Bridge technical condition prediction method based on fusion of category clustering and trend modeling

Abstract

The invention relates to a bridge technical condition prediction method based on category clustering and trend modeling fusion, belongs to the technical field of bridge state prediction and maintenance management, and solves the problems of limited generalization ability and low prediction accuracy of an existing prediction model. The method comprises the following steps: firstly, collecting regular detection data of a plurality of bridges and preprocessing; after feature coding and dimension reduction are carried out, a K-means clustering algorithm is adopted to group bridges, and an intra-class degradation trend model is established by using a Huber regression method; an XGBoost regression method is adopted to establish an intra-class technical condition prediction model; according to the intra-class technical condition prediction model and the intra-class degradation trend model, bridge technical condition prediction results of the target bridge in multiple prediction years in the future are generated through a dynamic weighted fusion formula; and finally performing model performance evaluation. According to the method, the accuracy of bridge technical condition prediction can be effectively improved, reliable long-term degradation trend prediction is provided, and the maintenance decision of the bridge is optimized.

Description

Technical Field

[0001] The present invention relates to the technical field of bridge status prediction and maintenance management, and in particular to a bridge technical condition prediction method based on the fusion of category clustering and trend modeling. The method is specifically used for predicting the technical condition level of bridges and analyzing degradation trends under limited inspection data conditions, providing intelligent technical support for bridge maintenance decision-making and repair optimization. Background Art

[0002] With the continued growth of road traffic and the increasing aging of infrastructure, the safety and durability of bridge structures face significant challenges. Bridge technical condition assessment and prediction of future degradation trends are key components of daily bridge operation and maintenance, as well as long-term health management. These assessments are crucial for ensuring smooth traffic flow, extending bridge service life, and developing appropriate maintenance strategies.

[0003] At present, bridge management systems at home and abroad mainly rely on regular inspections and manual inspections to obtain the technical condition level of bridges, and predict the future technical condition of bridges through expert experience, regression models or single machine learning models.

[0004] For example, the paper "Prediction of Bridge Technical Condition Based on Regression Analysis Method" published in the Journal of Beijing University of Civil Engineering and Architecture proposed using regression analysis method to fit the bridge technical condition decline model, and then predict the development trend of bridge technical condition.

[0005] For example, the paper "Research on Prediction Methods for Development Trends in Technical Condition Grades of In-Service Bridges" published in the journal "Highway and Transportation Research and Development (Applied Technology Edition)" established a development trend model for the technical condition grades of various types of bridges over operating time by performing linear regression on historical data on highway bridge technical condition grade assessments.

[0006] For example, the paper "Research on Grey Markov Chain Model for Bridge Technical Condition Prediction" published in the Journal of Wuhan University of Technology (Transportation Science and Engineering Edition) combines the grey model GM(1,1) with the Markov chain to consider the overall changes and local fluctuations in the technical condition of the bridge.

[0007] For example, the paper "Research on the Degradation Law of Beam Bridges Based on Markov Process" published in the journal "Journal of Jiamusi University (Natural Science Edition)" uses Markov theory and genetic algorithms of nonlinear programming to study the time-varying laws of the preventive maintenance technology of concrete highway beam bridges.

[0008] For example, the Chinese patent publication number CN113569908A, "A method for predicting the technical condition and defects of bridges based on deep learning," proposes a method for predicting the technical condition and defects of bridges based on a CPSO-BP neural network model. By optimizing the neural network structure and parameters, the location, type, and degree of bridge defects can be accurately predicted.

[0009] For example, the Chinese patent "Method for Predicting Service Performance Degradation of Reinforced Concrete Bridges" with publication number CN111160528A proposes using data from inspection reports over the years to train and adjust parameters of multiple LSTM neural networks to accurately predict the technical condition scores of reinforced concrete bridges in the next few years.

[0010] For example, the Chinese patent publication number CN114036258A, “A method for rapidly identifying the technical condition level of bridges based on natural language processing,” provides a method for rapidly identifying the technical condition level of bridges based on natural language processing. The method includes organizing bridge health status information into text descriptions, classifying the technical condition levels, and using machine learning algorithms to establish a training model to achieve rapid identification.

[0011] For example, the paper "Technical Condition Assessment of Small and Medium-Span Bridges Based on Machine Learning" published in the journal "Journal of Chang'an University (Natural Science Edition)" proposed an intelligent assessment method for the technical condition of bridges based on machine learning. By constructing a bridge status database and machine learning algorithm, it achieved an accurate assessment of the technical condition of bridges.

[0012] Most of these methods focus on modeling the historical data of a single bridge, failing to fully explore the common structural and degradation behaviors among similar bridges. This results in limited model generalization, especially in the case of limited data volumes and poor accuracy in long-term trend prediction. Furthermore, existing technologies often overlook the differences among bridge categories and the degradation patterns of different bridge types. Modeling degradation trends is primarily based on the average level of all bridges, making it difficult to reflect the true degradation process within specific categories. Furthermore, traditional methods often employ only a single short-term prediction model, failing to achieve an organic combination of short-term high-precision predictions and long-term trend extrapolation. Consequently, actual prediction results cannot effectively support refined bridge maintenance decisions.

[0013] Therefore, there is an urgent need for a new method for predicting the technical condition of bridges that can comprehensively utilize limited inspection data, combine bridge clustering with degradation trend modeling, and realize the dynamic fusion of short-term prediction and long-term trend excitation, so as to improve the intelligence and scientific level of bridge health management and make up for the shortcomings of existing technologies. Summary of the Invention

[0014] Given the limited bridge inspection data or incomplete historical records, conventional data-driven models have low prediction accuracy. There is an urgent need to develop a bridge technical condition prediction method that integrates structural mechanics mechanisms. This paper proposes a bridge technical condition prediction method based on the fusion of category clustering and trend modeling. This method, through data acquisition and preprocessing, category feature clustering and trend analysis, intra-category prediction model establishment, joint prediction and dynamic weighted fusion, result evaluation and comparative verification, ultimately achieves technical condition prediction and degradation trend assessment of bridges under limited data. This method can effectively improve the accuracy of bridge technical condition predictions, provide reliable long-term degradation trend predictions, and optimize bridge maintenance decisions.

[0015] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0016] A bridge technical condition prediction method based on the fusion of category clustering and trend modeling includes the following steps:

[0017] Step 1: Collect regular inspection data of multiple bridges, including bridge code, construction year, inspection year, assessment grade, structural parameters, and material properties. Preprocess the collected data, including missing value filling, outlier processing, feature calculation, and standardization. After preprocessing, structured preprocessed data is obtained.

[0018] Step 2: Feature encoding and dimensionality reduction are performed on the structured preprocessed data obtained in Step 1. The K-means clustering algorithm is then used to group the bridges into several categories with similar structures and degradation behaviors. Within each category, bridge samples with significant degradation are screened. The Huber regression method is then used to fit the degradation trend curves of bridge age and technical condition within each category to establish an intra-category degradation trend model.

[0019] Step 3: Within each bridge category, an XGBoost regression method is used to establish a technical condition prediction model based on historical ratings and static characteristics.

[0020] Step 4: Based on the intra-category technical condition prediction model and intra-category degradation trend model corresponding to the bridge category of the target bridge, a dynamic weighted fusion formula is used to generate the bridge technical condition prediction results for the target bridge in multiple future forecast years. In the dynamic weighted fusion formula, the intra-category technical condition prediction model has a higher weight for short-term prediction, and the intra-category degradation trend model has a higher weight for long-term prediction. The sum of the weights of the intra-category technical condition prediction model and the intra-category degradation trend model is 1.

[0021] Step 5: Use evaluation indicators to evaluate the model performance of the bridge technical condition prediction results obtained in step 4, where the evaluation indicators include accuracy, mean absolute error, and root mean square error.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] (1) Improved prediction accuracy and applicability: This invention overcomes the shortcomings of low prediction accuracy and poor generalization ability of traditional single models by combining category clustering and trend modeling, significantly improving the accuracy and robustness of bridge technical condition prediction, and is particularly suitable for actual engineering scenarios with limited scheduled inspection data;

[0024] (2) Realize short-term and long-term dynamic fusion prediction: adopt the dynamic weighted fusion of the intra-category technical status prediction model and the intra-category degradation trend model, taking into account the high accuracy of short-term prediction and the rationality of long-term trend extrapolation, effectively making up for the shortcomings of existing methods in long-term degradation prediction;

[0025] (3) Reflecting category differences and revealing degradation patterns: By clustering bridges based on structural and material characteristics and fitting degradation trends within the categories, the degradation differences and evolution patterns of bridges of different categories can be fully revealed, thus improving the engineering reference value of the prediction results.

[0026] (4) The method is scientific and reliable, and has strong engineering practicality: The present invention combines a variety of cross-validation and evaluation indicators to fully verify the scientific nature and superiority of the method, making it easy to integrate into the bridge health management system and improve the intelligence level and management efficiency of bridge maintenance decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the examples. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be considered as limiting the scope. A person of ordinary skill in the art can also derive other relevant drawings based on these drawings without inventive effort. In the drawings:

[0028] Figure 1 This is a flow chart of a bridge technical condition prediction method based on the fusion of category clustering and trend modeling according to one embodiment of the present invention;

[0029] Figure 2 This is the clustering result diagram of the bridge category;

[0030] Figure 3 A curve chart predicting the technical condition level of a bridge;

[0031] Figure 4 This is a forecast chart of the average future degradation trend of each type of bridge. DETAILED DESCRIPTION

[0032] To make the purpose, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. It should be understood that the specific embodiments described are only used to explain this application and are not used to limit this application.

[0033] The present invention proposes a method for joint prediction of bridge category clustering and trend modeling. The method mainly uses the K-means clustering algorithm to group bridges through preprocessed bridge data, and uses the Huber regression method to fit the degradation trend curve of bridge age-technical condition within the category. Then, within each bridge category, a machine learning regression method is used to establish an intra-category technical condition prediction model to achieve local prediction capabilities driven by short-term regular inspection data. Finally, joint prediction and dynamic weighted fusion are performed. For the target bridge, the intra-category technical condition prediction model and the intra-category degradation trend model are combined, and a dynamic weighted fusion scheme is used to generate bridge technical condition prediction results for many years in the future. Through this method of clustering first and then joint prediction, the characteristics of different bridge categories are fully considered, the accuracy and scientificity of the prediction are improved, and a reliable basis is provided for the intelligent maintenance and management of bridges.

[0034] Specifically, if Figure 1 As shown, this embodiment provides a bridge technical condition prediction method based on the fusion of category clustering and trend modeling, which mainly includes the following steps 1 to 5.

[0035] Step 1 (S100), data acquisition and preprocessing: collect periodic inspection data of multiple bridges, the periodic inspection data including bridge code, construction year, inspection year, assessment grade, structural parameters and material properties, and preprocess the collected periodic inspection data, including missing value filling, outlier processing, feature calculation and standardization processing, to obtain structured preprocessed data after preprocessing.

[0036] The periodic inspection data collected in this embodiment, i.e., bridge data, can be obtained from a bridge inspection database. The bridge inspection database stores inspection data for multiple bridges. Taking the periodic inspection data for five consecutive years (e.g., 2020-2024) as an example, a bridge data file is collected. The main fields include: bridge code, construction year, inspection year, assessment grade, total span length, maximum span of a single span, clear width of the bridge deck, name of the superstructure material, design load level, etc. The assessment grade text is mapped to a numerical type (such as "Class I" corresponds to 1, Class II corresponds to 2, and so on). For the periodic inspection data obtained, first, if fields such as the construction year are missing, the median of similar bridges is used to fill in the missing fields. Afterwards, based on the original data fields, the span ratio (total span length / maximum span of a single span), width-to-span ratio (clear width of the bridge deck / maximum span of a single span), material degradation coefficient (assigned by looking up the table according to the structure type), and load factor (numerized according to the design load level) are automatically calculated. For all numerical features, the mean and standard deviation of the same feature of all bridges are used for standardization to ensure the consistent distribution of features between different years and different bridges; for categorical features, one-hot encoding is used to generate 0-1 variables to adapt to model training. The box plot method shown in the following formula is used for correction:

[0037] (1);

[0038] in, For the The value of the sample after outlier correction; and are the 1st and 3rd quartiles, .

[0039] Step 2 (S200), category feature clustering and trend analysis: Feature encoding and dimensionality reduction are performed on the preprocessed bridge data, i.e., the structured preprocessed data obtained in step 1. The bridges are then grouped using the K-means clustering algorithm to obtain several bridge categories with similar structures and degradation behaviors. Bridge samples with significant degradation are screened within each bridge category. The Huber regression method is then used to fit the degradation trend curve of bridge age-technical condition within each bridge category to establish an intra-category degradation trend model.

[0040] Step 2.1: Construct a bridge feature matrix based on the preprocessed bridge data , which includes all the characteristics of bridge structure characteristics, material characteristics, operation and static characteristics (such as bridge age, span ratio, material degradation coefficient, load factor, recent assessment level, etc.) after coding and standardization. The bridge characteristic matrix is ​​analyzed by principal component analysis (PCA) method. Perform dimensionality reduction, reduce dimensionality Determined based on the cumulative variance contribution rate, the goal is to capture the main difference information. The PCA dimensionality reduction process is expressed as follows:

[0041] (2);

[0042] in, For the front The projection matrix composed of principal components; It is the feature matrix after dimensionality reduction by principal component analysis (PCA), representing the projection characteristics of the bridge in the principal component space.

[0043] Step 2.2: Perform K-means clustering on the PCA results to group the bridges. The number of cluster categories can be adaptively selected based on the actual number of bridges and feature differences, preferably 3-6 categories. In this embodiment, the number of categories is set to , the K-means clustering algorithm is used to group the bridges, and four bridge categories with similar structures and degradation behaviors are obtained, namely cluster 0, cluster 1, cluster 2 and cluster 3. The objective function of the K-means clustering algorithm is:

[0044] (3);

[0045] in, For the bridge PCA dimensionality reduction features, , For the Cluster centers, is the number of categories, is the number of bridges participating in clustering, express and cluster centers The square of the Euclidean distance between them.

[0046] The clustering results can be visualized by the scatter plot of the first two dimensions of the principal components, such as Figure 2 shown.

[0047] Step 2.3: Next, model the degradation trend within the category and establish an intra-category degradation trend model. First, within each bridge category, select bridge samples with significant degradation. The selection criteria are that the maximum rating of bridges within the category in the past five years is greater than or equal to 4, and the rating fluctuation range exceeds 1 point. Then, for the selected bridge samples, a regression sample is constructed. The Huber regression method is used to fit the degradation trend curve of bridge age-technical condition within each bridge category. The intra-category degradation trend model is obtained. Its calculation formula is:

[0048] (4);

[0049] in, For the bridge In the year The technical condition assessment grade of the bridge; For the bridge In the year Age of the bridge; is the regression coefficient; is the residual.

[0050] Step 3 (S300), establish a technical condition prediction model within the category: within each bridge category, based on the historical assessment grades and static characteristics over many years, use the XGBoost regression method to establish a technical condition prediction model within the category, and realize local prediction capabilities driven by short-term inspection data.

[0051] For each bridge category, within each category, the bridge The history of four consecutive years of rating and static characteristics For input, the following year's assessment grade The output is static features, which refer to the bridge's inherent properties corresponding to a particular inspection year and do not change with the inspection year. These include, but are not limited to, bridge age, span ratio, material degradation coefficient, load factor, structural material type, etc. These features are typically used as model inputs along with historical dynamic scores (such as technical condition grades). The XGBoost regression model is used to establish a within-category technical condition prediction model. The prediction model formula is as follows:

[0052] (5);

[0053] in, Indicates the bridge exist The predicted value of the annual technical condition level; is the number of weak learners; For the regression trees; is the input feature vector, which is composed of bridge Ratings for four consecutive years before the forecast year and static characteristics of bridges The spliced ​​components serve as the complete input to the in-category technology status prediction model.

[0054] Step 4 (S400), joint prediction and dynamic weighted fusion: for the target bridge, the technical condition prediction model and the degradation trend model within the category corresponding to the bridge category of the target bridge are combined, and the bridge technical condition prediction results of the target bridge in multiple future prediction years are generated through a dynamic weighted fusion formula. In the dynamic weighted fusion formula, the weight of the technical condition prediction model within the category is higher in the short-term prediction, that is, the short-term prediction is mainly based on the technical condition prediction model within the category, and the weight of the degradation trend model within the category is higher in the long-term prediction, that is, the long-term prediction gradually enhances the weight of the degradation trend model within the category and is mainly based on the degradation trend model within the category, and the sum of the weight of the technical condition prediction model within the category and the weight of the degradation trend model within the category is always 1.

[0055] Assume that the forecast year is the future In 2018, the output of the technology status prediction model within the category is , the output of the intra-class degradation trend model is ,in For the The bridge will be the first If the bridge age is years, the dynamic weighted fusion formula can be expressed as:

[0056] (6);

[0057] in, To build a bridge for the future The predicted value of the bridge technical condition grade in 2017; is a dynamic weight, which is based on the forecast year The adaptive determination of historical data fitting results is used to increase the weight of the technical status prediction model output within the category in the short-term prediction and to increase the weight of the degradation trend model output within the category in the long-term prediction. The setting of dynamic weights enables the fusion model to flexibly adjust the contribution of each component according to the forecast year, so that the short-term forecast relies on the data-driven model (XGBoost) to obtain high accuracy, while the long-term forecast focuses on the robustness of the degradation trend model to prevent extrapolation distortion. The short-term forecast in this embodiment refers to the forecast year. When the year threshold is less than or equal to 3, the technical condition grade of the target bridge is predicted. Long-term prediction refers to the prediction of the technical condition grade of the target bridge in the forecast year. When the age is greater than the year threshold, the technical condition grade of the target bridge is predicted.

[0058] Optionally, in the dynamic weighted fusion formula, the weight of the intra-category technical status prediction model decreases monotonically with the increase of the forecast year, and the weight of the intra-category degradation trend model increases monotonically with the increase of the forecast year. For example, in the first three years of the forecast year, the intra-category technical status prediction model is mainly used, so the weights of the intra-category technical status prediction model are set to 0.9, 0.7, and 0.5 respectively. After three years, the weight of the intra-category degradation trend model gradually increases, and the dynamic weight The formula is as follows:

[0059] (7);

[0060] In this example, the weights for the intra-category technical status forecast model are set at 0.9, 0.7, and 0.5 for the first three years, reflecting a high degree of confidence in short-term historical data. In the fourth year and thereafter, the weights are reduced by 0.1 each year until they reach 0, ensuring that the forecast results maintain a reasonable physical evolution trend over the long term. This weighting scheme ensures the model's applicability across different forecast years and allows for flexible adjustment based on actual needs.

[0061] After each prediction, the bridge technical condition prediction results are added to the historical data and recursively predicted for subsequent years. Figure 3 The graph shows a forecast curve of the technical condition level of a bridge in the next 10 years. The "fusion prediction" shown by the red curve in the figure is the forecast result of the technical condition of the bridge obtained by using the method of the present invention.

[0062] Step 5 (S500), result evaluation and comparative verification: The model performance of the bridge technical condition prediction results obtained in step 4 is evaluated using evaluation indicators, wherein the evaluation indicators include accuracy, mean absolute error, and root mean square error.

[0063] Accuracy ( ) is calculated as:

[0064] (8);

[0065] in, is the indicator function, For the The predicted value of the technical condition level of each bridge, For the The actual value of the technical condition grade of each bridge;

[0066] Mean absolute error ( ) is calculated as:

[0067] (9);

[0068] Root mean square error ( ) is calculated as:

[0069] (10);

[0070] in, is the total number of prediction samples.

[0071] Furthermore, using regular inspection data from 2,711 bridges as raw data, we conducted a leave-one-out cross-validation exercise on all 2,711 bridges. The results were compared with those of the traditional mean and single regression methods to verify the effectiveness and superiority of the proposed combined prediction method. The experimental results are shown in Table 1.

[0072] Table 1 Experimental results

[0073]

[0074] Figure 4 The chart shows the average degradation trend forecast for each bridge category. This average degradation trend is a trend curve derived by averaging the combined predicted technical condition grades of all bridges within a cluster over multiple years (e.g., a dynamic weighted fusion prediction). The "fused mean" in the figure is the annual average of the predicted bridge technical condition grades for all bridges in that category, reflecting the overall degradation evolution of that bridge category.

[0075] In summary, the present invention provides a bridge technical condition prediction method based on the fusion of category clustering and trend modeling. This method, through data acquisition and preprocessing, category feature clustering and trend analysis, intra-category prediction model establishment, joint prediction and dynamic weighted fusion, and result evaluation and comparative verification, ultimately enables bridge technical condition prediction and degradation trend assessment using limited data. By clustering bridge models of different categories, the degradation trends of bridges within each category are fully explored. Finally, a joint prediction method is used to fuse the intra-category technical condition prediction model with the intra-category degradation trend model. This effectively improves the accuracy of bridge technical condition prediction and provides reliable long-term degradation trend prediction, optimizing bridge maintenance decisions. Compared with traditional methods, the present invention offers superior practicality, efficiency, and broad application prospects. It fully exploits the structural information of limited detection data, integrates category degradation evolution patterns, balances short-term and long-term predictions, and is highly versatile, providing a scientific basis for bridge maintenance decision-making and risk assessment.

[0076] The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes based on the technical solution and inventive concept of the present invention within the scope disclosed by the present invention, which falls within the scope of protection of the present invention.

Claims

1. A bridge technical condition prediction method based on the fusion of category clustering and trend modeling is characterized by: The following steps are involved: Step 1: Collect regular inspection data of multiple bridges, including bridge code, construction year, inspection year, assessment grade, structural parameters, and material properties. Preprocess the collected data, including missing value filling, outlier processing, feature calculation, and standardization. After preprocessing, structured preprocessed data is obtained. Step 2: Feature encoding and dimensionality reduction are performed on the structured preprocessed data obtained in Step 1. The K-means clustering algorithm is then used to group the bridges into several categories with similar structures and degradation behaviors. Within each category, bridge samples with significant degradation are screened. The Huber regression method is then used to fit the degradation trend curves of bridge age and technical condition within each category to establish an intra-category degradation trend model. Step 3: Within each bridge category, an XGBoost regression method is used to establish a technical condition prediction model based on historical ratings and static characteristics. Step 4: Based on the intra-category technical condition prediction model and intra-category degradation trend model corresponding to the bridge category of the target bridge, a dynamic weighted fusion formula is used to generate the bridge technical condition prediction results for the target bridge in multiple future forecast years. In the dynamic weighted fusion formula, the intra-category technical condition prediction model has a higher weight for short-term prediction, and the intra-category degradation trend model has a higher weight for long-term prediction. The sum of the weights of the intra-category technical condition prediction model and the intra-category degradation trend model is 1. Step 5: Use evaluation indicators to evaluate the model performance of the bridge technical condition prediction results obtained in step 4, where the evaluation indicators include accuracy, mean absolute error, and root mean square error.

2. The bridge technical condition prediction method based on the fusion of category clustering and trend modeling according to claim 1 is characterized in that: The preprocessing in step 1 includes: For numerical features, mean filling is used; Perform one-hot encoding on categorical features to generate 0-1 variables for model training; For outliers The box plot method shown in the following formula is used for correction: （1）； in, For the The value of the sample after outlier correction; and are the 1st and 3rd quartiles, .

3. The bridge technical condition prediction method based on the fusion of category clustering and trend modeling according to claim 1 or 2 is characterized in that: Step 2 includes: Step 2.1: Construct a bridge feature matrix based on the structured preprocessed data obtained in step 1 , the principal component analysis method is used to analyze the bridge characteristic matrix To reduce the dimension, the PCA dimension reduction process is: （2）； in, For the front The projection matrix composed of principal components; is the feature matrix after dimensionality reduction by principal component analysis method; Step 2.2: Use the K-means clustering algorithm to group the bridges. The objective function of the K-means clustering algorithm is: （3）； in, For the bridge PCA dimensionality reduction features, For the cluster centers, is the number of categories, is the number of bridges participating in clustering, express and cluster centers The square of the Euclidean distance between them; Step 2.3: Within each bridge category, screen out bridge samples with significant degradation. For these bridge samples with significant degradation, use the Huber regression method to fit the degradation trend curve of bridge age-technical condition within the bridge category to obtain the degradation trend model within the category.

4. The bridge technical condition prediction method based on the fusion of category clustering and trend modeling according to claim 3 is characterized in that: The intra-class degradation trend model is: （4）； in, For the bridge In the year The technical condition assessment grade of the bridge; For the bridge In the year Age of the bridge; is the regression coefficient; is the residual.

5. The bridge technical condition prediction method based on the fusion of category clustering and trend modeling according to claim 3 is characterized in that: The screening criteria for selecting bridge samples with significant degradation within a bridge category are: the maximum rating of bridges in this bridge category in the past five years is greater than or equal to 4, and the rating fluctuation range exceeds 1 point.

6. The bridge technical condition prediction method based on the fusion of category clustering and trend modeling according to claim 1 or 2 is characterized in that: The technology status prediction model within the category is: （5）； in, Indicates the bridge exist The predicted value of the annual technical condition level; is the number of weak learners; For the regression trees; is the input feature vector.

7. The bridge technical condition prediction method based on the fusion of category clustering and trend modeling according to claim 1 or 2 is characterized in that: The dynamic weighted fusion formula is: （6）； in, For the target bridge in the future The predicted value of the bridge technical condition grade in 2017; is the dynamic weight; The output of the prediction model for the technical status within the category; is the output of the within-class degradation trend model.

8. The bridge technical condition prediction method based on the fusion of category clustering and trend modeling according to claim 7 is characterized in that: Dynamic Weight The formula is as follows: （7）； in, Forecast year.

9. The bridge technical condition prediction method based on the fusion of category clustering and trend modeling according to claim 1 or 2 is characterized in that: The calculation formula of the accuracy is: （8）； in, is the indicator function, For the The predicted value of the technical condition level of each bridge, For the The actual value of the technical condition grade of each bridge; The calculation formula of the mean absolute error is: （9）； The calculation formula of the root mean square error is: （10）； in, is the total number of prediction samples.

10. The bridge technical condition prediction method based on the fusion of category clustering and trend modeling according to claim 1 or 2, characterized in that: The periodic inspection data is collected from a bridge inspection database.

Images