AI-based bridge design optimization system

Abstract

The application relates to the technical field of bridge design optimization, in particular to a bridge design optimization system based on AI, which comprises a structural feature identification module, a span stiffness linkage module, a shear sequence extraction module, a path strength screening module and an intelligent modeling input module.In the application, the bridge pier stiffness is normalized, a parameter comparison sequence is constructed by combining the span layout, the bridge type attribution is identified by using the symmetry trend, the span length and the stiffness difference are paired to form an offset trend mapping, the dynamic connection between the size distribution and the stress balance is established, the path screening is based on the difference between the continuous paragraph growth rate and the overall average, the path with uneven internal force distribution is identified, the logical closed loop is formed by linking the structural parameter trend, the path stress fluctuation and the bridge type attribute, the redundant path structure combination is removed, the feature expression integrity of the modeling input and the structural adaptability of the path evaluation are improved, and the AI can effectively enhance the automatic classification and path optimization capability in the bridge design.

Description

Technical Field

[0001] This invention relates to the field of bridge design optimization technology, and in particular to an AI-based bridge design optimization system. Background Technology

[0002] The field of bridge design optimization technology encompasses research and applications related to improving the functionality and optimizing resource allocation of bridge structures using engineering design theories, structural mechanics analysis methods, and parameter optimization techniques. Its core content involves mathematical modeling and numerical simulation of factors such as the size, shape, materials, and stress performance of bridge structural components to achieve rationality and economy in bridge structural design. Bridge design optimization technology covers static analysis, dynamic analysis, load combination calculations, constraint definition, and structural safety assessment. The overall technical approach typically relies on finite element analysis, sensitivity analysis, and multi-objective optimization theory for integrated decision support during the design process.

[0003] Among them, the AI-based bridge design optimization system refers to the introduction of artificial intelligence methods into the bridge design optimization process to achieve adaptive search of design parameters and intelligent recommendation of structural schemes. It covers key issues such as structural form identification, parameter combination generation, performance evaluation index correlation modeling, and design optimization path decision-making. Specifically, it uses neural networks to train on historical bridge design data, establishes the mapping relationship between structural performance and design input through classification and regression models, and uses evolutionary search methods to perform global design parameter optimization, thereby assisting in the quantitative decision-making and structural generation process in the bridge design process.

[0004] In existing bridge design processes, structural classification is typically based on engineering experience or standard structural templates, lacking a robust and scalable attribution mechanism. This makes it difficult to categorize bridge types not included in the standard templates, leading to unreasonable path selection when the relationship between structural type and design path is unclear. The comparison of structural dimensions and stiffness often relies on static segment comparisons, failing to establish span continuity indicators and resulting in a lack of trend-based expression in judging bridge symmetry and component balance. Path analysis frequently uses point-valued shear stress for evaluation, neglecting the sequential relationship of shear response between nodes and ignoring the impact of continuously rising shear stress in local areas on path stability. Path selection is also largely dependent on historical path templates or structural experience scores, lacking computational logic directly related to the current structural response. This makes it difficult to selectively eliminate paths based on their mechanical adaptability under different bridge types, resulting in missing indicators in path evaluation and insufficient generalization ability of model training results. In the data preprocessing process before AI modeling, issues such as fuzzy structural parameter feature labels and unclear path weights weaken the ability to identify structural information in intelligent models and limit the data-driven efficiency of bridge optimization. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing an AI-based bridge design optimization system.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: An AI-based bridge design optimization system includes:

[0007] The structural feature recognition module obtains the longitudinal stiffness value of each pier, combines the bridge span information to combine adjacent pier pairs, calculates the relative stiffness difference between adjacent pier pairs, extracts the maximum and minimum values ​​and calculates the span range, determines the bridge structure, and generates a bridge structure classification label.

[0008] The span stiffness linkage module extracts the span length of adjacent spans based on the bridge structure classification labels, calculates the span stiffness ratio of adjacent spans, pairs it with the stiffness difference of the bridge piers in each span, calculates the cumulative offset of stiffness difference, and determines whether it conforms to linear change, generating span stiffness linkage trend analysis results.

[0009] Based on the span stiffness linkage trend analysis results, the shear sequence extraction module collects the initial shear stress and peak shear stress and calculates the shear stress growth rate, constructs the shear ratio change sequence, extracts the segments where the continuous node increase rate is continuously rising, records the number of segments and the maximum amplitude value, and generates a path shear trend sequence set.

[0010] The path strength screening module, based on the path shear trend sequence set, statistically analyzes the average shear stress growth rate and the overall average shear stress growth rate, compares and analyzes to determine and eliminate paths with uneven stress distribution within the structure, and generates a path stress screening list.

[0011] As a further aspect of the present invention, the bridge structure classification label includes a structure type identifier, a symmetry evaluation coefficient, and a stiffness distribution level; the span stiffness linkage trend analysis results include a span ratio mapping relationship, a stiffness offset trend, and a structural equilibrium matching degree; the path shear trend sequence set includes the number of shear growth segments, the peak increase index, and the path trend continuity; the path stress screening list includes path screening marks, stress concentration identification levels, and path retention numbers.

[0012] As a further aspect of the present invention, the structural feature recognition module includes:

[0013] The stiffness normalization submodule obtains the longitudinal stiffness values ​​of each pier under the bridge design conditions. Combining the standard parameters for the longitudinal stiffness of piers in the bridge design specifications, it selects the maximum longitudinal stiffness of the entire bridge as the pier stiffness benchmark value. It then performs division operations between the longitudinal stiffness of all piers and the pier stiffness benchmark value in sequence to obtain the dimensionless normalized stiffness value of each pier. Finally, it sorts the piers according to their numbers to generate a pier normalized stiffness sequence.

[0014] The span combination submodule, based on the normalized stiffness sequence of the piers and combined with the length and numbering order of each span in the bridge span information, combines adjacent piers into pier pairs according to the adjacent numbering method. For each pier pair, it calls the normalized stiffness values ​​corresponding to the two piers at both ends, calculates the relative stiffness difference, extracts the maximum and minimum values ​​among all relative stiffness differences, calculates the difference, and outputs the span of the adjacent stiffness difference.

[0015] The structural classification submodule organizes the length and distribution order of each continuous span based on the bridge span information. For the span of the adjacent stiffness difference within each span, it determines whether there is a symmetrical trend along the center position of the bridge span. If there is a symmetrical trend in multiple continuous spans, it is marked as a continuous beam structure. If there is no symmetrical trend, it is classified as a composite or cantilever structure, and a bridge structure classification label is generated.

[0016] As a further aspect of the present invention, the span stiffness linkage module includes:

[0017] The stiffness ratio calculation submodule obtains bridge data marked as continuous structure in the bridge structure classification label, extracts the span length of adjacent spans and the longitudinal stiffness value of the corresponding piers, calculates the span stiffness ratio of adjacent spans, and arranges them in the order of span number to establish a span stiffness ratio value sequence.

[0018] The difference pairing submodule, based on the segment stiffness ratio sequence, pairs the comparison values ​​with the stiffness differences item by item according to the segment number order, calls the stiffness ratio in each pair of paired values ​​and sorts them in ascending order, extracts the corresponding stiffness differences according to the sorting order, calculates the cumulative offset of stiffness differences, and establishes a sequence of cumulative offset of stiffness differences.

[0019] The trend judgment submodule extracts the segment number and corresponding offset value of each span based on the cumulative offset sequence of stiffness differences, and judges whether the offset value in the sequence shows a linear growth trend with the increase of the span number. If the trend is true, it is marked as a linear enhancement relationship; if it is not true, it is marked as a nonlinear relationship, and the span stiffness linkage trend analysis result is generated.

[0020] As a further aspect of the present invention, the cumulative offset of stiffness difference is calculated using the following formula:

[0021] ;

[0022] Calculations are performed, in which, Representing the The cumulative offset of stiffness difference of the term, Represents the sorted order The stiffness difference of the term, This represents the stiffness difference of the preceding term. Represents the sorted order The stiffness ratio of the terms, Represents the sorted order The stiffness ratio of the terms, This indicates the cumulative total from the first item to the second item. The summation operation of the terms uses a constant 1 in the denominator to avoid the case where the denominator is zero.

[0023] As a further aspect of the present invention, the cleavage sequence extraction module includes:

[0024] Based on the span stiffness linkage trend analysis results, the shear stress extraction submodule collects the initial shear stress value and peak shear stress value in all path nodes, calculates the shear stress growth rate of all nodes, arranges the corresponding growth rates in the order of node number, and generates a node shear stress growth rate sequence.

[0025] The shear ratio generation submodule, based on the node shear stress growth rate sequence, calls the node number order and connects the corresponding growth rates in sequence to construct a continuous change sequence, calculates the change magnitude between adjacent growth rates, arranges all changes in the node number order, and generates a shear ratio change sequence.

[0026] The trend segment identification submodule identifies segments in three or more consecutive nodes where the shear ratio change value is continuously increasing positively based on the shear ratio change sequence, extracts all segments that meet the conditions and counts their number, and at the same time extracts the amplitude value of the interval with the largest increase in all segments to establish a path shear trend sequence set.

[0027] As a further aspect of the present invention, the variation range between adjacent growth rates is determined by the following formula:

[0028] ;

[0029] Calculations are performed, in which, The growth rate of shear stress is represented in the first... With the The intensity of the shear ratio change between nodes The representative number is the The rate of increase of shear stress at each node, Representing the subsequent The rate of increase of shear stress at each node, This represents the absolute value of the sum of the growth rates of adjacent nodes. This is the square of the difference in the rate of increase of shear stress. Used to regularize the current node's growth rate and avoid division by zero.

[0030] As a further aspect of the present invention, the path strength screening module includes:

[0031] The growth rate statistics submodule extracts the shear stress growth rate of three consecutive nodes in each rising segment based on the continuous rising segments contained in each path shear trend sequence set, calculates the average shear stress growth rate of each segment in turn, and pairs it with the average value of the shear stress growth rate of all nodes in the corresponding path to generate path shear stress growth ratio data.

[0032] The abnormal path identification submodule extracts the average value of the rising segment and the overall average value of the corresponding path from each pair of paired values ​​based on the path shear stress growth ratio data. It then determines whether there is a path where the average value of any rising segment is greater than twice the overall average value. If the condition is met, the corresponding path is marked as an abnormal stress concentration path, and a list of stress concentration path numbers is obtained.

[0033] The stress path elimination submodule removes corresponding paths from the path shearing trend sequence based on the stress concentration path number list, retains the remaining path numbers and corresponding sequences, and organizes and generates a path stress screening list.

[0034] As a further aspect of the present invention, the system also includes an intelligent modeling input module;

[0035] The intelligent modeling input module binds the corresponding path shear change information and the bridge structure classification label according to the path stress screening list. It inputs the structure classification, span ratio trend and shear sequence characteristics into the AI ​​model. Through training data, it learns the correlation between bridge structure parameters and mechanical response, performs bridge AI modeling analysis, and obtains AI bridge design optimization results.

[0036] The AI ​​bridge design optimization results include the structural attribution input set, the path optimization level group, and the AI ​​training feedback data.

[0037] As a further aspect of the present invention, the intelligent modeling input module includes:

[0038] The structural data binding submodule extracts the bridge structure classification label and corresponding path shear ratio change data for each path based on the node information of the retained paths in the path stress screening list, calls the structural label and path shear sequence for association matching, and generates a path structure shear combination set.

[0039] The feature data integration submodule, based on the path structure shear combination set, combines the structural attribution labels of each path, the trend data of the span stiffness ratio, and the shear stress growth sequence to establish the structural and mechanical response feature combination items of each path, and constructs the input data matrix in a unified format to obtain the modeling input feature matrix.

[0040] The model association construction submodule performs interactive construction of bridge structural feature items and shear response data items corresponding to the path number based on the modeling input feature matrix. It compares the shear stress trend distribution and stiffness change sequence between each group of inputs according to the structural type, extracts the shear response concentration under each structural category, performs structure-response parameter mapping, and establishes AI bridge design optimization results.

[0041] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0042] In this invention, a parameter comparison sequence is constructed by normalizing the pier stiffness and combining it with the span layout. The symmetry trend is used to identify the bridge type, enhancing the ability to distinguish non-standard structures. The pairing of span length and stiffness difference forms an offset trend mapping, establishing a dynamic connection between size distribution and stress equilibrium. The shear growth sequence constructs a trend path based on the initial value and peak increase of nodal shear stress, avoiding local failures in path evaluation. Path screening is based on the difference between the growth rate of continuous segments and the overall mean, identifying paths with uneven internal force distribution. By linking structural parameter trends, path stress fluctuations, and bridge type attributes to form a logical closed loop, redundant path structure combinations are eliminated, and a responsive parameter sequence for AI models is constructed. This improves the completeness of feature expression of modeling input and the structural adaptability of path evaluation. Based on AI, the automatic classification and path selection capabilities in bridge design can be effectively enhanced. Attached Figure Description

[0043] Figure 1 This is a system flowchart of the present invention;

[0044] Figure 2 This is a system module diagram of the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0046] In the description of this invention, it should be understood that the terms length, width, up, down, front, back, left, right, vertical, horizontal, top, bottom, inside, outside, etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "multiple" means two or more, unless otherwise explicitly specified.

[0047] Please see Figure 1 and Figure 2The AI-based bridge design optimization system includes:

[0048] The structural feature recognition module obtains the longitudinal stiffness values ​​of each pier under the bridge design conditions (standard parameters, referring to bridge design specifications), normalizes the stiffness of all piers based on the maximum value of the entire bridge, combines the bridge span information (span length and distribution) to combine adjacent pier pairs, calculates the relative stiffness difference between adjacent pier pairs (stiffness difference between adjacent piers, a dimensionless parameter used to measure the uniformity of stiffness distribution), extracts the maximum and minimum values ​​and calculates the span range, and determines whether the span range is symmetrical along the center of the bridge span in multiple continuous spans. If the trend is true, it is marked as a continuous beam structure; otherwise, it is identified as a composite or cantilever structure, and a bridge structure classification label is generated.

[0049] The span stiffness linkage module extracts the span length (actual design parameter) of adjacent spans based on the continuous structural attributes marked by the bridge structure classification label, calculates the span stiffness ratio of adjacent spans (the proportional coefficient of span length and corresponding pier stiffness), pairs it with the span pier stiffness difference item by item, arranges them in ascending order of proportion, calculates the cumulative offset of stiffness difference, and determines whether it conforms to the linear change of stiffness symmetry enhancement when the span increases, generating span stiffness linkage trend analysis results.

[0050] Based on the span stiffness linkage trend analysis results, the shear sequence extraction module collects the initial shear stress (initial load value) and peak shear stress (load limit value) in all path nodes, calculates the shear stress growth rate, constructs a shear ratio change sequence according to the node order, extracts the segment with three or more consecutive nodes where the increase rate continues to rise, records the number of segments and the maximum amplitude value, and generates a path shear trend sequence set.

[0051] The path strength screening module calculates the average shear stress growth rate based on the continuous rising segment of each path in the path shear trend sequence set, and compares it with the overall average shear stress growth rate of the path. If the average shear stress growth rate of any three consecutive nodes is greater than twice the overall average shear stress growth rate, the corresponding path is determined to be a path with uneven stress distribution within the structure (high risk of local stress concentration) and is removed, generating a path stress screening list.

[0052] The intelligent modeling input module binds the corresponding path shear change information and bridge structure classification labels to the structural characteristics of the paths retained in the path stress screening list. It inputs the structural classification, span ratio trend, and shear sequence characteristics into the AI ​​model, learns the correlation between bridge structural parameters and mechanical response through training data, performs bridge AI modeling analysis, and obtains AI bridge design optimization results.

[0053] The bridge structure classification labels include structural type identifier, symmetry evaluation coefficient, and stiffness distribution level; the span stiffness linkage trend analysis results include span ratio mapping relationship, stiffness offset trend amount, and structural equilibrium matching degree; the path shear trend sequence set includes the number of shear growth segments, peak increase index, and path trend continuity; the path stress screening list includes path screening marks, stress concentration identification level, and path retention number; the AI ​​bridge design optimization results include structural classification input set, path optimization level group, and AI training feedback data.

[0054] Please see Figure 2 The structural feature recognition module includes a stiffness normalization submodule, a cross-segment combination submodule, and a structural classification submodule.

[0055] The stiffness normalization submodule obtains the longitudinal stiffness values ​​of each pier under the bridge design conditions. Combining the standard parameters for the longitudinal stiffness of piers in the bridge design specifications, it selects the maximum longitudinal stiffness of the entire bridge as the pier stiffness benchmark value. It then performs division operations between the longitudinal stiffness of all piers and the pier stiffness benchmark value in sequence to obtain the dimensionless normalized stiffness value of each pier. Finally, it sorts the piers according to their numbers to generate a pier normalized stiffness sequence.

[0056] In the stiffness normalization submodule, the bridge design condition document retrieves the longitudinal stiffness parameters of each pier. For example, if the stiffness values ​​of three piers of a certain bridge are 3.0×10^5 kN / m, 4.0×10^5 kN / m, and 5.0×10^5 kN / m, respectively, according to the longitudinal stiffness standard parameter requirements in Article 5.3.2 of JTGD60-2015, the system iterates through all pier stiffness values, compares them, and obtains the maximum stiffness value of 5.0×10^5 kN / m as the benchmark value. Then, it performs a division operation between each pier stiffness value and the benchmark value: 3.0×1 0^5 / 5.0×10^5=0.6, 4.0×10^5 / 5.0×10^5=0.8, 5.0×10^5 / 5.0×10^5=1.0, and the sequence [0.6, 0.8, 1.0] is generated by arranging the pier numbers P1-P3 in order. In a certain case, the benchmark value of a five-span continuous beam bridge is taken from the stiffness of its No. 3 pier, which is 8.5×10^5 kN / m. The normalized results of the other piers are 0.71, 0.82, 1.0, 0.94, and 0.88, respectively, generating the normalized stiffness sequence of the piers.

[0057] The span combination submodule is based on the normalized stiffness sequence of the piers. It combines the length and numbering order of each span in the bridge span information, and combines adjacent piers into pier pairs according to the adjacent numbering method. For each pier pair, it calls the normalized stiffness values ​​corresponding to the two piers, calculates the relative stiffness difference, and extracts the maximum and minimum values ​​among all relative stiffness differences. It calculates the difference and outputs the span of the adjacent stiffness difference.

[0058] In the cross-segment combination submodule, based on the previous normalized sequence [0.6, 0.8, 1.0], and combined with the three cross-segment lengths recorded in the cross-segment information table (30m, 30m, 30m respectively), adjacent piers P1-P2 and P2-P3 are combined into pier pairs. For P1-P2, the relative stiffness difference is calculated using 0.6 and 0.8: |0.8-0.6|=0.2. For P2-P3, the relative stiffness difference is calculated using 0.8 and 1.0: |1.0-0.6|=0.2. 8|=0.2, traverse the set of all pier pair difference values ​​[0.2, 0.2], take the maximum value 0.2 and the minimum value 0.2 to calculate the difference span: 0.2-0.2=0. In a certain case, a bridge measured the normalized sequence [0.5, 0.7, 0.9, 0.6], the difference values ​​of the combined pier pairs are 0.2, 0.2, 0.3, the difference span between the maximum difference 0.3 and the minimum difference 0.2 is 0.1, output the adjacent stiffness difference span.

[0059] The structural classification submodule organizes the length and distribution order of each continuous span based on the bridge span information. It determines whether the difference in stiffness between adjacent spans within each span is symmetrical along the center of the bridge span. If it is symmetrical in multiple continuous spans, it is marked as a continuous beam structure. If it is not symmetrical, it is classified as a composite or cantilever structure, and a bridge structure classification label is generated.

[0060] In the structural classification submodule, the bridge span information table shows that the length of each of the three spans is 30m, distributed in the order of span 1 to span 3. The set of adjacent stiffness difference spans [0, 0, 0] corresponding to each span is extracted. When judging symmetry, the system sets the center position of the span to 15m of span 2 and compares whether the difference spans of span 1 and span 3 are equal. In this example, the difference span of span 1 is 0 and the difference span of span 3 is 0, which is judged as a symmetrical trend. When the difference span sequence of a certain case is [0.1, 0.3, 0.1], it is also judged as symmetrical and marked as a continuous beam structure. When the difference span sequence of a certain bridge is [0.2, 0.5, 0.3], the difference of span 1 is 0.2 and the difference of span 3 is 0.3, which is not equal, and it is judged as not showing a symmetrical trend. It is classified as a composite structure, and a bridge structure classification label is generated.

[0061] Please see Figure 2 The span stiffness linkage module includes a stiffness ratio calculation submodule, a difference pairing submodule, and a trend judgment submodule.

[0062] The stiffness ratio calculation submodule obtains bridge data marked as continuous structure in the bridge structure classification label, extracts the span length of adjacent spans and the longitudinal stiffness value of the corresponding piers, calculates the span stiffness ratio of adjacent spans, and arranges them in the order of span number to establish a span stiffness ratio value sequence.

[0063] In the stiffness ratio calculation submodule, the database retrieves bridge IDs marked with continuous structure in the bridge structure classification tags. For example, for a continuous beam bridge with ID B-2035, its span information table records the span lengths of spans 1 to 3 as 40m, 50m, and 40m respectively. The corresponding longitudinal stiffness values ​​of piers P1 to P4 are 2.5×10^5kN / m, 3.0×10^5kN / m, 3.8×10^5kN / m, and 2.8×10^5kN / m. The span lengths of adjacent spans are extracted along with the corresponding pier stiffness values. The span length ratio of span 1 to span 2 is 40 / 50=0.8, and the stiffness ratio of piers P1 to P2 is 2.5×10^5 / 3.0×10^5≈0.833. The span length ratio of span 2 to span 3 is 50 / 40 = 1.25, corresponding to the stiffness ratio of piers P2 and P3 of 3.0×10^5 / 3.8×10^5≈0.789. The span length ratio of span 3 to span 4 (assuming span 4 is a virtual extension) is 40 / 0 (no actual span) set to 1, corresponding to the stiffness ratio of 3.8×10^5 / 2.8×10^5≈1.357. The system arranges the stiffness ratio values ​​in the order of span numbering as [0.833, 0.789, 1.357]. In a certain actual case, the span length ratio of a five-span continuous bridge is [0.6, 1.2, 0.9, 1.1], corresponding to the stiffness ratio sequence of [0.75, 0.85, 1.2, 0.95]. Establish the span stiffness ratio value sequence.

[0064] The difference pairing submodule is based on the cross-segment stiffness ratio sequence. It compares the values ​​and stiffness differences item by item according to the cross-segment number order, calls the stiffness ratio in each pair of values ​​and sorts them in ascending order, extracts the corresponding stiffness difference in sorting order, calculates the cumulative offset of stiffness difference, and establishes a sequence of cumulative offset of stiffness difference.

[0065] The cumulative offset due to stiffness differences is calculated using the following formula:

[0066] ;

[0067] Calculations are performed, in which, Representing the The cumulative offset of stiffness difference of the term, Represents the sorted order The stiffness difference of the term, This represents the stiffness difference of the preceding term. Represents the sorted order The stiffness ratio of the terms, Represents the sorted order The stiffness ratio of the terms, This indicates the cumulative total from the first item to the second item. The summation operation of the terms uses a constant 1 in the denominator to avoid the case where the denominator is zero.

[0068] Stiffness difference and The longitudinal stiffness values ​​of adjacent bridge piers are measured using on-site monitoring equipment, and the stiffness difference between adjacent bridge piers is calculated.

[0069] Stiffness ratio and The longitudinal stiffness values ​​of adjacent bridge piers are measured using on-site monitoring equipment, and the stiffness ratio between adjacent bridge piers is calculated.

[0070] Taking a certain bridge structure as an example, the following data has been obtained:

[0071] The longitudinal stiffness values ​​of piers P1 to P4 are as follows: kN / m, kN / m, kN / m, kN / m.

[0072] Stiffness difference between adjacent bridge piers The calculation is as follows:

[0073] kN / m;

[0074] kN / m;

[0075] kN / m;

[0076] Stiffness ratio between adjacent bridge piers The calculation is as follows:

[0077] ;

[0078] ;

[0079] ;

[0080] Calculation process:

[0081] Calculate the cumulative offset of the first item. :

[0082] because and It does not exist, let it be. kN / m, ;

[0083] ;

[0084] Calculate the cumulative offset of the second term. :

[0085] ;

[0086] Calculate the cumulative offset of the third item. :

[0087] ;

[0088] This result indicates that the cumulative offset The cumulative change in stiffness from item 1 to item 3 is 1.2821, reflecting the trend of stiffness change in the bridge structure across these three segments.

[0089] The formula's operational logic lies in comprehensively evaluating the changing trend of stiffness differences through two key dimensions. The first part is the absolute value of the rate of change of adjacent stiffness differences, which normalizes the relative change between the current stiffness difference and the previous difference. By adding 1 to the denominator, the division by zero is avoided, and numerical stability is controlled. This term measures the degree of abrupt stiffness changes. The second part is the square root of the square of the stiffness ratio difference divided by 2, reflecting the relative fluctuation amplitude of adjacent stiffness ratios. The squaring operation ensures the value is non-negative and amplifies the difference, while the square root operation is used to reduce the nonlinear effect of fluctuations. The sum of the two parts can achieve a joint expression of the change amplitude of stiffness differences and the changing trend of stiffness ratios, thus forming a comprehensive offset measure of the current segment stiffness state. This cumulative sum sequence can be further used by the trend judgment submodule to analyze the structural change characteristics.

[0090] The trend judgment submodule extracts the span number and corresponding offset value based on the cumulative offset sequence of stiffness differences, and judges whether the offset value in the sequence shows a linear growth trend as the span number increases. If the trend is valid, it is marked as a linear enhancement relationship; if not, it is marked as a nonlinear relationship, and the span stiffness linkage trend analysis results are generated.

[0091] In the trend judgment submodule, taking the previous calculation results as an example, the cumulative offset sequence of stiffness differences for spans 1 to 3 has been obtained as [0, 0.6311, 1.2821]. Further, the offset increment difference between each adjacent span is extracted, and the offset difference between span 2 and span 1 is calculated to be 0.6311, and the offset difference between span 3 and span 2 is 0.651. The rate of change of the adjacent offset difference is (0.651-0.6311) / 0.6311≈3.16%, which is within the threshold range of ±20%. Therefore, it is determined that the offset value in this sequence shows a linear growth trend with the increase of the span number, and is marked as a linear enhancement relationship. The span stiffness linkage trend analysis result is a linear enhancement relationship.

[0092] Please see Figure 2 The shear sequence extraction module includes a shear stress extraction submodule, a shear ratio generation submodule, and a trend segment identification submodule;

[0093] Based on the span stiffness linkage trend analysis results, the shear stress extraction submodule collects the initial shear stress value and peak shear stress value in all path nodes, calculates the shear stress growth rate of all nodes, arranges the corresponding growth rates in the order of node number, and generates a node shear stress growth rate sequence.

[0094] In the shear stress extraction submodule, based on the bridge ID list in the span stiffness linkage trend analysis results, bridge data marked as linear or nonlinear relationships are selected. For example, for a continuous beam bridge with bridge ID C-3072, the initial shear stress values ​​of nodes N1 to N5 recorded in its path node table are 50kPa, 55kPa, 60kPa, 58kPa, and 62kPa, respectively, and the peak shear stress values ​​are 75kPa, 80kPa, 85kPa, 78kPa, and 88kPa. The shear stress growth rate of each node is calculated as follows: Node N1 growth rate = (75-50) / 50 = 50%, Node N2 growth rate = (80-55) / 55 ≈ 45.45%, and Node N3 growth rate... = (85-60) / 60≈41.67%, the growth rate of node N4 = (78-58) / 58≈34.48%, the growth rate of node N5 = (88-62) / 62≈41.94%, and the growth rate sequence [50, 45.45, 41.67, 34.48, 41.94] is generated according to the node number order. In a certain case, the initial shear stress of nodes N10 to N15 of an arch bridge is [30, 32, 35, 33, 38] kPa, and the peak shear stress is [45, 48, 50, 47, 52] kPa. The growth rate sequence [50, 50, 42.86, 42.42, 36.84] is calculated, and the node shear stress growth rate sequence is generated.

[0095] The shear ratio generation submodule is based on the node shear stress growth rate sequence. It calls the node number order and connects the corresponding growth rates in sequence to construct a continuous change sequence. It calculates the change magnitude between adjacent growth rates, arranges all changes in the node number order, and generates a shear ratio change sequence.

[0096] The magnitude of change between adjacent growth rates is determined using the formula:

[0097] ;

[0098] Calculations are performed, in which, The growth rate of shear stress is represented in the first... With the The intensity of the shear ratio change between nodes The representative number is the The rate of increase of shear stress at each node, Representing the subsequent The rate of increase of shear stress at each node, This represents the absolute value of the sum of the growth rates of adjacent nodes. This is the square of the difference in the rate of increase of shear stress. Used to regularize the current node's growth rate and avoid division by zero.

[0099] Taking path nodes N1 to N5 of bridge number C-3072 as an example, the initial and peak shear stress data collected for this node are as follows (unit: kPa):

[0100] N1: Initial shear stress 50, peak stress 75;

[0101] N2: Initial shear stress 55, peak stress 80;

[0102] N3: Initial shear stress 60, peak stress 85;

[0103] N4: Initial shear stress 58, peak stress 78;

[0104] N5: Initial shear stress 62, peak stress 88;

[0105] Calculate the shear stress growth rate:

[0106] ;

[0107] ;

[0108] Will , Substitute into the formula to calculate :

[0109] ;

[0110] The results indicate that the shear ratio change intensity between nodes 1 and 2 is 0.03826, with a larger value representing a more pronounced abrupt change in the growth rate. This shear ratio change intensity value will be used as an element in the shear ratio change sequence and for the extraction and analysis of continuous positive growth segments in the trend segment identification submodule.

[0111] This formula comprehensively expresses the intensity of the change in the shear stress growth rate of adjacent nodes through a two-part structure. The first part is the ratio of the absolute value of the difference between adjacent growth rates to the square root of the current node's growth rate plus 1. This structure is used to standardize the rate of change of the growth rate. The constant 1 is added to the denominator to avoid division by zero anomalies, and the square root operation can suppress the exaggerated influence of the change amplitude when the current node's growth rate is large. The second part is the ratio of the square of the difference in growth rates to the absolute value of the sum of the growth rates of the two nodes plus 1. This part is used to measure the energy amplitude of the change in the shear stress growth rate. The squaring operation enhances the response to the change amplitude, and adding 1 to the denominator also plays a normalization role. The absolute value ensures balanced treatment of positive and negative fluctuations. The sum of the two terms in the overall formula structure is a comprehensive evaluation that takes into account both the rate of change and the amplitude of the trend change, so that the intensity of the shear ratio change remains dynamically sensitive and stable under different combinations of growth rates.

[0112] The trend segment identification submodule identifies segments in three or more consecutive nodes where the shear ratio change value is continuously increasing positively based on the shear ratio change sequence, extracts all segments that meet the conditions and counts them, and at the same time extracts the amplitude value of the interval with the largest increase in all segments to establish a path shear trend sequence set.

[0113] The trend segment identification submodule, based on the above shear stress growth rate data sequence [50, 45.45, 41.67, 34.48, 41.94], calculates the first two terms of the shear ratio change sequence as nodes N1-N2: Subsequent nodes are constructed sequentially to obtain the shear ratio change value. Corresponding to N2-N3, N3-N4, and N4-N5 respectively, the results are obtained by substituting the shear stress growth rate of adjacent nodes into the formula, and the rearranged sequence is [ ]; Traverse the shear ratio change sequence and check if there are three consecutive nodes where the shear ratio change values ​​show a positive increasing trend, for example, if , , If nodes N1-N4 constitute a continuous growth segment that satisfies positive increasing, its cumulative growth rate is extracted as the sum of the three differences. If the growth rate of this segment is the maximum value among all positive growth segments, it is recorded as the maximum growth segment in the path shearing trend sequence. Finally, the number of all shearing trend growth segments that meet the conditions is counted, and a path shearing trend sequence set is established.

[0114] Please see Figure 2 The path strength screening module includes a growth rate statistics submodule, an abnormal path identification submodule, and a stress path elimination submodule.

[0115] The growth rate statistics submodule extracts the shear stress growth rate of three consecutive nodes in each rising segment based on the continuous rising segments contained in each path shear trend sequence set. It then calculates the average shear stress growth rate of each segment in turn and pairs it with the average shear stress growth rate of all nodes in the corresponding path to generate path shear stress growth ratio data.

[0116] In the growth rate statistics submodule, the continuously rising segment data of path P-102 in the path shear trend sequence set is called. For example, this path contains a continuous three-node shear stress growth rate sequence [50%, 45%, 55%] for nodes N6-N8. The average value of this segment is calculated as (50+45+55) / 3=50%. At the same time, the growth rate sequence [50, 45, 55, 40, 35] for all nodes N6-N10 of path P-102 is extracted, and the overall average value is calculated as (50+45+55+40+35) / 5=45. The average value of the rising segment (50%) and the average value of the overall segment (45%) are paired and recorded as (50, 45). In a cable-stayed bridge case, path Q-205 contains rising segments [30%, 32%, 34%] with an average value of 32% and an overall path average value of 28.5%, and the paired record is (32, 28.5). In another suspension bridge path R-307, the average value of the rising segment [22%, 25%, 28%] is 25% and the overall path average value is 20.3%, and the paired record is (25, 20.3). The path shear stress growth ratio data are generated.

[0117] The abnormal path identification submodule extracts the average value of the rising segment and the overall average value of the corresponding path from the path shear stress growth ratio data, and determines whether there is a path where the average value of any rising segment is greater than twice the overall average value. If the condition is met, the corresponding path is marked as an abnormal stress concentration path, and a list of stress concentration path numbers is obtained.

[0118] In the abnormal path identification submodule, the paired values ​​in the growth ratio data are traversed. For example, for path P-102, the paired value is (50, 45). The ratio of the average value of the rising segment (50) to the overall average value (45) is calculated to be 50 / 45≈1.11 times, which does not exceed the two-times threshold, and is therefore determined to be a normal path. In a certain case, the paired value of path Q-205 (32, 28.5) has a ratio of 32 / 28.5≈1.12 times, which is still below the threshold. When the average value of the rising segment of a certain arch bridge path S-409 is 48%, the ratio of the average value of the rising segment to the overall average value is 2... When the pairing is 0%, the ratio of 48 / 20 = 2.4 times is calculated, which exceeds the threshold of 2 times. The path is marked as an abnormal stress concentration path. In another case, the pairing value (43, 18) of path T-503 is 2.39 times, which also triggers the marking. After the system traverses all paths, for example, if a project contains 10 paths, 3 of them have a ratio of more than 2 times (paths S-409, T-503, U-601), a list of stress concentration path numbers [S-409, T-503, U-601] is generated.

[0119] The stress path elimination submodule removes the corresponding paths from the path shear trend sequence based on the stress concentration path number list, retains the remaining path numbers and corresponding sequences, and organizes and generates a path stress screening list.

[0120] In the stress path elimination submodule, the list of stress concentration path numbers [S-409, T-503, U-601] is read. The corresponding numbers are deleted from the original path shear trend sequence set (including paths P-102, Q-205, R-307, S-409, T-503, U-601, V-704). Paths P-102, Q-205, R-307, and V-704 are retained. In a bridge engineering case, the original sequence set contains 15 paths. After identifying 5 abnormal paths, the elimination operation leaves 10 paths. For example, the shear ratio change sequence [1.2, 1.5, 1.8] of path V-704 is retained, and the sequence [0.8, 1.1, 1.3] of path W-805 is also retained because it is not marked. The system outputs the path stress screening list [P-102, Q-205, R-307, V-704].

[0121] Please see Figure 2 The intelligent modeling input module includes a structural data binding submodule, a feature data integration submodule, and a model association construction submodule;

[0122] The structural data binding submodule extracts the bridge structure classification label and corresponding path shear ratio change data for each path based on the node information of the retained paths in the path stress screening list. It then calls the structural label and path shear sequence for association matching to generate a path structure shear combination set.

[0123] In the structural data binding submodule, path numbers P-102, Q-205, R-307, and V-704 are read from the path stress filtering list. For example, the bridge structure corresponding to path P-102 is classified as a continuous beam structure, and its shear ratio change data is a sequence [0.5, 0.6, 0.7]. The label for path Q-205 is a composite structure, corresponding to the shear sequence [1.2, 1.3, 1.4]. The system calls the label field and the shear sequence to perform key-value pairing. The association generates a path structure shearing combination set {P-102: [continuous beam, [0.5, 0.6, 0.7]], Q-205: [combined, [1.2, 1.3, 1.4]]}. In a real-world case, path R-307 is labeled as a cantilever-type associated shearing sequence [0.8, 0.9, 1.0], and path V-704 is labeled as a continuous beam associated sequence [0.4, 0.5, 0.6]. The combination set is expanded into an associated dataset containing four paths.

[0124] The feature data integration submodule is based on the path structure shear combination set. It combines the structural attribution labels of each path, the trend data of the span stiffness ratio, and the shear stress growth sequence to establish the structural and mechanical response feature combination items of each path. It also constructs the input data matrix in a unified format to obtain the modeling input feature matrix.

[0125] In the feature data integration submodule, for a single data item in the path structure shear combination set, such as the feature items of path P-102 which include the structural label continuous beam, span stiffness ratio trend data (taken from historical module output) [0.8, 0.9, 1.0], and shear stress growth sequence [50%, 45%, 55%], it is converted into a feature vector [continuous beam, 0.8, 0.9, 1.0, 50, 45, 55]. After all paths are converted according to this rule, the input data is formed. According to the matrix: the rows represent paths (P-102, Q-205, etc.), and the columns are the structure type code (continuous beam = 1, combined = 2), stiffness ratio trend data (3 columns), and shear stress growth rate (3 columns). For example, the data in the first row of the matrix is ​​[1, 0.8, 0.9, 1.0, 50, 45, 55]. The data in the corresponding row of path Q-205 in a certain case is [2, 1.2, 1.1, 0.9, 60, 58, 62]. The generated modeling input feature matrix contains four rows and seven columns of data.

[0126] The model association construction submodule performs interactive construction of bridge structural feature items and shear response data items corresponding to the path number based on the modeling input feature matrix. It compares the shear stress trend distribution and stiffness change sequence between each group of inputs according to structural type, extracts the shear response concentration under each structural category, performs structure-response parameter mapping, and establishes AI bridge design optimization results.

[0127] In the model association construction submodule, each row of the input feature matrix is ​​traversed. For example, the structure type of path P-102 is 1 (continuous beam), its shear stress growth rate sequence is [50, 45, 55], and its span stiffness ratio sequence is [0.8, 0.9, 1.0]. The calculated shear stress concentration is (50+45+55) / 3=50, and the stiffness change range is 1.0-0.8=0.2. The structure type of path Q-205 is 2 (combined), and the shear stress concentration is (60+58+62) / 3=60, with a stiffness change range of 0.9-1.2=-0.3. Structural type grouping statistics: The average shear stress concentration of continuous beam paths is (50+40) / 2=45 (assuming the shear stress of another path V-704 is 40), the average for composite beams is 60, and the average for cantilever beams is (55+58) / 2=56.5 (assuming the shear stress of path R-307 is 55, and another path is 58). A structure-response parameter mapping table is established: {continuous beam: 45, composite beam: 60, cantilever beam: 56.5}. In a certain actual case, the mapping table is {continuous beam: 38, composite beam: 52, cantilever beam: 48}. The AI ​​bridge design optimization results are generated.

[0128] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. An AI-based bridge design optimization system, characterized by: The system comprises: The structural feature recognition module obtains the longitudinal stiffness values of each pier, combines the adjacent pier pairs according to the bridge span information, calculates the relative stiffness difference of the adjacent pier pairs, extracts the maximum and minimum values and calculates the span range, judges the bridge structure, generates the bridge structure attribution classification label, and generates the bridge structure attribution classification label; The span stiffness linkage module extracts the span length of the adjacent spans according to the bridge structure attribution classification label, calculates the span stiffness ratio of the adjacent spans, and pairs each item with the span pier stiffness difference value, calculates the cumulative offset of the stiffness difference, and judges whether it meets the linear change, generates the span stiffness linkage trend analysis result, and generates the span stiffness linkage trend analysis result; The shear sequence extraction module collects the initial shear stress and peak shear stress according to the span stiffness linkage trend analysis result, calculates the shear stress growth rate, constructs the shear ratio change sequence, extracts the section with continuously rising node amplitude, records the section number and maximum amplitude value, and generates the path shear trend sequence set; The path strength screening module calculates the average shear stress growth rate based on the path shear trend sequence set, compares and analyzes the structure stress distribution uneven path and eliminates it, and generates the path stress screening list. 2.The AI-based bridge design optimization system of claim 1, wherein: The bridge structure attribution classification label includes a structure type identifier, a symmetry evaluation coefficient, and a stiffness distribution level; The span stiffness linkage trend analysis result includes a span ratio mapping relationship, a stiffness offset trend quantity, and a structure balance matching degree; The path shear trend sequence set includes a shear growth section number, an amplitude peak index, and a path trend continuity degree; The path stress screening list includes a path screening mark, a stress concentration identification level, and a path retention number. 3.The AI-based bridge design optimization system of claim 1, wherein: The structural feature recognition module comprises: The stiffness normalization submodule obtains the longitudinal stiffness values of each pier under the bridge design working condition, selects the maximum longitudinal stiffness value of the whole bridge as the pier stiffness reference value according to the standard parameters of the pier longitudinal stiffness in the bridge design specification, performs division operation on all pier longitudinal stiffness values with the pier stiffness reference value in sequence, obtains the dimensionless normalized stiffness values of each pier, sorts them in the order of pier number, and generates the pier normalized stiffness sequence; The span combination submodule combines adjacent piers into pier pairs according to the adjacent number mode based on the pier normalized stiffness sequence and the length and number order of each span in the bridge span information, calls the normalized stiffness values of the two end piers for each pier pair, calculates the relative stiffness difference, extracts the maximum and minimum values of all relative stiffness differences, calculates the difference, and outputs the adjacent stiffness difference span; The structure classification submodule sorts the length and distribution order of each continuous span according to the bridge span information, judges whether the adjacent stiffness difference span in each span is symmetric along the bridge span center, marks it as a continuous beam structure if it is symmetric in multiple continuous spans, or classifies it as a combined or cantilever structure if it is not symmetric, and generates the bridge structure attribution classification label. 4.The AI-based bridge design optimization system of claim 1, wherein: The span stiffness linkage module comprises: The rigidity ratio calculation submodule obtains bridge data marked as a continuous structure in the bridge structure belonging classification label, extracts the span length of adjacent spans and the longitudinal rigidity value of the corresponding piers, calculates the span rigidity ratio of the adjacent spans, and arranges the rigidity ratio in sequence according to the span number, to establish a span rigidity ratio value sequence; The difference pairing submodule pairs the rigidity difference value with the rigidity ratio value in sequence according to the span number based on the span rigidity ratio value sequence, sorts the rigidity ratio in each pair of values in ascending order, extracts the corresponding rigidity difference value in the sorting order, calculates the rigidity difference cumulative offset, and establishes a rigidity difference cumulative offset sequence; The trend judgment submodule extracts the span number and the corresponding offset value according to the rigidity difference cumulative offset sequence, judges whether the offset value in the sequence increases linearly with the span number, and if the trend is established, marks it as a linear enhancement relationship, and if not, marks it as a nonlinear relationship, to generate a span rigidity linkage trend analysis result.

5. The AI-based bridge design optimization system of claim 4, wherein: The rigidity difference cumulative offset uses the formula: ； The calculation is performed, wherein, represents the stiffness difference of the first item after sorting, represents the cumulative offset of the stiffness difference of the first item after sorting, represents the stiffness difference of the first item after sorting, represents the stiffness ratio of the first item after sorting, represents the stiffness ratio of the first item after sorting, represents the summation operation from the first item to the first 6.The AI-based bridge design optimization system of claim 1, wherein: The shear sequence extraction module includes: The shear stress extraction submodule collects the initial shear stress value and the peak shear stress value in all path nodes according to the span rigidity linkage trend analysis result, calculates the shear stress growth rate of all nodes, arranges the corresponding growth rate in sequence according to the node number, and generates a node shear stress growth rate sequence; The shear ratio generation submodule constructs a continuous change sequence by sequentially connecting the corresponding growth rate according to the node number order based on the node shear stress growth rate sequence, calculates the change amplitude between adjacent growth rates, arranges all change amounts in sequence according to the node number, and generates a shear ratio change sequence; The trend section identification submodule identifies a section in which the shear ratio change value continuously increases positively in three or more consecutive nodes according to the shear ratio change sequence, extracts all sections that meet the condition and counts the number, and extracts the amplitude value of the largest growth interval in all sections to establish a path shear trend sequence set.

7. The AI-based bridge design optimization system of claim 6, wherein: The change amplitude between adjacent growth rates uses the formula: ； Compute where, represents the shear stress growth rate at the and the shear ratio change intensity value between the nodes, represents the shear stress growth rate of the node, represents the shear stress growth rate of the node thereafter, is the absolute value of the growth rate of adjacent nodes and is the square of the difference in shear stress growth rate, is used to regularize the current node growth rate and avoid division by zero operations.

8. The AI-based bridge design optimization system of claim 1, wherein: The path strength screening module includes: The growth rate statistics submodule extracts the shear stress growth rate of three consecutive nodes in each rising section based on the continuous rising section contained in each path in the path shear trend sequence set, sequentially calculates the average shear stress growth rate of each section, and pairs the average value of each rising section with the average value calculated from the shear stress growth rate of all nodes in the corresponding path to generate path shear stress growth ratio data; The abnormal path identification submodule extracts the average value of each pair of values in the path shear stress growth ratio data, judges whether there is any path in which the average value of any rising section is greater than twice the overall average value, and if the condition is met, marks the corresponding path as a stress concentration abnormal path to obtain a stress concentration path number list; The stress path elimination submodule eliminates the corresponding path from the path shear trend sequence set according to the stress concentration path number list, retains the remaining path numbers and corresponding sequences, and generates a path stress screening list. 9.The AI-based bridge design optimization system of claim 1, wherein: The system further includes an intelligent modeling input module; The intelligent modeling input module binds the corresponding path shear change information and the bridge structure attribution classification label according to the path stress screening list, inputs the structure attribution, span ratio trend, and shear sequence characteristics into the AI model, learns the correlation law of the bridge structure parameters and the mechanical response through the training data, performs bridge AI modeling analysis, and obtains AI bridge design optimization results; The AI bridge design optimization results include a structure attribution input set, a path optimization level group, and AI training feedback data.

10. The AI-based bridge design optimization system of claim 9, wherein: The intelligent modeling input module includes: A structure data binding submodule extracts the bridge structure attribution classification label and the corresponding path shear ratio change data of each path according to the node information of the retained path in the path stress screening list, calls the structure label and the path shear sequence for association matching, and generates a path structure shear combination set; A feature data integration submodule establishes the structure and mechanical response feature combination item of each path based on the path structure shear combination set, combines the path structure attribution label, the span stiffness ratio trend data, and the shear stress growth sequence, uniformly formats the input data matrix, and obtains a modeling input feature matrix; A model association construction submodule performs interactive construction of the bridge structure feature item corresponding to the path number and the shear response data item according to the modeling input feature matrix, compares the shear stress trend distribution and the stiffness change sequence between each group of inputs according to the structure type, extracts the shear response concentration in each structure category, and performs structure-response parameter mapping, and establishes AI bridge design optimization results.

Images