A method and system for evaluating a large-span bridge structure system

Abstract

This invention discloses a method and system for evaluating the structural system of long-span bridges. The method includes: collecting and preprocessing data from manual inspections, non-destructive testing, and health monitoring of long-span bridges to form a standardized reference dataset; identifying typical defects and structural characteristics through field investigations, literature reviews, and expert surveys; establishing a multi-level evaluation framework; determining indicators for each level; and forming an evaluation system covering safety, durability, and applicability dimensions. The method employs a group hierarchical analysis approach, combining expert opinions with judgment information collected on a five-level scale, optimizing the consistency of the judgment matrix, calculating expert weights, and weighting the indicators accordingly. Quantitative indicator types are categorized, and a nonlinear dimensionless model is constructed for normalization. For sequential indicators, standardized scores are obtained through measurement point evaluation, uniform score calculation, and non-uniform coefficient analysis. Evaluation results are calculated hierarchically to identify structural risks and hazards, and maintenance plans are output. This achieves a scientific and reasonable assessment of the service status of long-span bridges.

Description

Technical Field

[0001] This invention belongs to the field of bridge structure evaluation technology, and more specifically, relates to an evaluation method and system for long-span bridge structures. Background Technology

[0002] With the continuous expansion of infrastructure, the number and size of bridges have grown rapidly, making bridge structural safety assessments and maintenance decisions a core aspect of transportation management. As crucial nodes in transportation networks, the operational status of bridges directly impacts traffic safety and transportation efficiency. Therefore, conducting scientific and standardized bridge technical condition assessments is of great significance for ensuring safe bridge operation and extending their service life.

[0003] Currently, bridge assessment in China is mainly based on standards such as the "Specifications for Maintenance of Highway Bridges and Culverts" (JTG5120-2021), the "Standards for Technical Condition Assessment of Highway Bridges" (JTG / TH21-2011), and the "Regulations for Testing and Assessment of Load-Bearing Capacity of Highway Bridges" (JTG / TJ21-2011). It is implemented through a combination of manual inspection, portable testing instruments, and the local application of bridge health monitoring systems. Through a standardized indicator system and regular testing, the technical condition of small and medium-span bridges can be basically grasped, forming a standardized and systematic management model to a certain extent.

[0004] However, as bridges develop towards longer spans and more complex structures, existing assessment systems have gradually revealed several shortcomings. First, traditional manual inspections rely on the experience of inspectors, resulting in significant subjectivity and difficulty in comprehensively revealing potential internal structural damage. Second, in practical applications, health monitoring systems suffer from frequent sensor malfunctions and complex data anomalies, affecting the effectiveness and continuity of monitoring data. Third, inconsistent data standards from different sources hinder effective fusion and dynamic tracking, making it difficult to accurately grasp long-term performance evolution trends. Furthermore, for bridges with complex structures and diverse components, such as cable-stayed bridges and suspension bridges, existing assessment methods remain insufficient in identifying detailed component damage and comprehensively analyzing the overall structural condition, affecting the scientific rigor and timeliness of maintenance decisions. In summary, existing bridge assessment technologies cannot fully meet the practical needs of refined management and dynamic assessment of long-span, complex structural bridges, and further technological improvements are urgently needed in areas such as data standardization, information fusion analysis, and long-term performance evolution tracking. Summary of the Invention

[0005] The purpose of this application is to design a method for evaluating the service status of long-span bridge structural systems, taking into account the complexity and typical defects of long-span bridge structures, and to develop a digital platform system to achieve a scientific and reasonable evaluation of the service status of long-span bridges.

[0006] In view of the above-mentioned defects or improvement needs of the existing technology, as a first aspect of the present invention, the present invention provides an evaluation method for long-span bridge structural systems, comprising: S1. Collect manual inspection data, non-destructive testing data, and long-term health monitoring data of long-span bridges. Preprocess these multi-source data to form a standardized reference dataset. Second, through field investigation, literature review, and expert survey, identify the typical types of defects and structural characteristics of long-span bridges, build a multi-level evaluation framework of "component layer - component layer - defect layer", and then determine the specific evaluation indicators for each level based on the standardized reference data, ultimately forming a complete evaluation indicator system covering the dimensions of safety, durability, and applicability. S2. Using the group hierarchical analysis method, combined with the professional opinions of industry authorities on the evaluation indicators of long-span bridges, expert judgment information was collected according to a five-level scale; then the collected expert judgment matrix was optimized for consistency to reduce subjective judgment bias; finally, the weight of each expert was calculated based on the optimized expert judgment matrix, and then the final weight of each level of indicators in the evaluation indicator system was obtained through weighted calculation. S3. For quantitative evaluation indicators, they are classified into positive indicators, inverse indicators, and appropriate indicators based on their attributes. For each type of indicator, a nonlinear dimensionless model consisting of six standard values ​​and five line segments is constructed to complete the normalization process of the quantitative indicators. For sequential evaluation indicators, the line segment method appropriate indicator model is first used to obtain the evaluation value of a single measuring point. Then, the uniform change score is calculated by combining the measuring point weights. The non-uniform change coefficient is obtained through slope grey relational analysis. Finally, the two results are integrated to obtain the standardized score of the sequential indicators. S4. Calculations are carried out step by step according to the hierarchy of "component score - part score - structure score - overall bridge score" to obtain the evaluation results at each level; based on these multi-level evaluation results, the critical risks and potential safety hazards of the bridge structure are identified, and targeted maintenance plans are output.

[0007] Furthermore, the data preprocessing process in S1 includes: Wavelet transform denoising technology is used to remove high-frequency random noise from the monitoring data. Wavelet multi-scale decomposition is used to decouple the structural response components induced by temperature, traffic and wind load. Then, sensor network topology features are extracted based on graph convolutional network. Combined with cross-domain attention mechanism, cross-domain features of multiple subsystems are fused to obtain standardized reference data.

[0008] Furthermore, the calculation process for the expert weights in S2 is as follows: Based on the Weber-Fechner law, a five-level scale data collection method conforming to the laws of psychological perception is used. The evaluation system for long-span cable-stayed bridges, including the professional judgment information of authoritative industry experts on the indicator weights and standards, is structured around the expertise of each expert. Initial judgment matrix ,in Values ​​range from 1 to integers, To evaluate the number of indicators, the matrix satisfies reciprocity. and reflexivity , The scale value corresponds to the fifth-order scale; To address the minor biases that experts may exhibit during the scaling process, and with the goal of improving the consistency of the judgment matrix, a genetic algorithm is employed to refine each initial judgment matrix. Optimization is achieved through genetic operations including encoding, selection, crossover, and mutation. This process corrects scaling bias caused by subjective oversights and reduces inconsistencies in the judgment matrix while preserving the core information of the original expert judgments to the greatest extent possible, resulting in an optimized judgment matrix. , The initial value is 0, representing the result after the first optimization. Employing a peer-to-peer consensus model to quantify any two experts and The degree of consensus among opinions is calculated using the following formula: In the formula For experts and The value is a quantitative measure of the degree of agreement between the two experts. The smaller the value, the higher the agreement between the two experts, and vice versa. It is the identity matrix; Experts , After the first The optimized judgment matrix; , These are the scale elements in the corresponding matrix; o represents the Hadamard product of the matrix. Indicates matrix transpose; Filter out The two experts with the highest degree of disagreement. and The judgment matrix is ​​iteratively corrected according to the following formula: In the formula This represents the number of iterations. , The parameters are used to adjust the fusion ratio of the scaling elements from the two experts. , Experts , After the first The scaled elements after the second correction; after correction, they still need to satisfy the matrix reciprocity and reflexivity constraints; Repeat the steps of quantifying expert opinion consensus and dynamically adjusting the judgment matrix as described above until all experts are consistent in pairs. If the preset consistency requirement is met, the weights of each expert are calculated according to the following formula. : In the formula For the first The weight of each expert; For the first The expert and the first A quantitative value representing the degree of consistency in the opinions of the experts; This represents the total number of experts who participated in the survey.

[0009] Furthermore, in the final weighting of S2, the weight ratio of key components, including cable-stayed systems, steel box girders, and cable towers, is increased, while indicators for long-span bridges without corresponding structures are removed.

[0010] Furthermore, the nonlinear dimensionless model in S3 is specifically as follows: The detection value of the set quantity index , Actual test data representing quantitative indicators related to long-span bridge structures; dimensionless score value. , It is the detection value The evaluation score obtained after normalization has a range of values. , used to quantify the structural performance level corresponding to the indicator; The six preset grading standard values ​​are arranged in ascending order as follows: , , , , , These six standard values ​​are grading thresholds determined based on the structural type, component characteristics, and relevant industry standards and specifications of long-span bridges, and are used to classify the test values. Different performance ranges; Define interval mapping coefficients ,in The interval number, ranging from 1 to 5, corresponds to five consecutive intervals formed by dividing the data using six standard values. Standard value The corresponding benchmark score, Standard value The corresponding benchmark score, For the first Each grading standard value, mapping coefficient Used to characterize the detection value In the When varying within a range, the dimensionless score value The linear rate of change; The unified expression for the model is: When the detection value Less than or equal to the first grading standard value At that time, the dimensionless score value Take the benchmark score corresponding to the first standard value. When the detection value In the first Each grading standard value With the Each grading standard value Between, dimensionless score value Based on benchmark scores and mapping coefficients Deviation of detection value The linear superposition calculation is obtained; when the detected value Greater than or equal to the sixth grading standard value At that time, the dimensionless score value Take the benchmark score corresponding to the sixth standard value. .

[0011] Furthermore, the positive, negative, and moderate indicators in S3 are as follows: Positive indicators are those whose larger test values ​​indicate better bridge structural performance. Their benchmark scores are configured in a gradient-increasing manner, and the mapping coefficient is positive to achieve a positive correlation between the score and the performance of the indicator; they are suitable for evaluation scenarios of positive performance indicators, including concrete strength and structural stiffness. The inverse index is an index in which the smaller the detection value, the better the bridge structure performance. Its benchmark score is configured in a gradient decreasing manner, and the mapping coefficient is negative to achieve the inverse correlation between the score and the degree of index deterioration; it is suitable for evaluation scenarios of deterioration-type indicators, including concrete carbonation depth and steel corrosion rate. The appropriate index is the index that represents the optimal performance of the bridge structure when the detected value is within a reasonable range. Its benchmark score is configured with a logic of low at both ends and high in the middle. It is suitable for evaluation scenarios of indicators that need to be controlled within a reasonable range, including the main beam elevation and cable force deviation.

[0012] Furthermore, the calculation process for the standardized scoring results of the sequence-type indicators in S3 is as follows: First, the evaluation value of a single measuring point is calculated. For the detection data of each measuring point of the sequential index, the "line segment method" appropriate index model is used for dimensionless processing to obtain the first... Evaluation value of each measuring point Secondly, calculate the score for uniform variation. The uniform variation score is used to characterize the overall average state of the sequential index detection data, corresponding to the standard value curve translation characteristics of the detection data. If all measurement points in the sequence have equal importance, the uniform variation score is taken as the arithmetic mean of the estimated values ​​of each measurement point, calculated using the following formula: If the importance of the measuring points differs, then each measuring point is assigned a weight based on the criticality of its structural location. The uniform variation score is the weighted average of the estimated values ​​from each assessment, calculated using the following formula: In the formula, The total number of measurement points. For the first Evaluation value of each measuring point For the first The weight of each measurement point; Then, calculate the non-uniform variation coefficient. The non-uniform variation coefficient is used to emphasize the changing trend of sequence-type indicators, corresponding to the variation characteristics of the detection data around the standard value axis. It is calculated through slope grey relational analysis, and this relational degree is used to evaluate the similarity of the curve shape between the compared sequence and the reference sequence. For the standard or design value of this test item in the completed state of the bridge, the expression is: The sequence being compared For the bridge in operation The measured value of the second test is expressed as follows: ; First, calculate the slope of each adjacent point in the reference sequence and the compared sequence. The slope of the reference sequence at the nth... The slope of the adjacent points is The sequence being compared is the first one. The slope of the adjacent points is , Then, the slope grey relational degree is calculated using the following formula. And use it as the coefficient of non-uniform variation. : Finally, the standardized score of the sequential indicator is calculated. Scoring for uniform changes Coefficient of non-uniform variation Multiply by each other to obtain the standardized score result. The calculation formula is as follows: The scoring results comprehensively reflect the consistency between the overall average state and the trend of change of the sequential indicators, and achieve a comprehensive quantitative evaluation of the sequential indicators.

[0013] Furthermore, in S4, the component score is quantified based on the deduction value of each detection index; the component score is calculated by combining the average score, the lowest score of the component, and the component quantity adjustment coefficient; the structural score is obtained by dynamically adjusting the component weights through a variable weighting mechanism and then summing them up; the overall bridge score is obtained by weighting various structural scores according to their corresponding weights.

[0014] As a second aspect of the present invention, an evaluation system for long-span bridge structures is also provided, comprising: The data processing and indicator construction unit is used to collect manual inspection data, non-destructive testing data, and long-term health monitoring data of long-span bridges. These multi-source data are preprocessed to form a standardized reference dataset. Secondly, through field investigation, literature review, and expert survey, the typical types of defects and structural characteristics of long-span bridges are identified, and a multi-level evaluation framework of "component layer - component layer - defect layer" is built. Then, based on the standardized reference data, specific evaluation indicators for each level are determined, and finally a complete evaluation indicator system covering the dimensions of safety, durability, and applicability is formed. The weight determination unit is used to collect expert judgment information according to a five-level scale by using the group hierarchical analysis method and combining the professional opinions of industry authorities on the evaluation indicators of long-span bridges. Then, the collected expert judgment matrix is ​​optimized for consistency to reduce subjective judgment bias. Finally, the weight of each expert is calculated based on the optimized expert judgment matrix, and then the final weight of each level of indicators in the evaluation indicator system is obtained through weighted calculation. The evaluation index standardization processing unit is used to classify quantitative evaluation indicators into positive, negative, and moderate indicators based on their attributes. It constructs a nonlinear dimensionless model consisting of six standard values ​​and five line segments for each type of indicator to complete the normalization process of the quantitative indicators. For sequential evaluation indicators, the moderate indicator model of the line segment method is first used to obtain the evaluation value of a single measuring point. Then, the uniform change score is calculated by combining the measuring point weights. The non-uniform change coefficient is obtained through slope grey relational analysis. Finally, the two results are integrated to obtain the standardized score of the sequential indicator. The scoring calculation and maintenance output unit is used to perform calculations step by step according to the hierarchy of "component scoring - part scoring - structural scoring - overall bridge scoring" to obtain the evaluation results at each level. Based on these multi-level evaluation results, the critical risks and potential safety hazards of the bridge structure are identified, and targeted maintenance plans are output.

[0015] As a third aspect of the invention, a computer-readable storage medium is also provided, on which a computer program is stored, which is executed by a processor to provide an evaluation method for a long-span bridge structure system as described in any one of the claims.

[0016] In summary, compared with the prior art, the above-described technical solutions conceived by this invention can achieve the following beneficial effects: 1. The present invention provides an evaluation method for long-span bridge structural systems. By constructing a multi-level and systematic evaluation index system, it covers six major components, twenty-one types of components, and one hundred and twenty-six typical defect indicators, comprehensively covering key dimensions such as safety, durability, and applicability. It can accurately reflect the defect development trend and structural performance degradation characteristics of complex bridges such as long-span cable-stayed bridges, and provide effective support for bridge maintenance decisions.

[0017] 2. The present invention provides an evaluation method for long-span bridge structural systems. By constructing a dimensionless "line segment" model based on the characteristics of three types of quantitative indicators—positive, inverse, and moderate—it achieves nonlinear normalization of quantitative indicators. Simultaneously, for sequential indicators, by integrating uniform and non-uniform change analysis, it proposes a slope grey relational scoring mechanism, which can effectively capture the temporal evolution characteristics of component states and achieve dynamic quantitative evaluation of structural performance trends.

[0018] 3. The present invention provides an evaluation method for long-span bridge structural systems, which progressively advances through four levels: component scoring, part scoring, structural scoring, and overall bridge scoring. Combined with a variable weighting mechanism to dynamically adjust the importance of components, it can effectively reflect the contribution of different parts of the structure to the overall performance, thus giving the evaluation results higher engineering guidance value. Attached Figure Description

[0019] Figure 1 This is a flowchart of an evaluation method for a long-span bridge structural system according to an embodiment of the present invention; Figure 2 This is the service status assessment process for long-span bridge structural systems according to an embodiment of the present invention; Figure 3 This is a flowchart illustrating the method for establishing an evaluation index system for bridge structural systems according to an embodiment of the present invention. Figure 4 This is the calculation process for the weights of evaluation indicators for bridge structural system assessment in this embodiment of the invention. Figure 5 This is a schematic diagram of a dimensionless model according to an embodiment of the present invention; Figure 6 This is a flowchart illustrating the multi-level evaluation score calculation process according to an embodiment of the present invention. Figure 7 This is a schematic diagram of the system units in an embodiment of the present invention. Detailed Implementation

[0020] 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. Furthermore, the technical features involved in the various embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0021] Example 1 Please refer to Figure 1 This embodiment 1 provides a method for evaluating long-span bridge structural systems, including: S1. Collect manual inspection data, non-destructive testing data, and long-term health monitoring data of long-span bridges. Preprocess these multi-source data to form a standardized reference dataset. Second, through field investigation, literature review, and expert survey, identify the typical types of defects and structural characteristics of long-span bridges, build a multi-level evaluation framework of "component layer - component layer - defect layer", and then determine the specific evaluation indicators for each level based on the standardized reference data, ultimately forming a complete evaluation indicator system covering the dimensions of safety, durability, and applicability. S2. Using the group hierarchical analysis method, combined with the professional opinions of industry authorities on the evaluation indicators of long-span bridges, expert judgment information was collected according to a five-level scale; then the collected expert judgment matrix was optimized for consistency to reduce subjective judgment bias; finally, the weight of each expert was calculated based on the optimized expert judgment matrix, and then the final weight of each level of indicators in the evaluation indicator system was obtained through weighted calculation. S3. For quantitative evaluation indicators, they are classified into positive indicators, inverse indicators, and appropriate indicators based on their attributes. For each type of indicator, a nonlinear dimensionless model consisting of six standard values ​​and five line segments is constructed to complete the normalization process of the quantitative indicators. For sequential evaluation indicators, the line segment method appropriate indicator model is first used to obtain the evaluation value of a single measuring point. Then, the uniform change score is calculated by combining the measuring point weights. The non-uniform change coefficient is obtained through slope grey relational analysis. Finally, the two results are integrated to obtain the standardized score of the sequential indicators. S4. Calculations are carried out step by step according to the hierarchy of "component score - part score - structure score - overall bridge score" to obtain the evaluation results at each level; based on these multi-level evaluation results, the critical risks and potential safety hazards of the bridge structure are identified, and targeted maintenance plans are output.

[0022] Please refer to Figure 2 This embodiment 1 further elaborates on the above steps.

[0023] (1) Data processing and indicator construction like Figure 3As shown, a bridge structure evaluation index system is constructed, comprehensively utilizing multi-source data such as manual inspection, non-destructive testing, and long-term health monitoring. This system integrates defect records from different sources, measured structural response results, and periodic inspection reports to form comprehensive data support. Based on the characteristics of bridge structures, typical defect types are analyzed and identified through on-site investigations, literature reviews, and expert surveys. Combining the characteristics of these defects, the upper-level framework of the evaluation system is constructed. Then, the lower-level indicators of the evaluation system are determined based on manual inspection data, health monitoring data, and non-destructive testing data.

[0024] When introducing health monitoring data, an optimized data processing and fusion method was used to eliminate interference from external factors such as the environment and sensors. The specific process is as follows: Noise in bridge monitoring data is mostly high-frequency random signals (such as sensor electronic noise and environmental electromagnetic interference), while structural response signals are mostly low-frequency or specific frequency band useful signals. Wavelet transform decomposes the time-domain signal into wavelet coefficients of different scales to achieve frequency band separation. Thresholding is then used to remove wavelet coefficients corresponding to noise, and finally, the denoised clean signal is reconstructed. Considering the non-stationary nature of bridge monitoring data, the db4 wavelet is selected as the basis function, and the signal sampling frequency is considered. Determine the number of decomposition layers based on the noise frequency band. ,satisfy .

[0025] For the original monitoring signal Wavelet decomposition is performed to obtain approximate coefficients. and Detail coefficients ,Right now: A soft thresholding function is used to denoise the detail coefficients of each layer: in, The threshold is determined by the formula. calculate( For long-span bridges, the coefficient can be taken as 1.175 in this example. The standard deviation of noise. (This refers to the signal length).

[0026] Retain approximation coefficients The processed detail coefficients are then subjected to inverse wavelet transform to obtain the denoised signal. .

[0027] Furthermore, the structural responses of long-span bridges (such as main girder deflection and pylon offset) are the coupled results of temperature, traffic, and wind loads. Directly using these responses for evaluation would obscure the impact of different loads on structural performance. Wavelet decomposition technology can separate components based on the frequency differences of different load responses. Structural responses induced by different loads exhibit significant frequency characteristics. For example, temperature load responses change slowly with ambient temperature and are ultra-low frequency components; traffic load responses are correlated with traffic flow frequency and are low frequency components; wind load responses contain high-frequency components of pulsating wind and are mid-to-high frequency components. This method combines historical health monitoring data of long-span bridges to determine the cutoff frequencies of various structural responses. Through wavelet multi-scale decomposition, the coupled response signals can be layered by frequency to achieve decoupling of each load component. Furthermore, the accuracy of the separation results is verified through correlation analysis using concurrent environmental monitoring data (temperature, wind speed, and traffic flow). If the correlation coefficient between a component and the corresponding environmental monitoring data is high... If necessary, adjust the wavelet decomposition level or wavelet basis, and perform component separation again.

[0028] Meanwhile, bridge sensor networks have a well-defined topology, and the spatial location and connectivity of sensor points directly affect the correlation of monitoring data. Graph Convolutional Networks (GCNs) abstract the sensor network into a graph structure, aggregating neighborhood node information through graph convolution operations to extract and fuse structural response features based on spatial topological relationships.

[0029] Using sensor measurement points as nodes, node characteristics ( For the number of measurement points, (Assuming a single measurement point feature dimension), an adjacency matrix is ​​constructed using the spatial distance between measurement points as the weight. If the measuring point and If the distance between them is less than a set threshold, then Otherwise, the value is 0. The adjacency matrix is ​​then normalized. in, for( The degree matrix of ).

[0030] Using first-order graph convolution operation, the... The node characteristics of the layer are: in, The first , Layer feature matrix, These are the weight parameters for this layer. The ReLU activation function is used. By stacking GCNs, the final output is a global feature that fuses topological relationships. The topological features of subsystems such as stress, displacement, and vibration are linearly mapped to obtain embedded features with uniform dimensions. in, For embedding weights, This is a bias term.

[0031] Construct a cross-domain attention module to compute the query vector Q, key vector K, value vector V, and attention weights: in, Let K be the dimension of the key vector.

[0032] By employing a multi-head attention mechanism, the correlation weights of features from different domains are learned in parallel. The multi-head results are then concatenated and linearly transformed to obtain preliminary fused features. These features are then transformed nonlinearly through a feedforward neural network (FFN) to finally obtain cross-domain fused global features, which are then used as reference data input into the subsequent bridge evaluation model.

[0033] Furthermore, based on its structural composition, the bridge is divided into six major components and twenty-one categories of parts, further refined into one hundred and twenty-six specific defect indicators. Each indicator is clearly categorized, covering dimensions of safety, durability, and usability, ensuring a comprehensive and systematic assessment system.

[0034] (2) Weight determination After establishing the indicator system, it is necessary to determine the weight of each indicator in the comprehensive evaluation. For example... Figure 4 As shown, this scheme combines expert survey opinions and employs Group Analytic Hierarchy Process (GHP) to scientifically determine the weights of indicators. The expert survey primarily collects professional opinions from authoritative industry experts on the evaluation system, indicator weights, and standards for long-span cable-stayed bridges. Based on the Weber-Fechner law, a five-level scale conforming to psychological perception principles is used to collect these opinions. The evaluation system for long-span cable-stayed bridges, including the professional judgment information of authoritative industry experts on the indicator weights and standards, is structured around the expertise of each expert. Initial judgment matrix ,in Values ​​range from 1 to integers, To evaluate the number of indicators, the matrix satisfies reciprocity. and reflexivity , The scale value corresponds to the fifth-order scale; To address the minor biases that experts may exhibit during the scaling process, and with the goal of improving the consistency of the judgment matrix, a genetic algorithm is employed to refine each initial judgment matrix. Optimization is achieved through genetic operations including encoding, selection, crossover, and mutation. This process corrects scaling bias caused by subjective oversights and reduces inconsistencies in the judgment matrix while preserving the core information of the original expert judgments to the greatest extent possible, resulting in an optimized judgment matrix. , The initial value is 0, representing the result after the first optimization. Employing a peer-to-peer consensus model to quantify any two experts and The degree of consensus among opinions is calculated using the following formula: In the formula For experts and The value is a quantitative measure of the degree of agreement between the two experts. The smaller the value, the higher the agreement between the two experts, and vice versa. It is the identity matrix; Experts , After the first The optimized judgment matrix; , These are the scale elements in the corresponding matrix; o represents the Hadamard product of the matrix. Indicates matrix transpose; Filter out The two experts with the highest degree of disagreement. and The judgment matrix is ​​iteratively corrected according to the following formula: In the formula This represents the number of iterations. , The parameters are used to adjust the fusion ratio of the scaling elements from the two experts. , Experts , After the first The scaled elements after the second correction; after correction, they still need to satisfy the matrix reciprocity and reflexivity constraints; Repeat the steps of quantifying expert opinion consensus and dynamically adjusting the judgment matrix as described above until all experts are consistent in pairs. If the preset consistency requirement is met, the weights of each expert are calculated according to the following formula. : In the formula For the first The weight of each expert; For the first The expert and the first A quantitative value representing the degree of consistency in the opinions of the experts; This represents the total number of experts who participated in the survey.

[0035] Based on the above algorithm, in a specific application implementation process, the weights of the evaluation indicators for long-span cable-stayed bridges can be obtained as shown in Table 1 below; Meanwhile, compared with the current standards, the weighting algorithm of this evaluation index appropriately increases the weight of key components such as cable-stayed systems, steel box girders, and cable towers, and removes structures that do not exist in long-span bridges, thus optimizing the evaluation system.

[0036] (3) Standardization of evaluation indicators For quantitative indicators, based on their classification in positive, inverse, and moderate forms, three nonlinear dimensionless models are constructed: positive, inverse, and moderate indicators. Positive indicators are those whose larger test values ​​indicate better bridge structural performance. Their benchmark scores are configured in a gradient-increasing manner, and the mapping coefficient is positive to achieve a positive correlation between the score and the performance of the indicator; they are suitable for evaluation scenarios of positive performance indicators, including concrete strength and structural stiffness. The inverse index is an index in which the smaller the detection value, the better the bridge structure performance. Its benchmark score is configured in a gradient decreasing manner, and the mapping coefficient is negative to achieve the inverse correlation between the score and the degree of index deterioration; it is suitable for evaluation scenarios of deterioration-type indicators, including concrete carbonation depth and steel corrosion rate. The appropriate index is the index that represents the optimal performance of the bridge structure when the detected value is within a reasonable range. Its benchmark score is configured with a logic of low at both ends and high in the middle. It is suitable for evaluation scenarios of indicators that need to be controlled within a reasonable range, including the main beam elevation and cable force deviation.

[0037] like Figure 5 As shown, the nonlinear dimensionless curve of the evaluation index is simplified into a model consisting of six standard values ​​and five line segments to adapt to existing standards and facilitate the determination of standard values ​​for each evaluation index. This dimensionless model is called the "line segment" dimensionless model, and it is used for standardized mapping. Its specific construction process is as follows: The detection value of the set quantity index , Actual test data representing quantitative indicators related to long-span bridge structures; dimensionless score value. , It is the detection value The evaluation score obtained after normalization has a range of values. , used to quantify the structural performance level corresponding to the indicator; The six preset grading standard values ​​are arranged in ascending order as follows: , , , , , These six standard values ​​are grading thresholds determined based on the structural type, component characteristics, and relevant industry standards and specifications of long-span bridges, and are used to classify the test values. Different performance ranges; Define interval mapping coefficients ,in The interval number, ranging from 1 to 5, corresponds to five consecutive intervals formed by dividing the data using six standard values. Standard value The corresponding benchmark score, Standard value The corresponding benchmark score, For the first Each grading standard value, mapping coefficient Used to characterize the detection value In the When varying within a range, the dimensionless score value The linear rate of change; The unified expression for the model is: When the detection value Less than or equal to the first grading standard value At that time, the dimensionless score value Take the benchmark score corresponding to the first standard value. When the detection value In the first Each grading standard value With the Each grading standard value Between, dimensionless score value Based on benchmark scores and mapping coefficients Deviation of detection value The linear superposition calculation is obtained; when the detected value Greater than or equal to the sixth grading standard value At that time, the dimensionless score value Take the benchmark score corresponding to the sixth standard value. .

[0038] For sequential indicators, considering both uniform variation (shifting of the standard value curve) and non-uniform variation (variation around the standard value axis) of the test data, the following formula is used for calculation: In the formula, It is a rating value; It is a score for uniform variation, and for a certain detection point, it is a score for technical status. This is the non-uniform variation coefficient, emphasizing the changing trend of the sequence index. In the formula... and The specific calculation method is as follows: First, the evaluation value of a single measuring point is calculated. For the detection data of each measuring point of the sequential index, the "line segment method" appropriate index model is used for dimensionless processing to obtain the first... Evaluation value of each measuring point ; Secondly, calculate the score for uniform variation. The uniform variation score is used to characterize the overall average state of the sequential index detection data, corresponding to the standard value curve translation characteristics of the detection data. If all measurement points in the sequence have equal importance, the uniform variation score is taken as the arithmetic mean of the estimated values ​​of each measurement point, calculated using the following formula: If the importance of the measuring points differs, then each measuring point is assigned a weight based on the criticality of its structural location. The uniform variation score is the weighted average of the estimated values ​​from each assessment, calculated using the following formula: In the formula, The total number of measurement points. For the first Evaluation value of each measuring point For the first The weight of each measurement point; Then, calculate the non-uniform variation coefficient. The non-uniform variation coefficient is used to emphasize the changing trend of sequence-type indicators, corresponding to the variation characteristics of the detection data around the standard value axis. It is calculated through slope grey relational analysis, and this relational degree is used to evaluate the similarity of the curve shape between the compared sequence and the reference sequence. For the standard or design value of this test item in the completed state of the bridge, the expression is: The sequence being compared For the bridge in operation The measured value of the second test is expressed as follows: ; First, calculate the slope of each adjacent point in the reference sequence and the compared sequence. The slope of the reference sequence at the nth... The slope of the adjacent points is The sequence being compared is the first one. The slope of the adjacent points is , Then, the slope grey relational degree is calculated using the following formula. And use it as the coefficient of non-uniform variation. : (4) Scoring calculation and maintenance output Figure 6 This is a flowchart of the multi-level evaluation score calculation process, including four steps: component scoring, sub-component scoring, structural scoring, and overall bridge scoring, as detailed below: Bridge component scoring is calculated using the following formula: when hour, ; when hour, (in ); when hour, ; In the formula, It is the first Class structure Class of components The score for each component ranges from 0 to 100. It is the first Class of components The number of types of indicators that result in point deductions for a component; These are the introduced variables; It is a component category, for example Indicates supports, expansion joints, etc.; It is the first Class of components The first component Classification of detection indicators; It is the first Class of components The first component The deduction values ​​for each type of testing indicator are calculated based on the deduction values ​​for each testing indicator of the component.

[0039] The bridge component score is calculated using the following formula: In the formula, It is the first Class structure The score for each type of component ranges from 0 to 100. (i.e., non-bridge deck system and ancillary facilities), the score of a certain component of the main components in this part. exist When the interval is reached, the corresponding component score value ; It is the first Class structure The average score of each component in the category, with a value range of 0-100; It is the first Class structure The lowest score among all components in a given category; It is a coefficient that varies with the number of components. In a specific embodiment, The values ​​are shown in Table 2 below, and also in other implementation processes, The values ​​can be defined by the technical personnel themselves according to the actual project requirements.

[0040] The bridge structure score is calculated using the following formula: In the formula, This is the score for the c-th type of structure, with a value range of 0 to 100 points; It is the first The number of component types in a class structure; It is the first Class structure The weight of a component class is calculated using the following formula: In the formula, It is the first Class structure Initial weights of components before weighting changes. For components not assigned in the bridge, their weights should be allocated to existing components based on their affiliation, according to the proportion of each existing component's weight to the total weight of all existing components.

[0041] The overall score for the bridge is calculated using the following formula: In the formula, This is the overall score for the bridge, with a value range of 0-100 points; It is the first The weight of the class structure in the full bridge.

[0042] Simultaneously, based on the results of multi-level assessments, a three-dimensional risk identification system is first constructed, consisting of "indicator anomaly degree - structural hierarchy correlation - risk transmission path." This system, combined with the deviation of scores from thresholds for each component and part, identifies key structural areas where critical risks exist. Furthermore, through coupled analysis of historical disease development data and real-time monitoring trends, the evolution direction and impact range of potential safety hazards are predicted. Based on this, maintenance priorities are divided according to risk levels (high, medium, and low), and differentiated maintenance strategies are matched for different risk levels: emergency response plans are developed for high-risk areas, specifying the technical parameters and implementation timelines for emergency repairs; routine or preventative maintenance measures are planned for medium- and low-risk hazards, detailing maintenance processes, material selection, and schedules. Ultimately, a comprehensive and targeted maintenance plan is formed, encompassing risk details, maintenance objectives, technical solutions, and implementation plans. Example 2 Please refer to Figure 7 This embodiment 2 provides an evaluation system for long-span bridge structures, including: The data processing and indicator construction unit is used to collect manual inspection data, non-destructive testing data, and long-term health monitoring data of long-span bridges. These multi-source data are preprocessed to form a standardized reference dataset. Secondly, through field investigation, literature review, and expert survey, the typical types of defects and structural characteristics of long-span bridges are identified, and a multi-level evaluation framework of "component layer - component layer - defect layer" is built. Then, based on the standardized reference data, specific evaluation indicators for each level are determined, and finally a complete evaluation indicator system covering the dimensions of safety, durability, and applicability is formed. The weight determination unit is used to collect expert judgment information according to a five-level scale by using the group hierarchical analysis method and combining the professional opinions of industry authorities on the evaluation indicators of long-span bridges. Then, the collected expert judgment matrix is ​​optimized for consistency to reduce subjective judgment bias. Finally, the weight of each expert is calculated based on the optimized expert judgment matrix, and then the final weight of each level of indicators in the evaluation indicator system is obtained through weighted calculation. The evaluation index standardization processing unit is used to classify quantitative evaluation indicators into positive, negative, and moderate indicators based on their attributes. It constructs a nonlinear dimensionless model consisting of six standard values ​​and five line segments for each type of indicator to complete the normalization process of the quantitative indicators. For sequential evaluation indicators, the moderate indicator model of the line segment method is first used to obtain the evaluation value of a single measuring point. Then, the uniform change score is calculated by combining the measuring point weights. The non-uniform change coefficient is obtained through slope grey relational analysis. Finally, the two results are integrated to obtain the standardized score of the sequential indicator. The scoring calculation and maintenance output unit is used to perform calculations step by step according to the hierarchy of "component scoring - part scoring - structural scoring - overall bridge scoring" to obtain the evaluation results at each level. Based on these multi-level evaluation results, the critical risks and potential safety hazards of the bridge structure are identified, and targeted maintenance plans are output.

[0043] This system also utilizes a combination of main menus and tree menus to support quick access to and management of the assessment process. It supports standardized data entry and structured storage, constructing multiple database forms covering assessment standards, inspection and testing information, and assessment calculation parameters. The system clearly defines the data table structure and field relationships, and designs data exchange interfaces to ensure data integrity and smooth flow, providing data support for the assessment calculation module. Furthermore, the platform has query, statistical, and output functions for inspection content, disease information, testing data, and assessment results, facilitating quick retrieval, archiving, and analysis by users.

[0044] In response to the disease development process, the platform combines time-series factors and user interaction needs to design a multi-dimensional point-line-surface-volume-time axis display method, including disease time sequence, single component, overall component, and component time sequence. It supports the statistics and visualization of disease development trends, and helps users to grasp information such as component detection coverage, inspection timeliness, maintenance effect, and disease evolution trend in a timely manner.

[0045] This system integrates both standardized and refined assessment methods. Based on defect indicators, inspection and testing data, and structural grading standards, it completes an intelligent assessment process from underlying defects to the overall bridge structure. Using recent inspection data of the Sutong Bridge as a case study, the platform has completed functional testing and performance verification, ensuring that the system possesses good maintainability, scalability, and high fault tolerance to meet engineering application requirements. Simultaneously, a user manual has been compiled to guide users in mastering the functions and operating procedures of each module, improving efficiency and ensuring stable and reliable system operation.

[0046] Example 3 This embodiment 3 also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can implement any step of a method for evaluating a long-span bridge structure system.

[0047] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0048] For a description of the computer-readable storage medium provided in this application, please refer to the above method embodiments; further details will not be repeated here.

[0049] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for evaluating long-span bridge structural systems, characterized in that, include: S1. Collect manual inspection data, non-destructive testing data and long-term health monitoring data of long-span bridges, preprocess these multi-source data, and form a standardized reference dataset; Secondly, through on-site investigation, literature review and expert survey, the typical types of defects and structural characteristics of long-span bridges were identified, and a multi-level evaluation framework of "component layer - component layer - defect layer" was established. Then, based on standardized reference data, specific evaluation indicators for each level were determined, and finally a complete evaluation indicator system covering the dimensions of safety, durability and applicability was formed. S2. Using the group hierarchical analysis method, combined with the professional opinions of industry authorities on the evaluation indicators of long-span bridges, expert judgment information was collected according to a five-level scale; then the collected expert judgment matrix was optimized for consistency to reduce subjective judgment bias; finally, the weight of each expert was calculated based on the optimized expert judgment matrix, and then the final weight of each level of indicators in the evaluation indicator system was obtained through weighted calculation. S3. For quantitative evaluation indicators, they are classified into positive indicators, inverse indicators, and appropriate indicators based on their attributes. For each type of indicator, a nonlinear dimensionless model consisting of six standard values ​​and five line segments is constructed to complete the normalization process of the quantitative indicators. For sequential evaluation indicators, the line segment method appropriate indicator model is first used to obtain the evaluation value of a single measuring point. Then, the uniform change score is calculated by combining the measuring point weights. The non-uniform change coefficient is obtained through slope grey relational analysis. Finally, the two results are integrated to obtain the standardized score of the sequential indicators. S4. Calculations are carried out step by step according to the hierarchy of "component score - part score - structure score - overall bridge score" to obtain the evaluation results at each level; based on these multi-level evaluation results, the critical risks and potential safety hazards of the bridge structure are identified, and targeted maintenance plans are output.

2. The evaluation method for a long-span bridge structural system according to claim 1, characterized in that, The data preprocessing process in S1 includes: Wavelet transform denoising technology is used to remove high-frequency random noise from the monitoring data. Wavelet multi-scale decomposition is used to decouple the structural response components induced by temperature, traffic and wind load. Then, sensor network topology features are extracted based on graph convolutional network. Combined with cross-domain attention mechanism, cross-domain features of multiple subsystems are fused to obtain standardized reference data.

3. The evaluation method for a long-span bridge structural system according to claim 1, characterized in that, The calculation process for the expert weights in S2 is as follows: Based on the Weber-Fechner law, a five-level scale data collection method conforming to the laws of psychological perception is used. The evaluation system for long-span cable-stayed bridges, including the professional judgment information of authoritative industry experts on the indicator weights and standards, is structured around the expertise of each expert. Initial judgment matrix ,in Values ​​range from 1 to integers, To evaluate the number of indicators, the matrix satisfies reciprocity. and reflexivity , The scale value corresponds to the fifth-order scale; To address the minor biases that experts may exhibit during the scaling process, and with the goal of improving the consistency of the judgment matrix, a genetic algorithm is employed to refine each initial judgment matrix. Optimization is achieved through genetic operations including encoding, selection, crossover, and mutation. This process corrects scaling bias caused by subjective oversights and reduces inconsistencies in the judgment matrix while preserving the core information of the original expert judgments to the greatest extent possible, resulting in an optimized judgment matrix. , The initial value is 0, representing the result after the first optimization. Employing a peer-to-peer consensus model to quantify any two experts and The degree of consensus among opinions is calculated using the following formula: In the formula For experts and The value is a quantitative measure of the degree of agreement between the two experts. The smaller the value, the higher the agreement between the two experts, and vice versa. It is the identity matrix; Experts , After the first The optimized judgment matrix; , These are the scale elements in the corresponding matrix; o represents the Hadamard product of the matrix. Indicates matrix transpose; Filter out The two experts with the highest degree of disagreement. and The judgment matrix is ​​iteratively corrected according to the following formula: In the formula This represents the number of iterations. , The parameters are used to adjust the fusion ratio of the scaling elements from the two experts. , Experts , After the first The scale element after the second correction; Even after correction, the matrix reciprocity and reflexivity constraints must still be satisfied; Repeat the steps of quantifying expert opinion consensus and dynamically adjusting the judgment matrix as described above until all experts are consistent in pairs. If the preset consistency requirement is met, the weights of each expert are calculated according to the following formula. : In the formula For the first The weight of each expert; For the first The expert and the first A quantitative value representing the degree of consistency in the opinions of the experts; This represents the total number of experts who participated in the survey.

4. The evaluation method for a long-span bridge structural system according to claim 1, characterized in that, In the final weighting of S2, the weighting ratio of key components, including cable-stayed systems, steel box girders, and cable towers, is increased, while indicators for long-span bridges without corresponding structures are removed.

5. The evaluation method for a long-span bridge structural system according to claim 1, characterized in that, The nonlinear dimensionless model in S3 is specifically as follows: The detection value of the set quantity index , Actual test data representing quantitative indicators related to long-span bridge structures; dimensionless score value. , It is the detection value The evaluation score obtained after normalization has a range of values. , used to quantify the structural performance level corresponding to the indicator; The six preset grading standard values ​​are arranged in ascending order as follows: , , , , , These six standard values ​​are grading thresholds determined based on the structural type, component characteristics, and relevant industry standards and specifications of long-span bridges, and are used to classify the test values. Different performance ranges; Define interval mapping coefficients ,in The interval number, ranging from 1 to 5, corresponds to five consecutive intervals formed by dividing the data using six standard values. Standard value The corresponding benchmark score, Standard value The corresponding benchmark score, For the first Each grading standard value, mapping coefficient Used to characterize the detection value In the When varying within a range, the dimensionless score value The linear rate of change; The unified expression for the model is: When the detection value Less than or equal to the first grading standard value At that time, the dimensionless score value Take the benchmark score corresponding to the first standard value. When the detection value In the first Each grading standard value With the Each grading standard value Between, dimensionless score value Based on benchmark scores and mapping coefficients Deviation of detection value The linear superposition calculation is obtained; when the detected value Greater than or equal to the sixth grading standard value At that time, the dimensionless score value Take the benchmark score corresponding to the sixth standard value. .

6. The evaluation method for a long-span bridge structural system according to claim 5, characterized in that, The positive, negative, and moderate indicators in S3 are as follows: Positive indicators are those whose larger test values ​​indicate better bridge structural performance. Their benchmark scores are configured in a gradient-increasing manner, and the mapping coefficient is positive to achieve a positive correlation between the score and the performance of the indicator; they are suitable for evaluation scenarios of positive performance indicators, including concrete strength and structural stiffness. The inverse index is an index that indicates better bridge structural performance with a smaller detection value. Its benchmark score is configured in a gradient decreasing manner, and the mapping coefficient is negative to achieve an inverse correlation between the score and the degree of index deterioration. It is suitable for assessment scenarios of deterioration indicators, including concrete carbonation depth and steel corrosion rate. The appropriate index is the index that represents the optimal performance of the bridge structure when the detected value is within a reasonable range. Its benchmark score is configured with a logic of low at both ends and high in the middle. It is suitable for evaluation scenarios of indicators that need to be controlled within a reasonable range, including the main beam elevation and cable force deviation.

7. The evaluation method for a long-span bridge structural system according to claim 1, characterized in that, The calculation process for the standardized scoring results of the sequence-type indicators in S3 is as follows: First, the evaluation value of a single measuring point is calculated. For the detection data of each measuring point of the sequential index, the "line segment method" appropriate index model is used for dimensionless processing to obtain the first... Evaluation value of each measuring point Secondly, calculate the score for uniform variation. The uniform variation score is used to characterize the overall average state of the sequential index detection data, corresponding to the standard value curve translation characteristics of the detection data. If all measurement points in the sequence have equal importance, the uniform variation score is taken as the arithmetic mean of the estimated values ​​of each measurement point, calculated using the following formula: If the importance of the measuring points differs, then each measuring point is assigned a weight based on the criticality of its structural location. The uniform variation score is the weighted average of the estimated values ​​from each assessment, calculated using the following formula: In the formula, The total number of measurement points. For the first Evaluation value of each measuring point For the first The weight of each measurement point; Then, calculate the non-uniform variation coefficient. The non-uniform variation coefficient is used to emphasize the changing trend of sequence-type indicators, corresponding to the variation characteristics of the detection data around the standard value axis. It is calculated through slope grey relational analysis, and this relational degree is used to evaluate the similarity of the curve shape between the compared sequence and the reference sequence. For the standard or design value of this test item in the completed state of the bridge, the expression is: The sequence being compared For the bridge in operation The measured value of the second test is expressed as follows: ; First, calculate the slope of each adjacent point in the reference sequence and the compared sequence. The slope of the reference sequence at the nth... The slope of the adjacent points is The sequence being compared is the first one. The slope of the adjacent points is , Then, the slope grey relational degree is calculated using the following formula. And use it as the coefficient of non-uniform variation. : Finally, the standardized score of the sequential indicator is calculated. Scoring for uniform changes Coefficient of non-uniform variation Multiply by each other to obtain the standardized score result. The calculation formula is as follows: The scoring results comprehensively reflect the consistency between the overall average state and the trend of change of the sequential indicators, and achieve a comprehensive quantitative evaluation of the sequential indicators.

8. The evaluation method for a long-span bridge structural system according to claim 1, characterized in that, In S4, the component score is quantified based on the deduction value of each detection index; the component score is calculated by combining the average score, the lowest score of the component and the component quantity adjustment coefficient; the structural score is obtained by dynamically adjusting the component weights through a variable weighting mechanism and then summing them up; the overall bridge score is obtained by weighting the various structural scores according to their corresponding weights.

9. An evaluation system for long-span bridge structures, characterized in that, include: The data processing and indicator construction unit is used to collect manual inspection data, non-destructive testing data and long-term health monitoring data of long-span bridges, and to preprocess these multi-source data to form a standardized reference dataset. Secondly, through on-site investigation, literature review and expert survey, the typical types of defects and structural characteristics of long-span bridges were identified, and a multi-level evaluation framework of "component layer - component layer - defect layer" was established. Then, based on standardized reference data, specific evaluation indicators for each level were determined, and finally a complete evaluation indicator system covering the dimensions of safety, durability and applicability was formed. The weight determination unit is used to collect expert judgment information according to a five-level scale by combining the group hierarchical analysis method with the professional opinions of authoritative industry experts on the evaluation indicators of long-span bridges. The collected expert judgment matrix is ​​then optimized for consistency to reduce subjective judgment bias. Finally, the weights of each expert are calculated based on the optimized expert judgment matrix, and the final weights of each level of indicators in the evaluation index system are obtained through weighted calculation. The evaluation index standardization processing unit is used to classify quantitative evaluation indicators into positive, negative, and moderate indicators based on their attributes. It constructs a nonlinear dimensionless model consisting of six standard values ​​and five line segments for each type of indicator to complete the normalization process of the quantitative indicators. For sequential evaluation indicators, the moderate indicator model of the line segment method is first used to obtain the evaluation value of a single measuring point. Then, the uniform change score is calculated by combining the measuring point weights. The non-uniform change coefficient is obtained through slope grey relational analysis. Finally, the two results are integrated to obtain the standardized score of the sequential indicator. The scoring calculation and maintenance output unit is used to perform calculations step by step according to the hierarchy of "component scoring - part scoring - structural scoring - overall bridge scoring" to obtain the evaluation results at each level. Based on these multi-level evaluation results, the critical risks and potential safety hazards of the bridge structure are identified, and targeted maintenance plans are output.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer program is executed by a processor according to any one of claims 1-8, which describes a method for evaluating long-span bridge structural systems.

Images