Bridge structure crack resistance optimization method and system based on finite element analysis

Abstract

The invention provides a bridge structure anti-cracking performance optimization method and system based on finite element analysis, and relates to the technical field of finite element analysis and bridge structure engineering, and the method comprises the steps: detecting an optimization auxiliary demand generated by a user in a bridge structure anti-cracking performance optimization process based on the finite element analysis; matching an initial assistance capability required for satisfying an optimization assistance demand; evaluating an auxiliary effect corresponding to the auxiliary capability to be selected of the initial auxiliary capability under different reduction strategies and a user dependence possibility; taking the auxiliary effect and the to-be-selected auxiliary capability of which the user dependence possibility reaches the optimal balance at present as a target auxiliary capability; and performing assistance on the user according to the target assistance capability. By intelligently detecting the requirement of a user in bridge structure anti-cracking performance optimization, the auxiliary capability is matched and dynamically adjusted, the optimization efficiency and precision of finite element analysis are remarkably improved, and meanwhile, the auxiliary effect and user independence are effectively balanced.

Description

Technical Field

[0001] This invention relates to the fields of finite element analysis and bridge structural engineering technology, and in particular to a method and system for optimizing the crack resistance performance of bridge structures based on finite element analysis. Background Technology

[0002] Currently, bridges, as a crucial component of transportation infrastructure, have their structural crack resistance performance directly impacting their safety, durability, and service life. Traditional bridge structural design and optimization largely rely on empirical formulas and static analysis methods, making it difficult to comprehensively consider the influence of complex loads, material nonlinearity, and environmental factors on crack development. In recent years, finite element analysis (FEA) technology has been widely used for crack resistance optimization due to its ability to accurately simulate the mechanical behavior of bridge structures. However, users often face challenges in practical applications of FEA, such as complex modeling, difficulty in parameter selection, and unclear optimization objectives, leading to low optimization efficiency or unsatisfactory results. Furthermore, existing optimization aids are mostly general-purpose, lacking customized support for bridge crack resistance optimization, making it difficult to dynamically adapt to changes in user needs or balance the aid's effectiveness with user independence.

[0003] Therefore, there is an urgent need for an optimization method that can intelligently detect user needs, match personalized assistance capabilities, and achieve the optimal balance between assistance effect and user dependence, so as to improve the efficiency and accuracy of bridge structure crack resistance optimization, while reducing users' excessive dependence on tools. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for optimizing the crack resistance performance of bridge structures based on finite element analysis, so as to solve the problems pointed out in the background art.

[0005] In a first aspect, the bridge structure crack resistance optimization method based on finite element analysis provided in the embodiments of the present invention includes: S1. Detect the optimization auxiliary needs generated by users during the process of optimizing the crack resistance performance of bridge structures based on finite element analysis; S2. Match the initial auxiliary capabilities required to meet the optimization auxiliary needs; S3. Evaluate the auxiliary effects and user dependency potential of the candidate auxiliary capabilities under different reduction strategies for the initial auxiliary capabilities; S4. Select the auxiliary capability that achieves the optimal balance between auxiliary effect and user dependency probability at present as the target auxiliary capability; S5. Provide assistance to the user based on the target assistance capabilities.

[0006] Secondly, the bridge structure crack resistance optimization system based on finite element analysis provided in this embodiment of the invention includes: The detection module is used to detect the optimization assistance needs generated by users during the process of optimizing the crack resistance performance of bridge structures based on finite element analysis. The matching module is used to match the initial assistance capabilities required to meet the optimization assistance needs; The evaluation module is used to evaluate the auxiliary effects and user dependency potential of the candidate auxiliary capabilities under different reduction strategies. The selection module is used to select the auxiliary capabilities that achieve the best balance between the auxiliary effect and the user's reliance probability at present as the target auxiliary capability. The assistance module is used to provide assistance to the user based on the target assistance capabilities.

[0007] The present invention has achieved the following beneficial effects: This method intelligently detects the user's needs in optimizing the crack resistance performance of bridge structures, matches and dynamically adjusts auxiliary capabilities, significantly improving the optimization efficiency and accuracy of finite element analysis, while effectively balancing the auxiliary effect with user independence.

[0008] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0009] 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: Figure 1 This is a schematic diagram of a bridge structure crack resistance optimization method based on finite element analysis in an embodiment of the present invention; Figure 2 This is a schematic diagram of a bridge structure crack resistance optimization system based on finite element analysis in an embodiment of the present invention. Detailed Implementation

[0010] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0011] Figure 1 A flowchart of a method for optimizing the crack resistance of bridge structures based on finite element analysis is provided for embodiments of this application, such as... Figure 1 As shown, the method includes: S1. Detect the optimization assistance needs generated by users during the process of optimizing the crack resistance performance of bridge structures based on finite element analysis. Step S1 specifically includes the following sub-steps: S11. Real-time monitoring of user modeling operation sequences and parameter modification behavior in finite element analysis software, identifying user operation patterns of repeatedly adjusting stress parameters in key parts of bridge structure; among which, key parts of bridge structure include at least the mid-span of beam, support connection and cantilever end.

[0012] Modeling operation sequence refers to the sequence of modeling actions performed by the user in finite element analysis software, including geometric modeling, mesh generation, boundary condition setting, and material property definition. Parameter modification behavior refers to the user's adjustment of stress-related parameters in the finite element model, such as beam stiffness, material strength, or load distribution. Key parts of the bridge structure include the mid-span of the beam (the central region of the main span, bearing the maximum bending moment), the support connection (the connection area between the beam and the supporting structure, prone to stress concentration), and the cantilever end (the free end of the cantilever beam, susceptible to shear force). Operation mode refers to the user's behavior of repeatedly adjusting stress parameters in these key parts, such as repeatedly modifying the material elastic modulus at the mid-span of the beam or the constraint stiffness at the support connection. The specific steps for obtaining the operation sequence and parameter modification behavior are as follows: Capture each user's operation command through the logging interface of the finite element analysis software (such as ANSYS's APDL log or Abaqus' Python script interface), recording the operation time, operation type (e.g., "modify material properties"), and specific parameter values ​​(e.g., adjusting the elastic modulus from 200> GPa to 210> GPa); perform time series analysis on the log data to extract the user's operation frequency and parameter adjustment range at key locations; analyze the operation sequence using pattern recognition algorithms (e.g., behavior clustering algorithms based on k-means clustering) to identify repeated adjustment behavior patterns (e.g., the user repeatedly adjusts the elastic modulus more than 3 times in the middle of the beam span, with an adjustment range greater than 5%). The pattern recognition algorithm is constructed as follows: First, user operation logs are collected, and operation timestamps, operation objects (such as mid-span of a beam), and parameter change values ​​are extracted to construct a time series dataset. Then, the k-means clustering algorithm is used to cluster the operation sequences according to the operation object and adjustment frequency. The number of clusters is set to k=5 (corresponding to different operation modes). The similarity between operation sequences is calculated through Euclidean distance, the cluster centers are iteratively optimized, and finally the user's operation mode is output, such as "high-frequency adjustment of stress parameters at mid-span of the beam".

[0013] Here is a specific implementation example: In a bridge crack resistance optimization project, the user used ANSYS software for finite element modeling, aiming to optimize the crack resistance of a 100-meter span prestressed concrete continuous beam bridge. Through ANSYS's APDL log interface, the system recorded user operations in real time, including geometric modeling (defining the beam's mid-span cross-section dimensions as 2m × 1.5m), mesh generation (setting the global mesh size to 0.1m), and material property settings (initial value of the concrete's elastic modulus was 200 GPa). The log showed that the user repeatedly adjusted the elastic modulus at the mid-span of the beam, from 200 GPa to 205 GPa, 210 GPa, and 208 GPa, making four adjustments with magnitudes ranging from 2.5% to 5%. At the support connections, the user modified the constraint stiffness three times, from 1 × 10^6 N / m to 1.2 × 10^6 N / m. The system analyzes log data using the k-means clustering algorithm, setting the cluster number k=5. It clusters the operation sequences by object (mid-span of the beam, support connection, etc.) and adjustment frequency, calculates the Euclidean distance between operations, and identifies the user's "high-frequency stress parameter adjustment" pattern (adjustment times ≥ 3, amplitude ≥ 2%) at the mid-span of the beam and the "mid-frequency constraint stiffness adjustment" pattern (adjustment times = 3, amplitude ≈ 20%) at the support connection. Finally, the system generates an operation mode report, recording the mid-span of the beam and the support connection as key crack-resistant areas of user concern for subsequent analysis.

[0014] S12. Extract the key parameter adjustment trajectory and simulation restart frequency in the operation mode to construct a user behavior feature vector.

[0015] The key parameter adjustment trajectory refers to the sequence of parameter adjustments (such as elastic modulus and constraint stiffness) made by the user at key locations (such as mid-span of a beam or support connection), including the parameter value and time point of each adjustment. The simulation restart frequency refers to the number of times the user reruns the finite element simulation after parameter adjustments, reflecting the iterative needs of modeling or solving. The steps to obtain the key parameter adjustment trajectory are: extracting parameter modification records corresponding to the operation mode from the finite element analysis software logs, parsing the parameter name (such as elastic modulus), adjustment value (such as from 200> GPa to 210> GPa), and time point (such as an operation time interval of 10 minutes) for each modification; performing time series analysis on the parameter adjustment sequence to generate adjustment trajectory curves (such as the elastic modulus changing over time). The steps to obtain the simulation restart frequency are: counting the number of times the user calls the solver (such as the ANSYS "SOLVE" command) through the log interface, and calculating the number of restarts per unit time (such as per hour). The user behavior feature vector is constructed as follows: the key parameter adjustment trajectory is quantified into feature values ​​(such as the average adjustment amplitude and adjustment frequency), and the simulation restart frequency is quantified into restart times / hour. Combined with the operation weights of key parts (0.4 for mid-span of beam, 0.3 for support connection, and 0.3 for cantilever end), a multi-dimensional vector is constructed. The vector construction algorithm is based on a weighted feature extraction algorithm, specifically: the mean and standard deviation of the parameter adjustment trajectory of each key part are calculated (such as the mean of elastic modulus adjustment 205 > GPa, and the standard deviation 5 > GPa), and the average number of restarts per hour is calculated for the simulation restart frequency (such as 2 times / hour). The vectors are then summed according to their weights to generate a feature vector (such as [205, > 5, > 2, > 0.4, > 0.3, > 0.3]).

[0016] In the aforementioned bridge optimization project, the system extracted the mid-span elastic modulus adjustment trajectory from the ANSYS logs, recording that the user adjusted it 4 times within 1 hour (200> GPa → 205> GPa → 210> GPa → 208> GPa), with a calculated mean of 205.75> GPa and a standard deviation of 4.03> GPa; the constraint stiffness at the support connection was adjusted 3 times (1×10^6> N / m → 1.1×10^6> N / m → 1.2×10^6> N / m), with a mean of 1.1×10^6> N / m and a standard deviation of 0.1×10^6> N / m. The simulation restart frequency was recorded by statistically analyzing the "SOLVE" command calls, showing 3 restarts within 1 hour, a frequency of 3 times / hour. The system uses a weighted feature extraction algorithm, setting weights of 0.4 at mid-span of the beam, 0.3 at support connections, and 0.3 at the cantilever end (the weights are only used for normalization because no adjustments were detected at the cantilever end). Feature values ​​are calculated as follows: mean elastic modulus adjustment 205.75 > GPa, standard deviation 4.03 > GPa; mean constraint stiffness adjustment 1.1 × 10^6 > N / m, standard deviation 0.1 × 10^6 > N / m; restart frequency 3 times / hour. The final generated user behavior feature vector [205.75, > 4.03, > 1.1 × 10^6, > 0.1 × 10^6, > 3, > 0.4, > 0.3, > 0.3] reflects the user's frequent adjustments and high-simulation restart needs at mid-span and support connections, providing a basis for subsequent scenario matching and optimization.

[0017] S13. Match the user behavior feature vector with the pre-set typical crack resistance optimization scenario template to determine the user's weak links and computational bottlenecks in crack resistance performance optimization, and generate optimization auxiliary requirements that include optimization objectives and resource requirements.

[0018] Typical crack resistance optimization scenario templates refer to pre-set bridge crack resistance optimization scenario datasets, containing feature vectors and solutions corresponding to different bridge types (such as continuous beam bridges and suspension bridges), material properties (such as concrete and steel), and optimization objectives (such as reducing the peak stress at mid-span of the beam). The matching process is based on the cosine similarity algorithm, specifically: calculating the cosine similarity between the user behavior feature vector and the template feature vector, and selecting the template with the highest similarity; the template includes weak points (such as "uneven stress distribution at mid-span of the beam") and computational bottlenecks (such as "excessive solution time due to overly fine mesh"). Optimization auxiliary requirements include optimization objectives (such as reducing the peak stress at mid-span of the beam to below 50 MPa) and resource requirements (such as needing 4 parallel computing nodes). The steps to obtain templates are: extracting successful cases from historical bridge crack resistance optimization projects, recording the feature vectors (such as the average value of elastic modulus adjustment, simulation restart frequency) and solutions (such as mesh granularity of 0.1m, convergence threshold of 0.001) for each scenario, and constructing a template library. The cosine similarity algorithm is constructed as follows: standardize the user behavior feature vector and the template feature vector (mean is set to zero, variance is set to one), calculate the dot product of the two vectors and divide by the magnitude, and output the similarity value (e.g., 0.9 indicates a high match); if the similarity is lower than the threshold of 0.7, it is marked as a new scene and recorded.

[0019] For example, in a bridge optimization project, the system loads a template library from the crack resistance optimization knowledge base, containing templates for continuous beam bridges (feature vector: [200> GPa, >5> GPa, >1×10^6> N / m, >0.1×10^6> N / m, >2.5 times / hour, >0.4, >0.3, >0.3], optimization target: peak stress at mid-span of the beam 0.7). Matching results show the user's weak link is "uneven stress distribution at mid-span of the beam" (due to frequent adjustments to the elastic modulus), and the computational bottleneck is "high simulation restart frequency leading to long solution time" (3 times / hour). The system generates optimization assistance requirements: the optimization target is to reduce the peak stress at mid-span of the beam to below 50 MPa, and the resource requirements are 4 parallel computing nodes, a mesh size of 0.1m, and a convergence threshold of 0.001. This requirement reflects the user's need for assistance in optimizing the material parameter settings at mid-span of the beam and improving simulation efficiency, providing resources for subsequent steps.

[0020] S2. Match the initial assistance capabilities required to meet the optimization assistance needs. Step S2 specifically includes the following sub-steps: S21. Based on the optimization objectives in the optimization auxiliary requirements, retrieve the crack-resistant design rule set from the crack-resistant optimization knowledge base that matches the current bridge structure type and material properties.

[0021] Based on the optimization objectives in the optimization assistance requirements (e.g., reducing the peak stress at mid-span of the beam to below 50 MPa), the crack-resistant design rule set matching the bridge structure type (e.g., continuous beam bridge) and material properties (e.g., concrete elastic modulus 200 GPa) is retrieved from the crack-resistant optimization knowledge base to ensure targeted design guidance. The crack-resistant design rule set refers to a pre-set set of crack-resistant optimization rules, including material parameter optimization rules (e.g., concrete elastic modulus range 190-210 GPa), geometric optimization rules (e.g., beam section height adjustment range of 10%), and boundary condition optimization rules (e.g., support stiffness range 0.8-1.2×10^6 N / m). The retrieval process is based on keyword matching and rule priority ranking algorithms, specifically: extracting the bridge type (obtained through user input or model metadata, such as "continuous beam bridge"), material properties (parsed from model parameters, such as concrete C50), and optimization objective (peak stress) from the optimization assistance requirements. <50> MPa); The matching rule set is queried in the crack resistance optimization knowledge base, and the priority ranking is based on the historical success rate of the rule (success rate = number of successful optimization cases / total number of cases). The knowledge base is constructed as follows: rules are extracted from historical bridge optimization projects, and the applicable scenarios (such as continuous beam bridges, concrete materials) and optimization effects (such as stress reduction of 20%) of each rule are recorded. After expert review, the rules are stored. The keyword matching algorithm is constructed as follows: bridge type, material properties and optimization objectives are converted into keyword vectors (such as ["continuous beam bridge", > "concrete C50", > "peak stress"). <50> ‎MPa”]), calculate the Jaccard similarity with the rules in the knowledge base, and select the rule set with the highest similarity and a success rate greater than 80%.

[0022] For example, in a continuous beam bridge optimization project, the optimization auxiliary requirements specify the bridge type as "prestressed concrete continuous beam bridge", the material property as "concrete C50 (elastic modulus 200 > GPa)", and the optimization objective as "peak stress at mid-span of the beam". <50> MPa. The system retrieves the rule set from the crack resistance optimization knowledge base, with the keyword vector being ["continuous beam bridge", > "concrete C50", > "peak stress"]. <50> The knowledge base contains rule set A (applicable to continuous beam bridges, concrete C40-C60, rule: adjust elastic modulus to 195-205>GPa, increase beam height by 5-10%, success rate 85%) and rule set B (applicable to suspension bridges, steel, success rate 90%). Using Jaccard similarity calculation, rule set A has a similarity of 0.9, while rule set B has a similarity of 0.3; rule set A is selected. Rule set A includes: elastic modulus optimization rules (range 195-205>GPa, step size 2>GPa), beam section height optimization rules (increase by 5-10%, step size 2%), and support stiffness optimization rules (0.9-1.1×10^6>N / m). The system outputs rule set A as the matching result for subsequent resource allocation, ensuring users follow high-success-rate crack-resistant design guidelines when optimizing mid-span stress in beams.

[0023] S22. Based on the resource requirements in the optimization auxiliary requirements, allocate an initial computing resource configuration scheme; wherein, the initial computing resource configuration scheme shall at least include the finite element mesh granularity, the iteration convergence threshold, and the number of parallel computing nodes.

[0024] Based on the resource requirements in the optimization auxiliary requirements (e.g., 4 parallel computing nodes, mesh granularity of 0.1m, convergence threshold of 0.001), an initial computing resource configuration scheme is allocated to ensure the computational efficiency and accuracy of the finite element simulation. The initial computing resource configuration scheme includes the finite element mesh granularity (controlling the mesh element size, e.g., 0.1m), the iterative convergence threshold (controlling the solver convergence accuracy, e.g., residual <0.001), and the number of parallel computing nodes (allocating the number of CPU cores, e.g., 4 nodes). The allocation process is based on a resource optimization algorithm, specifically: extracting resource requirement parameters from the optimization auxiliary requirements; calculating the required computing resources based on the complexity of the bridge model (estimated through the number of nodes and degrees of freedom, e.g., 100,000 nodes, 300,000 degrees of freedom); and allocating the mesh granularity, convergence threshold, and number of nodes based on the hardware environment (e.g., 16 CPU cores and 64GB of memory on a server). The mesh granularity is determined as follows: The initial mesh size (e.g., 0.1m) is determined based on the model's geometric dimensions (e.g., beam length 100m) and material homogeneity (e.g., uniform concrete). Mesh independence is verified through trial calculations (stress change <5% after halving the mesh size). The convergence threshold is determined as follows: A residual threshold (e.g., 0.001) is set based on the optimization objective's accuracy requirements (e.g., stress error <1%). The number of parallel nodes is determined as follows: The number of nodes is allocated based on the model's degrees of freedom and the number of hardware cores (e.g., degrees of freedom / 100,000 = number of nodes). The resource optimization algorithm is constructed as follows: The resource allocation scheme is solved using linear programming, with computation time and accuracy as the objective functions and hardware resource limits as the constraint.

[0025] For example, in a continuous beam bridge project, the optimization auxiliary requirements specify resource requirements of 4 parallel computing nodes, a mesh size of 0.1m, and a convergence threshold of 0.001. The system analyzes the bridge model (beam length 100m, 100,000 nodes, 300,000 degrees of freedom), with a hardware environment of 16-core CPU and 64+ GB of memory. The mesh size was determined through trial and error: the initial mesh was 0.1m, and after trial meshing to 0.05m, the stress change was 3% (<5%), confirming that 0.1m was appropriate. The convergence threshold was set to 0.001 based on the stress error requirement of <1%. The number of parallel nodes was adjusted to 4 nodes based on the requirement of 300,000 / 100,000 degrees of freedom = 3. The resource optimization algorithm aims to minimize computation time (target <1 hour) and maximize accuracy (error <1%), with constraints of 16 cores and 64+ GB of memory. It is solved through linear programming, outputting a configuration scheme with a mesh size of 0.1m, a convergence threshold of 0.001, and 4 parallel nodes. The solution ensures simulation efficiency (estimated calculation time 50 minutes) and accuracy (stress error 0.8%), providing a foundation for subsequent fusion.

[0026] S23. Integrate the crack-resistant design rule set and the computational resource allocation scheme to form initial auxiliary capabilities.

[0027] Initial auxiliary capability refers to an executable optimization scheme that combines design rules (such as the range of elastic modulus adjustment) and computational resources (such as a mesh size of 0.1m). The fusion process is based on a rule-resource mapping algorithm, specifically: associating the rules of the crack-resistant design rule set (such as elastic modulus 195-205> GPa) with the parameters of the computational resource configuration scheme (such as a mesh size of 0.1m) to generate an optimization instruction set; the instruction set includes parameter adjustment instructions (such as "set elastic modulus to 200> GPa") and computation setting instructions (such as "set mesh size to 0.1m"). The rule-resource mapping algorithm is constructed as follows: taking the priority of the rules (based on success rate) and resource constraints (such as the number of nodes ≤ 16) as input, rule-resource pairs (such as "elastic modulus adjustment rule → 4 nodes") are constructed, and the execution efficiency of the instruction set is optimized through a greedy algorithm (prioritizing rules with high success rates and low resource consumption configurations). The verification method for the fusion results is: checking whether the instruction set meets the optimization objective (such as peak stress) through finite element analysis. <50> MPa). The steps for generating the instruction set are: parse the rule set and extract the rule parameters; parse the resource configuration and extract the calculation parameters; map the two into instructions that can be recognized by the finite element software (such as ANSYS's APDL commands).

[0028] For example, in a continuous beam bridge project, the crack-resistant design rule set A includes elastic modulus adjustment rules (195-205> GPa), beam height adjustment rules (increase by 5-10%), and support stiffness rules (0.9-1.1×10^6> N / m); the resource configuration scheme is a mesh size of 0.1m, a convergence threshold of 0.001, and 4 nodes. The system uses a rule-resource mapping algorithm to map the elastic modulus adjustment rules of the rule set to 4-node calculations (due to high precision requirements), and the beam height adjustment rules to a mesh size of 0.1m (to ensure geometric accuracy). Through greedy algorithm optimization, the elastic modulus rules with a success rate of 85% are prioritized, generating an instruction set: parameter adjustment instruction ("SET>E>200>GPa") and calculation setting instruction ("MESH>SIZE>0.1"). Trial calculations show that the peak stress is reduced to 48>MPa ( <50> (MPa), satisfying the optimization objective. The final initial auxiliary capabilities include the instruction set ["SET>E>200>GPa", >"MESH>SIZE>0.1", >"SOLVE>4>NODES"], to support subsequent evaluation.

[0029] S3. Evaluate the assistive effects and user dependency probability of the candidate assistive capabilities under different reduction strategies. Step S3 specifically includes the following sub-steps: S31. Based on the initial auxiliary capability, construct multiple reduction strategies; wherein each reduction strategy includes at least a combination of mesh coarsening ratio, upper limit of iteration steps and rule simplification degree.

[0030] Reduction strategies refer to the simplification and combination of resources or rules in the initial auxiliary capabilities. These include the mesh coarsening ratio (reducing the number of mesh cells, e.g., from 0.1m to 0.2m), the maximum number of iterations (limiting the maximum number of iterations, e.g., 5000), and the degree of rule simplification (reducing the range of rule parameters, e.g., reducing the elastic modulus from 195-205 GPa to 198-202 GPa). The construction process is based on a combinatorial optimization algorithm, specifically: extracting resource parameters (mesh granularity 0.1m, 4 nodes) and rule parameters (elastic modulus range) from the initial auxiliary capabilities; defining the reduction range (e.g., mesh coarsening ratio 20%-50%, maximum number of iterations 3000-5000, degree of rule simplification 10%-30%); and generating a strategy set through permutations and combinations (e.g., Strategy 1: mesh coarsening 20%, maximum number of iterations 5000, rule simplification 10%). The combinatorial optimization algorithm is constructed as follows: aiming to minimize resource consumption and maximize optimization effect, with constraints including resource cap and rule validity, a grid search algorithm is used to traverse and reduce parameter combinations, generating at least 5 strategies. Parameters are obtained as follows: the grid coarsening ratio is calculated based on the model's geometric complexity (number of nodes), the upper limit of the number of iteration steps is estimated based on the convergence speed (residual change rate), and the degree of rule simplification is adjusted based on the rule success rate.

[0031] In the continuous beam bridge project, the initial auxiliary capabilities included a mesh size of 0.1m, 4 nodes, and an elastic modulus of 195-205 > GPa. The system defined the reduction range as follows: mesh coarsening ratio 20%-50% (0.12m-0.15m), maximum number of iterations 3000-5000, and rule simplification degree 10%-20% (elastic modulus of 198-202 > GPa). Five reduction strategies were generated using a mesh search algorithm: Strategy 1 (mesh size 0.12m, iteration limit 5000, rule simplification 10%), Strategy 2 (mesh size 0.15m, iteration limit 4000, rule simplification 15%), etc. The mesh coarsening ratio was based on a model with 100,000 nodes, calculating the number of nodes after coarsening (0.12m corresponds to 80,000 nodes); the iteration limit was estimated based on a residual change rate of 0.001 / step; the rule simplification degree was adjusted based on a rule success rate of 85%, narrowing the range to reduce complexity. Five strategies are ultimately output, ensuring that each strategy differs in resource consumption (e.g., reducing the number of nodes from 4 to 3) and rule complexity (e.g., narrowing the parameter range), providing diverse options for subsequent evaluation.

[0032] S32. Perform finite element simulations for each reduction strategy, extract the stress peak value and crack development index of key parts of the beam, and calculate the quantitative value of the auxiliary effect.

[0033] Peak stress refers to the maximum principal stress at the mid-span of the beam, the support connection, and the cantilever end (e.g., 50 > MPa), reflecting the structural stress state. Crack propagation index refers to the probability of crack propagation based on material fracture mechanics (e.g., 0.1 represents a 10% probability), reflecting crack resistance. The auxiliary effect quantification value refers to the weighted score of the combined peak stress reduction rate and crack propagation index reduction rate (e.g., 0.8 indicates good optimization effect). The simulation execution steps are as follows: load the resource configuration (e.g., mesh size 0.12m) and rule parameters (e.g., elastic modulus 200 > GPa) of the reduction strategy into the finite element software; run the solver to obtain stress distribution and crack propagation data for key parts; calculate the peak stress (extracting the maximum principal stress through post-processing) and crack propagation index (calculating the crack propagation rate based on fracture mechanics models, such as Paris's law, combined with material parameters). The auxiliary effect quantification value is calculated as follows: peak stress reduction rate = (initial stress - optimized stress) / initial stress; the crack propagation index reduction rate is similarly calculated using a weighted sum (weight: stress 0.6, crack 0.4). The fracture mechanics model is constructed as follows: based on Paris's law (crack propagation rate da / dN=C(ΔK)^m, where da / dN represents the crack propagation rate, C is a constant related to material properties, ΔK is the stress intensity factor amplitude, and m is an exponential constant describing the relationship between crack propagation rate and stress intensity factor variation), the crack propagation probability is estimated through material parameters (such as the C and m values ​​of concrete C50) and stress intensity factor ΔK (calculated from simulated stress).

[0034] For example, in a continuous beam bridge project, for reduction strategy 1 (mesh size 0.12m, iteration limit 5000, elastic modulus 198-202 > GPa), the system was configured in ANSYS, with a mesh size of 0.12m and an elastic modulus of 200 > GPa, and the solver was run. Simulation results showed that the peak stress at mid-span of the beam decreased from 60 > MPa to 48 > MPa, a reduction rate of 20%; the stress at the support connection decreased from 55 > MPa to 45 > MPa, a reduction rate of 18%. The crack propagation index was calculated using Paris's law. Inputting concrete C50 parameters (C=1×10^-10, m=3), the stress intensity factor ΔK was calculated from the simulated stress distribution, resulting in a crack propagation probability reduction from 0.15 to 0.09, a reduction rate of 40%. The quantified value of the auxiliary effect = 0.6×20%+0.4×40%=28%. Similarly, Strategy 2 (mesh size 0.15m, iteration limit 4000) calculated a 15% reduction in peak stress and a 30% reduction in crack propagation index, with a quantified value of 0.6 × 15% + 0.4 × 30% = 21%. Simulations were performed for each strategy, data from key components were extracted, and quantified values ​​(28%, 21%, etc.) were output to provide a basis for subsequent evaluation.

[0035] S33. Based on the user's historical operation data, analyze the user's tendency to adapt to different candidate auxiliary capabilities, the frequency of autonomous adjustment, and the ability to independently achieve anti-cracking optimization goals, and comprehensively predict the user's dependence probability.

[0036] Adaptability refers to a user's willingness to accept alternative auxiliary capabilities (e.g., mesh coarsening of 0.12m), calculated by the acceptance rate of similar configurations in historical operations. Autonomous adjustment frequency refers to the frequency with which users manually adjust parameters even under the guidance of auxiliary capabilities (e.g., twice per hour). Independent achievement capability refers to the probability that a user can complete the optimization goal without assistance (e.g., 30%). User dependency probability refers to the degree of user dependence on auxiliary capabilities (0-1, 1 being complete dependence). The analysis steps are: extracting user operation records for similar configurations from historical operation logs and calculating the acceptance rate (e.g., number of times the 0.12m mesh was accepted / total number of times); counting the number of manual adjustments and calculating the frequency (e.g., twice per hour); and determining the independent achievement rate (e.g., the percentage of successful optimization cases) based on historical optimization results. Dependency probability prediction is based on a logistic regression model, specifically: using adaptation tendency, adjustment frequency, and independent achievement capability as input features, with a training dataset from historical user operations (1000 records), and outputting dependency probability (e.g., 0.7). The construction of the logistic regression model involves: collecting historical user operation data, labeling it with dependency tags (dependent / non-dependent), extracting features (acceptance rate, adjustment frequency, achievement rate), optimizing the logistic regression parameters using gradient descent, and outputting the probability.

[0037] For example, in a continuous beam bridge project, the system extracted data from user historical logs (500 records): the user acceptance rate for the 0.12m grid configuration was 80% (40 acceptances / 50 recommendations), the self-adjustment frequency was 1.5 times / hour (based on manually modifying the elastic modulus records), and the independent achievement rate was 25% (10 successful optimizations / 40 attempts). For strategy 1 (0.12m grid), the acceptance rate was 80%, the adjustment frequency was 1.5 times / hour, and the achievement rate was 25%; for strategy 2 (0.15m grid), the acceptance rate was 70%, the adjustment frequency was 2 times / hour, and the achievement rate was 20%. The logistic regression model was trained based on 1000 historical user data records, with features [acceptance rate, > adjustment frequency, > achievement rate], and labels of dependent / independent. Gradient descent optimization was used, predicting a dependency probability of 0.65 for strategy 1 and 0.75 for strategy 2. The results show that users are more adaptable to strategy 1 but have lower dependency, while strategy 2 requires more adjustments due to the coarser grid and has higher dependency. The system outputs the dependency probability of each strategy (0.65, 0.75, etc.) to provide data for the evaluation matrix.

[0038] S34. Establish an evaluation matrix based on the dimensions of auxiliary effect and dependency probability, and evaluate the auxiliary effect and user dependency probability of each candidate auxiliary capability.

[0039] The evaluation matrix uses the quantified auxiliary effect value (from S32, e.g., 28%) and user dependency probability (from S33, e.g., 0.65) as two-dimensional coordinates, with matrix elements representing candidate auxiliary capabilities (e.g., strategy 1, strategy 2). The matrix construction steps are as follows: collect the quantified auxiliary effect value and dependency probability of each reduction strategy, normalize them to the [0, 1] interval (effect value / 100, dependency value is 0-1); construct a two-dimensional matrix, with the horizontal axis representing the auxiliary effect and the vertical axis representing the dependency probability, with elements representing strategy numbers; evaluate the strategy distribution through visualization analysis (e.g., scatter plot). The matrix construction algorithm is based on a two-dimensional interpolation algorithm, mapping the effect value and dependency value to a grid, marking the coordinates of each strategy (e.g., [0.28, >0.65]). The evaluation method prioritizes strategies with high effect and low dependency (e.g., effect >25%, dependency <0.7). Parameter acquisition involves extracting the effect value from the S32 simulation results and the dependency probability from the S33 logistic regression model output.

[0040] For example, in a continuous beam bridge project, the auxiliary effect quantification values ​​and dependency probabilities of the five reduction strategies are as follows: Strategy 1 (28%, >0.65), Strategy 2 (21%, >0.75), Strategy 3 (25%, >0.70), Strategy 4 (30%, >0.80), and Strategy 5 (20%, >0.60). The system normalizes the data (effect value / 100) and constructs an evaluation matrix, with the horizontal axis representing effect [0.2, >0.3] and the vertical axis representing dependency [0.6, >0.8], and the elements representing strategies 1-5. A two-dimensional interpolation algorithm maps the data to a grid, generating a scatter plot. This shows that Strategy 4 has the highest effect (30%) but a relatively high dependency (0.80), while Strategy 5 has the lowest dependency (0.60) but a poorer effect (20%). The evaluation prioritizes strategies with effects >25% and dependencies <0.7, thus Strategy 1 (28%, >0.65) and Strategy 3 (25%, >0.70) are included in the candidate range. The matrix clearly shows the advantages and disadvantages of each strategy, providing data support for subsequent Pareto optimization.

[0041] S4. Select the candidate auxiliary capability that achieves the optimal balance between auxiliary effect and user dependency probability as the target auxiliary capability. Step S4 specifically includes the following sub-steps: S41. Using the Pareto front search algorithm, identify the non-dominated solution set of auxiliary effects and user dependency probability in the evaluation matrix.

[0042] A non-dominated solution set refers to the set of strategies for which no other strategy is superior to it in both auxiliary effect and dependency probability (e.g., strategy A's effect and dependency are not inferior to B). The Pareto front search algorithm is constructed as follows: With auxiliary effect (maximization) and dependency probability (minimization) as objectives, input the strategy coordinates of the evaluation matrix (e.g., [0.28, >0.65]); determine whether a strategy is dominated by a pairwise comparison (if strategy A's effect ≥ B and dependency ≤ B, then A dominates B); output the set of non-dominated strategies. The specific steps of the algorithm are: traverse the strategy pairs in the evaluation matrix, compare effect values ​​and dependency values, and mark non-dominated strategies (e.g., strategy 1 is not dominated by any strategy); repeat until all strategies are classified. Parameter acquisition involves extracting strategy coordinates (effect values, dependency values) from the evaluation matrix of S34, ensuring data normalization (effect values ​​0-1, dependency values ​​0-1). The algorithm complexity is O(n^2), where n is the number of strategies, ensuring efficient screening.

[0043] For example, in a continuous beam bridge project, the evaluation matrix contains five strategies: Strategy 1 (0.28, >0.65), Strategy 2 (0.21, >0.75), Strategy 3 (0.25, >0.70), Strategy 4 (0.30, >0.80), and Strategy 5 (0.20, >0.60). The Pareto front search algorithm compares each strategy pairwise: Strategy 1 (0.28, >0.65) is more effective and less dependent than Strategy 2 (0.21, >0.75), thus dominating Strategy 2; Strategy 4 (0.30, >0.80) is more effective than Strategy 1 but has a higher dependency, so it does not dominate; Strategy 5 (0.20, >0.60) has a lower dependency than Strategy 1 but is less effective, so it does not dominate. The final non-dominated solution set is {strategy1, >strategy4, >strategy5}, corresponding to coordinates [0.28, >0.65], [0.30, >0.80], and [0.20, >0.60]. The results show that strategies 1, 4, and 5 have no absolute disadvantage in terms of effect and dependency, and are therefore included in the subsequent weight calculation, excluding strategies 2 and 3.

[0044] S42. Based on the user's current operation stage and computational environment constraints, dynamically adjust the weight coefficients of the auxiliary effect and the dependency probability, and calculate the comprehensive utility value of each non-dominated solution in the non-dominated solution set.

[0045] Based on the user's current operational stage (e.g., modeling, solving) and computational environment constraints (e.g., 16 CPU cores), the weight coefficients of auxiliary effects and dependency probabilities are dynamically adjusted to calculate the comprehensive utility value of the non-dominated solution set and select the optimal strategy. The weight coefficients indicate the relative importance of auxiliary effects and dependency probabilities (e.g., effect weight 0.7, dependency weight 0.3). The operational stage is determined through log analysis (e.g., focusing on effect during modeling, focusing on efficiency during solving); computational environment constraints are obtained from hardware parameters (e.g., number of cores, memory). The weight adjustment steps are as follows: if the user is in the modeling stage, the effect weight is increased (e.g., 0.8); if in the solving stage, the dependency weight is increased (e.g., 0.5); if the number of cores is 8 → effect weight 0.8), the weights are output through fuzzy inference. The utility value is used to rank non-dominated solutions. The formula for calculating the comprehensive utility value is: the comprehensive utility value of a non-dominated solution equals the auxiliary effect value multiplied by the effect weight coefficient, minus the dependency probabilities value multiplied by the dependency weight coefficient. That is, comprehensive utility value = auxiliary effect value × effect weight coefficient - dependency probabilities value × dependency weight coefficient.

[0046] For example, in a continuous beam bridge project, the user is in the modeling phase (logs show frequent modifications to the elastic modulus), and the hardware consists of 16 cores and 64GB of memory. The fuzzy logic algorithm takes "modeling phase, 16 cores" as input and the rule "modeling → effect weight 0.8, dependency weight 0.2" as output, with a weight of [0.8, >0.2]. The utility values ​​of the non-dominated solution set {strategy1(0.28, >0.65), >strategy4(0.30, >0.80), >strategy5(0.20, >0.60)} are calculated as follows: Strategy 1 utility = 0.28 × 0.8 - 0.65 × 0.2 = 0.094; Strategy 4 utility = 0.30 × 0.8 - 0.80 × 0.2 = 0.08; Strategy 5 utility = 0.20 × 0.8 - 0.60 × 0.2 = 0.04. Strategy 1 has the highest utility and is prioritized as a candidate. The system records weights and utility values ​​to provide a basis for subsequent verification.

[0047] S43. Select the auxiliary capability with the highest comprehensive utility value as the candidate capability, and verify the stability and robustness of the crack resistance optimization effect of the candidate capability through multi-condition finite element simulation.

[0048] Multi-condition loading refers to a combination of various loads and boundary conditions (e.g., vehicle load 10>kN / m, wind load 5>kN / m). The verification steps are as follows: Load the resource configuration (e.g., mesh 0.12m) and rules (e.g., elastic modulus 200>GPa) for candidate capabilities; set the load combination (e.g., Load 1: vehicle load 10>kN / m + fixed support; Load 2: wind load 5>kN / m + sliding support); run finite element simulation, extract the peak stress and crack propagation index of key components; calculate stability (standard deviation <5%) and robustness (effect change <10% after parameter perturbation ±5%). The simulation is based on finite element software (e.g., ANSYS), and the load combination is automatically generated by a script (e.g., APDL script for cyclic loading). The stability and robustness evaluation algorithm is as follows: calculate the mean and standard deviation of the multi-condition effect, repeat the simulation after perturbing the parameters, and compare the rate of change of the effect.

[0049] For example, in a continuous beam bridge project, Strategy 1 (utility 0.094, mesh 0.12m, elastic modulus 200>GPa) is the candidate capability. The system sets three load cases in ANSYS: Load Case 1 (vehicle load 10>kN / m, fixed support), Load Case 2 (wind load 5>kN / m, sliding support), and Load Case 3 (combined load 12>kN / m). Simulation results: Load Case 1: peak stress 48>MPa, crack index 0.09; Load Case 2: stress 49>MPa, crack index 0.10; Load Case 3: stress 47>MPa, crack index 0.08. The mean (stress 48.3>MPa, crack 0.09) and standard deviation (stress 0.82>MPa, crack 0.008) are <5%, indicating stability. The robustness test showed that the perturbation elastic modulus was ±5% (190-210> GPa), the stress change was <8%, and the crack index change was <9%, thus the robustness test was passed. Strategy 1 was approved as a candidate capability and proceeded to final confirmation.

[0050] S44. If the verification is successful, the candidate capability is determined as the target auxiliary capability; otherwise, the weight coefficients are iteratively adjusted and re-evaluated until the target auxiliary capability that meets the requirements is obtained.

[0051] If a candidate capability passes stability and robustness verification, it is confirmed as the target auxiliary capability; otherwise, iteratively adjust the weight coefficients and re-execute S41-S43 until the requirements are met. The target auxiliary capability refers to the optimization scheme ultimately used for user assistance (e.g., mesh size 0.12m, elastic modulus 200 > GPa). The verification pass conditions are: stability (effect standard deviation 5%), adjust the effect weight (decrease by 0.1) or dependency weight (increase by 0.1); rerun the Pareto search and utility calculation. The gradient descent method is constructed as follows: with the verification pass rate as the objective, the weights as the optimization variables, calculate the gradient of the utility value with respect to the weights, and iteratively update the weights until verification is passed or the maximum number of iterations (e.g., 10 times) is reached. The reasons for failure are extracted from the simulation results (e.g., stress standard deviation 6%), and the weight adjustment range is [0.1, > 0.9].

[0052] For example, in a continuous beam bridge project, Strategy 1, after verification (stability standard deviation 0.82 > MPa < 5%, robustness variation < 9%), is confirmed as the target auxiliary capability: mesh 0.12m, elastic modulus 200 > GPa, 4 nodes. If verification fails (e.g., stress standard deviation 6%), the system analyzes the cause (mesh too coarse), reduces the effect weight by 0.1 (0.8 → 0.7), increases the dependency weight by 0.1 (0.2 → 0.3), and reruns S41-S43. Gradient descent, with pass rate as the objective, iterates 3 times to obtain new weights [0.6, > 0.4]. The new candidate strategy 2 (mesh 0.15m) passes verification and is confirmed as the target auxiliary capability. Finally, Strategy 1 is output as the target auxiliary capability, proceeding to S5.

[0053] S5. Provide assistance to the user based on the target assistance capabilities. Step S5 specifically includes the following sub-steps: S51. Convert the target auxiliary capability into an executable finite element analysis instruction set; wherein the finite element analysis instruction set includes at least the mesh generation scheme, material constitutive model parameters, and convergence control conditions.

[0054] The finite element analysis instruction set includes a mesh generation scheme (defining mesh size, such as 0.12m), material constitutive model parameters (defining material properties, such as elastic modulus 200 > GPa), and convergence control conditions (defining solver parameters, such as residual < 0.001). The conversion process is based on an instruction generation algorithm, specifically: analyzing the resource configuration (e.g., mesh 0.12m) and rule parameters (e.g., elastic modulus 200 > GPa) of the target auxiliary capability; generating corresponding instructions (e.g., "SET > E > 200") according to the instruction format of the finite element software (e.g., ANSYS > APDL or Abaqus > Python); and verifying the integrity of the instruction set (checking parameter coverage, such as 100% coverage of mesh, material, and convergence parameters). The instruction generation algorithm is constructed by using the software API as a template, mapping the target capability parameters to instructions, and prioritizing highly compatible instructions (e.g., ANSYS general commands). Parameter acquisition involves extracting mesh granularity, material parameters, and convergence thresholds from the target auxiliary capability and directly mapping them to software parameters.

[0055] For example, in a continuous beam bridge project, the target auxiliary capability is a mesh size of 0.12m, an elastic modulus of 200 > GPa, 4 nodes, and a convergence threshold of 0.001. The system generates a command set based on the ANSYS > APDL format: mesh generation command ("MESH > SIZE > 0.12"), material parameter command ("SET > E > 200 > GPa"), and convergence control command ("SOLVE > TOL > 0.001 > NODES > 4"). The command generation algorithm parses the target capability, maps it to APDL commands, and verifies 100% coverage (including mesh, material, and convergence parameters). The command set is saved as an executable script, ensuring that users can run it directly in ANSYS to optimize the mid-span stress of the beam to 48 > MPa, meeting the crack resistance target.

[0056] S52. By embedding an auxiliary engine into the finite element analysis software, the finite element analysis instruction set is dynamically loaded, guiding the user to adjust the modeling and solution settings in real time.

[0057] An auxiliary engine refers to a plug-in module embedded in the software (such as ANSYS's Python plug-in), which supports dynamically loading instructions and provides user interface prompts. The dynamic loading process involves: loading the instruction set into the software via an API (such as ANSYS's Python interface); monitoring user operations in real time, and if deviations from the instruction set occur (e.g., the user sets the mesh size to 0.2m instead of 0.12m), providing adjustment suggestions via pop-ups or logs (e.g., "Suggested mesh size 0.12m"). The auxiliary engine is constructed by: developing a Python plug-in, integrating the software API, defining the instruction loading interface and user interaction module, and real-time parsing of differences between user operations and the instruction set. Parameter acquisition involves: extracting mesh, material, and convergence parameters from the instruction set, and obtaining the user's current settings from the software logs.

[0058] For example, in a continuous beam bridge project, the auxiliary engine is an ANSYS Python plugin, loading the instruction set ["MESH>SIZE>0.12", > "SET>E>200>GPa", > "SOLVE>TOL>0.001>NODES>4"]. The plugin dynamically executes instructions via ANSYS>Python>API, monitoring user operations: when the user sets the mesh size to 0.2m, the engine pops up a message suggesting "a mesh size of 0.12m is recommended to improve accuracy." After the user adjusts, the engine loads material instructions, sets the elastic modulus to 200>GPa, and finally guides the solution settings (convergence threshold 0.001, 4 nodes). Simulation results show that the mid-span stress of the beam drops to 48>MPa, the user completes the optimization, and the engine records the operation log for future optimization.

[0059] This method intelligently detects user needs in optimizing the crack resistance performance of bridge structures, matches and dynamically adjusts auxiliary capabilities, significantly improving the optimization efficiency and accuracy of finite element analysis, while effectively balancing the auxiliary effect with user independence. Compared with traditional methods, this technical solution can provide customized support according to the actual needs of users, reducing the blindness of modeling and parameter optimization, and shortening the optimization cycle. By evaluating the auxiliary effects and dependency possibilities under different reduction strategies, it ensures that users obtain efficient assistance without over-reliance on tools, maintaining technical autonomy. Ultimately, based on the optimal auxiliary capabilities, it improves the optimization quality of bridge crack resistance performance, providing a reliable guarantee for safe and durable bridge design.

[0060] Specifically, S11-S13 accurately identify user optimization needs and computational bottlenecks in key areas (such as mid-span beams and support connections) by real-time monitoring of user operations, constructing behavioral feature vectors, and matching optimization scenario templates. This reduces blind adjustments and improves the targeting of needs analysis. S21-S23 provides efficient optimization guidance and resource support by retrieving crack-resistant design rule sets, allocating computational resources, and integrating them to form initial auxiliary capabilities, ensuring the accuracy and efficiency of finite element analysis. S31-S34 constructs reduction strategies, evaluates auxiliary effects and user dependency possibilities, builds an evaluation matrix, comprehensively quantifies the performance of different strategies, balances optimization effects and user independence, and avoids over-reliance on tools. S41-S44 uses Pareto front search and dynamic weight adjustment to select target auxiliary capabilities with the optimal balance between effects and dependencies, and verifies their stability and robustness through multi-condition simulations to ensure the reliability of the optimization scheme.

[0061] Furthermore, in the field of bridge structural design and maintenance, crack resistance is a key factor in ensuring the long-term durability and safety of bridges. Traditional methods for optimizing bridge crack resistance largely rely on finite element analysis (FEM), which simulates the structural response under different loads and environmental conditions to adjust design parameters and reduce the probability of crack occurrence. However, existing methods have significant limitations in practical applications: First, the optimization process is highly dependent on the engineer's experience and lacks a systematic support mechanism, resulting in low efficiency in generating innovative ideas. Second, during complex FEM analyses, users struggle to accurately capture the correlation between key areas (such as areas of stress mutation or abnormal displacement) and innovative inspiration, limiting the depth and breadth of design optimization. Third, existing tools lack dynamic perception and support for the user's cognitive state and cannot adaptively provide innovative guidance based on the user's behavioral patterns, making it easy for the optimization process to get stuck in local optima and hindering breakthrough innovation. In recent years, with the advancement of artificial intelligence and human-computer interaction technologies, data-driven cognitive enhancement methods have gradually been applied to the field of engineering optimization. However, existing technologies have not yet fully integrated user behavior analysis, temporal attention mechanisms, and multi-objective optimization algorithms, failing to achieve dynamic and personalized innovative assistance in optimizing bridge crack resistance.

[0062] Therefore, in some embodiments, during the process of providing assistance to the user based on the target assistive capabilities, an innovative cognitive enhancement mechanism is further integrated, specifically including: S6. Based on the process expectation of assisted execution, the future time period is divided into multiple idea-inspiring time windows. Among them, the process expectation includes the correlation pattern between structural response, parameter adjustment trajectory and historical innovation breakthroughs during the finite element analysis iteration process. Through a window partitioning algorithm based on the temporal attention mechanism, the duration of innovative ideas generated by users after optimization is predicted in different time periods. With the goal of maximizing the duration of innovative ideas generated by users in each idea-inspiring time window, the window boundary and window length of each idea-inspiring time window are dynamically adjusted. When constructing the window partitioning algorithm, a structural response sensitivity factor is introduced so that the idea-inspiring time window is automatically extended in the time period corresponding to the stress mutation or displacement anomaly region of the finite element model, so as to adapt to the characteristic that users are more likely to have innovative thinking in key areas.

[0063] Process prediction refers to predicting the time period during which users may generate innovative ideas by analyzing the correlation patterns between structural responses (such as stress distribution and displacement changes), parameter adjustment trajectories (such as elastic modulus adjustment sequences), and historical innovation breakthroughs (such as proposing new beam cross-section designs) during finite element analysis iterations. The idea-inspiring time window refers to a segmented future time period, with each time window corresponding to the duration during which users may generate innovative ideas (e.g., 30 minutes). Window segmentation is based on a temporal attention mechanism-based window segmentation algorithm, specifically: extracting structural responses (such as the rate of change of peak stress at mid-span of the beam), parameter adjustment trajectories (such as elastic modulus adjustment frequency), and historical innovation breakthroughs (such as users proposing new designs during stress abrupt changes) from the finite element analysis log; constructing a temporal dataset, inputting it into the attention mechanism model, and calculating the innovation potential weight at each time point (based on structural response sensitivity and adjustment frequency); and using a dynamic programming algorithm to segment time windows with the goal of maximizing innovation potential, adjusting the window boundaries (such as start time) and window length (e.g., from 30 minutes to 60 minutes). The structural response sensitivity factor is defined as the weight of the influence of structural parameters (such as stress and displacement) on innovative behavior. It is calculated by analyzing the correlation between stress abrupt changes (e.g., a 10% increase in peak stress) or displacement anomalies (e.g., a 5mm exceedance in displacement) and innovative breakthroughs in historical data. The method involves extracting stress and displacement time series from finite element simulation results, calculating the rate of change (e.g., stress change rate = Δσ / Δt, where Δσ is the amount of stress change and Δt is the time of change), and determining the sensitivity factor (e.g., 0.8) through correlation analysis (e.g., Pearson correlation coefficient). The window partitioning algorithm is constructed based on a temporal attention mechanism using the Transformer model. It takes the structural response and parameter adjustment sequences as input, encodes time-point features, calculates attention weights (reflecting innovative potential), and optimizes window partitioning through dynamic programming. The objective is to maximize the sum of innovative potential for each window, with a constraint that the window length is 30-120 minutes. The specific steps of the algorithm are as follows: collect finite element log data, extract stress, displacement, and parameter adjustment sequences; normalize the data, input it into the Transformer model, and calculate the attention score; based on the score, dynamically plan and divide the time window, extending the window length corresponding to stress mutation or displacement anomaly regions (e.g., extending it by 20%). Finally, multiple time windows are output (e.g., [10:00-10:40, 10:40-11:20]), ensuring that users have more time to think in critical areas (e.g., stress mutation at the mid-span of the beam).

[0064] Here is a specific implementation example: In a crack resistance optimization project for a prestressed concrete continuous beam bridge, the system, based on ANSYS log analysis, anticipated the process and divided the time window for idea activation. The log recorded that the mid-span stress of the beam increased from 50 MPa to 60 MPa within one hour (rate of change 0.167 MPa / min), and the elastic modulus was adjusted three times (200 GPa → 205 GPa → 202 GPa). Historical data showed a correlation of 0.85 between stress mutations (>10%) and the user-proposed new cross-section design (calculated using Pearson correlation coefficient). The system extracted the stress change rate (0.167 MPa / min) and displacement changes (mid-span displacement increased from 2 mm to 2.5 mm), calculating a structural response sensitivity factor of 0.8 (based on correlation analysis). A temporal attention mechanism model (based on Transformer, 4 layers, 8 attention heads) was used as input for stress, displacement, and parameter adjustment sequences. After normalization, time-point features were encoded, and attention weights were calculated (e.g., a weight of 0.9 corresponds to a stress mutation at 10:20). The dynamic programming algorithm aims to maximize the sum of weights, dividing the time window into three time windows: 10:00-10:40 (40 minutes, stress stable), 10:40-11:30 (50 minutes, stress abrupt change, extended by 20%), and 11:30-12:00 (30 minutes, return to stable). The window length is adjusted based on a sensitivity factor to ensure that the window length is extended to 50 minutes in the stress abrupt change region (10:40-11:30). Finally, the system outputs a list of time windows, prompting users to pay close attention to the stress changes at the mid-span of the beam between 10:40 and 11:30, potentially leading to innovative ideas such as "adjusting the cross-sectional height" or "optimizing the prestressed structure."

[0065] S7. Whenever an idea-generating time window is entered, the intent-aware module is activated to continuously detect whether the user shows an intention to generate innovative ideas. The intention includes: the user's repeated correction of the crack development path in the finite element model, the nonlinear adjustment of material parameters, and the key annotation of local stress concentration areas.

[0066] The intent-driven actions include repeated corrections to crack propagation paths in the finite element model (e.g., multiple modifications to crack propagation parameters), nonlinear adjustments to material parameters (e.g., irregular changes in the elastic modulus adjustment range), and focused annotations of local stress concentration areas (e.g., marking high-stress points at the mid-span of a beam). The intent-aware module is based on a behavior sequence analysis algorithm, specifically: extracting user operation sequences from finite element analysis software logs (e.g., ANSYS APDL logs), including correction actions (recording crack path parameter modifications, such as adjusting the crack propagation rate from 0.1 to 0.08), adjustment trials (recording material parameter changes, such as the elastic modulus changing from 200 GPa to 210 GPa and back to 205 GPa), and annotation actions (recording annotations added by the user to the model, such as "stress concentration"). Parameters are acquired by capturing operation commands through the software log interface, parsing the operation type and parameter values; calculating the correction frequency (e.g., modifying the crack path 3 times per hour), the root mean square deviation of the adjustment range (e.g., the standard deviation of the elastic modulus adjustment is 5 GPa), and the number of annotations (e.g., annotating 5 stress points). The behavioral sequence analysis algorithm is constructed as follows: based on a Long Short-Term Memory (LSTM) network, the input is an operation sequence (timestamp, operation type, parameter value), the training dataset is historical user operations (1000 records, labeled with innovative intent / non-intent), and the output is the intent probability (e.g., 0.8 indicates high intent). The algorithm steps are: preprocessing log data, extracting operation feature vectors (e.g., [adjusting frequency, standard deviation, number of labels]); inputting into the LSTM model, encoding sequence features, and predicting the intent probability; if the probability > 0.7, it is determined to be an innovative intent. The module monitors each time window in real time (e.g., 10:40-11:30) to ensure timely detection of user intent.

[0067] For example, in a continuous beam bridge project, the system activated the intent-aware module during the 10:40-11:30 time window, detecting user behavior based on ANSYS logs. The logs showed that the user modified the crack propagation rate three times in the mid-span of the beam (from 0.1 to 0.09, 0.08, and 0.085, 3 times / 50 minutes), adjusted the elastic modulus four times (200GPa→210GPa→205GPa→208GPa, standard deviation 4.79GPa), and added five stress concentration annotations ("high stress zone") at the mid-span and support connections. The intent-aware module, based on an LSTM model (2 layers, 128 hidden units, 1000 training data points, 80% intent annotation accuracy), input a feature vector [3 times / 50 minutes, 4.79GPa, 5 values], encoded the operation sequence, and output an intent probability of 0.82 (>0.7). The system determined that the user exhibited innovative intent, possibly attempting things like "optimizing crack control reinforcement configuration" or "adjusting material mix proportions." The module records the intent trigger time (10:45) and behavioral details (crack modification, parameter adjustment, annotation), and passes the results to S8 to activate the confidence enhancement strategy. The entire process ensures that the user's innovative intent is captured in real time within the time window, providing accurate assistance for subsequent actions.

[0068] S8. If the user is detected to have expressed intention within the current idea-inspiring time window, the confidence enhancement decision engine is activated to determine the best time to enhance the user's confidence in innovative ideas and the corresponding confidence enhancement strategy. The confidence enhancement strategy includes: presenting structural analogy animations of successful crack-resistant bridge cases, deconstructing the logical mind map of crack causes and countermeasures in the current model in stages, and dynamically displaying the impact path of parameter adjustment on structural stress redistribution.

[0069] Confidence-enhancing strategies include presenting structural analogies to successful bridge crack-resistant cases (e.g., animations showing optimized stress distribution), deconstructing the causes and countermeasures of cracks in stages using logical diagrams (e.g., diagrams breaking down stress concentration causes), and dynamically displaying the impact of parameter adjustments on structural stress redistribution (e.g., graphs showing stress reduction corresponding to changes in elastic modulus). The decision engine, based on a multi-objective optimization algorithm, comprehensively considers sustained benefits, cross-window attenuation effects, and cognitive absorption potential to determine the optimal timing (e.g., a specific minute within a time window) and strategy. Sustained benefits refer to the effect of the strategy within a preset decay time (e.g., 30 minutes) after execution (e.g., a 5% increase in stress optimization rate); cross-window attenuation effects refer to the reduction in the effect of the current strategy by subsequent local auxiliary processes (e.g., mesh adjustment) within a time window, quantified by inter-window coupling coefficients (e.g., 0.3, indicating a 30% attenuation) and attention decay factors (e.g., 0.9, indicating 90% attention retention); cognitive absorption potential refers to the user's learning ability at a specific time, calculated through the user's historical behavior and task complexity. The parameters are obtained as follows: operation frequency and task complexity (e.g., 100,000 nodes) are extracted from finite element logs; coupling coefficient (based on inter-window operation overlap rate) and decay factor (based on attention duration) are calculated from historical data. The multi-objective optimization algorithm is constructed as follows: with the objectives of maximizing sustained benefits, minimizing cross-window attenuation, and maximizing cognitive absorption potential, candidate timings (e.g., per minute) and strategies are input, and the non-dominated sorting genetic algorithm (NSGA-II) is used for iterative optimization (population size 100, 50 iterations), outputting the optimal timing and strategy combination.

[0070] For example, in a continuous beam bridge project, user innovation intent was detected at 10:45, triggering a confidence-enhancing decision engine. Candidate strategies included: animation showcasing successful cases (stress reduction in a continuous beam bridge to 45 MPa), a logic map decomposing crack causes (stress concentration → material optimization), and dynamic curves demonstrating stress changes corresponding to elastic modulus adjustments (200 GPa → 48 MPa). Sustained benefits were estimated through simulation (e.g., animation improving optimization rate by 5%), and cross-window attenuation was calculated using historical data (inter-window operation overlap rate 0.2, coupling coefficient 0.3, attenuation factor 0.9). Cognitive absorption potential was calculated as 0.75 based on user operation frequency (3 times / 50 minutes) and task complexity (100,000 nodes). The NSGA-II algorithm (population 100, iterations 50) input candidate timings (10:45-11:30 per minute) and three strategies, with optimization objectives of sustained benefits > 5% and attenuation of 0.7, outputting the optimal timing of 10:50 and a strategy "logic map". The mind map analyzes the causes of stress concentration at mid-span of the beam (high load, insufficient material stiffness) and proposes countermeasures (optimizing the elastic modulus, increasing the cross-sectional height). The system records the decision results and transmits them to S9 for execution, ensuring that users gain effective confidence enhancement during high-potential periods.

[0071] Specifically, step S8 includes the following sub-steps: S81. Construct a sequence model of auxiliary processes within the current idea-generating time window, predict the sustained benefits of any candidate confidence enhancement strategy executed at any candidate time point within the subsequent preset increase decay time, and evaluate the impact of cross-window weakening on the sustained benefits of local auxiliary processes within the subsequent idea-generating time window; the cross-window weakening impact is quantified by the inter-window coupling coefficient and the attention decay factor.

[0072] The auxiliary process sequence model is based on a time series prediction algorithm. It takes as input the operation sequence (e.g., parameter adjustment, simulation restart) from the finite element analysis log and outputs the sustained benefits (optimization rate improvement within 30 minutes after strategy execution, e.g., 5%) and cross-window attenuation effects (the attenuation of the effect by subsequent time windows, e.g., 20%) for each candidate timing (e.g., 10:45, 10:50). The sustained benefits are obtained through historical data regression analysis. Specifically, the operation sequence (e.g., elastic modulus adjustment 3 times / 50 minutes) is extracted from the log, and the stress reduction rate after strategy execution is calculated. The cross-window attenuation effect is calculated using the inter-window coupling coefficient (based on operation overlap rate, e.g., 0.3) and the attention decay factor (based on user attention duration, e.g., 0.9). Parameters are obtained by extracting operation frequency and simulation results (e.g., stress 48 MPa) from the log; the coupling coefficient is calculated using inter-window operation similarity (Jaccard coefficient); and the decay factor is estimated using the user operation interval time (average 10 minutes). The time series prediction algorithm is based on a gated recurrent unit (GRU) and is constructed as follows: input is an operation sequence (features: operation type, parameter value, timestamp), training data consists of historical operations (1000 operations, labeled with benefits and attenuation values), and output is the predicted benefits and attenuation values. The algorithm steps are: normalize the operation sequence, input it into the GRU model (2 layers, 64 hidden units), predict the benefits and attenuation values ​​at each time point; verify the prediction error (mean squared error < 0.05).

[0073] For example, in a continuous beam bridge project, within the time window of 10:40-11:30, the system constructs an auxiliary process sequence model. The logs show the user operation sequence: elastic modulus adjustment 3 times (200GPa→205GPa→202GPa), simulation restart 2 times, and stress annotation 5 times. Using a GRU model (2 layers, 64 hidden units, 1000 training data points, mean squared error 0.03) as input, the system predicts that the sustained benefit of executing the logic map strategy at 10:50 is a stress reduction rate of 6% (within 30 minutes), and a cross-window attenuation effect of 20% (overlap rate of 0.2, coupling coefficient 0.3, and attenuation factor 0.9 for the next window 11:30-12:00). At other times (e.g., 10:45), the predicted benefit is 5.5% and the attenuation is 22%. The model is trained based on historical data (1000 data points, 80% prediction accuracy) to ensure prediction reliability. The results output the benefit and attenuation values ​​for each time point, for use in S83 optimization.

[0074] S82. Based on users' historical behavior data, use a hidden Markov model to predict the user's cognitive state transition path within the current thought-stimulating time window, and combine the structural complexity of the current finite element analysis task to calculate the cognitive absorption potential value corresponding to each candidate opportunity.

[0075] Cognitive states include focus (high attention), exploration (trying new parameters), and fatigue (reduced operation frequency). Transition paths refer to the probabilities of state transitions (e.g., focus → exploration probability 0.6). Cognitive absorption potential refers to a user's learning ability at a given time (0-1, e.g., 0.75), calculated using state probabilities and task complexity. Parameters are obtained by extracting operation frequency (3 times / 50 minutes), parameter adjustment amplitude (standard deviation 5 GPa), and task complexity (100,000 nodes) from logs. The HMM input features are operation sequences, and the output is a state sequence and transition probabilities. The HMM is constructed as follows: defining states (focus, exploration, fatigue), observations as operation features (frequency, amplitude), and training data as historical operations (1000 entries, labeled with states). The transition matrix is ​​optimized using the Baum-Welch algorithm to predict state sequences. The potential value is calculated as: Potential = Focus probability × 0.5 + Exploration probability × 0.3 - Fatigue probability × 0.2, multiplied by the complexity weight (number of nodes / 100,000). The algorithm steps are as follows: preprocess logs and extract features; train HMM and output state probabilities; calculate potential values ​​based on complexity.

[0076] For example, in a continuous beam bridge project, the system trains an Hidden Markov Model (HMM) based on 1000 historical operation data points (labeled as 50% focus, 30% exploration, and 20% fatigue). Within the 10:40-11:30 window, the log shows an operation frequency of 3 times / 50 minutes, an adjustment amplitude standard deviation of 4.79 GPa, and a task complexity of 100,000 nodes and a weight of 1. The HMM (3 states, observed features [frequency, amplitude]) predicts the state probability at 10:50: focus 0.6, exploration 0.3, fatigue 0.1, and the transition probability (e.g., focus → exploration 0.5). The cognitive absorption potential is calculated as 0.6 × 0.5 + 0.3 × 0.3 - 0.1 × 0.2 × 1 = 0.37. The potential for other times (e.g., 10:45) is 0.35. The results output the potential values ​​for each time period for optimization by the S83 algorithm.

[0077] S83. With the goals of maximizing sustained benefits, minimizing the cross-window attenuation effect, and maximizing cognitive absorption potential, a multi-objective optimization algorithm is used to solve for the optimal timing point and the corresponding confidence enhancement strategy from candidate timing and candidate confidence enhancement strategies.

[0078] The sustained benefits, cross-window attenuation effects, and cognitive absorption potential are obtained from S81 and S82, respectively (e.g., benefits 6%, attenuation 20%, potential 0.37). The multi-objective optimization algorithm, based on NSGA-II, is constructed as follows: input candidate timings (e.g., 10:45-11:30 per minute) and strategies (e.g., animation, mind map), with the objectives of maximizing benefits, minimizing attenuation, and maximizing potential; iterative optimization (population size 100, 50 iterations) is performed through non-dominated sorting and crowding comparison, outputting a Pareto optimal solution set (e.g., [10:50, mind map]). Parameter acquisition involves extracting benefit and attenuation values ​​from S81 and potential values ​​from S82. The algorithm steps are: constructing a population (timing-strategy pairs), calculating target values ​​(benefits, attenuation, potential), non-dominated sorting, retaining the optimal solution; generating a new population through crossover and mutation, and iterative optimization. Finally, the optimal timing and strategy are output, ensuring a balance between effectiveness and user acceptance.

[0079] For example, in a continuous beam bridge project, NSGA-II inputs 60 candidate times from 10:45 to 11:30 and three strategies (animation, mind map, and curve), with target values ​​derived from S81 (6% benefit, 20% reduction) and S82 (0.37 potential). The algorithm (population 100, 50 iterations) calculates the benefit of executing the mind map strategy at 10:50 as 6.2%, reduction as 19%, and potential as 0.38, which is the optimal solution after non-dominated sorting. Other solutions (such as the 10:45 animation, benefit 5.8%, reduction as 22%, and potential as 0.35) are dominated. The system outputs the optimal combination [10:50, mind map], ensuring users gain efficient confidence enhancement during high-potential times.

[0080] S9. At the optimal time, implement confidence enhancement strategies to users through the human-computer interaction interface. The implementation of confidence enhancement strategies integrates the structural performance data generated in real time by the finite element analysis engine, and dynamically adjusts the parameter values ​​and visual effects in the presented cases or mind maps to make them on the same scale as the user's current model, thereby reducing the user's cognitive leap burden.

[0081] At the optimal time, confidence-enhancing strategies are implemented to users through the human-computer interaction interface. This involves integrating real-time structural performance data generated by the finite element analysis engine and dynamically adjusting the presented content (such as case animations and logic diagrams) to reduce the user's cognitive burden. Confidence-enhancing strategies include structural analogy animations, logic diagrams, and parameter influence path displays, executed through an interactive interface (such as an ANSYS plugin pop-up). Parameter acquisition involves extracting real-time stress (48MPa) and crack index (0.09) from the finite element engine; and extracting presentation parameters (such as stress concentration causes) from the optimal strategy (such as the diagram). Dynamic adjustment involves aligning the presentation parameters with real-time data (e.g., setting the diagram stress value to 48MPa) and generating visual effects through an interface rendering algorithm (based on OpenGL). The interface rendering algorithm is constructed by inputting strategy content and real-time data, mapping them to graphical elements (such as diagram nodes and animation curves), adjusting values ​​to the user's model scale (e.g., adjusting stress from 50MPa to 48MPa), and rendering pop-ups or panels. The execution steps are: loading strategy content, integrating real-time data, rendering the interactive interface, and prompting the user to view (e.g., "View stress optimization diagram").

[0082] For example, in a continuous beam bridge project, the optimal timing is 10:50, and the strategy is a logical mind map. The finite element engine provides real-time data: mid-span stress of 48 MPa and crack index of 0.09. The mind map content (stress concentration → material optimization) is adjusted to the current stress value (48 MPa), and an interactive pop-up window is generated through the ANSYS plugin (OpenGL rendering) to display the mind map: node 1 "stress concentration 48 MPa" and node 2 "optimize elastic modulus to 200 GPa". When the user clicks to view, the interface prompts "It is recommended to increase the section height by 5%", reducing the cognitive burden. The system records execution logs to ensure that the strategy is consistent with the user's model.

[0083] S10. Activate the feedback tracking mechanism to continuously monitor the feedback generated by users' innovative ideas after the implementation of the confidence enhancement strategy; feedback generated by innovative ideas includes: proposing new structural connection forms, trying non-standard material combinations, and adjusting load arrangement schemes.

[0084] Feedback generated from innovative ideas includes proposing new structural connection forms (such as novel support designs), experimenting with non-standard material combinations (such as adding fiber-reinforced concrete), and adjusting load distribution schemes (such as optimizing vehicle load distribution). The feedback tracking mechanism is based on a behavior detection algorithm, specifically: extracting user operations (such as adding a new support model, modifying material parameters) from the finite element log, and determining whether it is innovative feedback through a classification algorithm. Parameter acquisition involves: logging the operation type (model modification, parameter adjustment), parameter values ​​(such as material strength increasing from 30MPa to 35MPa), and timestamps. The classification algorithm is based on a random forest, constructed as follows: input operation features (type, value, frequency), training data consists of historical operations (1000 entries, labeled innovative / non-innovative), and output feedback category (such as "new connection form"). The algorithm steps are: extracting the operation sequence, constructing a feature vector (such as [modification type, parameter change, frequency]), inputting it into the random forest (100 trees), and outputting the feedback probability (>0.7 for innovative). The mechanism continuously monitors the strategy execution for 30 minutes after execution to ensure all feedback is captured.

[0085] For example, in a continuous beam bridge project, after the map-based strategy was executed at 10:50, the feedback tracking mechanism was activated. Logs showed that the user added a new support model (sliding support stiffness 1.1 × 10^6 N / m) at 10:55 and adjusted the material to fiber-reinforced concrete (strength 35 MPa) at 11:00, twice every 10 minutes. A random forest model (100 trees, 1000 training data points, 85% accuracy) was used as input for the feature ["support modification", 1.1 × 10^6, twice every 10 minutes], and output feedback categories "new connection form" and "non-standard material combination" with a probability of 0.78. The system recorded the feedback, confirmed that the strategy triggered innovative behavior, and transferred it to S11 storage.

[0086] S11. If a user is detected to have generated feedback on an innovative idea, the feedback, along with the time of its generation, the associated structural part, and the information of the finite element sub-model to which it belongs, will be recorded in the idea spatiotemporal axis. The idea spatiotemporal axis is a multi-dimensional dynamic data structure used to integrate and record the type, spatiotemporal attributes, and structural context of the user's innovative ideas. The content recorded in the idea spatiotemporal axis will be used to train the user's personalized cognitive model, which will be used to more accurately predict the user's innovative behavior patterns in subsequent window segmentation and strategy generation.

[0087] The idea spatiotemporal axis is a multi-dimensional dynamic data structure that records innovative idea types (e.g., new connection forms), generation times (e.g., 10:55), structural locations (e.g., support connections), and sub-model information (e.g., support mesh model with 10,000 nodes). The recording process is based on a data structure management algorithm, specifically: extracting feedback categories, timestamps, locations, and sub-model data from S10; constructing spatiotemporal axis records ([type, time, location, sub-model]); and storing them in a dynamic database. Parameter acquisition involves: obtaining feedback categories from the S10 classification algorithm, extracting timestamps from logs, and parsing locations and sub-models from finite element model metadata (e.g., ANSYS model tree). The algorithm is constructed by defining a record format (vector form), storing data through key-value pairs (e.g., "time: 10:55, location: support"), and supporting dynamic updates. These records are used to train a personalized cognitive model (subsequently in S12), predicting user innovation patterns by analyzing the spatiotemporal axis data.

[0088] For example, in a continuous beam bridge project, S10 detected feedback of "new connection type" (sliding bearing) and "non-standard material combination" (fiber-reinforced concrete) at times 10:55 and 11:00, respectively. The associated location was the bearing connection, and the sub-model was the bearing mesh (10,000 nodes). The system constructed records: ["new connection type", 10:55, "bearing", "10,000 nodes"] and ["non-standard material", 11:00, "bearing", "10,000 nodes"], and stored them in the spatiotemporal axis database. The data structure management algorithm ensured consistent record format and supported subsequent queries. The records were used to train the user cognitive model and analyze the user's innovative tendencies at the bearing location (such as a preference for novel connections).

[0089] S12. Based on the idea spacetime axis, mine users’ implicit cognitive preferences and unmet exploration needs in real time, generate new optimization and assistance needs, and trigger the iterative execution process of steps S2 to S5.

[0090] Based on the idea spatiotemporal axis, the system mines users' implicit cognitive preferences (such as preference for support optimization) and unmet exploratory needs (such as trying more material combinations) in real time, generating new optimization assistance needs and triggering S2-S5 iterations to continuously optimize the assistance process. Cognitive preferences are obtained by analyzing the distribution of idea types and locations in the spatiotemporal axis (e.g., support-related ideas account for 60%); unmet needs are extracted by detecting unimplemented ideas (e.g., proposed but not implemented load adjustments). The mining is based on clustering and pattern mining algorithms, specifically: extracting records from the spatiotemporal axis and constructing feature vectors ([idea type, location, frequency]); using the DBSCAN clustering algorithm to identify preference patterns (e.g., support optimization clusters); and detecting unimplemented ideas (e.g., load adjustments not reflected in the model). Parameter acquisition involves: extracting idea frequency (2 times / hour) and location distribution (support 60%) from the spatiotemporal axis; and verifying implementation status from logs. The DBSCAN algorithm is constructed as follows: inputting feature vectors, setting the neighborhood radius ε=0.5 and the minimum number of points 5, clustering high-frequency ideas, and outputting preferences and unmet needs. The new requirements include optimization objectives (such as optimizing support stiffness) and resource requirements (such as 2 nodes). The algorithm steps are: extract spatiotemporal axis data, perform cluster analysis, generate requirements, and trigger S2.

[0091] For example, in a continuous beam bridge project, the spatiotemporal axis records "new connection forms" and "non-standard material combinations" (supports, 2 times / hour). The DBSCAN algorithm (ε=0.5, minimum number of points 5) inputs the feature vector and clusters to show support optimization preferences (60% of ideas are concentrated on supports). The log shows that users proposed load adjustments but did not implement them, which are identified as unmet requirements. The system generates a new optimization auxiliary requirement: the optimization goal is "reduce support stress to 40MPa", and the resource requirement is "2 nodes, 0.1m mesh". The requirement triggers S2, retrieves the support optimization rule set, allocates resources, and continuously assists users in optimizing support design.

[0092] This method significantly improves the efficiency and innovation of bridge structure crack resistance optimization by integrating finite element analysis with an innovative cognitive enhancement mechanism. It dynamically divides the idea-generating time window using a temporal attention mechanism and combines it with structural response sensitivity factors to accurately capture the timing of users generating innovative ideas in key areas. Through an intent-aware module and a confidence-enhancing decision engine, it detects user behavior in real time and provides personalized confidence-enhancing strategies, reducing cognitive burden and stimulating breakthrough design ideas. The introduction of an idea spatiotemporal axis and the training of personalized user cognitive models further enable dynamic iteration of optimization assistance requirements, ensuring the continuity and depth of the innovation process, thereby effectively improving the quality and innovation level of bridge crack resistance optimization.

[0093] Figure 2 A schematic diagram of a bridge structure crack resistance optimization system based on finite element analysis is provided for embodiments of this application, such as... Figure 2 As shown, the system includes: Detection module 1 is used to detect the optimization auxiliary needs generated by users during the process of optimizing the crack resistance performance of bridge structures based on finite element analysis; Matching module 2 is used to match the initial auxiliary capabilities required to meet the optimization assistance needs; Evaluation module 3 is used to evaluate the auxiliary effects and user dependency probability of the candidate auxiliary capabilities under different reduction strategies. Select module 4, which is used to select the auxiliary capabilities that achieve the best balance between the auxiliary effect and the user's reliance probability at present as the target auxiliary capability; Assistance module 5 is used to provide assistance to the user based on the target assistance capabilities.

[0094] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for optimizing the crack resistance performance of bridge structures based on finite element analysis, characterized in that, include: S1. Detect the optimization auxiliary needs generated by users during the process of optimizing the crack resistance performance of bridge structures based on finite element analysis; S2. Match the initial auxiliary capabilities required to meet the optimization auxiliary needs; S3. Evaluate the auxiliary effects and user dependency potential of the candidate auxiliary capabilities under different reduction strategies for the initial auxiliary capabilities; S4. Select the auxiliary capability that achieves the optimal balance between auxiliary effect and user dependency at present as the target auxiliary capability; S5. Provide assistance to the user based on the target assistance capabilities.

2. The method for optimizing the crack resistance performance of bridge structures based on finite element analysis as described in claim 1, characterized in that, The detection of user-generated optimization assistance needs during the process of optimizing the crack resistance performance of bridge structures based on finite element analysis includes: Real-time monitoring of user modeling operation sequences and parameter modification behavior in finite element analysis software, identifying user operation patterns of repeatedly adjusting stress parameters in key parts of bridge structure; wherein, the key parts of bridge structure include at least the mid-span of the beam, the support connection and the cantilever end; Extract key parameter adjustment trajectories and simulation restart frequencies from the operation mode to construct user behavior feature vectors; By matching user behavior feature vectors with pre-set typical crack resistance optimization scenario templates, the weak links and computational bottlenecks of users in crack resistance performance optimization are identified, and optimization auxiliary requirements containing optimization objectives and resource requirements are generated.

3. The method for optimizing the crack resistance performance of bridge structures based on finite element analysis as described in claim 2, characterized in that, The initial assistance capabilities required to meet the matching optimization assistance needs include: Based on the optimization objectives in the optimization auxiliary requirements, retrieve the crack-resistant design rule set from the crack-resistant optimization knowledge base that matches the current bridge structure type and material properties; Based on the resource requirements in the optimization auxiliary requirements, an initial computing resource configuration scheme is allocated; wherein, the initial computing resource configuration scheme includes at least the finite element mesh granularity, the iteration convergence threshold, and the number of parallel computing nodes; By integrating the crack-resistant design rule set with the computational resource allocation scheme, an initial auxiliary capability is formed.

4. The method for optimizing the crack resistance performance of bridge structures based on finite element analysis as described in claim 3, characterized in that, The evaluation of the initial assistive capabilities under different reduction strategies, including the assistive effects and user dependency probability of the candidate assistive capabilities, includes: Based on the initial auxiliary capabilities, multiple reduction strategies are constructed; each reduction strategy includes at least a combination of: mesh coarsening ratio, upper limit of iteration steps, and rule simplification degree; Finite element simulations were performed for each reduction strategy to extract the peak stress and crack development index of key parts of the beam and to calculate the quantitative value of the auxiliary effect. Based on users' historical operation data, we analyze users' tendency to adapt to different candidate auxiliary capabilities, the frequency of autonomous adjustment, and their ability to independently achieve anti-cracking optimization goals, and comprehensively predict the likelihood of user dependence. Establish an evaluation matrix based on the dimensions of assistive effect and dependency probability to evaluate the assistive effect and user dependency probability of each candidate assistive capability.

5. The method for optimizing the crack resistance performance of bridge structures based on finite element analysis as described in claim 4, characterized in that, The candidate assistive capabilities that achieve the optimal balance between assistive effect and user dependency probability at present are selected as target assistive capabilities, including: The Pareto front search algorithm is used to identify the non-dominated solution set of aid effects and user dependency probability in the evaluation matrix; Based on the user's current operation stage and computing environment constraints, the weight coefficients of the auxiliary effect and the dependency probability are dynamically adjusted, and the comprehensive utility value of each non-dominated solution in the non-dominated solution set is calculated. The auxiliary capability with the highest comprehensive utility value was selected as the candidate capability, and the stability and robustness of the crack resistance optimization effect of the candidate capability were verified by multi-condition finite element simulation. If the verification passes, the candidate capability is determined as the target auxiliary capability; otherwise, the weight coefficients are iteratively adjusted and re-evaluated until the target auxiliary capability that meets the requirements is obtained.

6. The method for optimizing the crack resistance performance of bridge structures based on finite element analysis as described in claim 5, characterized in that, The provision of assistance to the user based on the target assistance capability includes: The target auxiliary capability is converted into an executable finite element analysis instruction set; wherein, the finite element analysis instruction set includes at least a mesh generation scheme, material constitutive model parameters, and convergence control conditions; By embedding an auxiliary engine into the finite element analysis software, the finite element analysis instruction set is dynamically loaded, guiding users to adjust modeling and solution settings in real time.

7. The method for optimizing the crack resistance performance of bridge structures based on finite element analysis as described in claim 1, characterized in that, In the process of providing assistance to users based on target assistance capabilities, an innovative cognitive enhancement mechanism is further integrated, specifically including: Based on the process expectation of assisted execution, the future time period is divided into multiple idea-inspiring time windows. Among them, the process expectation includes the correlation pattern between the structural response, parameter adjustment trajectory and historical innovation breakthroughs during the finite element analysis iteration process. Through the window division algorithm based on the temporal attention mechanism, the duration of innovative ideas generated by users after optimization in different time periods is predicted. With the goal of maximizing the duration of innovative ideas generated by users in each idea-inspiring time window, the window boundary and window length of each idea-inspiring time window are dynamically adjusted. Whenever an idea-generating window is entered, the intent-aware module is activated to continuously detect whether the user shows an intention to generate innovative ideas. If a user is detected to have expressed an intention to perform within the current idea-generating time window, the confidence enhancement decision engine is activated to determine the best time to enhance the user's confidence in innovative ideas and the corresponding confidence enhancement strategy. At the optimal time, implement confidence-enhancing strategies to users through the human-computer interaction interface; Initiate a feedback tracking mechanism to continuously monitor feedback on innovative ideas generated by users after the implementation of the confidence enhancement strategy; If a user is detected to have generated feedback on an innovative idea, the innovative idea feedback, along with its generation time, associated structural part, and finite element sub-model information, will be recorded in the idea spacetime axis. Based on the idea spacetime axis, the system mines users’ implicit cognitive preferences and unmet exploration needs in real time, generates new optimization and assistance needs, and triggers the iterative execution process of steps S2 to S5.

8. The method for optimizing the crack resistance performance of bridge structures based on finite element analysis as described in claim 7, characterized in that, The specific steps for determining the optimal timing and corresponding confidence-boosting strategies include: Construct a sequence model of auxiliary processes within the current idea-inspiring time window, predict the sustained benefits of any candidate confidence enhancement strategy executed at any candidate time point within the subsequent preset increase decay time, and evaluate the impact of cross-window weakening of the sustained benefits on the local auxiliary processes within the subsequent idea-inspiring time window. Based on users' historical behavior data, a hidden Markov model is used to predict the transition path of users' cognitive state within the current thought-stimulating time window. Combined with the structural complexity of the current finite element analysis task, the cognitive absorption potential value corresponding to each candidate opportunity is calculated. With the goals of maximizing sustained benefits, minimizing the cross-window attenuation effect, and maximizing cognitive absorption potential, a multi-objective optimization algorithm is used to solve for the optimal timing point and the corresponding confidence enhancement strategy from candidate timing and candidate confidence enhancement strategies.

9. The method for optimizing the crack resistance performance of bridge structures based on finite element analysis as described in claim 7, characterized in that, When constructing the window partitioning algorithm, a structural response sensitivity factor is introduced, which automatically extends the time window for idea generation within the time period corresponding to the stress mutation or displacement anomaly region of the finite element model, in order to adapt to the characteristic that users are more likely to generate innovative thinking in key areas.

10. A bridge structure crack resistance optimization system based on finite element analysis, characterized in that, include: The detection module is used to detect the optimization assistance needs generated by users during the process of optimizing the crack resistance performance of bridge structures based on finite element analysis. The matching module is used to match the initial assistance capabilities required to meet the optimization assistance needs; The evaluation module is used to evaluate the auxiliary effects and user dependency potential of the candidate auxiliary capabilities under different reduction strategies. The selection module is used to select the auxiliary capabilities that achieve the best balance between the auxiliary effect and the user's reliance probability at present as the target auxiliary capability. The assistance module is used to provide assistance to the user based on the target assistance capabilities.

Images