A digital twin driven method for safety assessment of truss structure in coastal environment

Abstract

The present application provides a kind of digital twin driven truss structure safety evaluation method in coastal environment, and truss is the space truss structure that upper and lower chord continuous and intermediate web is fixed at chord;Evaluation method includes the following steps;Step S1, establish shape material integrated model, i.e., based on the data monitored by multiple source sensors arranged at truss, establish the fine model of truss structure in coastal environment in multiple scales;Step S2, establish reduced order model, i.e., based on the model data of step S1, fill in the missing structure response data in monitoring process;Step S3, establish safety state evaluation model, directly calculate the failure probability of structure in combination with twin data recovered in step S2;Step S4, establish real-time dynamic visualization platform, based on the data of step S1-S3, the structural change characteristics and performance evolution trend of truss are presented in image;The present application can maximize the use of limited monitoring data, intuitively and real-time reflect the change of truss structure, and accurately and quantitatively evaluate the state of truss structure.

Description

Technical Field

[0001] This invention relates to the field of structural safety assessment, and in particular to a digital twin-driven method for assessing the safety of truss structures in a coastal environment. Background Technology

[0002] Coastal environments present a unique challenge for truss structures during service. Compared to other service environments, the atmosphere in these environments contains higher concentrations of CL (Clean Air). - Sea breezes accelerate the propagation of ions such as SO2, NO2, and H2S. Steel is constantly exposed to a cycle of wet and dry conditions, and intense ultraviolet radiation damages the ozone layer, exacerbating corrosion. Furthermore, coastal areas experience abundant sunshine and large diurnal temperature variations. Combined with the shading effect between structural members and differences in their heat absorption and dissipation capacities, this creates a severely uneven temperature field within the structure. In addition, coastal areas experience much higher wind speeds and pressures than ordinary areas, and are also at risk of typhoon landfall. All of these factors contribute to a decrease in the stiffness and strength of structural members, leading to increased stress and displacement, and ultimately, overall failure. Therefore, it is necessary to conduct a safety assessment specifically for truss structures in coastal environments.

[0003] Structural safety assessment is an integral part of Structural Health Monitoring (SHM). Traditional SHM relies solely on cumbersome analytical calculation software and periodic monitoring data, resulting in delayed structural information reception, a lack of systematic and flexible data utilization, and an inability to accurately and intuitively reflect structural changes. Combining Digital Twin (DT) with SHM, along with finite element analysis software, various intelligent algorithms, and 3D dynamic visualization software, can effectively overcome the shortcomings of data inconsistencies between different SHM processes. By establishing a virtual model system, including a structural behavior proxy model and a dynamic display model of structural behavior changes, the characteristics of structural changes and performance evolution trends can be quickly and intuitively perceived, enabling timely assessment of structural safety status and corresponding prognostic decisions.

[0004] Therefore, in order to prevent accident risks, it is necessary to combine the concept of digital twins and conduct safety assessments for truss structures in coastal environments. Summary of the Invention

[0005] This invention proposes a digital twin-driven method for assessing the safety of truss structures in coastal environments. This method maximizes the use of limited monitoring data, intuitively and in real-time reflects changes in the truss structure, and accurately and quantitatively assesses the state of the truss structure.

[0006] A digital twin-driven method for safety assessment of truss structures in coastal environments, wherein the truss is a spatial truss structure with continuous upper and lower chords and intermediate web members fixed at the chords; the assessment method includes the following steps;

[0007] Step S1: Establish an integrated model of the truss structure, that is, based on the data monitored by multi-source sensors located at the truss, establish a multi-scale refined model of the truss structure in the coastal environment;

[0008] Step S2: Establish a reduced-order proxy model, that is, based on the model data in step S1, fill in the missing structural response data during the monitoring process;

[0009] Step S3: Establish a safety status assessment model and directly calculate the failure probability of the structure by combining the twin data retrieved in step S2.

[0010] Step S4: Establish a real-time dynamic visualization platform to vividly present the structural change characteristics and performance evolution trend of the truss based on the data from steps S1 to S3.

[0011] The intermediate web member is welded and fixed to the chord member; the coastal environment is an environment with severe corrosion inducing factors and often has strong sea winds and strong sunlight and high temperature.

[0012] The specific implementation method of step S1 is as follows:

[0013] Step A1: Monitor the ion concentration, temperature, humidity, pH value, wind speed, irradiance value, and stress and strain of the truss structure in the coastal atmospheric environment through multi-source sensors, and transmit the monitoring data using Internet of Things technology.

[0014] Step A2: Based on random field theory, complete the morphological mapping reconstruction of the random field model of corrosion depth and the random distribution model of rust pits, and realize the establishment of the micro-geometric scale model of the truss structure.

[0015] Step A3: Based on the monitoring data and the morphology obtained from 3D scanning, fit a model for predicting the corrosion depth of the web steel along the chord direction, expressed as a formula:

[0016] D i =D(T) i W i H i N i F i Z i Formula 1

[0017] In the formula, D i Let T be the corrosion depth of the i-th web member. i Let W be the corrosion time of the i-th web member. i Let H be the temperature of the i-th web member. i Let N be the humidity of the i-th web member. i Let F be the ion concentration of the i-th web member. i Z represents the wind speed experienced by the i-th web member. iLet be the irradiance value of the i-th web member;

[0018] Step A4: Based on the monitoring data and the morphology obtained from the 3D scan, fit the rust pit prediction model as follows.

[0019] γ i =γ(t) i w i h i n i f i , z i Formula 2

[0020] In the formula, γ i Let t be the diameter-to-depth ratio of the rust pit on the i-th member. i Let w be the corrosion time of the i-th member. i Let h be the temperature of the i-th rod. i Let n be the humidity of the i-th rod. i Let f be the ion concentration of the i-th rod. i Let z be the wind speed experienced by the i-th member. i Let be the irradiance value of the i-th rod;

[0021] Step A5: Based on random field theory, randomly generate the coordinates of rust pit locations on the steel surface;

[0022] Step A6: Based on the material property test, obtain the material property degradation formula of the corroded steel, and then combine it with the steel rust pit depth prediction model formula to determine the time-varying relationship of the steel material properties to establish a physical material scale model, which is used to accurately calculate the yield strength, tensile strength, elastic modulus and elongation of the steel after corrosion.

[0023] Step A7: Combining the data from the physical material scale model and the mesoscopic geometric model of the truss, construct a refined integrated profile model, and iteratively update the model by fitting relevant parameters through implicit functions under the drive of real-time data.

[0024] The specific implementation method of step S2 is as follows:

[0025] Step B1: For environmental factors that may affect truss safety, sample the truss safety in environments with these factors to generate several parameter samples that conform to a uniform distribution. Each parameter sample consists of the magnitude of the structural load, the magnitude of the change in the structural ambient temperature, and the corrosion depth of the members. The structural load includes dead load, live load, and wind load.

[0026] Step B2: Using the multi-scale refined model from step S1, calculate the structural response corresponding to each parameter sample, i.e., the maximum stress and maximum displacement of the members, and build the twin database required for the surrogate model.

[0027] Step B3: Establish a corresponding reduced-order surrogate model for each parameter sample of the missing link response. Select a surrogate model with higher nonlinearity, stronger robustness and stability, and shorter prediction time. Use 85% of the sample size as the training set and reserve 15% as the validation set to verify the accuracy of the surrogate model. The reduced-order surrogate model for each missing response satisfies the following relationship:

[0028] y i =y(x load1 x load2 , ..., x loadn x temp x cor1 x cor2 , ...x corn y1, ...y i-1 y i+1 , ...y n )

[0029] Formula 3;

[0030] In the formula, y i The response to the missing i-th web member is expressed as stress or displacement, x. loadi Let x be the load borne by the i-th web member. temp x represents the magnitude of the environmental temperature change of the structure. cori Let represent the corrosion morphology variation of the i-th web member, where i = 1, 2, ..., n.

[0031] The specific implementation method of step S3 is divided into two steps. The first step is:

[0032] Based on the stress characteristics and sensor layout, a limited number of areas requiring focused attention are selected from the complete structure.

[0033] For areas requiring special attention, resampling is performed to generate several parameter samples where all members are subjected to the same load, temperature rise and fall, and corrosion. The mean absolute stress value of the members is calculated using the multi-scale refined model from step S1, generating new training samples. A reduced-order surrogate model is used to fit this sample set, and a large number of parameter samples conforming to a normal distribution are generated. The trained reduced-order surrogate model is then used to predict the mean absolute stress value of the members corresponding to all samples. Sensitivity analysis is performed on the large number of samples to obtain the sensitivity indices of each influencing factor.

[0034] The sensitivity indices of each order corresponding to the different influencing factors on each region are statistically analyzed, and the influence coefficient k of each region is obtained by weighted averaging.

[0035] The second step in the specific implementation method of step S3 is as follows:

[0036] Use Bayesian networks to build a security assessment network for each region of the structure;

[0037] The specific connection of the first layer of the Bayesian safety assessment network is that the state nodes of each influencing factor are the parent nodes of the stress state nodes and displacement state nodes of a single member, and the state nodes of each influencing factor, the stress state nodes of a single member, and the displacement state nodes of a single member together serve as the parent nodes of the corresponding member safety state nodes.

[0038] In the coastal environment described, each influencing factor has 2 to 4 states. The value range of the influencing factors is determined based on the specific circumstances and the accuracy of the network assessment; specifically:

[0039] The stress state of a member node can be either elastic or inelastic, and is determined by dividing the stress range of the member based on the actual yield strength of the steel used in the member.

[0040] The displacement state nodes of members can be divided into two states: exceeding the limit and not exceeding the limit. This division is based on the provisions of different specifications for the ultimate displacement of members.

[0041] The safety status node of a member has two states: damaged and undamaged.

[0042] Input all the evidence obtained through step S2 into the first layer of the Bayesian completeness evaluation network to obtain the probability of safe state or failure state of each member.

[0043] The second layer of the Bayesian security assessment network consists of the pole safety status nodes of the first layer. The connection method is that there is a topological relationship between the connected poles. The importance ranking of the poles is obtained based on the change of the overall bearing capacity of the area after the removal of a single pole. The parent node is the pole with high importance and the child node is the pole with low importance.

[0044] The specific connection of the third layer of the Bayesian security assessment network is that all nodes of the safety status of poles in the region directly serve as parent nodes pointing to the child nodes of the region's safety status. The nodes of the region's safety status have two states: failed and not failed.

[0045] The overall structural failure probability is expressed by the formula:

[0046]

[0047] In the formula, β j Let k be the failure probability of the j-th region. j Let be the influence coefficient of the j-th region, and m be the number of regions. The corresponding safety level and operation and maintenance recommendations for the structure are given.

[0048] The specific implementation method of step S4 is as follows:

[0049] Step C1: Extract the basic physical geometry information of the structure from the multi-scale refined model in step S1, set the corresponding script, and generate the basic structural model on the real-time dynamic visualization platform.

[0050] Step C2: Write a Socket server program for a real-time dynamic visualization platform to receive data collected from sensors, then call the "rbf_pre" function to load a reduced-order surrogate model for prediction, calculate the failure probability of the structure, and send the results through a Socket connection.

[0051] Step C3: In the real-time dynamic visualization platform, write the corresponding Socket network communication, initialize the Socket connection, and store the server's IP address and port number; create the corresponding instance, specify the address protocol, socket type and communication protocol of the multi-source sensor, pass in the IP address and port number, and call the Socket's "Connect" to establish a connection with the server of the real-time dynamic visualization platform;

[0052] Step C4: After ensuring the sensor is alive and operating normally, start the Socket program on the server side of the real-time dynamic visualization platform. The platform can dynamically present the physical performance evolution of the structure, response change characteristics, overall service status degradation degree, and related maintenance suggestions in real time.

[0053] This invention discloses a digital twin-driven method for safety assessment of truss structures in coastal environments, comprising a high-fidelity twin, a reduced-order surrogate model, a safety status assessment model, and a real-time dynamic visualization platform. The integrated truss model can construct a refined structural model based on multi-source sensor data. The reduced-order surrogate model can quickly and accurately predict the response of missing members simultaneously subjected to corrosion, temperature, and load. The safety status assessment model can accurately quantify the structural service status. The real-time dynamic visualization platform can intuitively and vividly display structural changes and performance based on transmitted data. This invention overcomes the shortcomings of traditional structural safety assessments, which lack precision, intuitiveness, and timeliness, achieving efficient assessment combining qualitative and quantitative methods.

[0054] Compared with the prior art, the beneficial effects of the present invention are:

[0055] Combined with the concept of digital twins, a complete system for safety assessment of truss structures in the special service environment of coastal areas has been formed. Multi-source sensors monitor the corrosion status of truss structures in coastal environments, and data-driven high-fidelity dynamic evolution of structural mechanical properties is achieved. A reduced-order proxy model fully utilizes monitoring data to fit many unmonitored structural responses. Based on all available information, the safety status of the structure is accurately quantified, addressing the owners' most pressing concerns. Finally, the changes in the structural entity and corresponding performance information are presented in real time on the client side, making maintenance and operation more visual, intuitive, and timely. Attached Figure Description

[0056] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0057] Appendix Figure 1 This is a schematic diagram of the framework of the present invention;

[0058] Appendix Figure 2 This is a schematic diagram of the specific process of the present invention;

[0059] Appendix Figure 3 This is a schematic diagram illustrating an example of a truss structure according to an embodiment of the present invention;

[0060] Appendix Figure 4 This is a schematic diagram of a Bayesian network example according to an embodiment of the present invention;

[0061] In the diagram: A - upper chord of the truss structure to be evaluated; B, D - web members of the truss structure to be evaluated; C, E - lower chord of the truss structure to be evaluated.

[0062] ① Load state node of member A; ② Temperature state node of member A; ③ Corrosion state node of member A; ④ Stress state node of member A; ⑤ Displacement state node of member A; ⑥ Safety state node of member A; ⑦ Load state node of member B; ⑧ Temperature state node of member B; ⑨ Corrosion state node of member B; (⑩0 Stress state node of member B; -Displacement state node of member B; -Safety status node of member B; -The temperature state node of member C; - Corrosion status node of member C; -Stress state node of member C; -Displacement state node of member C; -Safety state node of member C; -The temperature state node of member D; - Corrosion status node of member D; -Stress state node of member D; -Displacement state node of member D; -Safety state node of member D; -The temperature state node of member E; - Corrosion status node of member E; -Stress state node of member E; -Displacement state node of member E; -Safety status node of member E; -Regional security status node. Detailed Implementation

[0063] As shown in the figure, a digital twin-driven method for safety assessment of truss structures in a coastal environment is described. The truss is a spatial truss structure with continuous upper and lower chords and intermediate web members fixed at the chords. The assessment method includes the following steps.

[0064] Step S1: Establish an integrated model of the truss structure, that is, based on the data monitored by multi-source sensors located at the truss, establish a multi-scale refined model of the truss structure in the coastal environment;

[0065] Step S2: Establish a reduced-order proxy model, that is, based on the model data in step S1, fill in the missing structural response data during the monitoring process;

[0066] Step S3: Establish a safety status assessment model and directly calculate the failure probability of the structure by combining the twin data retrieved in step S2.

[0067] Step S4: Establish a real-time dynamic visualization platform to vividly present the structural change characteristics and performance evolution trend of the truss based on the data from steps S1 to S3.

[0068] The intermediate web member is welded and fixed to the chord member; the coastal environment is an environment with severe corrosion inducing factors and often has strong sea winds and strong sunlight and high temperature.

[0069] The specific implementation method of step S1 is as follows:

[0070] Step A1: Monitor the ion concentration, temperature, humidity, pH value, wind speed, radiation value, and stress and strain of the truss structure in the coastal atmospheric environment through multi-source sensors, and transmit the monitoring data using Internet of Things technology.

[0071] Step A2: Based on random field theory, complete the morphological mapping reconstruction of the random field model of corrosion depth and the random distribution model of rust pits, and realize the establishment of the micro-geometric scale model of the truss structure.

[0072] Step A3: Based on the monitoring data and the morphology obtained from 3D scanning, fit a model for predicting the corrosion depth of the web steel along the chord direction, expressed as a formula:

[0073] D i =D(T) i W i H i N i F i Z i Formula 1

[0074] In the formula, D i Let T be the corrosion depth of the i-th web member. i Let W be the corrosion time of the i-th web member. i Let H be the temperature of the i-th web member. i Let N be the humidity of the i-th web member. i Let F be the ion concentration of the i-th web member. i Z represents the wind speed experienced by the i-th web member. i Let be the irradiance value of the i-th web member;

[0075] Step A4: Based on the monitoring data and the morphology obtained from the 3D scan, fit the rust pit prediction model as follows.

[0076] γ i =γ(t) i w i h i n i f i , z i Formula 2

[0077] In the formula, γ i Let t be the diameter-to-depth ratio of the rust pit on the i-th member. i Let w be the corrosion time of the i-th member. i Let h be the temperature of the i-th rod. i Let n be the humidity of the i-th rod. i Let f be the ion concentration of the i-th rod. i Let z be the wind speed experienced by the i-th member. i Let be the irradiance value of the i-th rod;

[0078] Step A5: Based on random field theory, randomly generate the coordinates of rust pit locations on the steel surface;

[0079] Step A6: Based on the material property test, obtain the material property degradation formula of the corroded steel, and then combine it with the steel rust pit depth prediction model formula to determine the time-varying relationship of the steel material properties to establish a physical material scale model, which is used to accurately calculate the yield strength, tensile strength, elastic modulus and elongation of the steel after corrosion.

[0080] Step A7: Combining the data from the physical material scale model and the mesoscopic geometric model of the truss, construct a refined integrated profile model, and iteratively update the model by fitting relevant parameters through implicit functions under the drive of real-time data.

[0081] The specific implementation method of step S2 is as follows:

[0082] Step B1: For environmental factors that may affect truss safety, sample the truss safety in environments with these factors to generate several parameter samples that conform to a uniform distribution. Each parameter sample consists of the magnitude of the structural load, the magnitude of the change in the structural ambient temperature, and the corrosion depth of the members. The structural load includes dead load, live load, and wind load.

[0083] Step B2: Using the multi-scale refined model from step S1, calculate the structural response corresponding to each parameter sample, i.e., the maximum stress and maximum displacement of the members, and build the twin database required for the surrogate model.

[0084] Step B3: Establish a corresponding reduced-order surrogate model for each parameter sample of the missing link response. Select a surrogate model with higher nonlinearity, stronger robustness and stability, and shorter prediction time. Use 85% of the sample size as the training set and reserve 15% as the validation set to verify the accuracy of the surrogate model. The reduced-order surrogate model for each missing response satisfies the following relationship:

[0085] y i =y(x load1 ,x load2 ,...,x loadn ,x temp ,x cor1 ,x cor2 ,...x corn ,y1,...y i-1 ,y i+1 ,...y n )

[0086] Formula 3;

[0087] In the formula, y i The response to the missing i-th web member is expressed as stress or displacement, x. loadi Let x be the load borne by the i-th web member. temp x represents the magnitude of the environmental temperature change of the structure. cori Let represent the corrosion morphology variation of the i-th web member, where i = 1, 2, ..., n.

[0088] The specific implementation method of step S3 is divided into two steps. The first step is:

[0089] Based on the stress characteristics and sensor layout, a limited number of areas requiring focused attention are selected from the complete structure.

[0090] For areas requiring special attention, resampling is performed to generate several parameter samples where all members are subjected to the same load, temperature rise and fall, and corrosion. The mean absolute stress value of the members is calculated using the multi-scale refined model from step S1, generating new training samples. A reduced-order surrogate model is used to fit this sample set, and a large number of parameter samples conforming to a normal distribution are generated. The trained reduced-order surrogate model is then used to predict the mean absolute stress value of the members corresponding to all samples. Sensitivity analysis is performed on the large number of samples to obtain the sensitivity indices of each influencing factor.

[0091] The sensitivity indices of each order corresponding to the different influencing factors on each region are statistically analyzed, and the influence coefficient k of each region is obtained by weighted averaging.

[0092] The second step in the specific implementation method of step S3 is as follows:

[0093] Use Bayesian networks to build a security assessment network for each region of the structure;

[0094] The specific connection of the first layer of the Bayesian safety assessment network is that the state nodes of each influencing factor are the parent nodes of the stress state nodes and displacement state nodes of a single member, and the state nodes of each influencing factor, the stress state nodes of a single member, and the displacement state nodes of a single member together serve as the parent nodes of the corresponding member safety state nodes.

[0095] In the coastal environment described, each influencing factor has 2 to 4 states. The value range of the influencing factors is determined based on the specific circumstances and the accuracy of the network assessment; specifically:

[0096] The stress state of a member node can be either elastic or inelastic, and is determined by dividing the stress range of the member based on the actual yield strength of the steel used in the member.

[0097] The displacement state nodes of members can be divided into two states: exceeding the limit and not exceeding the limit. This division is based on the provisions of different specifications for the ultimate displacement of members.

[0098] The safety status node of a member has two states: damaged and undamaged.

[0099] Input all the evidence obtained through step S2 into the first layer of the Bayesian completeness evaluation network to obtain the probability of safe state or failure state of each member.

[0100] The second layer of the Bayesian security assessment network consists of the pole safety status nodes of the first layer. The connection method is that there is a topological relationship between the connected poles. The importance ranking of the poles is obtained based on the change of the overall bearing capacity of the area after the removal of a single pole. The parent node is the pole with high importance and the child node is the pole with low importance.

[0101] The specific connection of the third layer of the Bayesian security assessment network is that all nodes of the safety status of poles in the region directly serve as parent nodes pointing to the child nodes of the region's safety status. The nodes of the region's safety status have two states: failed and not failed.

[0102] The overall structural failure probability is expressed by the formula:

[0103]

[0104] In the formula, β j Let k be the failure probability of the j-th region. j Let be the influence coefficient of the j-th region, and m be the number of regions. The corresponding safety level and operation and maintenance recommendations for the structure are given.

[0105] The specific implementation method of step S4 is as follows:

[0106] Step C1: Extract the basic physical geometry information of the structure from the multi-scale refined model in step S1, set the corresponding script, and generate the basic structural model on the real-time dynamic visualization platform.

[0107] Step C2: Write a Socket server program for a real-time dynamic visualization platform to receive data collected from sensors, then call the "rbf_pre" function to load a reduced-order surrogate model for prediction, calculate the failure probability of the structure, and send the results through a Socket connection.

[0108] Step C3: In the real-time dynamic visualization platform, write the corresponding Socket network communication, initialize the Socket connection, and store the server's IP address and port number; create the corresponding instance, specify the address protocol, socket type and communication protocol of the multi-source sensor, pass in the IP address and port number, and call the Socket's "Connect" to establish a connection with the server of the real-time dynamic visualization platform;

[0109] Step C4: After ensuring the sensor is alive and operating normally, start the Socket program on the server side of the real-time dynamic visualization platform. The platform can dynamically present the physical performance evolution of the structure, response change characteristics, overall service status degradation degree, and related maintenance suggestions in real time.

[0110] Example:

[0111] In this example, the specific implementation of S3 is in two steps. The first step is:

[0112] Based on the stress characteristics of the truss and the arrangement of sensors, a limited number of areas requiring key attention are selected from the complete truss structure.

[0113] Resample the areas that require special attention to generate 300 parameter samples of all members subjected to the same load, temperature rise and fall, and corrosion. Then, use the multi-scale refinement model in step S1 to calculate the mean value of the absolute value of the stress in the members and generate new learning samples.

[0114] The sample set was fitted using a reduced-order surrogate model. 10,000 parameter samples that conform to a normal distribution were generated using Latin hypercube sampling. The mean absolute value of the stress of the members corresponding to all samples was predicted using the trained reduced-order surrogate model.

[0115] SOBOL sensitivity analysis was performed on 10,000 well-organized samples to obtain the SOBOL indices of each order for the structural influencing factors.

[0116] The SOBOL indices of each order corresponding to the different influencing factors affecting each region are statistically analyzed, and the weighted average is used to obtain the influence coefficient k of each region of the structure.

[0117] The specific implementation of S3 is in two steps, the second step being:

[0118] Use Bayesian networks to build a security assessment network for each region of the structure, such as... Figure 4 As shown;

[0119] The first layer of the Bayesian safety assessment network represents the relationship between various influencing factors, stress in individual members, displacement in individual members, and the safety state of individual members.

[0120] like Figure 3 As shown, a key area of ​​concern in the structure includes members A, B, C, D, and E. Members A and B are subjected to load, temperature, and corrosion, while members C, D, and E are subjected to temperature and corrosion.

[0121] like Figure 4 As shown, the load state node ①, temperature state node ②, and corrosion state node ③ of member A are the parent nodes of the stress state node ④ and displacement state node ⑤ of member A. Nodes ①, ②, ③, ④, and ⑤ are all the parent nodes of the safety state node ⑥ of member A; the temperature state node of member C... Corrosion level status node The stress state node of member C Displacement state node The parent node, the node All are safety state nodes of member C. The parent node;

[0122] The load-action state node, temperature-action state node, and corrosion degree state node each have two states: exceeding the limit and not exceeding the limit. The dividing index is 80% of the extreme value of the sampling. The stress state node of the member has two states: elastic and inelastic. The dividing index is the yield strength of the steel of the member. The displacement state node of the member has two states: exceeding the limit and not exceeding the limit. The dividing index is the allowable displacement value of the member specified in the "Steel Structure Design Standard" GB50017-2017. The safety state node of the member has two states: failure and non-failure. The dividing basis is whether the member has fractured or failed.

[0123] Input all the evidence obtained through step S2 into the first layer of the Bayesian completeness evaluation network to obtain the probability of safe state or failure state of each member.

[0124] The second layer of the Bayesian safety assessment network represents the relationships between members within a single region. It consists of the member safety state nodes mentioned in the first layer. The connection method is such that there is a topological relationship between connected members, with parent nodes representing members with higher risk and child nodes representing members with lower risk. For example... Figure 4 As shown, member A is connected to members B and D, member B is connected to members C and D, and member D is connected to member E.

[0125] Based on the change in the overall bearing capacity of the area after removing individual members, the importance of the members is ranked, and the second-layer network is obtained by connecting them, such as... Figure 4 As shown, the order of the links from the parent node to the child node is link A, link B, link D, link C, and link E;

[0126] The specific connections of the third layer of the Bayesian security assessment network are as follows: Figure 4 As shown, node ⑥ represents the safety status of all members within the area. The parent node points directly to the child node whose region is safe. Regional security nodes have two states: failed and not failed.

[0127] The overall structural failure probability is

[0128]

[0129] In the formula, β j Let k be the failure probability of the j-th region. j Let be the impact coefficient of the j-th region, and m be the number of regions. The corresponding structural safety level and operation and maintenance recommendations are as follows:

[0130]

[0131] In this example, the sensor is located on or at the location of the truss.

Claims

1. A digital twin-driven method for safety assessment of truss structures in coastal environments, characterized in that: The truss is a spatial truss structure with continuous upper and lower chords and intermediate web members fixed to the chords; the evaluation method includes the following steps; Step S1: Establish an integrated model of the truss structure, that is, based on the data monitored by multi-source sensors located at the truss, establish a multi-scale refined model of the truss structure in the coastal environment; Step S2: Establish a reduced-order proxy model, that is, based on the model data in step S1, fill in the missing structural response data during the monitoring process; Step S3: Establish a safety status assessment model and directly calculate the failure probability of the structure by combining the twin data retrieved in step S2. Step S4: Establish a real-time dynamic visualization platform to vividly present the structural change characteristics and performance evolution trend of the truss based on the data from steps S1 to S3. The specific implementation method of step S1 is as follows: Step A1: Monitor the ion concentration, temperature and humidity, pH value, wind speed, radiation value and stress and strain of the truss structure in the coastal atmospheric environment through multi-source sensors, and transmit the monitoring data using Internet of Things technology; Step A2: Based on random field theory, complete the morphological mapping reconstruction of the random field model of corrosion depth and the random distribution model of rust pits, and realize the establishment of the micro-geometric scale model of the truss structure. Step A3: Based on the monitoring data and the morphology obtained from 3D scanning, fit a model for predicting the corrosion depth of the web steel along the chord direction, expressed as a formula: Formula 1, In the formula, Let be the corrosion depth of the i-th web member. Let be the corrosion time of the i-th web member. Let i be the temperature of the i-th web member. Let the humidity of the i-th web member be . Let be the ion concentration of the i-th web member. Let be the wind speed experienced by the i-th web member. Let be the irradiance value of the i-th web member; Step A4: Based on the monitoring data and the morphology obtained from the 3D scan, fit the rust pit prediction model as follows. Formula 2, In the formula, Let be the diameter-to-depth ratio of the rust pit on the i-th member. Let i be the corrosion time of the i-th rod. Let i be the temperature of the i-th rod. Let the humidity of the i-th rod be , The ion concentration of the i-th rod Let be the wind speed experienced by the i-th member. Let be the irradiance value of the i-th rod; Step A5: Based on random field theory, randomly generate the coordinates of rust pit locations on the steel surface; Step A6: Based on the material property test, obtain the material property degradation formula of the corroded steel, and then combine it with the steel rust pit depth prediction model formula to determine the time-varying relationship of the steel material properties to establish a physical material scale model, which is used to accurately calculate the yield strength, tensile strength, elastic modulus and elongation of the steel after corrosion. Step A7: Combine the data from the physical material scale model and the mesoscopic geometric model of the truss to construct a refined integrated profile model, and iteratively update the model by fitting relevant parameters through implicit functions under the drive of real-time data; The specific implementation method of step S2 is as follows: Step B1: For environmental factors that may affect truss safety, sample the truss safety in environments with these factors to generate several parameter samples that conform to a uniform distribution. Each parameter sample consists of the magnitude of the structural load, the magnitude of the change in the structural ambient temperature, and the corrosion depth of the members. The structural load includes dead load, live load, and wind load. Step B2: Using the multi-scale refined model from step S1, calculate the structural response corresponding to each parameter sample, i.e., the maximum stress and maximum displacement of the members, and build the twin database required for the surrogate model. Step B3: Establish a corresponding reduced-order surrogate model for the parameter samples of each missing link response. Select a surrogate model with higher nonlinearity, stronger robustness and stability, and shorter prediction time. Use 85% of the sample size as the training set and reserve 15% of the sample size as the validation set to verify the accuracy of the surrogate model. The reduced-order surrogate model for each missing response satisfies the following relation: Formula 3; In the formula, Let represent the response when the i-th web member is missing, and the response can be either stress or displacement. Let be the load borne by the i-th web member. The magnitude of the environmental temperature change of the structure. The corrosion morphology of the i-th web member is characterized by its variation. .

2. The method for assessing the safety of truss structures in a coastal environment driven by digital twins according to claim 1, characterized in that: The intermediate web member is welded and fixed to the chord member; the coastal environment is an environment with severe corrosion inducing factors and often has strong sea winds and strong sunlight and high temperature.

3. The method for assessing the safety of truss structures in a coastal environment driven by digital twins according to claim 1, characterized in that: The specific implementation method of step S3 is divided into two steps. The first step is: Based on the stress characteristics and sensor layout, a limited number of areas requiring focused attention are selected from the complete structure. For areas requiring special attention, resampling is performed to generate several parameter samples where all members are subjected to the same load, temperature rise and fall, and corrosion. The mean absolute stress value of the members is calculated using the multi-scale refined model from step S1, generating new training samples. A reduced-order surrogate model is used to fit this sample set, and a large number of parameter samples conforming to a normal distribution are generated. The trained reduced-order surrogate model is then used to predict the mean absolute stress value of the members corresponding to all samples. Sensitivity analysis is performed on the large number of samples to obtain the sensitivity indices of each influencing factor. The sensitivity indices of each order corresponding to the different influencing factors on each region are statistically analyzed, and the influence coefficient k of each region is obtained by weighted averaging.

4. The method for assessing the safety of truss structures in a coastal environment driven by digital twins according to claim 3, characterized in that: The second step in the specific implementation method of step S3 is as follows: Use Bayesian networks to build a security assessment network for each region of the structure; The specific connection of the first layer of the Bayesian safety assessment network is that the state nodes of each influencing factor are the parent nodes of the stress state nodes and displacement state nodes of a single member, and the state nodes of each influencing factor, the stress state nodes of a single member, and the displacement state nodes of a single member together serve as the parent nodes of the corresponding member safety state nodes. In the coastal environment described, each influencing factor has 2 to 4 states. The value range of the influencing factors is determined based on the specific circumstances and the accuracy of the network assessment; specifically: The stress state of a member node can be either elastic or inelastic, and is determined by dividing the stress range of the member based on the actual yield strength of the steel used in the member. The displacement state nodes of members can be divided into two states: exceeding the limit and not exceeding the limit. This division is based on the provisions of different specifications for the ultimate displacement of members. The safety status node of a member has two states: damaged and undamaged. Input all the evidence obtained through step S2 into the first layer of the Bayesian completeness evaluation network to obtain the probability of safe state or failure state of each member. The second layer of the Bayesian security assessment network consists of the pole safety status nodes of the first layer. The connection method is that there is a topological relationship between the connected poles. The importance ranking of the poles is obtained based on the change of the overall bearing capacity of the area after the removal of a single pole. The parent node is the pole with high importance and the child node is the pole with low importance. The specific connection of the third layer of the Bayesian security assessment network is that all nodes of the safety status of poles in the region directly serve as parent nodes pointing to the child nodes of the region's safety status. The nodes of the region's safety status have two states: failed and not failed. The overall structural failure probability is expressed by the formula: Formula 4; In the formula, Let j be the failure probability of the j-th region. Let be the influence coefficient of the j-th region, and m be the number of regions. The corresponding safety level and operation and maintenance recommendations for the structure are given.

5. A digital twin-driven method for safety assessment of truss structures in coastal environments according to any one of claims 2, 3, and 4, characterized in that: The specific implementation method of step S4 is as follows: Step C1: Extract the basic physical geometry information of the structure from the multi-scale refined model in step S1, set the corresponding script, and generate the basic structural model on the real-time dynamic visualization platform. Step C2: Write a Socket server program for a real-time dynamic visualization platform to receive data collected from sensors, then call the "rbf_pre" function to load a reduced-order surrogate model for prediction, calculate the failure probability of the structure, and send the results through a Socket connection. Step C3: In the real-time dynamic visualization platform, write the corresponding Socket network communication, initialize the Socket connection, and store the server's IP address and port number; create the corresponding instance, specify the address protocol, socket type and communication protocol of the multi-source sensor, pass in the IP address and port number, and call the Socket's "Connect" to establish a connection with the server of the real-time dynamic visualization platform; Step C4: After ensuring the sensor is alive and operating normally, start the Socket program on the server side of the real-time dynamic visualization platform. The platform can dynamically present the physical performance evolution of the structure, response change characteristics, overall service status degradation degree, and related maintenance suggestions in real time.

Images