Multi-model parallel prediction method for bridge structure deformation
Through the multi-model parallel prediction method, combined with environmental monitoring data and multi-factor bridge displacement prediction model, the complexity problem of structural deformation prediction of low tower cable-stayed bridges is solved, and bridge safety prediction under the combined effect of multiple factors is achieved, which improves the Prediction accuracy and efficiency.
Patent Information
- Application Number
- CN202411490454.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-24
- Publication Date
- 2024-11-26
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art is difficult to effectively predict the structural deformation of the cable-stayed bridge of the short tower under the combined action of multiple factors, especially under the influence of external complex factors, traditional methods have serious subjective influence and prediction difficulties.
Multi-model parallel prediction method is adopted to obtain the environmental monitoring data of the bridge, use the method of reentry sampling to obtain training samples, establish a multi-factor bridge displacement prediction model, and conduct parallel prediction based on these models, considering wind speed, historical displacement, and snowfall intensity. , rainfall intensity and vehicle dynamic load, establish a mapping relationship between the prediction path and the prediction results, and achieve accurate prediction of bridge displacement.
It improves the accuracy and accuracy of bridge safety prediction, can effectively handle multi-dimensional nonlinear features, reduces the risk of overfitting of a single model, improves the robustness and efficiency of prediction, and reduces the calculation cost.
Smart Images

Figure CN119026482A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of bridge risk control, and in particular to a multi-model parallel prediction method for bridge structure deformation. Background Art
[0002] The multi-model parallel prediction method for the deformation of the in-service low-tower cable-stayed bridge structure is the result of the joint action of multiple factors. The current evaluation method is mostly based on the on-site conditions during construction to make artificial safety predictions, which has serious subjective influences. In fact, there is a certain mapping relationship between the influencing factors such as wind speed, historical bridge displacement, snowfall degree, rainfall degree, and vehicle dynamic load and bridge displacement. Under the influence of complex external factors, this mapping relationship manifests as a complex multi-dimensional nonlinear characteristic, which is difficult to express using general explicit equations. The traditional prediction of the deformation of bridge engineering structures is mostly based on one-dimensional nonlinear characteristics. It is difficult to predict the safety of in-service low-tower cable-stayed bridges under the combined action of multiple factors. Summary of the invention
[0003] The invention discloses a multi-model parallel prediction method for bridge structure deformation to overcome the above technical problems.
[0004] In order to achieve the above object, the technical solution of the present invention is: A multi-model parallel prediction method for bridge structure deformation includes the following steps: S1. Obtain the first ~ C The environmental monitoring data set of the day and the C +1 day of bridge displacements to obtain a sample database; C represents the total number of days in history for which environmental monitoring data were obtained, and C >1; S2. According to the sample database, a sampling method with replacement is adopted to obtain n Group training samples; among them, n is the number of training sample groups; S3: Establish a multi-factor bridge displacement prediction model and n The training samples are grouped and the multi-factor bridge displacement prediction model is trained respectively to obtain n A trained multi-factor bridge displacement prediction model; S4: According to n The trained multi-factor bridge displacement prediction model is based on the first ~ C Acquire environmental monitoring data for the day n The bridge displacement on the date to be predicted is calculated to obtain the final predicted bridge displacement on the date to be predicted, thus completing the prediction of the bridge displacement on the date to be predicted.
[0005] Furthermore, the method for establishing the multi-factor bridge displacement prediction model is as follows: S31: Based on the environmental monitoring data i , i =1, … , I The value range of the environmental factors determines the i Multiple value intervals of environmental factors; among them, i is the index number of the environmental factor, I is the total number of environmental factors; S32: Determine the sequence of environmental factors in the environmental monitoring data; S33: According to i The value range of each environmental factor is determined, and a prediction path based on the environmental factor sequence is obtained to establish a mapping relationship between the prediction path and the prediction result, thereby completing the establishment of a multi-factor bridge displacement prediction model.
[0006] Furthermore, the method for establishing the mapping relationship between the prediction path and the prediction result is as follows: S331: Determine the first i , i =1, … , I The first of the environmental factors c The monitoring value of the day belongs to i The value range of each environmental factor; S332: Get the c,c= 1, … , C The prediction path based on the environmental factor sequence for the day is Lq 1 -L q 2 -…- Lq i -…-Lq I ; in, q i For the i The first of the environmental factors c The monitoring value of the day belongs to i The index number of the value range of each environmental factor; Lq i For the i The first of the environmental factors c The monitoring value of the day belongs to i The value range of each environmental factor; c The number of days in history when environmental monitoring data were obtained; S333: Establishment c,c= 1,… , C The predicted path based on the environmental factor sequence of the first C +1 day bridge displacement, and obtain the c The mapping relationship between the prediction path and the prediction results of the previous day is obtained to obtain the mapping relationship between the final prediction path and the prediction results.
[0007] Furthermore, the environmental monitoring data in the environmental monitoring data set includes multiple environmental factors related to bridge structure deformation.
[0008] Further, the environmental factors include wind speed, historical bridge displacement, snowfall intensity, rainfall intensity and vehicle dynamic load; Among them, the historical bridge displacement is used to C +1 day bridge displacement prediction 1~ C The bridge displacement of the day.
[0009] Furthermore, the calculation formula for the total number of training samples is as follows: , Where: Indicates the number of training sample groups; Represents the total number of data samples in the sample database; Indicates the number of samplings with replacement.
[0010] Furthermore, the formula used to obtain the final predicted displacement of the bridge on the date to be predicted is as follows: , Where: predict displacements for the final bridge; It is Prediction results of a trained multi-factor bridge displacement prediction model.
[0011] Beneficial effects: The multi-model parallel prediction method for bridge structure deformation of the present invention is based on the first to C The environmental monitoring data set of the day and the C +1 day bridge displacement formed samples, using the method of sampling with replacement to obtain multiple groups of training samples, respectively trained the established multi-factor bridge displacement prediction model, and finally used the bridge displacement obtained by multiple trained multi-factor bridge displacement prediction models to obtain the final bridge prediction displacement, and completed the prediction of the bridge displacement on the predicted date. Among them, the established multi-factor bridge displacement prediction model can fully consider the environmental factors related to the deformation of the bridge structure, predict the multi-dimensional nonlinear characteristics, and has a high prediction accuracy and precision for the prediction of the safety of the in-service bridge under the combined action of multiple factors. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.
[0013] Figure 1 is a flow chart of the prediction method of the present invention; Figure 2 A flow chart of a multi-model parallel prediction method for bridge structure deformation in an embodiment of the present invention; Figure 3 A multi-factor bridge displacement prediction flow chart in an embodiment of the present invention; Figure 4 A comparison diagram of actual bridge displacement and model-predicted bridge displacement in an embodiment of the present invention; Figure 5 4 is a graph showing the error value of the prediction model in an embodiment of the present invention. DETAILED DESCRIPTION
[0014] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution 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 creative work are within the scope of protection of the present invention.
[0015] This embodiment introduces a multi-model parallel prediction method for bridge structure deformation. Figure 1-Figure 3 As shown, the following steps are included: S1. Obtain the first ~ C The environmental monitoring data set of the day and the C +1 day of bridge displacements to obtain a sample database; C is a natural number, indicating the total number of days in which environmental monitoring data were obtained in history, and C >1; Specifically, this embodiment is for the first to the second cable-stayed bridges in service. C The environmental monitoring data collected on the day and the corresponding C +1 day bridge displacement data is statistically analyzed to form a sample database, and each data sample consists of the 1st~ C The environmental monitoring data collected on the day and the corresponding C+1 day of bridge displacement data; the bridge has passed the past 1~ C The environmental monitoring data including bridge displacement monitored on the first day is used to predict the bridge displacement on the next day, i.e. C +1 day of bridge displacement.
[0016] Preferably, the environmental monitoring data in the environmental monitoring data set includes multiple environmental factors related to the deformation of the bridge structure; the environmental factors include wind speed, historical bridge displacement, snowfall intensity, rainfall intensity and vehicle dynamic load, etc.; wherein the historical bridge displacement is used for the bridge to C +1 day bridge displacement prediction 1~ C displacement of the sky; Specifically, the build capacity is E Sample database of D =( x , y ),in: x =( x 1, x 2,…, x C ), x C Indicates C Environmental monitoring data for the day, x represents an environmental monitoring dataset; y = ( y 1, y 2,…, y C ), y C Indicates C +1 day of bridge displacement, y represents the bridge displacement set; in this embodiment, x 1, x 2,…, x C It is the first consecutive time in history. C The environmental monitoring data of the day, similarly, y 1, y 2,…, y C For the corresponding 2~ C +1 day of bridge displacement, which is also continuous C In this embodiment, the environmental monitoring data in the environmental monitoring data set are derived from the daily monitoring of the safety detection system of the low-tower cable-stayed bridge in service. The environmental factors in this embodiment include wind speed, historical bridge displacement, snowfall intensity, rainfall intensity, and vehicle dynamic load.x 1=(F 1风 , F 1桥 , F 1雪 , F 1雨 , F 1车 ), x 2=(F 2风 , F 2桥 , F 2雪 , F 2雨 , F 2车 ),…, x c =(F c风 , F c桥 , F c雪 ,F c雨 , F c车 ),…, x C =(F C风 , F C桥 , F C雪 , F C雨 , F C车 ), where F c风 Indicates the history of c Wind speed of the day α c Impact degree on bridge, F c桥 Indicates that based on the history of c The influence of the bridge displacement monitored on the day on the bridge; c雪 Indicates the history of c Snowfall intensity per day Impact degree on bridges; F c雨 Indicates the history of c Rainfall intensity per day δ c Impact degree on bridges; F c车 Indicates the history of c Vehicle dynamic load per day φ c The degree of impact on the bridge; In this embodiment, snowfall intensity monitoring is carried out through a weighing precipitation sensor, which is placed outdoors to measure precipitation by quickly responding to mass changes, and the weight change caused by precipitation is converted into an electrical signal, which is then converted and processed to obtain accurate snowfall data. Rainfall intensity detection is to collect rainwater and measure its volume through a rain gauge to obtain rainfall data. Usually 24 hours is used as a boundary to determine the intensity of snowfall and rainfall intensity; wind speed is measured by an anemometer and a wind vane; vehicle dynamic load is measured by on-site dynamic load sensing to obtain data, and the vehicle dynamic load sensor obtains real-time data by measuring the weight and axle weight information of the vehicle during operation in real time. y 1,y 2,…, y C The value of comes from the field measurement on site, where y C =F (C+1)桥 The significance of this embodiment is that according to the history of continuous 1~ C The external environmental factors of the day (including historical bridge displacement) are used to predict the C +1-day bridge displacement, thereby knowing the deformation degree of the low-tower cable-stayed bridge to determine the safety status of the bridge body and achieve rapid, accurate and intelligent prediction of bridge safety.
[0017] S2. According to the sample database, a sampling method with replacement is adopted to obtain n Group training samples; among them, n is the total number of training sample groups; The training samples in this embodiment are obtained by performing a sampling method with replacement on the data samples in the sample database. Specifically, the data samples are sampled. a Sampling with replacement, and forming n Group training samples and m Group verification samples.
[0018] Preferably, the total number of training samples is calculated as follows: , Where: Indicates the number of training sample groups; Represents the total number of data samples in the sample database; Indicates the number of sampling with replacement; Specifically, this embodiment performs a sample database D The data samples in a Extract with replacement, get n The number of samples in each training set is smaller than the sample database. D The number of data samples in the sample database; at the same time, the unextracted data is recorded as the verification sample, and the number of samples in each group of verification samples is also smaller than the sample database D The number of data samples in the formula for calculating the number of verification samples is: , Where: The number of samples to be verified.
[0019] This embodiment obtains training samples by implementing a sampling method with replacement, which can effectively expand the sample size in a small-scale data set, thereby making the model training process more robust. This sampling method not only ensures the diversity of samples, but also because it puts the samples back to the original set after each extraction, the number of times each sample is selected may be different, thereby increasing the possibility of sample selection. Such diversity helps to enhance the stability and generalization ability of the model.
[0020] S3: Establish a multi-factor bridge displacement prediction model and n The training samples are grouped and the multi-factor bridge displacement prediction model is trained respectively to obtain n A trained multi-factor bridge displacement prediction model; Preferably, the multi-factor bridge displacement prediction model is established as follows: S31: Based on the environmental monitoring data i , i =1, … , I The value range of the environmental factors determines the i Multiple value intervals of environmental factors; among them, i is the index number of the environmental factor, I is the total number of environmental factors; Specifically, this embodiment will i The value range of each environmental factor is evenly divided, and we get Q i No. i The value range of each environmental factor; Q i For the i The total number of value intervals of environmental factors; for example, the wind speed in the environmental monitoring data set, within the 30-day monitoring data, the actual wind speed measured is 0-20km / h, therefore, the range of 0-20km / h is evenly divided, and the range of each value interval of the wind speed is determined to be 0.1km / h, then i The wind speed range is [( i -1)* 0.1, i *0.1].
[0021] S32: Determine the sequence of environmental factors in the environmental monitoring data; S33: According to i The value interval of each environmental factor is used to obtain a prediction path based on the sequence of environmental factors, so as to establish a mapping relationship between the prediction path and the prediction result; Preferably, the method for establishing the mapping relationship between the prediction path and the prediction result is as follows: S331: Determine the firsti , i =1, … , I The first of the environmental factors c The monitoring value of the day belongs to i The value range of each environmental factor; S332: Get the c,c= 1, … , C The prediction path based on the environmental factor sequence for the day is Lq 1 -L q 2 -…- Lq i -…-Lq I ; in, q i For the i The first of the environmental factors c The monitoring value of the day belongs to i The index number of the value range of each environmental factor; Lq i For the i The first of the environmental factors c The monitoring value of the day belongs to i The value range of each environmental factor; c The number of days in history when environmental monitoring data were obtained; S333: Establishment c,c= 1, … , C The predicted path based on the environmental factor sequence of the first C +1 day bridge displacement, and obtain the c The mapping relationship between the prediction path and the prediction results of the previous day is obtained to obtain the mapping relationship between the final prediction path and the prediction results.
[0022] Specifically, in this embodiment, each sample data in the training sample is processed according to the above steps. c, c= 1, … , C The predicted path based on the environmental factor sequence of the first C The mapping relationship between the bridge displacements on day +1 is obtained U * C The mapping relationship between prediction paths and prediction results is obtained by fitting the mapping relationships between these prediction paths and prediction results, and the final mapping relationship between the prediction paths and prediction results can be obtained; wherein, Uis the number of data samples in each set of training samples, that is, the number of environmental monitoring data in each set of training samples; Specifically, this embodiment is n The training samples are grouped, and the multi-factor bridge displacement prediction model is trained by executing S31 to S33, that is, n A trained multi-factor bridge displacement prediction model. This embodiment constructs multiple multi-factor bridge displacement prediction models for training, and performs parallel prediction based on multiple different trained models, which can significantly improve the prediction accuracy of the parallel model. This multi-model parallel method can not only capture the complex patterns and nonlinear relationships in the data by integrating the prediction results of multiple models, thereby enhancing the generalization ability of the model, but also effectively reduce the impact of abnormal data on the prediction results by taking the average when there are problems with a small number of models, thereby reducing the prediction error. In addition, since each multi-factor bridge displacement prediction model is trained independently, this method can reduce the random fluctuations that a single model may encounter and improve the stability of the overall model. At the same time, the parallel processing method also speeds up the training speed of the model, making it more efficient when processing large-scale data sets. Overall, this multi-model parallel prediction method not only improves the accuracy and robustness of the prediction by combining the power of multiple independent models, but also optimizes the efficiency of model training.
[0023] Specifically, the sequence of environmental factors in this embodiment is determined by technical personnel in the field based on discussions among 10 experts with relevant working experience. After discussion, it is found that the vehicle dynamic load has the greatest impact, followed by historical bridge displacement, wind speed, rainfall degree and snowfall degree. Therefore, the order of environmental factor prediction is the same as the order of influence above; it can also be determined in ascending order according to the customized pruning coefficient; The calculation formula of the trimming coefficient is: , Where: represents the pruning coefficient of environmental factors; Indicates the total number of environmental factor intervals; q i The index number representing the environmental factor interval; Pq i Indicates that the data sample belongs to q i The probability of an environmental factor interval; Specifically, a specific implementation of the present invention is as follows: Taking the real data collected within 35 days as an example, all inputs and outputs are evenly divided into 200 intervals according to the value range of the data sample set of each environmental factor, and each interval represents an equal amount of data value range. Among them, assuming that the influencing factor 1 of the first environmental monitoring data (sample 1) is divided into the 4th interval of influencing factor 1 according to the interval, recorded as g 1 4 , and then continue to divide the second influencing factor; the influencing factor 2 of sample 1 is divided into the third interval of the influencing factor according to the interval, recorded as g 2 3 . The division continues until all influencing factors are divided. The output corresponding to sample 1 at this time is y The 9th interval is recorded as y 1 9 , then the mapping relationship between the input value and the output value of the first sample of the multi-factor bridge displacement prediction model is obtained, that is, , and then use this logic to divide other data samples in turn and find the corresponding mapping relationship and fit it through the existing method, so that the first model can be obtained and The corresponding value of , at this time, a multi-factor bridge displacement prediction model is formed. Among them, and All are fitting coefficients.
[0024] S4: According to n The trained multi-factor bridge displacement prediction model is based on the first ~ C Acquire environmental monitoring data for the day n The bridge displacement on the date to be predicted is calculated to obtain the final predicted bridge displacement on the date to be predicted, thus completing the prediction of the bridge displacement on the date to be predicted.
[0025] Specifically, a corresponding number of multi-factor bridge displacement prediction models are constructed according to the number of training sample groups. A multi-model parallel prediction method is formed by combining multiple multi-factor bridge displacement models in parallel. The prediction result of this prediction method is the sum of the prediction results of the multi-factor bridge displacement prediction model and the average value. The prediction result is recorded as ; The formula used to obtain the final predicted displacement of the bridge on the date to be predicted is as follows: , Where: Predict the displacement of the final bridge, i.e. the parallel method prediction result; It is n Prediction results of a trained multi-factor bridge displacement prediction model.
[0026] Specifically, this embodiment adopts a sampling method with replacement to obtain n After the training samples are formed, the remaining data samples in the sample database are used as verification samples; when the error between the prediction result obtained by the verification sample and the actual measurement result is less than the set threshold, the training of the multi-factor bridge displacement prediction model is completed, otherwise the multi-factor bridge displacement prediction model is retrained.
[0027] Specifically, this embodiment verifies the prediction method, and outputs the result if the verification is successful. If the verification fails, restart from the sampling step: bring the verification sample into the multi-model parallel prediction method for verification, and calculate the prediction result and the true value of the verification sample by the formula. If it is less than or equal to 0.1, it means that the prediction accuracy of the model is good, and the obtained multi-model parallel prediction method can be used; if the value is greater than 0.1, it means that the prediction effect of the model is poor, and it is necessary to restart from the sample extraction step. The prediction accuracy calculation formula is as follows: , in: Model feasibility value; To verify the number of samples; Predict the displacement of the final bridge, i.e. the sample prediction result; The actual results of the samples.
[0028] Thus, this embodiment realizes direct, rapid and accurate calculation and prediction of future bridge displacements through multiple environmental factors related to bridge structure deformation from external environmental monitoring data measured by the bridge safety monitoring system, thereby greatly improving the safety of the bridge.
[0029] Specifically, the embodiment of the present invention predicts the bridge displacement of the next day through 35 days of historical bridge displacement and external environmental factors, so as to determine whether the bridge is safe and can continue to operate. Compared with the existing prediction through a single model, this method greatly reduces the risk of overfitting of the model through the prediction results of multiple multi-factor bridge displacement models. The multi-factor bridge displacement model can handle high-dimensional data and a large number of samples, and the multi-factor bridge prediction model is established in parallel and is independent of each other, which not only improves the speed of model prediction, but also has better robustness when processing complex data sets and large-scale data sets, thereby obtaining prediction results with higher prediction accuracy.
[0030] When sampling data, a certain amount of sample data will not be extracted and can be directly used as test samples to test the model. There is no need to obtain additional samples to test the model prediction, which significantly reduces the computing cost.
[0031] The environmental monitoring data set of this embodiment is derived from the data monitored by the monitoring system of the Xiong'an Bridge, a low-tower cable-stayed bridge across the Daguan Expressway, within 35 days, as shown in Table 1:
[0032] Among them: the above monitoring data are data obtained for 35 consecutive days. Wind speed, snowfall intensity, rainfall intensity, vehicle dynamic load, and historical bridge displacement are divided into 135 intervals, 124 intervals, 196 intervals, 401 intervals, and 234 intervals according to the value range of the collected samples, and the interval length is 0.1. Snowfall intensity monitoring is through a weighing precipitation sensor, which is placed outdoors to measure the snowfall by quickly responding to mass changes, and the weight change caused by snowfall is converted into an electrical signal, and then converted and processed to obtain accurate snowfall data; rainfall intensity detection is to collect rainwater through a rain gauge and measure its volume to obtain rainfall data; usually 24 hours is used as the boundary to determine the size of snowfall intensity and rainfall intensity; wind speed is measured by anemometer and wind vane; vehicle dynamic load is measured by on-site dynamic load sensor measurement to obtain data, and the vehicle dynamic load sensor can obtain real-time data by measuring the weight and axle weight information of the vehicle in real time. The safe range of bridge displacement is usually limited to between 1 / 5000 and 1 / 2000 of the span. The length of this bridge is 440 meters, so the value is within 8.8cm-22cm.
[0033] This embodiment determines the sequence of environmental factors based on the trimming coefficient. After calculation, the minimum trimming coefficient of the vehicle dynamic load is 0.011, and the rest are wind speed, rainfall intensity, historical bridge displacement, and snowfall intensity. Therefore, the vehicle dynamic load is the preferred external factor, and the others are arranged in order.
[0034] The total sample set was sampled less than 35 times with replacement to form a group of training samples. The subsequent 99 groups of samples were formed in the same way as the first group. The remaining samples after each group of samples were drawn formed verification samples for testing the trained model. The number of multi-factor bridge prediction models was set to 100, and the number of nodes for each model division was set to 5.
[0035] By bringing 100 groups of training samples into the multi-factor bridge displacement prediction model for training, 100 trained multi-factor bridge prediction models are obtained. The prediction results of each model are obtained by predicting the 100 trained multi-factor bridge prediction models: , The prediction results calculated by each multi-factor bridge displacement prediction model are summed up and then the average is taken to obtain the parallel model prediction result.
[0036] Then, the model feasibility value Determine the prediction accuracy and finally determine whether the model is feasible. , which means that the model prediction accuracy is high and the prediction results are feasible.
[0037] In order to further prove the accuracy of the model, 10 verification samples were input into the parallel model for verification. The distribution of their influencing factors is shown in Table 2:
[0038] Table 3 summarizes the prediction errors of the 10 test sample groups shown in Table 2. It can be seen that in the calculation results of this embodiment, the maximum absolute error is 0.7, the maximum relative error is 9.7%, and the accuracy is high, which can meet the needs of rapid prediction of bridge safety;
[0039] Figure 4 The comparison between the actual bridge displacement of this embodiment and the bridge displacement predicted by the model is shown; the figure is a graph of the measured bridge displacement and the predicted bridge displacement. It can be seen from the figure that the bridge displacement prediction value of the model is relatively accurate, there is no mutation point, and the overall prediction trend is the same as the measured value, and the model prediction effect is good.
[0040] Figure 5 The absolute error and relative error value of the prediction model of this embodiment are shown; the picture is an analysis chart of the absolute error and relative error value of the prediction model. From this figure, it can be seen that the relative error of the prediction result is kept within 10%, and the highest absolute error is 0.7, indicating that the model has good prediction effect and low error, and can perform prediction work for such projects.
[0041] The bridge displacement safety prediction model of the in-service low-tower cable-stayed bridge of this embodiment can establish a mapping relationship between wind speed, historical bridge displacement, snowfall intensity, rainfall intensity, vehicle dynamic load and other important safety influencing factors and bridge predicted displacement based on the monitoring data set of the previous bridge safety damage value, and predict the bridge displacement of the next day through the historical bridge displacement and the external influencing factors of the day, so as to determine whether the bridge can be safely passed. The risk of overfitting of a single prediction model is greatly reduced. The multi-factor bridge displacement prediction model of this embodiment can process high-dimensional data and a large number of samples, and each multi-factor bridge prediction model is established in parallel and is independent of each other, which not only improves the speed of model prediction, but also has good robustness when processing complex data sets and large-scale data sets, thereby obtaining a prediction result with higher prediction accuracy. And when the multi-factor bridge prediction model is formed by the pruning algorithm, there are a small number of samples that cannot be extracted, which are directly used as test samples to test the model, and the performance of the model can be evaluated without additional calculation to obtain training samples, which significantly reduces the calculation cost, thereby providing a fast, accurate and low-cost calculation model for the dynamic safety prediction of the low-tower cable-stayed bridge in service.
[0042] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A multi-model parallel prediction method for bridge structure deformation, characterized in that: The steps include: S1. Obtain the first ~ C The environmental monitoring data set of the day and the C +1 day of bridge displacements to obtain a sample database; C represents the total number of days in history for which environmental monitoring data were obtained, and C >1; S2. According to the sample database, a sampling method with replacement is adopted to obtain n Group training samples; among them, n is the number of training sample groups; S3: Establish a multi-factor bridge displacement prediction model and n The training samples are grouped and the multi-factor bridge displacement prediction model is trained respectively to obtain n A trained multi-factor bridge displacement prediction model; S4: According to n The trained multi-factor bridge displacement prediction model is based on the first ~ C Acquire environmental monitoring data for the day n The bridge displacement on the date to be predicted is calculated to obtain the final predicted bridge displacement on the date to be predicted, thus completing the prediction of the bridge displacement on the date to be predicted.
2. The multi-model parallel prediction method for bridge structure deformation according to claim 1 is characterized in that: The method for establishing the multi-factor bridge displacement prediction model is as follows: S31: Based on the environmental monitoring data i , i =1, … , I The value range of the environmental factors determines the i Multiple value intervals of environmental factors; among them, i is the index number of the environmental factor, I is the total number of environmental factors; S32: Determine the sequence of environmental factors in the environmental monitoring data; S33: According to i , i =1, … , I The value range of each environmental factor is determined, and a prediction path based on the environmental factor sequence is obtained to establish a mapping relationship between the prediction path and the prediction result, thereby completing the establishment of a multi-factor bridge displacement prediction model.
3. The multi-model parallel prediction method for bridge structure deformation according to claim 2 is characterized in that: The method for establishing the mapping relationship between the prediction path and the prediction result is as follows: S331: Determine the first i , i =1, … , I The first of the environmental factors c The monitoring value of the day belongs to i The value range of each environmental factor; S332: Get the c,c= 1, … , C The prediction path based on the environmental factor sequence for the day is Lq 1 -L q 2 -…-Lq i -…- Lq I ; in, q i For the i The first of the environmental factors c The monitoring value of the day belongs to i The index number of the value range of each environmental factor; Lq i For the i The first of the environmental factors c The monitoring value of the day belongs to i The value range of each environmental factor; c The number of days in history when environmental monitoring data were obtained; S333: Establishment c,c= 1, … , C The predicted path based on the environmental factor sequence of the first C +1 day bridge displacement, and obtain the c The mapping relationship between the prediction path and the prediction results of the previous day is obtained to obtain the mapping relationship between the final prediction path and the prediction results.
4. The multi-model parallel prediction method for bridge structure deformation according to claim 1 is characterized in that: The environmental monitoring data in the environmental monitoring data set includes a plurality of environmental factors related to the deformation of the bridge structure.
5. The multi-model parallel prediction method for bridge structure deformation according to claim 4 is characterized in that: The environmental factors include wind speed, historical bridge displacement, snowfall intensity, rainfall intensity, and vehicle dynamic loads; Among them, the historical bridge displacement is used to C +1 day bridge displacement prediction 1~ C The bridge displacement of the day.
6. The multi-model parallel prediction method for bridge structure deformation according to claim 1 is characterized in that: The calculation formula for the total number of training samples is as follows: , Where: Indicates the number of training sample groups; Represents the total number of data samples in the sample database; Indicates the number of samplings with replacement.
7. The multi-model parallel prediction method for bridge structure deformation according to claim 1 is characterized in that: The formula used to obtain the final predicted displacement of the bridge on the predicted date is as follows: , Where: predict displacements for the final bridge; It is Prediction results of a trained multi-factor bridge displacement prediction model.
Citation Information
Patent Citations
Traffic accident severity prediction method and system based on convolutional neural network
CN115689040A
Port machine remote control method and system based on binocular camera
CN117935453A
Ultrahigh bridge tower intelligent measurement and control method and system based on camera array
CN118534809A
