Intelligent adjusting method and system for jack during pushing of bent slope inclined steel box girder bridge

Abstract

The invention relates to the field of bridge pushing, in particular to an intelligent adjusting method and system for a jack during pushing of a bent slope inclined steel box girder bridge. The method comprises the following steps: acquiring historical sample data of pushing of the curved slope inclined steel box girder bridge; constructing a jack control model, and training and evaluating the jack control model by using the historical sample data; acquiring real-time data of the bent slope inclined steel box girder bridge in the incremental launching construction period, and dynamically adjusting the weight and offset of the jack control model by using the real-time data; and introducing a noise alarm system, and intelligently adjusting the jack by using the noise alarm system and the dynamically adjusted jack control model. The problem that in the prior art, when a bent slope inclined steel box girder bridge is pushed, jack control is complex and not intelligent enough is solved. The self-adaptive adjustment of the weight and offset of the jack control model during the pushing construction of the curved slope inclined steel box girder bridge is realized, and the intelligent control and practicability of the method are improved.

Description

Technical Field

[0001] The invention relates to the field of bridge jacking, and in particular to an intelligent adjustment method and system for a jack during jacking of a curved slope inclined steel box girder bridge. Background Art

[0002] In bridge construction projects, curved and sloped steel box girder bridges are increasingly used in complex terrain and traffic environments due to their unique advantages. However, the jacking construction of such bridges faces many challenges, especially in terms of jack control.

[0003] When the curved and sloped steel box girder bridge is pushed, due to the complexity of its spatial form, the collaborative operation of the jacks is extremely demanding. The existing technology makes it difficult to achieve accurate coordination of multiple jacks in terms of force size, direction, and stroke extension, which makes it easy for the box girder to deflect and twist during the pushing process. At present, the control of the jacks mostly relies on traditional methods, using finite element analysis software combined with manual intervention to control the jacks, but the curved and sloped steel box girder bridge has a special structure. Its curved, sloped, and inclined characteristics make the force conditions at different positions extremely complex and changeable. Finite element analysis is also difficult to accurately determine the optimal force of each jack, which increases the difficulty and risk of construction. In addition, environmental factors at the construction site and subtle differences in material properties may cause the actual working conditions to deviate from the preset conditions, and the existing automation system is difficult to adjust flexibly, which further affects construction safety and progress.

[0004] With the development of the bridge construction industry, the requirements for construction safety, accuracy and efficiency are getting higher and higher. Traditional jack control technology can no longer meet the complex requirements of the top-pushing construction of curved and inclined steel box girder bridges. Therefore, a more intelligent jack control technology is urgently needed. Summary of the invention

[0005] In view of the defects in the prior art, the present invention provides an intelligent adjustment method and system for the jacks when pushing a curved inclined steel box girder bridge, which solves the problem in the prior art that the jack control is complex and not intelligent enough when pushing a curved inclined steel box girder bridge.

[0006] In order to achieve the above-mentioned purpose, one aspect of the present invention provides a method for intelligently adjusting the jacks during the jacking of a curved and inclined steel box girder bridge, the method comprising: obtaining historical sample data of the jacking of the curved and inclined steel box girder bridge; constructing a jack control model, and using the historical sample data to train and evaluate the jack control model; obtaining real-time data of the curved and inclined steel box girder bridge during the jacking construction, and using the real-time data to dynamically adjust the weights and biases of the jack control model; introducing a noise alarm system, and using the noise alarm system and the dynamically adjusted jack control model to intelligently adjust the jacks.

[0007] The present invention provides a solid data foundation for the training of the jack control model through the historical sample data of the jacking of the curved slope inclined steel box girder bridge, so that the jack control model has reliable jack control capabilities. The introduction of real-time data can dynamically adjust the model weights and biases, so that the model can adapt to various changes in the construction process and improve the accuracy and adaptability of the control. The use of the noise alarm system and the dynamically adjusted model to intelligently adjust the jack can effectively cope with the complex working conditions in the jacking construction of the curved slope inclined steel box girder bridge, and provide guarantees for safe construction, thereby improving the intelligence and practicality of the jack control model.

[0008] Optionally, the method of obtaining historical sample data of the jacking of the curved slope inclined steel box girder bridge includes: obtaining first historical data related to the jacking of the curved slope inclined steel box girder bridge; removing outliers and interpolating missing values ​​from the first historical data to obtain second historical data; dividing the second historical data into a plurality of historical construction stage data according to the characteristics of different construction stages of the jacking of the curved slope inclined steel box girder bridge; adding noise to the historical construction stage data according to the characteristics of the different construction stages to obtain noise fusion data; and preprocessing the noise fusion data to obtain the historical sample data.

[0009] The present invention ensures the quality and integrity of the data by eliminating outliers and interpolating missing values ​​in the historical data of the jacking of curved and sloped inclined steel box girders. The data is divided according to the construction stage and corresponding noise is added to simulate the uncertainty and variables in the real construction environment, thereby improving the diversity of the data. This not only makes the data closer to the actual operating environment, but also enhances the sensitivity and responsiveness of the model to changes in key parameters.

[0010] Optionally, the preprocessing of the noise fusion data to obtain the historical sample data includes: extracting features from the noise fusion data according to the characteristics of the different construction stages to obtain feature data of different construction stages; and normalizing the feature data of different construction stages to obtain the historical sample data.

[0011] The present invention performs feature extraction and normalization processing on the noise fusion data. The feature extraction targets the data characteristics of different construction stages and extracts key information. The normalization processing ensures that the data of different construction stages have a unified scale, thereby greatly improving the quality of the data as a whole and providing a solid foundation for the training of the jack control model.

[0012] Optionally, the construction of the jack control model includes: constructing a first loss function according to the force size, force direction and stroke extension of the jack; setting a constraint penalty item, and constructing a second loss function according to the first loss function and the constraint penalty item; based on a long short-term memory network, using the second loss function to construct the jack control model.

[0013] The present invention uses the actual function of the jack and the force, direction and stroke extension of the jack to form a multi-dimensional first loss function, which can enable the jack control to accurately capture the key performance indicators of the jack during the jacking process during training. The introduction of constraint penalty terms not only optimizes the prediction accuracy of the model, but also ensures structural safety and construction quality. The long short-term memory network (LSTM) is used to process time series data, so that the model can learn the time dependency in historical data, thereby making more accurate predictions and adjustments to the behavior of the jack.

[0014] Optionally, the first loss function satisfies the following formula: , in, is the loss function, is the number of predicted values ​​of the jack, For the The force of the jack, For the The predicted force of the jacks, No. The force direction of the jack, for The predicted force direction of the jacks, For the The stroke of the jack is extended. For the The predicted stroke extension of the jack.

[0015] The formula of the present invention fully considers the three key parameters of the jack: the force size, force direction and stroke extension, and provides a comprehensive performance evaluation standard for the jack control model. In addition, the penalty for large deviations is strengthened in the form of the sum of squares, thereby guiding the precise control of these key parameters in the model learning process and structurally encouraging the model to achieve smoother and more controllable operations during the jacking process, which helps to optimize construction efficiency, reduce safety risks, and ultimately improve the quality of the entire bridge jacking project.

[0016] Optionally, setting the constraint penalty item and constructing the second loss function according to the first loss function and the constraint penalty item includes: obtaining the maximum allowable stress in the box girder design, the maximum allowable offset in the box girder design, and the minimum allowable height of the guide beam according to relevant characteristics of the box girder; setting a stress equalization penalty item, a maximum offset penalty item, and a guide beam height penalty item according to the maximum allowable stress in the box girder design, the maximum offset in the box girder design, and the minimum allowable height of the guide beam; setting penalty weights for the stress equalization penalty item, the maximum offset penalty item, and the guide beam height penalty item respectively; setting the constraint penalty item according to the stress equalization penalty item, the maximum offset penalty item, the guide beam height penalty item, and the penalty weight; and constructing the second loss function in combination with the first loss function and the constraint penalty item.

[0017] The present invention can significantly improve the adaptability and accuracy of the jack control model in the actual construction environment by setting the constraint penalty term. At the same time, the setting of the weight can improve the accuracy of the decision-making strategy of the jack control model and provide a clear direction for the training of the model.

[0018] Optionally, the dynamic adjustment of the weights and bias of the jack control model using the real-time data includes: using the real-time data to calculate a complexity index of the real-time data; dynamically adjusting the penalty weight of the second loss function based on the complexity index; using the real-time data to calculate the loss value of the adjusted second loss function; based on the loss value, using a back propagation algorithm to calculate the gradient of the loss function; and dynamically adjusting the weights and bias of the jack control model based on the gradient.

[0019] The present invention can intelligently adjust the penalty weight in the second loss function by accurately calculating the complexity index of real-time data, so as to better meet the needs of the actual construction environment. By calculating the loss value of the loss function after adjusting the penalty weight and using the back propagation algorithm to obtain gradient information, the weight and bias of the jack control model are dynamically adjusted, which not only improves the response speed of the jack control model to changes in real-time data, but also enhances the accuracy of model prediction, improves the precision of jack control, and at the same time improves the adaptability of the jack control model to complex environments.

[0020] Optionally, the complexity index includes a stress complexity index, an offset complexity index and a guide beam height complexity index, and the stress complexity index satisfies the following formula: , in, is the stress complexity index, is the number of stress monitoring points, For the The value of each stress monitoring point, is the average value of the stress monitoring points, is the maximum value of the stress monitoring point, is the minimum value of the stress monitoring point; The offset complexity index satisfies the following formula: , in, is the offset complexity index, is the number of offset monitoring points, For the The value of the offset monitoring point, is the average value of the offset monitoring points, is the maximum value of the offset monitoring point, is the minimum value of the offset monitoring point, The value with the largest change in the offset monitoring point between adjacent real-time data collection moments; The guide beam height complexity index satisfies the following formula: ,in, is an indicator of the height complexity of the guide beam, is the number of monitoring points for the guide beam height, For the The value of the guide beam height monitoring point, is the average value of the guide beam height monitoring points, is the maximum value of the guide beam height monitoring point, is the minimum value of the guide beam height monitoring point, It is the value with the largest change in the height monitoring point of the guide beam at adjacent real-time data collection moments.

[0021] The complexity index formula of the present invention comprehensively considers the discrete degree of the monitoring point value and the average value, the fluctuation range of the monitoring point value, and the maximum change value of the monitoring point at adjacent moments. Through multi-dimensional calculation, the complexity of the offset can be comprehensively and accurately measured.

[0022] Optionally, dynamically adjusting the weights and biases of the jack control model according to the gradient includes: introducing an adaptive moment estimation, calculating a first-order moment estimation and a second-order moment estimation of the adaptive moment estimation according to the gradient; correcting the first-order moment estimation and the second-order moment estimation, and using the corrected first-order moment estimation and the second-order moment estimation to dynamically adjust the weights and biases of the jack control model.

[0023] The present invention introduces an adaptive moment estimation mechanism to adjust the weight and bias of the jack control model, thereby improving the robustness of the weight and bias of the jack control model in complex environments, reducing the need for manual intervention, achieving more precise and intelligent bridge jacking construction control, and further improving the flexibility and adaptability of the jack control model.

[0024] Another aspect of the present invention provides an intelligent adjustment system for jacks during jacking of a curved and inclined steel box girder bridge, comprising: a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, and the computer program includes program instructions, and the processor is configured to call the program instructions to execute the intelligent adjustment method for jacks during jacking of a curved and inclined steel box girder bridge as described in any one of the previous aspects of the present invention.

[0025] The intelligent adjustment system of the jacks during the jacking of a curved and inclined steel box girder bridge of the present invention has a compact structure, stable performance, high integration and simple composition, and can stably execute the intelligent adjustment method of the jacks during the jacking of a curved and inclined steel box girder bridge provided in the previous aspect of the present invention, further improving the overall applicability and practical application capability of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a flow chart of an intelligent adjustment method of a jack during top pushing of a curved and inclined steel box girder bridge according to an embodiment of the present invention; Figure 2 The present invention is a schematic structural diagram of an intelligent adjustment system for jacks during top pushing of a curved and inclined steel box girder bridge according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] The specific embodiments of the present invention will be described in detail below. It should be noted that the embodiments described herein are only for illustration and are not intended to limit the present invention. In the following description, a large number of specific details are set forth in order to provide a thorough understanding of the present invention. However, it is obvious to those of ordinary skill in the art that these specific details do not need to be adopted to implement the present invention. In other examples, in order to avoid confusing the present invention, known circuits, software or methods are not specifically described.

[0028] Throughout the specification, references to "one embodiment," "an embodiment," "an example," or "an example" mean that a particular feature, structure, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment of the present invention. Therefore, the phrases "in one embodiment," "in an embodiment," "an example," or "an example" appearing in various places throughout the specification do not necessarily all refer to the same embodiment or example. In addition, particular features, structures, or characteristics may be combined in one or more embodiments or examples in any suitable combination and / or subcombination. In addition, it should be understood by those of ordinary skill in the art that the figures provided herein are for illustrative purposes and that the figures are not necessarily drawn to scale.

[0029] Figure 1 This is a flow chart of an intelligent adjustment method for jacks during top-pushing of a curved and inclined steel box girder bridge according to an embodiment of the present invention, which solves the problem of complex and insufficient intelligence in the prior art of jack control during top-pushing of a curved and inclined steel box girder bridge. It realizes adaptive adjustment of the weight and bias of the jack control model through real-time data in the construction environment of top-pushing of box girders with curved design and overall slope of the bridge, thus improving the intelligent control and practicality of the present invention. Figure 1 The method shown comprises the following steps: Step S1, obtaining historical sample data of the jacking of the curved slope inclined steel box girder bridge.

[0030] Among them, obtaining the historical sample data of the top-pushing of the curved slope inclined steel box girder bridge specifically includes the following sub-steps: Step S101, obtaining first historical data related to the jacking of the curved slope inclined steel box girder bridge.

[0031] In an embodiment, the first historical data is historical data related to the jacking of the curved and sloped inclined steel box girder bridge. The first historical data includes but is not limited to structural design parameters, construction parameters, environmental conditions, monitoring data, and operation records. The structural design parameters include but are not limited to the length, width, height, curvature, slope, box girder strength, elastic modulus and the overall structure of the bridge. The construction parameters include but are not limited to the force direction, force size, and stroke elongation of the jack. The environmental conditions include but are not limited to the weather temperature, humidity, wind speed, and terrain conditions during the historical jacking construction of the curved and sloped inclined steel box girder bridge. The monitoring data includes but is not limited to the jacking path, speed control, and pause point.

[0032] Step S102: remove abnormal values ​​and interpolate missing values ​​from the first historical data to obtain second historical data.

[0033] In this embodiment, the second historical data is data obtained by eliminating abnormal values ​​and interpolating missing values ​​from the first historical data.

[0034] By removing outliers, outliers caused by equipment failure, measurement errors or extreme events can be eliminated as much as possible, thereby improving the reliability of data and the accuracy of analysis. The method for removing outliers can be the K-nearest neighbor distance method.

[0035] The principle of the K-nearest neighbor distance method is to calculate the Euclidean distance between each data point in the pumping data and other data points. The distribution of the K nearest neighbor data points around the data point is used to determine whether it is an outlier. If the distance between a data point and its K nearest neighbor data points is significantly greater than the distance between other points, then it may be an outlier.

[0036] First, calculate the Euclidean distance between two arbitrary monitoring points, satisfying the following formula: , in, For two different monitoring points and The Euclidean distance between is the number of features of the monitoring points, For monitoring point In the The value of the feature, For monitoring point In the The value of a feature.

[0037] Then, the average distance between the monitoring point and all other monitoring points is calculated to satisfy the following formula: , in, For the The average distance between a monitoring point and all other monitoring points, The number of monitoring points, For two different monitoring points and The Euclidean distance between .

[0038] Finally, the abnormal monitoring point is identified if the average distance between the monitoring point and other monitoring points is significantly greater than the average distance of all monitoring points. , then the monitoring point is determined to be an abnormal monitoring point, and the average distance of all monitoring points Satisfies the following formula: , in, is the average distance of all monitoring points, for The number of monitoring points, For the The average distance between a monitoring point and all other monitoring points.

[0039] By interpolating missing values, the integrity of the first historical data can be ensured, analysis bias or insufficient model training caused by missing first historical data can be avoided, and the availability of the first historical data can be enhanced. The method of interpolating missing values ​​can adopt the mean interpolation method. The principle is that the value of a data point is usually close to the values ​​of its adjacent points. Therefore, if the values ​​of the two points before and after a data point are known, the value of the point can be estimated by calculating the average of these two values. This method is particularly suitable for time series data or data sets with relatively uniform spatial distribution, where the differences between data points are usually not large.

[0040] The mean interpolation method interpolates missing values ​​to meet the following formula: , In the formula, Indicates The Missing values ​​for monitoring points, Indicates The The value of the monitoring point, Indicates The value of the monitoring point of the dimension.

[0041] Step S103, dividing the second historical data into a plurality of historical construction stage data according to the characteristics of different construction stages of the jacking of the curved slope inclined steel box girder bridge.

[0042] In this embodiment, the characteristics of different construction stages of the curved slope inclined steel box girder jacking construction project are that different project progresses have different concerns about construction safety, structural stress, linear control, equipment operation status, etc. The project progress is divided into early, middle and late stages, and the second historical data is correspondingly divided into early historical construction stage data, middle historical construction stage data and late historical construction stage data.

[0043] Step S104, adding noise to the historical construction stage data according to the characteristics of the different construction stages to obtain noise fusion data.

[0044] In this embodiment, the addition of noise mainly considers the influence of uncertain factors on the data. The greater the influence, the greater the noise added. Taking into account the different characteristics of the front, middle and late stages of the advancement, different noises should be added. The way to add noise is to first determine the proportion range of the error range caused by uncertain factors in the current stage of a data, use a random number generator to generate a random value within the range, and then add the value to the data to form data that is more in line with reality.

[0045] Taking the force parameters of the jack as an example, the early stage is the construction preparation and initial jacking stage, and the noise may be small because the equipment calibration and testing are mainly carried out. A smaller range of Gaussian noise can be added to simulate the random errors in the initial calibration and testing of the equipment. The mid-term stage is the main stage of the jacking process. The force parameters of the jack vary greatly. The wear of mechanical parts and the fluctuation of the hydraulic system may cause large noise. A medium range of uniformly distributed noise can be added to simulate the random fluctuations and impacts in the jacking process. The later stage is after the box girder is pushed into place, and the main work is structural adjustment and inspection. The noise may be small, but the accuracy of the final position needs to be ensured. A smaller range of normally distributed noise can be added to simulate the small errors in the final adjustment process.

[0046] In summary, adding noise to the historical construction stage data according to the characteristics of the forces in different construction stages can significantly improve the adaptability of the jack control model to various uncertainties in actual construction, thereby achieving more intelligent and precise adjustment of the jack and ensuring the quality and safety of the jacking construction of the curved and inclined steel box girder bridge.

[0047] Step S105, preprocessing the noise fusion data to obtain the historical sample data.

[0048] The process of preprocessing the noise fusion data to obtain the historical sample data specifically includes the following sub-steps: Step S10501, extracting features from the noise fusion data according to the characteristics of the different construction stages to obtain feature data of the different construction stages.

[0049] In this embodiment, due to the differences in data characteristics at different construction stages, different feature extraction methods need to be used to more accurately obtain feature data reflecting the construction status at each stage.

[0050] In the early construction stage, during the construction preparation and initial jacking stage, the force parameters of the jack are relatively stable, and the data changes are relatively small, but there may be some small fluctuations caused by factors such as the instability of the initial state of the equipment. At this time, discrete Fourier transform can be used to extract features from noise fusion data. Discrete Fourier transform can convert time domain signals into frequency domain signals. By analyzing the frequency domain characteristics, the periodic components and potential frequency characteristics in the data can be clearly identified. In the early construction stage, some small vibrations or unstable factors of the equipment may appear as components of specific frequencies in the frequency domain, and these features can be accurately captured using discrete Fourier transform. Its beneficial effect is that it can discover possible problems with the equipment in the initial state in advance, such as the resonant frequency of equipment components, etc., provide a reference for subsequent construction operations, and ensure that the equipment is in good operating condition before the formal jacking.

[0051] The mid-term construction stage is the main stage of the jacking process. The force parameters of the jack vary greatly. The wear of mechanical parts and the fluctuation of the hydraulic system may cause complex nonlinear and non-stationary characteristics in the data. In view of this situation, it is more appropriate to use wavelet transform for feature extraction. Wavelet transform has the characteristics of multi-resolution analysis and can analyze signals at different time and frequency scales. It has unique advantages in processing non-stationary signals. In the mid-term construction stage, the noise and useful signals in the data are often intertwined. Wavelet transform can effectively decompose and reconstruct the signal, highlight the characteristic components of the signal, and suppress noise interference. By analyzing the wavelet coefficients at different scales, detailed information on the change of jack force, the degree of wear of mechanical parts, etc. can be obtained, which is helpful to monitor the construction process in real time, adjust the construction parameters in time, and ensure the smooth progress of construction.

[0052] The later construction stage is when the box girder is pushed into place, and the main work is structural adjustment and inspection. At this time, the data is relatively stable, but the accuracy of the final position is required to be high. At this stage, the principal component analysis method can be used to extract features from the noise fusion data. The principal component analysis method is a dimensionality reduction technique that can convert multiple related variables into a few unrelated principal components, which contain the main information of the original data. In the later construction stage, the principal component analysis method can be used to remove redundant information in the data and extract key features related to structural adjustment and position accuracy. For example, by analyzing the data of multiple monitoring points using the principal component analysis method, the principal components that have the greatest impact on the final position of the structure can be found, thereby more accurately evaluating the state of the structure, providing an accurate basis for structural adjustment, and ensuring the accuracy and stability of the final position of the bridge.

[0053] Therefore, according to the characteristics of different stages of the top-pushing construction of the curved and sloped steel box girder bridge, the appropriate feature extraction method can more effectively obtain the feature data of different construction stages, provide more accurate input for the jack control model, thereby realizing the intelligent adjustment of the jack and improving the quality and safety of the bridge top-pushing construction. Step S10502, normalizing the characteristic data of different construction stages to obtain the historical sample data.

[0054] In this embodiment, normalization is a data preprocessing technology that aims to convert data of different ranges and magnitudes into a unified numerical range to make the data comparable and consistent. When processing the characteristic data of different stages of the top-pushing construction of the curved slope inclined steel box girder bridge, the normalization is mainly to adjust the numerical range of the characteristic data extracted by different methods in the early, middle and late stages.

[0055] First, the frequency domain features obtained by discrete Fourier transform in the early stage, the features extracted by wavelet transform in the middle stage, and the principal component data obtained by principal component analysis in the later stage will result in large differences in numerical range and magnitude. If normalization is not performed, when using these data to train the jack control model, more time and computing resources are required to adjust the weights and biases to adapt to data inputs of different magnitudes. After normalization, the data is in a relatively uniform range, the model can converge faster and improve the training effect, reduce training time, and improve training efficiency.

[0056] Secondly, normalization makes the data distribution more stable and consistent, avoiding the excessive or insufficient impact of certain feature dimensions on model training due to excessive or small values. The model can consider each feature more evenly when learning data features, thereby better capturing the inherent laws in the data. When faced with new actual construction data, the model can more accurately adjust the jack intelligently, improve the generalization ability of the model, and enable it to perform better in different construction scenarios and conditions.

[0057] Moreover, during the model training and calculation process, the original data of different magnitudes may lead to instability in numerical calculation. The normalized data is in a relatively small and stable numerical range, which can effectively avoid the occurrence of these numerical calculation problems and ensure the stability and reliability of the model training process.

[0058] Finally, the physical meaning and measurement units of the characteristic data at different construction stages may be different. Normalization eliminates these differences and allows data from different stages to be compared and analyzed at the same scale. For example, through normalization, the characteristic data reflecting the change in jack force and the characteristic data reflecting the accuracy of the structural position can be put together for comprehensive consideration, which facilitates a more comprehensive assessment of the construction status and precise control of the jack.

[0059] Step S2, constructing a jack control model, and using the historical sample data to train and evaluate the jack control model.

[0060] Among them, building the jack control model specifically includes the following sub-steps: Step S201, constructing a first loss function according to the force magnitude, force direction and stroke extension of the jack.

[0061] The first loss function satisfies the following formula: , in, is the loss function, is the number of predicted values ​​of the jack, For the The force of the jack, For the The predicted force of the jacks, No. The force direction of the jack, for The predicted force direction of the jacks, For the The stroke of the jack is extended. For the The predicted stroke extension of the jack.

[0062] Step S202: setting a constraint penalty term, and constructing a second loss function according to the first loss function and the constraint penalty term.

[0063] Among them, setting a constraint penalty term and constructing a second loss function according to the first loss function and the constraint penalty term specifically includes the following sub-steps: Step S20201, based on the relevant characteristics of the box girder, obtain the maximum allowable stress in the box girder design, the maximum allowable offset in the box girder design, and the minimum allowable height of the guide beam.

[0064] The maximum allowable stress in box girder design refers to the maximum stress value that the structure can withstand, which is determined through precise calculation and analysis during the box girder design stage. Once the actual stress exceeds this value, it will seriously affect the structural quality of the box girder and may even cause safety accidents.

[0065] The maximum allowable offset of the box girder design is the maximum offset allowed during the jacking process for construction safety. Excessive offset may cause the position of the box girder to deviate, affecting the overall structure of the bridge and subsequent construction.

[0066] The minimum allowable height of the guide beam is the minimum height between the guide beam at the front end of the box beam and the next support and the ground.

[0067] Step S20202, respectively setting a stress equalization penalty item, a maximum offset penalty item and a guide beam height penalty item according to the maximum allowable stress in the box beam design, the maximum allowable offset in the box beam design and the minimum allowable height of the guide beam.

[0068] The stress balance penalty term is set to ensure that the stress distribution of the key section remains balanced during the jacking process, so as to avoid the local stress exceeding the maximum stress allowed by the design, thereby preventing structural damage and failure. The penalty term calculates the deviation between the stress value of each monitoring point and the maximum stress allowed by the design, and averages the square sum, which is used as a penalty signal to feed back to the model.

[0069] The maximum offset penalty term is set to limit the horizontal offset of the structure during the jacking process to ensure that the offset does not exceed the maximum offset allowed by the design. The penalty term calculates the deviation between the offset value of each monitoring point and the maximum offset allowed by the design, and averages the square sum, which is used as a penalty signal to feed back to the model.

[0070] A guide beam height penalty item is set to ensure the stability and functionality of the structure by ensuring that the height of the guide beam is not lower than the minimum allowable height of the design. The penalty item calculates the deviation between the height value of each monitoring point and the minimum allowable height of the design, and averages the square sum of them, which is used as a penalty signal to feed back to the model. Among them, the minimum allowable height not lower than the design is dynamically determined based on the relative height of the guide beam and the next support point. The guide beam height should be 10 to 20 cm higher than the highest point of the next support point.

[0071] The stress equilibrium penalty term satisfies the following formula: , in, is the stress equilibrium penalty term, is the total number of stress monitoring points, For the The value of each stress monitoring point, Design the maximum allowable stress for the box girder.

[0072] The maximum deviation penalty term satisfies the following formula: , in, is the maximum deviation penalty term, is the number of offset monitoring points, For the The value of the offset monitoring point, Design the maximum allowable deflection for the box girder.

[0073] The guide beam height penalty term satisfies the following formula: , in, is the guide beam height penalty term, is the number of monitoring points for the guide beam height, For the The value of the guide beam height monitoring point, is the minimum allowable height of the guide beam.

[0074] Step S20203, setting penalty weights for the stress equalization penalty item, the maximum offset penalty item, and the guide beam height penalty item, respectively.

[0075] In this embodiment, respectively set , , The weights of the stress equilibrium penalty, maximum deflection penalty, and guide beam height penalty respectively. The weights should reflect the importance of each penalty in ensuring structural safety and meeting design requirements.

[0076] In an optional embodiment, from the perspective of safe construction, through empirical analysis, the weight of the initial stress balance penalty term is Set to 0.3, the weight of the initial maximum offset penalty term Set to 0.4, the weight of the initial guide beam height penalty term Set to 0.3. The jack control model can be combined for preliminary verification, and the weights can be manually adjusted according to the verification results to meet the training requirements of the model as much as possible.

[0077] Step S20204: setting the constraint penalty item according to the stress equalization penalty item, the maximum offset penalty item, the guide beam height penalty item, and the penalty weight.

[0078] In this embodiment, all penalty items are combined and processed, and the weight of each penalty item is combined to form a constraint penalty item. The jack control model can optimize the jacking process while meeting specific constraints. This method improves the adaptability and intelligence level of construction, helps to optimize the construction plan, reduce risks, and improve the overall efficiency and quality of bridge jacking construction. Dynamic adjustment of these weights can further ensure the adaptability of the model at different construction stages and ensure the stability and safety of the construction process.

[0079] Step S20205: construct the second loss function by combining the first loss function and the constraint penalty term.

[0080] The second loss function satisfies the following formula: , in, is the second loss function, is the first loss function, is the weight of the stress equilibrium penalty term, is the stress equilibrium penalty term, is the weight of the maximum offset penalty term, is the maximum deviation penalty term, is the weight of the guide beam height penalty term, It is the penalty term for the guide beam height.

[0081] Step S203: construct the jack control model based on the long short-term memory network using the second loss function.

[0082] In this embodiment, the long short-term memory network LSTM is an advanced recurrent neural network structure, which is specially designed to solve the gradient vanishing and gradient exploding problems encountered by traditional recurrent neural networks when processing time series data. It effectively controls the inflow and outflow of information by introducing a unique gating mechanism, including input gate, forget gate and output gate, so that it can capture and remember long-term dependencies. This ability enables LSTM to perform well in processing tasks such as natural language processing, speech recognition and time series prediction, with better flexibility and robustness, while alleviating the gradient problem in traditional RNN and improving the model's ability to process long sequence data. These advantages of LSTM make it the preferred model in the field of time series analysis and are widely used in various practical applications. Based on the long short-term memory network, a jack control model is constructed, and the second loss function is set as the loss function of the constructed jack control model. The selection of hyperparameters also plays a crucial role in the training of the jack control model.

[0083] In an optional embodiment, the hyperparameters are optimized using the sparrow algorithm, which can greatly improve the training effect of the jack control model.

[0084] First, determine the hyperparameters of the jack control model, and determine the value range of the hyperparameters based on experience and data investigation. The combination of each hyperparameter in the value range constitutes a solution space. Use the sparrow algorithm to find any parameter combination in the solution space, use the jack control model to configure the parameter combination, and then further use historical sample data for verification. Finally, find the set of production parameters with the best prediction effect through verification as the production parameters of the jack control model.

[0085] Step S3, obtaining real-time data of the curved slope inclined steel box girder bridge during the jacking construction, and using the real-time data to dynamically adjust the weight and bias of the jack control model.

[0086] In this embodiment, according to the structural characteristics and force analysis of the steel box girder, the installation position of the strain gauge or fiber grating sensor is accurately determined, and stress concentration or key stress-bearing parts should be selected, such as near welds, support points, etc. The installation surface needs to be cleaned and polished before installation to ensure that the sensor fits tightly against the surface of the structure to ensure the accuracy of the measured data. At the same time, the range and accuracy of the sensor should be considered to avoid data distortion due to overload or insufficient accuracy. The control points of the total station should be set in a stable position that is not disturbed by construction, and good line of sight conditions should be ensured between the control points. When installing a laser ranging sensor or displacement sensor, ensure that its measurement direction is consistent with the displacement direction to be monitored, and it should be firmly installed to prevent displacement or loosening during construction that affects the measurement results.

[0087] The installation and debugging of other equipment will not be elaborated here. After the equipment corresponding to the data to be obtained is installed, the equipment and the jack control model are integrated, and long-distance real-time communication is carried out using long-distance radio line of sight. Then the box girder enters the jacking stage, and the equipment sends real-time data to the jack control model. After feature extraction and normalization based on historical sample data, real-time data is formed.

[0088] The method of dynamically adjusting the weight and bias of the jack control model using the real-time data specifically includes the following sub-steps: Step S301, using the real-time data, calculating the complexity index of the real-time data.

[0089] The complexity index includes a stress complexity index, an offset complexity index and a guide beam height complexity index, and the stress complexity index satisfies the following formula: , in, is the stress complexity index, is the number of stress monitoring points, For the The value of each stress monitoring point, is the average value of the stress monitoring points, is the maximum value of the stress monitoring point, is the minimum value of the stress monitoring point; In this embodiment, the first part of the offset complexity index reflects the discreteness of the stress monitoring point data relative to the average value as a whole. The higher the degree of discreteness, the greater the contribution to the stress complexity index. The second part reflects the maximum fluctuation in the stress monitoring point data. The larger the fluctuation range, the greater the proportion in the stress complexity index.

[0090] By adding the first part and the second part, the discrete degree and fluctuation range of the stress monitoring point data are comprehensively reflected to evaluate the complexity of the stress distribution.

[0091] The offset complexity index satisfies the following formula: , in, is the offset complexity index, is the number of offset monitoring points, For the The value of the offset monitoring point, is the average value of the offset monitoring points, is the maximum value of the offset monitoring point, is the minimum value of the offset monitoring point, The value with the largest change in the offset monitoring point between adjacent real-time data collection moments; In this embodiment, the first part of the offset complexity index reflects the overall discreteness of the offset monitoring point data relative to the average value. The higher the discreteness, the greater the contribution to the offset complexity index. The second part reflects the maximum fluctuation in the offset monitoring point data. The larger the fluctuation range, the greater the proportion in the offset complexity index. The third part highlights the impact of the maximum change of the offset at adjacent moments on the offset complexity.

[0092] By adding the three parts of the migration complexity index, the discreteness of the migration monitoring point data, the fluctuation range, and the changes in adjacent moments are comprehensively considered, which can more comprehensively evaluate the complexity of the migration.

[0093] The guide beam height complexity index satisfies the following formula: , in, is an indicator of the height complexity of the guide beam, is the number of monitoring points for the guide beam height, For the The value of the guide beam height monitoring point, is the average value of the guide beam height monitoring points, is the maximum value of the guide beam height monitoring point, is the minimum value of the guide beam height monitoring point, It is the value with the largest change in the height monitoring point of the guide beam at adjacent real-time data collection moments.

[0094] In this embodiment, the first part of the guide beam height complexity index reflects the overall discreteness of the guide beam height monitoring point data relative to the average value. The higher the discreteness, the greater the contribution to the guide beam height complexity index. The second part reflects the maximum fluctuation in the guide beam height monitoring point data. The larger the fluctuation range, the greater the proportion in the guide beam height complexity index. The third part highlights the impact of the maximum change in the guide beam height at adjacent moments on the guide beam height complexity.

[0095] By adding the three parts of the guide beam height complexity index, the discrete degree, fluctuation range and changes in adjacent moments of the guide beam height monitoring point data are comprehensively considered, which can more comprehensively evaluate the complexity of the guide beam height and provide a reference for related construction and analysis.

[0096] Step S302: dynamically adjust the penalty weight of the second loss function according to the complexity index.

[0097] In this embodiment, the value ranges of the weights of the penalty weight stress equilibrium penalty item, the maximum offset penalty item and the guide beam height penalty item are pre-set according to actual engineering requirements and experience, and then the mapping relationship between the complexity index and the corresponding weight is further confirmed, and the calculated complexity index is substituted into the mapping relationship to obtain the corresponding penalty weight, which is applied to the second loss function.

[0098] In an optional embodiment, taking into account the complex construction environment, the curvature of the bridge, and the slope of the bridge, which have a significant impact on the force of the jack and the stress of the box girder, etc., to ensure that the model can quickly adapt to environmental changes and make timely responses, an exponential function is used to map the mapping relationship between the complexity index and the corresponding weight.

[0099] The penalty weight of the second loss function is dynamically adjusted to satisfy the following formula: , in, For the The penalty weights corresponding to the class complexity indexes include stress complexity index, offset complexity index and guide beam height complexity index, which are used to adjust the penalty degree for the corresponding construction conditions in the loss function of the jack control model. For the The minimum value of the class penalty weight, For the The maximum value of the class penalty weight, For the Class complexity index, For the The median value of the class complexity index, To control the The parameter of the curve shape of the function of the relationship between the class complexity index and the penalty weight can be positive or negative. A positive number indicates that as Increase, The growth rate is accelerating, and negative numbers indicate that Increase, The growth rate has slowed down.

[0100] in, For the The middle value of the complexity index of the class is obtained by collecting a large amount of corresponding complexity index data in the past construction process and sorting these data from small to large. If the amount of data is an odd number, the value in the middle position is taken as the middle value; if the amount of data is an even number, the average of the two middle values ​​is taken. For example, 100 sets of stress complexity index data were obtained in multiple bridge jacking constructions. After sorting, the average of the 50th and 51st data was taken as the middle value of the stress complexity index.

[0101] To control the The parameters of the shape of the function curve of the relationship between the class complexity index and the penalty weight are used. The complexity index and the corresponding appropriate penalty weight value in the historical construction data are used to find the parameter value that can minimize the error between the weight value calculated by the function and the actual required weight value through data fitting techniques such as the least squares method. For example, there is a set of stress complexity index data and its corresponding stress equalization penalty weight data that has been verified to be effective in practice. The parameters are adjusted through the least squares method to make the weight calculated by the formula as close as possible to the actual effective weight.

[0102] Step S303: using the real-time data, calculate the adjusted loss value of the second loss function.

[0103] In this embodiment, real-time data is substituted into the second loss function, and the loss value is obtained by calculation. The loss value measures the performance of the jack control model under the current data, that is, the degree of deviation between the actual situation and the model prediction, which can provide a quantitative basis for further optimizing the model or adjusting the construction strategy.

[0104] Step S304: Calculate the gradient of the loss function using a back propagation algorithm according to the loss value.

[0105] In this embodiment, the core idea of ​​the back propagation algorithm is to back propagate the output error of the network back to each layer in the network, so as to calculate the contribution of each weight to the final error.

[0106] Use the loss value to start the back propagation process. For the output layer of the model, the partial derivative of the output of the output layer is calculated according to the loss function. This step is the basis for subsequent calculations. Then, use the chain rule to back propagate this partial derivative to the previous layer. Specifically, if the previous layer uses an activation function, it is necessary to first derive the activation function, and then combine the weights from this layer to the output layer to calculate the partial derivative of the loss function with respect to the output of this layer. In this way, back propagate layer by layer, and each layer must calculate the partial derivative of the loss function with respect to the weight parameters of this layer according to the chain rule. For example, for a three-layer neural network, back propagate from the output layer to the hidden layer, and then back propagate from the hidden layer to the input layer, and continuously calculate the partial derivatives corresponding to the weight parameters of each layer. During the calculation process, the calculation logic and parameter relationship of each layer must be carefully handled to ensure the accuracy of the partial derivative calculation. Finally, the partial derivatives of the loss function with respect to each weight parameter can be obtained, and the vector composed of these partial derivatives is the gradient.

[0107] Step S305, dynamically adjusting the weight and bias of the jack control model according to the gradient.

[0108] The step of dynamically adjusting the weight and bias of the jack control model according to the gradient specifically includes the following sub-steps: Step S30501, introducing adaptive moment estimation, and calculating the first-order moment estimation and the second-order moment estimation of the adaptive moment estimation according to the gradient.

[0109] In this embodiment, the core of adaptive moment estimation is to substitute the calculated gradient into the first-order moment estimation formula and the second-order moment estimation formula to calculate the corresponding results. First, the key parameters are initialized, the time step is 0, and the first-order moment estimation and the second-order moment estimation of each weight parameter are set to 0. Assume that in the first The gradient of the second loss function with respect to the weight parameter after adjusting the penalty weight is obtained through the back propagation algorithm in the time step, and the first-order moment estimate satisfies the following formula: , in, is the first-order moment estimate, is the exponential decay rate of the first-order moment estimate, usually close to 0.9, is the first-order moment estimate of the previous time step, is the gradient.

[0110] The second-order moment estimate satisfies the following formula: , in, is the second-order moment estimate, is the exponential decay rate of the second-order moment estimate, which is generally close to 0.999. is the second-order moment estimate for the previous time step, is the gradient.

[0111] Step S30502, correcting the first-order moment estimate and the second-order moment estimate, and using the corrected first-order moment estimate and the second-order moment estimate to dynamically adjust the weight and bias of the jack control model.

[0112] Since the first-order moment estimate and the second-order moment estimate are both 0 at initialization, the first-order moment estimate and the second-order moment estimate obtained earlier will also be biased towards 0, so the first-order moment estimate and the second-order moment estimate need to be corrected to satisfy the following formula: , in, is the bias-corrected first-order moment estimate, is the unbias-corrected first-order moment estimate, is the exponential decay rate of the first-order moment estimate of Power, is the time step.

[0113] , is the bias-corrected second-order moment estimate, is the unbias-corrected second-order moment estimate, is the exponential decay rate of the second-order moment estimate of Power, is the time step.

[0114] In this embodiment, after correcting the first-order moment estimate and the second-order moment estimate, the update amount of the weight and bias can be calculated. Assuming that the learning rate is , a very small constant , the value is , the purpose is to prevent the denominator from being 0 during calculation, and update the amount , is the weight or bias.

[0115] The update amount satisfies the formula:

[0116] After obtaining the update amount, the weight and bias of the jack control model can be dynamically adjusted to meet the following formula: , in, For in time The updated weight matrix, is the weight matrix of the current time step, is the weight update amount.

[0117] , in, For in time The updated bias matrix, is the bias matrix of the current time step, is the offset update amount.

[0118] The above steps are repeated iteratively, and as the time step increases, the weights and biases of the jack control model are dynamically adjusted according to the gradients calculated in real time. In each iteration, the new gradients are calculated by the back propagation algorithm, and then the first-order moment estimate and the second-order moment estimate are updated. After correction, the update amount is calculated again and the weights and biases are adjusted, so that the jack control model can better adapt to different working conditions and real-time data, and improve the accuracy and stability of control.

[0119] Step S4, introducing a noise alarm system, and using the noise alarm system and the dynamically adjusted jack control model to intelligently adjust the jack.

[0120] Using the noise alarm system and the dynamically adjusted jack control model to intelligently adjust the jack includes: The noise alarm system is used to monitor the offset of the box beam and the height of the guide beam in real time; Controlling the jack to suspend the jacking construction according to the monitoring result, and manually correcting the jacking construction; Based on the result of the manual deviation correction, the jack is intelligently adjusted using the dynamically adjusted jack control model.

[0121] Intelligent regulation here means that, compared with the prior art, the present invention requires manual correction when the noise alarm system detects a major deviation. At other times, the jack is controlled by an adjusted jack control model, and the jack control model can dynamically adapt to complex jacking environments and intelligently provide a jack force plan, so it is intelligent regulation.

[0122] In this embodiment, a noise alarm system is introduced, and the jack is intelligently adjusted by using the noise alarm system and the dynamically adjusted jack control model, which can improve the adaptability and safety control of the jack control model to complex environments. For example, for a curved inclined steel box girder bridge, the bridge is characterized in that the bridge is bent into a certain arc, and the bridge itself has a certain slope. Calculating the force of the jack by traditional methods will undoubtedly add a lot of difficulty to the jacking construction. However, the jack control model of the present invention can be adaptively adjusted according to real-time data, and is easy to implement. Because during the jacking construction process, the arc and slope of the bridge will extend according to a rule, and due to the existence of this rule, the jack control model can more easily learn changes in the environment, thereby realizing intelligent control of the jack.

[0123] The noise alarm system is a system that monitors the offset of the box girder and the height of the guide beam in real time for the safety of the jacking construction, and issues a noise alarm for the box girder offset and guide beam height that exceed the safety threshold. After receiving the alarm signal, the jack control model controls the jack to stop working, and the jacking construction is manually corrected. After the manual correction is completed, the jack control model continues to control the jacking.

[0124] The addition of a noise alarm system can ensure safety in complex construction environments by preventing uncontrollable factors from causing excessive deviations in jacking construction, which can greatly improve the safety of jacking construction.

[0125] like Figure 2As shown, the present invention also provides an intelligent adjustment system for jacks during jacking of a curved and inclined steel box girder bridge, comprising: a processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute relevant steps of a relevant embodiment of a method for intelligent adjustment of jacks during jacking of a curved and inclined steel box girder bridge in the previous aspect of the present invention.

[0126] The present invention provides an intelligent adjustment system for jacks during top pushing of a curved and inclined steel box girder bridge, wherein each functional component can be integrated into one processing component, or each component can exist physically separately, or two or more components can be integrated into one component. The above-mentioned integrated components can be implemented in the form of hardware or in the form of software functions, further improving the overall applicability and practical application capability of the present invention.

Claims

1. An intelligent adjustment method for jacks during top pushing of a curved and sloped steel box girder bridge, characterized in that: The method comprises: Obtain historical sample data of top-thrusting of curved and sloped steel box girder bridges; Constructing a jack control model, and using the historical sample data to train and evaluate the jack control model; Acquire real-time data of the curved slope inclined steel box girder bridge during the jacking construction, and dynamically adjust the weight and bias of the jack control model using the real-time data; A noise alarm system is introduced, and the jack is intelligently adjusted by using the noise alarm system and the dynamically adjusted jack control model.

2. The intelligent adjustment method for jacks during top pushing of a curved and inclined steel box girder bridge according to claim 1 is characterized in that: The historical sample data of the jacking of the curved slope inclined steel box girder bridge are obtained as follows: Acquire first historical data related to the jacking of the curved slope inclined steel box girder bridge; Eliminate outliers and interpolate missing values ​​from the first historical data to obtain second historical data; According to the characteristics of different construction stages of the jacking of the curved slope inclined steel box girder bridge, the second historical data is divided into a plurality of historical construction stage data; According to the characteristics of the different construction stages, noise is added to the historical construction stage data to obtain noise fusion data; The noise fusion data is preprocessed to obtain the historical sample data.

3. The intelligent adjustment method for jacks during top pushing of a curved and inclined steel box girder bridge according to claim 2 is characterized in that: The preprocessing of the noise fusion data to obtain the historical sample data includes: Extracting features from the noise fusion data according to the characteristics of the different construction stages to obtain feature data of the different construction stages; The characteristic data of different construction stages are normalized to obtain the historical sample data.

4. The intelligent adjustment method for jacks during jacking of a curved and inclined steel box girder bridge according to claim 1 is characterized in that: The construction of the jack control model comprises: Constructing a first loss function according to the force magnitude, force direction and stroke extension of the jack; Setting a constraint penalty term, and constructing a second loss function according to the first loss function and the constraint penalty term; Based on the long short-term memory network, the jack control model is constructed using the second loss function.

5. The intelligent adjustment method for jacks during top pushing of a curved and inclined steel box girder bridge according to claim 4 is characterized in that: The first loss function satisfies the following formula: , in, is the loss function, is the number of predicted values ​​of the jack, For the The force of the jack, For the The predicted force of the jacks, No. The force direction of the jack, for The predicted force direction of the jacks, For the The stroke of the jack is extended. For the The predicted stroke extension of the jack.

6. The intelligent adjustment method for jacks during top pushing of a curved and inclined steel box girder bridge according to claim 4 is characterized in that: The setting of the constraint penalty term and constructing the second loss function according to the first loss function and the constraint penalty term comprises: According to the relevant characteristics of the box girder, the maximum allowable stress of the box girder design, the maximum allowable offset of the box girder design, and the minimum allowable height of the guide beam are obtained; According to the maximum allowable stress in the box girder design, the maximum allowable offset in the box girder design and the minimum allowable height of the guide beam, a stress equalization penalty item, a maximum offset penalty item and a guide beam height penalty item are respectively set; Setting penalty weights for the stress equalization penalty item, the maximum deviation penalty item, and the guide beam height penalty item respectively; The constraint penalty item is set according to the stress equalization penalty item, the maximum deviation penalty item, the guide beam height penalty item and the penalty weight; The second loss function is constructed by combining the first loss function and the constraint penalty term.

7. The intelligent adjustment method for jacks during jacking of a curved and inclined steel box girder bridge according to claim 6 is characterized in that: The dynamically adjusting the weight and bias of the jack control model using the real-time data includes: Utilizing the real-time data, calculating a complexity index of the real-time data; Dynamically adjusting the penalty weight of the second loss function according to the complexity index; Using the real-time data, calculating the adjusted loss value of the second loss function; According to the loss value, the gradient of the loss function is calculated using a back propagation algorithm; The weights and biases of the jack control model are dynamically adjusted according to the gradient.

8. The intelligent adjustment method for jacks during jacking of a curved and inclined steel box girder bridge according to claim 7 is characterized in that: The complexity index includes a stress complexity index, an offset complexity index and a guide beam height complexity index, and the stress complexity index satisfies the following formula: , in, is the stress complexity index, is the number of stress monitoring points, For the The value of each stress monitoring point, is the average value of the stress monitoring points, is the maximum value of the stress monitoring point, is the minimum value of the stress monitoring point; The offset complexity index satisfies the following formula: , in, is the offset complexity index, is the number of offset monitoring points, For the The value of the offset monitoring point, is the average value of the offset monitoring points, is the maximum value of the offset monitoring point, is the minimum value of the offset monitoring point, The value with the largest change in the offset monitoring point between adjacent real-time data collection moments; The guide beam height complexity index satisfies the following formula: , in, is an indicator of the height complexity of the guide beam, is the number of monitoring points for the guide beam height, For the The value of the guide beam height monitoring point, is the average value of the guide beam height monitoring points, is the maximum value of the guide beam height monitoring point, is the minimum value of the guide beam height monitoring point, It is the value with the largest change in the height monitoring point of the guide beam at adjacent real-time data collection moments.

9. The intelligent adjustment method for jacks during jacking of a curved and inclined steel box girder bridge according to claim 7 is characterized in that: The dynamically adjusting the weight and bias of the jack control model according to the gradient includes: Introducing an adaptive moment estimation, and calculating a first-order moment estimation and a second-order moment estimation of the adaptive moment estimation according to the gradient; The first-order moment estimate and the second-order moment estimate are corrected, and the weight and bias of the jack control model are dynamically adjusted using the corrected first-order moment estimate and the second-order moment estimate.

10. An intelligent adjustment system for jacks during top pushing of a curved and inclined steel box girder bridge, characterized in that: include: A processor, an input device, an output device and a memory, wherein the processor, the input device, the output device and the memory are interconnected, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is configured to call the program instructions to execute the intelligent adjustment method of the jack during the jacking of a curved slope inclined steel box girder bridge as described in any one of claims 1 to 9.

Images