.

[0015] In step 1, the uncertainties include: the estimated wind pressure changes of the structure based on the project location and construction season, the estimated structural node coordinates and component section deviations considering the accuracy of steel structure fabrication and installation, and the degree of asynchronous lifting of multiple lifting points of the steel structure.

[0016] In step 2, regarding deformation: the ratio of the overall tilt angle, deflection, and horizontal displacement of the steel structure to its predetermined threshold can be set as α; regarding stress level: the ratio of the overall maximum stress of the structure or the lifting cable force to its predetermined threshold can be set as β; when either α or β reaches 1, the comprehensive early warning function value is set to 1, that is, the steel structure lifting construction enters the early warning state, and the emergency plan is activated to adjust the lifting construction process in a timely manner.

[0017] In step 3, Latin hypercube sampling is performed on the uncertainties in the steel structure lifting process. The structural response under each working condition, i.e., the early warning function value, is calculated through the finite element model, and then sensitivity analysis is performed. Several key uncertainties that contribute significantly to the structural response are selected as the main control items, while the remaining uncertainties can be used as auxiliary control items.

[0018] In step 7, sensors for the main control project must be set up to capture the input factors and real-time changes of parameters of key uncertainties in the steel structure during the lifting process, forming multi-source monitoring data; sensors for the auxiliary control project are not set up or are selectively set up according to the actual working conditions to capture the input factors and real-time changes of parameters of non-critical uncertainties in the steel structure during the lifting process.

[0019] Compared with the prior art, the present invention has the following advantages:

[0020] 1. This invention comprehensively considers the structural response corresponding to key uncertainties such as deformation, displacement and component stress level, and flexibly and accurately reflects the risks in the overall lifting construction process of steel structure through the early warning function value corresponding to multi-source monitoring data, thereby facilitating the safety control of critical and dangerous projects.

[0021] 2. This invention establishes an evaluation model based on machine learning, which calculates and outputs early warning function values ​​in real time based on multi-source monitoring data. This avoids the problem of long calculation time in traditional finite element software. It can perform dynamic safety assessment of the overall lifting construction process of steel structures. The assessment results are highly timely and accurate, and can be used to effectively match and guide the lifting construction process, thereby improving the safety and digitalization level of large-span steel structure construction. Attached Figure Description

[0022] Figure 1 This is a flowchart of the method for dynamic safety assessment of steel structure lifting structures based on multi-source monitoring data, as described in this invention.

[0023] Figure 2 This is a finite element model diagram of Embodiment 1 of the dynamic evaluation method for the safety of steel structure lifting structure based on multi-source monitoring data of the present invention;

[0024] Figure 3 yes Figure 2 The structural displacement calculation results are shown in the figure (under dead load).

[0025] Figure 4 This is the sensitivity analysis result of Example 1 in the dynamic evaluation method for the safety of steel structure lifting structure based on multi-source monitoring data of the present invention;

[0026] Figure 5 This is a flowchart of the neural network training process in Embodiment 1 of the dynamic evaluation method for the safety of steel structure lifting structure based on multi-source monitoring data of the present invention.

[0027] Figure 6 This is a neural network verification fitting accuracy diagram of Embodiment 1 in the dynamic evaluation method for the safety of steel structure lifting structure based on multi-source monitoring data of the present invention;

[0028] Figure 7This is a schematic diagram of the sensor arrangement in Embodiment 1 of the method for dynamic evaluation of the safety of steel structure lifting structure based on multi-source monitoring data of the present invention. In the figure, (a) is a plan view and (b) is an elevation view.

[0029] Figure 8 This is the curve showing the change of the early warning function value during the construction process in Example 1 of the present invention, which is a dynamic evaluation method for the safety of steel structure lifting structures based on multi-source monitoring data. Detailed Implementation

[0030] The present invention will be further described below with reference to the accompanying drawings and specific embodiments.

[0031] Please see the appendix Figure 1 A dynamic assessment method for the safety of steel structure lifting structures based on multi-source monitoring data includes the following steps:

[0032] Step 1: Based on the actual situation of the project, determine the probability distribution model of the uncertainties in the structure during the steel structure lifting construction. According to the investigation results of each factor, the probability distribution model can be a normal distribution or an equal probability distribution function.

[0033] The uncertainties mentioned include: the estimated wind pressure changes of the structure based on the project location and construction season, the estimated structural node coordinates and component cross-sectional deviations considering the accuracy of steel structure fabrication and installation, and the degree of asynchronous lifting of multiple lifting points of the steel structure.

[0034] Step 2: Determine the specific safety criteria and early warning functions for steel structure lifting construction based on the specifications or project requirements.

[0035] Specifically, the early warning functions include: Regarding deformation, the ratio of the overall tilt angle, deflection, and horizontal displacement of the steel structure to a predetermined threshold can be set as α; regarding stress level, the ratio of the overall maximum stress of the structure or the lifting cable force to a predetermined threshold can be set as β. α and β correspond to different early warning function values. During the steel structure lifting process, when the comprehensive early warning function value reaches 1, the steel structure lifting construction enters an early warning state.

[0036] Preferably, for safety reasons, the maximum value of each warning value is taken. That is, when either α or β reaches 1, the comprehensive warning function value is taken as 1, which means that the steel structure lifting construction has entered the warning state. At this time, the emergency plan should be activated and the lifting construction process should be adjusted in a timely manner.

[0037] Step 3: Establish a finite element model based on the drawings, conduct sensitivity analysis, and determine the main control items and auxiliary control items.

[0038] Specifically, Latin hypercube sampling is used to identify uncertainties during the steel structure lifting process. The structural response under various working conditions, i.e., the early warning function value, is calculated using a finite element model, followed by sensitivity analysis. Several key uncertainties that significantly contribute to the structural response are selected as primary control items, while the remaining uncertainties are designated as secondary control items.

[0039] Step 4: Use the main control items and their corresponding early warning function values ​​from Step 3 as the dataset for model training, where the early warning function values ​​serve as the output values ​​of the multi-input data.

[0040] Step 5: Split the model's dataset into a training set, a validation set, and a test set.

[0041] Preferably, referring to general experience in neural network training, 70% of the dataset can be initially used as the training dataset, 20% as the validation set, and the remaining 10% as the test set.

[0042] Step 6: Use machine learning algorithms to train the model and obtain the evaluation model.

[0043] Taking multi-layer neural network training as an example, the key uncertainties in the training set are temperature changes, wind load conditions, and the degree of asynchronous suspension points. The output early warning function value corresponds to the structural response index. Hyperparameters such as the number of hidden layers are adjusted to train the model, and the model is tested through a test set.

[0044] Step 7: Install the corresponding sensors on the steel structure according to the main control items and auxiliary control items.

[0045] Specifically, sensors used for the main control project must be set up to capture the input factors and real-time changes of parameters of key uncertainties in the steel structure during the lifting process, forming multi-source monitoring data; sensors used for the auxiliary control project are not set up or can be selectively set up according to the actual working conditions to capture the input factors and real-time changes of parameters of non-critical uncertainties in the steel structure during the lifting process.

[0046] Preferably, with the goals of stress monitoring of key structural components, overall attitude description, and environmental perception, various sensors are deployed on the pre-lifting structure, such as static levels at the lifting points; a cable gauge is installed on one cable at each lifting point; wind speed is measured by anemometers installed at the top, middle, and bottom of the steel corridor; stress monitoring is performed by vibrating wire strain gauges installed on key structural components, which also serve as temperature monitoring at the corresponding locations; and tilt angle is tested by inclinometers, which can be placed in the middle of the structure.

[0047] Data from various sensors are input into the evaluation model, which then outputs a warning function value as a predictive structural response index for rapid safety assessment during construction.

[0048] Step 8: Set the warning threshold and formulate an emergency plan. When the warning function value reaches the warning threshold, execute the emergency plan; otherwise, continue to observe the warning function value and record its real-time changes.

[0049] Please see the appendix Figure 1 Example 1: Taking the overall lifting construction process of a super-heavy steel connecting corridor as an example, its dynamic safety assessment method includes the following steps:

[0050] Step 1: Based on the actual situation of the project, determine the probability distribution model of the uncertain factors of the structure during the steel structure lifting construction.

[0051] The probability distribution model can be a normal distribution or other probability distribution function, such as the average wind pressure value, standard deviation and corresponding probability distribution function (such as normal distribution) of wind load. Similarly, statistical analysis is performed on other loads that need to be considered. The resistance model of the steel structure itself during the construction of the steel corridor is determined, such as the average value, standard deviation and corresponding probability distribution function of the component section error. Similarly, statistical analysis is performed on other structural parameters that need to be considered.

[0052] The load probability model function is shown in equation (1):

[0053]

[0054] If no actual measurement data is available, refer to the data or specifications that are consistent with the actual situation of the project; similar to the load probability model, a probability distribution model of factors such as structural node deviation and cross-sectional dimension deviation is calculated based on on-site measured data.

[0055] Step 2: Determine the specific safety guidelines and early warning functions for the steel connecting corridor lifting construction based on the specifications or project requirements.

[0056] Specifically, safety factors are set for stress levels and deformation degrees. Regarding deformation: α can be set as the ratio of the overall tilt angle, deflection, and horizontal displacement of the steel structure to its predetermined threshold. Regarding stress levels: β can be set as the ratio of the maximum overall stress of the structure or the lifting cable force to its predetermined threshold. When either α or β reaches 1, the comprehensive early warning function value is set to 1, indicating that the steel corridor lifting construction has entered an early warning state. At this time, the emergency plan should be activated to adjust the lifting construction process in a timely manner.

[0057] Step 3: Based on the design drawings and hoisting process, use Midas software to build a finite element model of the steel connecting corridor lifting, conduct sensitivity analysis, and determine the main control items and auxiliary control items.

[0058] Please see the appendix Figure 2Latin hypercube sampling was used to investigate the uncertainties in the steel structure lifting process. Different levels of each uncertainty were assigned according to a probability distribution model. These values ​​were then substituted into the finite element model to obtain the calculation results and early warning function values, as shown in the attached figure. Figure 3 and attached Figure 4 As shown; based on the sensitivity analysis results, three key uncertainty factors that contribute significantly to the structural response were selected: the degree of vertical asynchrony of the suspension points, temperature change, and stress change of the members. Sensors corresponding to the key uncertainty factors were selected as the main control items, while the remaining uncertainty factors and sensors were used as auxiliary control items.

[0059] Step 4: Use the main control items and their corresponding early warning function values ​​from Step 3 as the dataset for model training, where the early warning function values ​​serve as the output values ​​of the multi-input data.

[0060] Please see the appendix Figure 5 Step 5: Split the model's dataset, using 70% as the training set, 20% as the validation set, and the remaining 10% as the test set.

[0061] Step 6: Use a multi-layer neural network machine learning algorithm to train the model on the training set, validate it on the validation set and adjust hyperparameters such as the number of hidden layers, train the neural network model, and test it on the test set to obtain the evaluation model.

[0062] Specifically, the temperature value during construction, the vertical asynchronous displacement, and the maximum stress of the monitored members are used as the input layer variables of the neural network. First, these uncertain factors are normalized. The level of each uncertain factor can be transformed into the interval [-1,1] according to Equation (2), which is used as the input for neural network training. Then, the warning function value is used as the output value. Parameters such as the maximum number of training rounds, training target error, and learning rate are set to optimize the accuracy of the final prediction result of the neural network model and verify it through the test set, while avoiding the occurrence of overfitting or underfitting.

[0063] It is worth noting that, considering the uncertainties of actual lifting conditions, the monitored members may not include those experiencing maximum stress during the actual lifting process. Therefore, to correspond to the actual construction process, the members monitored for each condition in the training model should remain consistent, using the maximum stress of these members as the input layer variable; the maximum stress occurring in the structure should always be used as the source for calculating the warning function value, rather than the maximum stress value of the monitored members. When the maximum stress value of a member appears in a group of monitored members, they are counted separately without affecting each other.

[0064]

[0065] In the formula, x represents the original data, x0 represents the normalized data, and x...max x min These are the maximum and minimum values ​​of the data before normalization.

[0066] Please see the appendix Figure 6 The model was validated in 19 rounds, with the 13th round having the best performance, having a mean squared error of 0.0049797.

[0067] The training and validation of the evaluation model can be carried out using mathematical modeling software such as MATLAB. The training of neural network models in machine learning is based on existing algorithms, and its detailed process will not be elaborated here.

[0068] Step 7: Install the corresponding sensors on the steel structure according to the main control items and auxiliary control items.

[0069] Based on the sensitivity analysis results, with the goals of stress monitoring of key structural components, overall attitude description, and environmental perception, various sensors were deployed on the pre-lifting structure, as shown in the attached figure. Figure 7 As shown, three key factors are the main control items: four static levels are set up at four lifting points, marked as J1 to J4, to control the attitude (i.e., the degree of vertical asynchrony of the lifting points); stress monitoring consists of n vibrating wire strain gauges arranged at 1 / 2 length of the key structural members, marked as YB1 to YBn, which are also used for temperature monitoring (stress level and temperature) at the corresponding locations.

[0070] Additional auxiliary control items can be selected as follows: use 8 steel cable gauges at the lifting points for cable tension control, corresponding to GS1 to GS8. Select 2 hoists at each lifting point, and select one cable for each hoist for testing; use 3 anemometers for wind speed monitoring, corresponding to FS1 to FS3. FS1 is located at the top of the steel corridor to test the horizontal wind speed, FS2 is located at the bottom of the steel corridor to test the vertical wind speed, and FS3 is located at the bottom of the steel corridor to test the horizontal wind speed.

[0071] Data from various sensors are input into the evaluation model, which then outputs a warning function value as a predictive structural response index for rapid safety assessment during construction.

[0072] Step 8: Set early warning thresholds and formulate emergency plans. Sensors are deployed to capture real-time changes in various input parameters during the lifting process of the steel connecting corridor. This data is then fed into a trained model. The model is evaluated to predict structural response indicators, and early warning function values ​​are calculated in real time for rapid safety assessment of the construction process.

[0073] When the warning function value reaches the warning threshold, the emergency plan is executed; otherwise, the warning function value is continuously monitored and its real-time changes are recorded, as shown in the attached figure. Figure 8 As shown.

[0074] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the invention. Therefore, any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A dynamic safety assessment method for steel structure lifting structures based on multi-source monitoring data, characterized by: Includes the following steps: Step 1: Based on the actual situation of the project, determine the probability distribution model of the structural uncertainties during the steel structure lifting construction; Step 2: Determine the specific safety criteria and early warning functions for steel structure lifting construction based on the specifications or project requirements; Step 3: Establish a finite element model based on the drawings, conduct sensitivity analysis, and determine the main control items and auxiliary control items; Step 4: Use the main control items and their corresponding early warning function values ​​from Step 3 as the dataset for model training, where the early warning function values ​​serve as the output values ​​of the multi-input data; Step 5: Split the model's dataset into a training set, a validation set, and a test set; Step 6: Train the model using machine learning algorithms to obtain the evaluation model; Step 7: Install corresponding sensors on the steel structure according to the main control items and auxiliary control items; Step 8: Set the warning threshold and formulate an emergency plan. When the warning function value reaches the warning threshold, execute the emergency plan; otherwise, continue to observe the warning function value and record its real-time changes. In step 2, the early warning function includes: in terms of deformation, setting the ratio of the overall tilt angle, deflection, and horizontal displacement deformation of the steel structure to its predetermined threshold value as α; Regarding stress levels: the ratio of the overall maximum stress of the structure or the lifting cable force to its predetermined threshold is set to β; when either α or β reaches 1, the comprehensive early warning function value is set to 1, which means that the steel structure lifting construction has entered the early warning state, and the emergency plan is activated to adjust the lifting construction process in a timely manner. In step 3, Latin hypercube sampling is performed on the uncertainties in the steel structure lifting process. The structural response under each working condition, i.e., the early warning function value, is calculated through the finite element model, and then sensitivity analysis is performed. Several key uncertainties that contribute significantly to the structural response are selected as the main control items, while the remaining uncertainties can be used as auxiliary control items. In step 7, sensors for the main control project must be set up to capture the input factors and real-time changes of parameters of key uncertainties in the steel structure during the lifting process, forming multi-source monitoring data; sensors for the auxiliary control project are not set up or are selectively set up according to the actual working conditions to capture the input factors and real-time changes of parameters of non-critical uncertainties in the steel structure during the lifting process.

2. The method for dynamic safety assessment of steel structure lifting structures based on multi-source monitoring data according to claim 1, characterized in that: In step 1, the uncertainties include: the estimated wind pressure changes of the structure based on the project location and construction season, the estimated structural node coordinates and component section deviations considering the accuracy of steel structure fabrication and installation, and the degree of asynchronous lifting of multiple lifting points of the steel structure.

Images