A bridge condition assessment method based on the slope of vehicle weight-strain mapping model

Abstract

The present invention discloses a bridge state assessment method based on the slope of a vehicle weight-strain mapping model, comprising the following steps: step 1: obtaining vehicle weight data and vehicle-induced strain data of a bridge to be assessed; step 2: clustering vehicle-induced strain and vehicle weight data respectively; step 3: generating vehicle-induced strain and vehicle weight random sampling data of consistent length; step 4: obtaining the slope of a vehicle weight-strain digital mapping model based on Bayesian linear regression; step 5: calculating the slope assessment threshold of each cluster obtained by clustering, and step 6: determining the performance change and safety status of the bridge in the subsequent monitoring process based on the slope assessment threshold. The method can effectively consider the safety state change of the bridge under daily traffic flow and timely provide abnormal bridge performance warning.

Description

Technical Field

[0001] The invention relates to the field of bridge health status assessment, and in particular to a bridge status assessment method based on the slope of a vehicle weight-strain mapping model. Background Art

[0002] With the continuous increase in transportation demand, the traffic loads that bridges bear during their service are also increasing. Among the many factors that affect the safety status and remaining life of existing bridges, vehicle loads are the main live load that causes bridge damage, especially for concrete bridges, which can easily cause bridge cracking, leading to safety problems such as bridge performance degradation, which may cause engineering accidents. In order to meet the requirements of transportation, it is necessary to find a balance between the increasing traffic flow and the degradation of structural safety performance. By analyzing the inherent relationship between bridge vehicle loads and structural responses, and considering the performance changes and performance of bridges under daily traffic flow, it is possible to monitor the safety status of bridge structures under traffic loads for a long time, providing a reliable basis for the subsequent safe operation and maintenance of bridges.

[0003] With the widespread application of dynamic weighing systems and health monitoring systems in bridge structures, massive amounts of monitoring data are used in bridge safety status assessment. With the advent of the big data era and the continuous development of machine learning, various artificial intelligence networks and clustering methods have made bridge status assessment methods based on monitoring data more accurate and comprehensive. However, dynamic weighing systems and health monitoring systems are usually used as two independent modules to achieve different functions in bridge services. The dynamic weighing system of the bridge is mainly aimed at traffic flow and can effectively collect information such as the weight and speed of vehicles passing through the bridge, while the health monitoring system is used to collect the response of the bridge under various environmental and human factors.

[0004] Therefore, in order to consider the performance changes of bridges under traffic loads, it is necessary to integrate the dynamic weighing system with the health monitoring system and find a new method with good feasibility, high accuracy and strong applicability to accurately evaluate the safety status of bridges under the influence of daily traffic flow. Summary of the invention

[0005] Technical problem: The technical problem to be solved by the present invention is to provide a method for evaluating the safety status of a bridge under the influence of increasing traffic loads, which method can accurately provide early warning of abnormal safety status of the bridge.

[0006] Technical solution: The present invention adopts a bridge condition assessment method based on the slope of a vehicle weight-strain mapping model. The method takes the long-term monitoring data of the weight of vehicles passing the bridge and the vehicle-induced strain of the bridge as the object. First, the vehicle weight and strain data are clustered by analyzing the statistical characteristics of the data. Then, the Metropolis-Hastings algorithm is used to perform random simulation sampling on the vehicle weight and strain data to obtain vehicle weight and strain sampling data with consistent length. Then, a vehicle weight-strain linear correlation mapping model is established using Bayesian linear regression. Finally, a method for evaluating the overall performance of the bridge during long-term service is formulated using the slope of the model as an indicator.

[0007] A bridge condition assessment method based on the slope of a vehicle weight-strain mapping model comprises the following steps:

[0008] Step 1: Obtain vehicle weight data and vehicle-induced strain data of the bridge to be evaluated;

[0009] Step 2: Cluster the vehicle-induced strain data and vehicle weight data respectively;

[0010] Step 3: For each cluster obtained by clustering in step 2, select N groups of data from the vehicle-induced strain and vehicle weight data of each cluster according to the set requirements to generate random sampling data of vehicle-induced strain and vehicle weight with the same length; where N is a positive integer;

[0011] Step 4: Take each set of randomly sampled data of vehicle-induced strain and vehicle weight as a set of samples, generate a vehicle weight-strain mapping model based on the Bayesian linear regression method, and obtain the slope of each vehicle weight-strain mapping model;

[0012] Step 5: Calculate the distance between the slope of each vehicle weight-strain mapping model in each cluster and the slope of the centroid of the vehicle weight-strain mapping model in the cluster, and take the maximum value of the distance calculated in each cluster as the evaluation threshold of each cluster;

[0013] Step 6: Periodically collect vehicle weight data and vehicle-induced strain data of the bridge to be evaluated in real time. If the data exceeds M consecutive periods, the distance between the slope of at least one vehicle weight-strain mapping model in each cluster and the centroid of the slope of the vehicle weight-strain mapping model in the cluster is greater than the evaluation threshold of the cluster, indicating that the bridge structure performance is abnormal (the bridge may be damaged and an alarm is required). Otherwise, it is considered that the bridge structure performance is normal; where M is a positive integer.

[0014] Furthermore, in step 1, the method for obtaining the bridge vehicle weight data and vehicle-induced strain data is as follows:

[0015] The bridge is equipped with a dynamic weighing system and a health monitoring system to collect real-time information on vehicles passing through the bridge structure and bridge strain response data at different time periods every day;

[0016] Extracting the vehicle weight data of the vehicle crossing the bridge from the vehicle information;

[0017] From the bridge strain response data, the temperature-induced strain of the bridge is separated, and the maximum value data of the vehicle-induced strain caused by the vehicle weight is extracted.

[0018] Furthermore, the method for clustering the vehicle-induced strain data and the vehicle weight data in step 2 is as follows:

[0019] Perform distribution fitting on the vehicle weight data to obtain the distribution statistical characteristics of the vehicle weight data, and cluster the vehicle weight data according to the distribution statistical characteristics of the vehicle weight data;

[0020] Performing distribution fitting on the vehicle-induced strain data to obtain the distribution statistical characteristics of the vehicle-induced strain data, and clustering the vehicle-induced strain data according to the distribution statistical characteristics of the vehicle-induced strain data;

[0021] The clustering of the above vehicle-induced strain data and vehicle weight data is consistent with the cluster categories, both of which are K clusters.

[0022] Furthermore, in step 3, the method for generating vehicle-induced strain and vehicle weight random sampling data with consistent length is as follows:

[0023] For each cluster obtained by clustering in step 2, according to the set requirements, select N groups of data from the vehicle-induced strain and vehicle weight data of each cluster as random sampling objects;

[0024] For each set of vehicle weight data and vehicle-induced strain data, the distribution fitting function is taken as the objective function, and the Gaussian distribution that can envelope the probability distribution of vehicle weight data and vehicle-induced strain data is taken as the recommended distribution function. The Metropolis-Hastings algorithm is used to perform random simulation sampling to generate a set of vehicle weight data and vehicle-induced strain random sampling data with the same data length that obey the distribution of the original vehicle weight data and vehicle-induced strain data.

[0025] Furthermore, in step 4, the method for obtaining the slope of the vehicle weight-strain mapping model based on Bayesian linear regression is as follows:

[0026] For each cluster, arrange each set of vehicle-induced strain and vehicle weight random sampling data obtained in step 3 in descending order and correspond one to one to form a sample set for each group;

[0027] Based on the sample set of each group, a vehicle weight-strain mapping model is established by using the Bayesian linear regression method, and the slope of the vehicle weight-strain mapping model is obtained.

[0028] Furthermore, in step 5, the Euclidean distance between the slope of each vehicle weight-strain mapping model in each cluster and the slope of the centroid of the vehicle weight-strain mapping model in the cluster is calculated.

[0029] Beneficial effects: Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. Accurately consider the safety status of the bridge under the influence of daily traffic flow

[0031] The present invention provides a performance evaluation method for a bridge under the influence of vehicle load by analyzing the intrinsic relationship between vehicle load and bridge strain; a more sophisticated mapping model can be established through clustering; the correspondence between vehicle load and the bridge strain caused by it can be ensured through random simulation sampling; the uncertainty of the mapping model can be effectively considered through Bayesian linear regression; in addition, by analyzing the change of the slope of the mapping model during the long-term monitoring of the bridge, the accuracy of the corresponding evaluation method is guaranteed;

[0032] 2. The evaluation method is simple and practical

[0033] The evaluation method of the present invention takes the long-term monitoring data of the dynamic weighing system and the health monitoring system as the object, is simple and practical, has good feasibility, and the data extraction and analysis process is simple and accurate, has good applicability, and can better integrate the dynamic weighing system and the health monitoring system so that the two no longer play a role in isolation, thereby maximizing the data value; at the same time, the measurement method of the present invention can also provide a theoretical basis for the inspection of bridge concrete cracks and the control of overweight vehicles crossing the bridge, and can be widely promoted and applied. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 is a flow chart of the method of the present invention;

[0035] Figure 2 It is a typical vehicle-induced strain for bridges;

[0036] Figure 3 The frequency histogram and fitting results of bridge vehicle-induced strain and vehicle weight data, where (a) is strain data and (b) is vehicle weight data;

[0037] Figure 4 The clustering results of bridge vehicle-induced strain and vehicle weight data, where (a) is strain data and (b) is vehicle weight data;

[0038] Figure 5 This is an example of the comparison between a group of vehicle-induced strain and vehicle weight sampling data and the original data in cluster 1, where (a) is strain data and (b) is vehicle weight data;

[0039] Figure 6 This is an example diagram of the vehicle weight-strain correlation mapping model for a group of vehicles in cluster 1. DETAILED DESCRIPTION

[0040] The technical solution of the present invention will be described in detail below through embodiments with reference to the accompanying drawings.

[0041] In one embodiment, a bridge condition assessment method based on the slope of a vehicle weight-strain mapping model is provided. Figure 1 As shown, the following steps are included:

[0042] Step 1: Obtain vehicle weight data and vehicle-induced strain data of the bridge to be evaluated;

[0043] Step 2: Cluster the vehicle-induced strain data and vehicle weight data respectively;

[0044] Step 3: For each cluster obtained by clustering in step 2, select N groups of data from the vehicle-induced strain and vehicle weight data of each cluster according to the set requirements to generate random sampling data of vehicle-induced strain and vehicle weight with the same length; where N is a positive integer;

[0045] Step 4: Take each set of randomly sampled data of vehicle-induced strain and vehicle weight as a set of samples, generate a vehicle weight-strain mapping model based on the Bayesian linear regression method, and obtain the slope of each vehicle weight-strain mapping model;

[0046] Step 5: Calculate the distance between the slope of each vehicle weight-strain mapping model in each cluster and the slope of the centroid of the vehicle weight-strain mapping model in the cluster, and take the maximum value of the distance calculated in each cluster as the evaluation threshold of each cluster;

[0047] Step 6: Periodically collect vehicle weight data and vehicle-induced strain data of the bridge to be evaluated in real time. If the data exceeds M consecutive cycles, the distance between the slope of at least one vehicle weight-strain mapping model in each cluster and the centroid of the slope of the vehicle weight-strain mapping model in the cluster is greater than the evaluation threshold of the cluster, indicating that the structural performance of the bridge is abnormal (the beam may be damaged and an alarm is required). Otherwise, it is considered that the structural performance of the bridge is normal.

[0048] In one embodiment, step 1 specifically includes:

[0049] The bridge is equipped with a dynamic weighing system and a health monitoring system to collect real-time information on vehicles passing through the bridge structure and bridge strain response data at different time periods every day;

[0050] Extracting the vehicle weight data of the vehicle crossing the bridge from the vehicle information;

[0051] From the bridge strain response data, the temperature-induced strain of the bridge is separated, and the maximum value data of the vehicle-induced strain caused by the vehicle weight is extracted.

[0052] In one embodiment, step 2 specifically includes:

[0053] Perform distribution fitting on the vehicle weight data to obtain the distribution statistical characteristics of the vehicle weight data, and cluster the vehicle weight data according to the distribution statistical characteristics of the vehicle weight data;

[0054] Performing distribution fitting on the vehicle-induced strain data to obtain the distribution statistical characteristics of the vehicle-induced strain data, and clustering the vehicle-induced strain data according to the distribution statistical characteristics of the vehicle-induced strain data;

[0055] The clustering of the above vehicle-induced strain data and vehicle weight data is consistent with the cluster categories, both of which are K clusters.

[0056] In one embodiment, step 3 specifically includes:

[0057] For each cluster obtained by clustering in step 2, according to the set requirements, select N groups of data from the vehicle-induced strain and vehicle weight data of each cluster as random sampling objects;

[0058] For each set of vehicle weight data and vehicle-induced strain data, the distribution fitting function is taken as the objective function, and the Gaussian distribution that can envelope the probability distribution of vehicle weight data and vehicle-induced strain data is taken as the recommended distribution function. The Metropolis-Hastings algorithm is used to perform random simulation sampling to generate a set of vehicle weight data and vehicle-induced strain random sampling data with the same data length that obey the distribution of the original vehicle weight data and vehicle-induced strain data.

[0059] In one embodiment, step 4 specifically includes:

[0060] For each cluster, arrange each set of vehicle-induced strain and vehicle weight random sampling data obtained in step 3 in descending order and correspond one to one to form a sample set for each group;

[0061] Based on the sample set of each group, a vehicle weight-strain mapping model is established by using the Bayesian linear regression method, and the slope of the vehicle weight-strain mapping model is obtained.

[0062] In one embodiment, step 5 is specifically:

[0063] The distance between the slope of each vehicle weight-strain mapping model in each cluster and the slope of the centroid of the vehicle weight-strain mapping model in the cluster is calculated, and the maximum value of the distance calculated in each cluster is taken as the evaluation threshold of each cluster.

[0064] In one embodiment, taking the Euclidean distance as an example, the evaluation threshold d of a cluster is max The calculation formula is:

[0065] d max =max|x i -x c |

[0066] Where: i∈[1,N], x i is the slope of the i-th vehicle weight-strain mapping model in the cluster, x c is the slope of the centroid of the vehicle weight-strain mapping model within the cluster.

[0067] In one embodiment, the specific implementation process of the present invention is described by taking the vehicle weight data collected by the dynamic weighing system of a bridge in Rugao, Jiangsu Province in a certain year and the bridge strain collected by the health monitoring system as an example.

[0068] Through the dynamic weighing system and health monitoring system installed on a bridge, the weight data and strain response of vehicles passing through the bridge in a certain year were extracted, and the temperature-induced strain component in the strain response was eliminated to obtain the vehicle-induced strain response of the bridge, such as Figure 2 shown.

[0069] The vehicle weight data and vehicle-induced strain data of that year were fitted with distributions. Figure 3 The frequency histograms and fitting results of vehicle weight and vehicle-induced strain data are shown in Figure 4. It can be seen from the figure that the long-term distribution of vehicle weight and vehicle-induced strain is similar. Based on the distribution statistical characteristics of vehicle weight data and vehicle-induced strain, clustering is performed separately. Vehicle weight and vehicle-induced strain are clustered into 3 clusters respectively. The clustering results are shown in Figure 4.

[0070] Considering the actual situation of a certain bridge, the data length of each group in each cluster is set to 5-6 days of data collection, and each cluster has 60 groups.

[0071] The Metropolis-Hastings algorithm is used to perform random simulation sampling on each group of vehicle weight data and vehicle-induced strain data, generating random sampling data of vehicle-induced strain and vehicle weight that are consistent with the distribution of the original vehicle-induced strain data and vehicle weight data and have the same data length. The vehicle-induced strain and vehicle weight data of a group in cluster 1 obtained by clustering are used as an example to demonstrate the comparison of the distribution of the original vehicle weight and strain data and the estimated value of the probability density function of the sampling data. Figure 5 .

[0072] The generated random sampling data of each group of vehicle-induced strain and vehicle weight are arranged in descending order and correspond one to one. The Bayesian linear regression is used to establish the vehicle weight-strain digital mapping correlation model of each group and extract the model slope. Figure 6 This is an example diagram of a linear correlation model between vehicle weight and strain for cluster 1. For the three clusters obtained by clustering, 60 models are generated for each cluster, and the slopes of these 60 models are calculated.

[0073] The evaluation threshold within each cluster is calculated as the basis for judging changes in bridge performance during subsequent monitoring.

[0074] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any person familiar with the technology can understand and think of any changes or substitutions within the technical scope disclosed by the present invention, which should be included in the scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A bridge condition assessment method based on the slope of a vehicle weight-strain mapping model, characterized in that: The steps include: Step 1: Obtain vehicle weight data and vehicle-induced strain data of the bridge to be evaluated; Step 2: Cluster the vehicle-induced strain data and vehicle weight data respectively; Step 3: For each cluster obtained by clustering in step 2, select N groups of data from the vehicle-induced strain and vehicle weight data of each cluster according to the set requirements to generate random sampling data of vehicle-induced strain and vehicle weight with the same length; where N is a positive integer; Step 4: Take each set of randomly sampled data of vehicle-induced strain and vehicle weight as a set of samples, generate a vehicle weight-strain mapping model based on the Bayesian linear regression method, and obtain the slope of each vehicle weight-strain mapping model; Step 5: Calculate the distance between the slope of each vehicle weight-strain mapping model in each cluster and the slope of the centroid of the vehicle weight-strain mapping model in the cluster, and take the maximum value of the distance calculated in each cluster as the evaluation threshold of each cluster; Step 6: Periodically collect vehicle weight data and vehicle-induced strain data of the bridge to be evaluated in real time. If the data are collected continuously for more than M periods, and the distance between the slope of at least one vehicle weight-strain mapping model in each cluster and the centroid of the slope of the vehicle weight-strain mapping model in the cluster is greater than the evaluation threshold of the cluster, it indicates that the bridge structure performance is abnormal, otherwise it is considered that the bridge structure performance is normal; where M is a positive integer; In step 3, the method for generating random sampling data of vehicle-induced strain and vehicle weight with consistent length is as follows: For each cluster obtained by clustering in step 2, according to the set requirements, select N groups of data from the vehicle-induced strain and vehicle weight data of each cluster as random sampling objects; For each set of vehicle weight data and vehicle-induced strain data, the distribution fitting function is used as the objective function, and the Gaussian distribution that can envelop the probability distribution of the vehicle weight data and vehicle-induced strain data is used as the recommended distribution function. The Metropolis-Hastings algorithm is used to perform random simulation sampling to generate a set of vehicle weight data and vehicle-induced strain random sampling data with the same data length that obey the distribution of the original vehicle weight data and vehicle-induced strain data. In step 4, the method for obtaining the slope of the vehicle weight-strain mapping model based on Bayesian linear regression is as follows: For each cluster, arrange each set of vehicle-induced strain and vehicle weight random sampling data obtained in step 3 in descending order and correspond one to one to form a sample set for each group; Based on the sample set of each group, a vehicle weight-strain mapping model is established by using the Bayesian linear regression method, and the slope of the vehicle weight-strain mapping model is obtained.

2. A bridge condition assessment method based on the slope of a vehicle weight-strain mapping model according to claim 1, characterized in that: In step 1, the method for obtaining the vehicle weight data and vehicle-induced strain data of the bridge to be evaluated is as follows: The bridge is equipped with a dynamic weighing system and a health monitoring system to collect real-time information on vehicles passing through the bridge structure and bridge strain response data at different time periods every day; Extracting the vehicle weight data of the vehicle crossing the bridge from the vehicle information; From the bridge strain response data, the temperature-induced strain of the bridge is separated, and the maximum value data of the vehicle-induced strain caused by the vehicle weight is extracted.

3. The bridge condition assessment method based on the slope of the vehicle weight-strain mapping model according to claim 2 is characterized in that: The method for clustering vehicle-induced strain data and vehicle weight data in step 2 is as follows: Perform distribution fitting on the vehicle weight data to obtain the distribution statistical characteristics of the vehicle weight data, and cluster the vehicle weight data according to the distribution statistical characteristics of the vehicle weight data; Performing distribution fitting on the vehicle-induced strain data to obtain the distribution statistical characteristics of the vehicle-induced strain data, and clustering the vehicle-induced strain data according to the distribution statistical characteristics of the vehicle-induced strain data; The clustering of the above vehicle-induced strain data and vehicle weight data is consistent with the cluster categories, both of which are K clusters.

4. The bridge condition assessment method based on the slope of the vehicle weight-strain mapping model according to claim 1 is characterized in that: In step 5, the Euclidean distance between the slope of each vehicle weight-strain mapping model in each cluster and the slope of the centroid of the vehicle weight-strain mapping model in the cluster is calculated.

Images