Safety evaluation method for initial installation state of highway bridge girder erection machine

Abstract

The invention relates to the technical field of safety evaluation of highway bridge girder erection machines, and discloses a safety evaluation method for the initial installation state of a highway bridge girder erection machine. Firstly, a multi-modal safety evaluation model including a structure layer, a dynamic parameter layer and a risk factor layer is established; dividing a dynamic parameter layer feature interval by using a fuzzy clustering algorithm, calculating a risk factor correlation coefficient by using a grey correlation analysis method, determining a dynamic parameter layer parameter objective weight by using an entropy weight method, and performing parameter fusion by combining the two to obtain a comprehensive weight; then multi-source sensing data installed by the bridge erecting machine are collected in real time and discretized; and finally calculating risk indexes layer by layer, and judging the safety level of the initial installation state of the bridge girder erection machine. The method comprehensively considers the structure, the dynamic parameters and the risk factors of the bridge girder erection machine, improves the evaluation accuracy by applying various algorithms, realizes the automatic judgment of the safety level, can effectively reduce the installation risk of the bridge girder erection machine, and guarantees the construction safety and the smooth engineering construction.

Description

Technical Field

[0001] The present invention relates to the technical field of safety evaluation of highway bridge erecting machines, and specifically to a safety evaluation method for the initial installation state of highway bridge erecting machines. Background Technique

[0002] In modern highway construction, bridge erecting machines, as key equipment, are widely used in bridge erection projects. With the continuous expansion of the scale of highway bridge construction and the increasing requirements for construction efficiency and quality, the usage frequency and complexity of bridge erecting machines are also continuously rising. However, when a bridge erecting machine is in the initial installation state, due to its complex structure, numerous installation links involved, and various dynamic parameter changes, there are relatively high safety risks. Once a safety accident occurs, it will not only cause serious casualties and property losses but also have a huge impact on the progress and cost of the entire highway construction project.

[0003] Currently, the traditional safety assessment methods for highway bridge erecting machines are relatively single, mostly relying on manual experience judgment and simple equipment inspections. This method has obvious limitations: on the one hand, manual experience judgment is highly subjective, and the judgment criteria of different personnel vary greatly, making it difficult to accurately evaluate complex safety conditions; on the other hand, simple equipment inspections can only detect some surface and obvious problems, and are unable to effectively monitor and evaluate potential risks inside the equipment and safety hazards during the dynamic change process.

[0004] From the perspective of the structure of the bridge erecting machine itself, it consists of multiple complex components, such as main girders, legs, brackets, etc. The physical topological relationships between the components are close and interact with each other. During the installation process, the state change of each component may trigger a chain reaction, affecting the overall safety. For example, inaccurate positioning of the main girder may lead to deviations during subsequent beam erection, problems with the leg synchronization will cause uneven stress on the whole machine, and insufficient load-bearing capacity of the bracket may lead to local structural damage. However, existing safety evaluation methods often cannot comprehensively consider these complex relationships between components and the resulting safety risks.

[0005] In terms of dynamic parameter monitoring, various dynamic parameters are generated during the installation process of the bridge erecting machine, such as displacement deviation, angle offset, stress distribution, etc. These parameters change continuously over time, reflecting the real-time state of the bridge erecting machine during the installation process. However, existing monitoring technologies are difficult to accurately and comprehensively collect and analyze these multi-dimensional dynamic parameters, and cannot timely detect the safety risks indicated by abnormal parameter changes. At the same time, there is a lack of scientific and effective methods for dividing the characteristic intervals of these dynamic parameters, making it lack an accurate basis when evaluating the safety state.

[0006] There are also problems with risk factor assessment. Due to the lack of a complete historical failure database for support, it is difficult to accurately determine the critical threshold intervals for each module, and potential risk factors cannot be effectively identified. When calculating the correlation coefficient of risk factors and determining the weights of each parameter, traditional methods also have problems with low accuracy, resulting in the final safety evaluation results being unable to truly reflect the actual safety status of the initial installation state of the bridge erecting machine. Summary of the Invention

[0007] The purpose of the present invention is to provide a safety evaluation method for the initial installation state of a highway bridge erecting machine to solve the problems raised in the above background technology.

[0008] To achieve the above purpose, the present invention provides the following technical solution: A safety evaluation method for the initial installation state of a highway bridge erecting machine, the method comprising: S1. Establish a multi-modal safety evaluation model for the highway bridge erecting machine, including a structure layer, a dynamic parameter layer, and a risk factor layer; S2. Divide the multi-dimensional feature intervals of the dynamic parameter layer based on the fuzzy clustering algorithm; S3. Calculate the correlation coefficient of each parameter in the risk factor layer using the grey relational analysis method; S4. Determine the objective weights of each parameter in the dynamic parameter layer through the entropy weight method to obtain the entropy weight; S5. Combine the correlation coefficient and the entropy weight to perform parameter fusion to generate the comprehensive weight of the dynamic parameter layer; S6. Real-time collect multi-source sensing data during the installation process of the bridge erecting machine and perform discretization processing; S7. Calculate the risk index layer by layer to determine the safety level of the initial installation state of the highway bridge erecting machine.

[0009] Preferably, in the step S1: The structure layer is divided into a main beam positioning module, a leg synchronization module, a support bearing module, and an overall machine balance module based on the physical topological relationship of the bridge erecting machine installation components; The dynamic parameter layer collects displacement deviation, angle deviation, and stress distribution parameters of each module through a sensor network; The risk factor layer extracts the critical threshold intervals of each module based on the historical failure database.

[0010] Preferably, the steps of dividing the feature intervals based on the fuzzy clustering algorithm in the step S2 include: Set the fuzzy membership function according to the parameter distribution characteristics of the dynamic parameter layer; Calculate the similarity matrix between different parameters through an iterative optimization algorithm; Perform feature clustering grouping on the dynamic parameter layer according to the similarity matrix.

[0011] Preferably, the implementation steps of the grey relational analysis method in step S3 include: Select the benchmark reference sequence for each parameter in the risk factor layer; Calculate the correlation coefficient between each parameter sequence and the benchmark reference sequence; Generate a correlation degree coefficient matrix according to the mean value of the correlation coefficients.

[0012] Preferably, the steps for determining the objective weight of the entropy weight method in step S4 include: Perform probability density function conversion on the historical data set of the dynamic parameter layer, calculate the information entropy value and the difference coefficient of each parameter, and generate the objective weight of the dynamic parameter layer through normalization of the difference coefficient.

[0013] Preferably, the steps of parameter fusion in step S5 include: Take the correlation degree coefficient as the subjective weight factor and the entropy weight as the objective weight factor; Adopt a Bayesian network model to perform probability fusion on the subjective and objective weights, and generate the comprehensive weight of the dynamic parameter layer through the posterior probability distribution.

[0014] Preferably, the discretization process in step S6 includes: Perform abnormal fluctuation detection and noise filtering on the multi-source sensing data, convert the data with different dimensions into a unified discrete interval through the range normalization method, and generate a discretization distribution map of the dynamic parameters according to the time series.

[0015] Preferably, the calculation of the risk index in step S7 includes: Multiply the comprehensive weight of the dynamic parameter layer by the discretized data to generate the risk index of each structural layer; Generate the total risk score by superimposing the preset structural layer weight coefficient and the risk index, and divide the safety level based on the score interval.

[0016] Preferably, the dynamic parameter layer includes at least one of the following parameters: Longitudinal displacement deviation, lateral torsion angle and track levelness of the main beam positioning module; Hydraulic pressure difference, telescopic speed synchronization rate and vertical tilt angle of the leg synchronization module; Difference value of the support forces of the front and rear legs and symmetry degree of the transverse movement track of the whole machine balance module.

[0017] Preferably, the steps for generating the discretization distribution map include: Divide the standardized sensing data into equal-length data segments according to the time window, adopt the wavelet packet decomposition algorithm to extract the frequency domain energy characteristics of each data segment, and calculate the characteristic difference degree matrix of the data segments in different time windows through the Mahalanobis distance.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: At the model construction level, a multi-modal safety evaluation model for highway bridge erecting machines is established, which includes a structure layer, a dynamic parameter layer, and a risk factor layer. The structure layer is carefully divided based on the physical topological relationship of the installation components of the bridge erecting machine, covering the main beam positioning module, the leg synchronization module, the support bearing module, and the overall machine balance module, which can comprehensively and accurately reflect the structural states of the key parts of the bridge erecting machine. The dynamic parameter layer collects parameters such as displacement deviation, angle deviation, and stress distribution of each module through a sensor network to achieve real-time monitoring of the dynamic characteristics of the bridge erecting machine during operation. The risk factor layer extracts the critical threshold intervals of each module based on the historical fault database, providing a scientific and accurate reference standard for risk assessment. This multi-modal model integrates the structure of the bridge erecting machine, dynamic operation parameters, and historical fault information, overcoming the drawback of traditional evaluation methods that only focus on a single factor, and greatly improving the comprehensiveness and accuracy of safety evaluation.

[0019] In terms of algorithm application, the multi-dimensional feature intervals of the dynamic parameter layer are divided based on the fuzzy clustering algorithm. According to the distribution characteristics of the dynamic parameters, a fuzzy membership function is set, and then the similarity matrix is calculated and clustered by means of an iterative optimization algorithm, which can effectively handle the complexity and uncertainty of the dynamic parameters, reasonably distinguish the parameter characteristics under different operating states, and lay a solid foundation for the subsequent accurate assessment of safety risks. The grey relational analysis method is used to calculate the correlation coefficient of each parameter in the risk factor layer. By selecting the reference sequence, calculating the correlation coefficient, and generating the correlation coefficient matrix, the correlation degree between each risk factor can be accurately analyzed, the key risk factors can be identified, and the evaluation result is more targeted. The entropy weight method is used to determine the objective weights of the parameters in the dynamic parameter layer. By performing probability density function conversion on the historical data set, calculating the information entropy value and the difference coefficient and normalizing them, the information of the data itself is fully utilized, avoiding the arbitrariness of subjective weight assignment, and ensuring the scientificity and objectivity of weight determination.

[0020] In the parameter fusion link, the parameters are fused by combining the correlation coefficient and the entropy weight. The correlation coefficient is used as the subjective weight factor, and the entropy weight is used as the objective weight factor. The Bayesian network model is used for probability fusion to generate the comprehensive weight. This fusion method takes into account both expert experience and data-driven, effectively integrates subjective and objective information, and makes the evaluation result more in line with the actual situation.

[0021] In the data processing and evaluation process, multi-source sensing data during the erection process of the bridge girder erecting machine is collected in real time and discretized. First, abnormal fluctuation detection and noise filtering are carried out, and then the data dimension is unified through the range normalization method to generate a discretized distribution map, ensuring that the data input into the evaluation model is true, reliable and easy to analyze. The risk index is calculated layer by layer to determine the safety level. The risk index of each structural layer is obtained by multiplying the comprehensive weight of the dynamic parameter layer by the discretized data, and the total risk score is generated by combining and superimposing the weight coefficients of the structural layer to divide the safety level, realizing the full process automation and standardization from data collection, processing to safety evaluation, which can timely and accurately reflect the safety status of the initial state of the bridge girder erecting machine installation, provide intuitive and effective safety warnings for construction personnel, help them take measures in time to eliminate potential safety hazards, greatly reduce the safety risks during the erection process of the bridge girder erecting machine, and ensure the life safety of construction personnel and the smooth progress of the project construction. At the same time, the application of this method also helps to improve the installation quality and construction efficiency of the bridge girder erecting machine, reduce the project construction cost, and is of great significance to promoting the technological progress and sustainable development of the highway bridge construction industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is the working principle diagram of the safety evaluation method for the initial state of the highway bridge girder erecting machine installation described in the present invention; Figure 2 is the step diagram for generating the correlation coefficient matrix by using the grey relational analysis method; Figure 3 is the working principle diagram for generating the comprehensive weight by combining the correlation coefficient and the entropy weight; Figure 4 is the flow chart of the discretization process of multi-source sensing data. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] Please refer to Figures 1 - 4 , the present invention provides a safety evaluation method for the initial state of the highway bridge girder erecting machine installation, and the specific implementation steps are as follows: Construct a multi-modal safety evaluation model including a structural layer, a dynamic parameter layer and a risk factor layer. The structural layer is divided according to the physical topological relationship of the erection components of the bridge girder erecting machine, the dynamic parameter layer collects relevant parameters through a sensor network, and the risk factor layer extracts the critical threshold interval based on the historical fault database.

[0025] Set the fuzzy membership function according to the parameter distribution characteristics of the dynamic parameter layer, calculate the similarity matrix between different parameters using the iterative optimization algorithm, and then perform feature clustering and grouping on the dynamic parameter layer based on the similarity matrix.

[0026] Select the reference sequence of each parameter in the risk factor layer, calculate the correlation coefficient between each parameter sequence and the reference sequence, and generate the correlation coefficient matrix according to the average value of the correlation coefficients.

[0027] Perform probability density function conversion on the historical data set of the dynamic parameter layer, calculate the information entropy value and the coefficient of variation of each parameter, and generate the objective weight of the dynamic parameter layer through the normalization of the coefficient of variation.

[0028] Take the correlation coefficient as the subjective weight factor and the entropy weight as the objective weight factor, use the Bayesian network model to perform probability fusion on the subjective and objective weights, and generate the comprehensive weight of the dynamic parameter layer through the posterior probability distribution.

[0029] Collect multi-source sensing data during the installation process of the bridge girder erecting machine in real time, perform abnormal fluctuation detection and noise filtering on the data, convert data with different dimensions into a unified discrete interval through the range normalization method, and generate a discretized distribution map of dynamic parameters according to the time series.

[0030] Multiply the comprehensive weight of the dynamic parameter layer by the discretized data to generate the risk index of each structural layer; superimpose the preset structural layer weight coefficient and the risk index to generate the total risk score, and divide the safety level based on the score interval.

[0031] The present invention will be further described below in conjunction with Embodiments 1 to 5: Embodiment 1: When establishing a multi-modal safety evaluation model for a highway bridge girder erecting machine, the structural layer is divided based on the physical topological relationship of the installation components of the bridge girder erecting machine. The installation components of the bridge girder erecting machine have complex physical connections and collaborative working relationships. Among them, the main beam positioning module is a key part to ensure the accurate position of the main beam during installation. Its stability and accuracy directly affect the quality and safety of subsequent bridge erection. In actual installation, the position deviation of the main beam may cause uneven stress on the bridge, leaving potential safety hazards. The leg synchronization module is mainly responsible for the synchronization of multiple legs during movement. If the legs are not synchronized, it will cause unbalanced stress on the bridge girder erecting machine during support and movement, and even lead to tipping accidents. The support bearing module bears the weight of the entire bridge girder erecting machine and the bridge to be erected, and its bearing capacity and stability are crucial. Once there is a problem with the support bearing, it may cause the collapse of the bridge girder erecting machine. The overall balance module ensures the balance state of the bridge girder erecting machine during installation as a whole. It comprehensively considers the stress and position relationships of each part to ensure the stable operation of the bridge girder erecting machine.

[0032] The dynamic parameter layer collects displacement deviation, angle offset and stress distribution parameters of each module through the sensor network. In the main beam positioning module, it is necessary to accurately collect parameters such as longitudinal displacement deviation, lateral torsion angle and track horizontality. These parameters can reflect the position and posture changes of the main beam in real time. In the outrigger synchronization module, parameters such as hydraulic pressure difference, telescopic speed synchronization rate and vertical tilt angle are collected to monitor whether the working status of the outrigger is synchronized. For the whole machine balance module, parameters such as the difference in support force of the front and rear outriggers and the symmetry of the lateral track are collected to provide a basis for judging the balance of the whole machine. The risk factor layer extracts the critical threshold interval of each module based on the historical fault database. By analyzing and sorting a large amount of historical fault data, the safety parameter range of each module under different working conditions is determined. When the real-time parameters collected exceed this range, it means that there is a certain safety risk. When building the model, it is necessary to ensure that the data interaction between the layers is smooth and can accurately reflect the actual situation during the installation of the bridge crane.

[0033] In the highway bridge-building machine installation scenario, taking a certain type of highway bridge-building machine installation project as an example, the bridge-building machine is used to build a cross-river highway bridge. The bridge is 500 meters long and consists of 20 prefabricated box girders, each of which weighs about 80 tons.

[0034] Before the actual construction, the structural layers are divided according to the physical topological relationship of the installation components of the bridge-erecting machine. The main beam of the bridge-erecting machine is a key component that carries the prefabricated box beam and realizes its lifting and installation. The accuracy of its installation position is directly related to the overall structural stability of the bridge. In this project, the main beam positioning module covers devices for accurately measuring the longitudinal, lateral position and height of the main beam. Among them, the longitudinal positioning uses a high-precision laser rangefinder, which is installed on one side of the bridge-erecting machine track to monitor the distance deviation between the front end of the main beam and the established installation position in real time; the lateral positioning uses a horizontal displacement sensor, which is installed at both ends of the main beam to monitor the lateral displacement of the main beam. For height adjustment, it is equipped with multiple high-precision pressure sensors distributed at the main beam support points, which reflect whether the height of the main beam is within the design requirements based on the pressure change.

[0035] The outrigger synchronization module is extremely important for the smooth operation of the bridge erection machine. The bridge erection machine uses four hydraulic outriggers, and the outrigger synchronization module mainly monitors the hydraulic system parameters and the outrigger movement status. A pressure sensor is installed on the hydraulic cylinder of each outrigger to collect the hydraulic pressure difference in real time to determine whether the force on each outrigger is balanced. At the same time, a displacement sensor is installed on the outrigger telescopic mechanism to ensure the coordination of the telescopic movements of each outrigger by calculating the synchronization rate of the telescopic speed. In addition, an inclination sensor is used to monitor the vertical inclination angle of the outrigger to prevent the outrigger from tilting too much and causing the bridge erection machine to become unstable.

[0036] The support bearing module is a key part that bears the weight of the bridge girder erecting machine itself and the load of lifting the box girder. Stress sensors are installed at key stress-bearing parts of the support, such as the connection between the column and the crossbeam, the bottom support points of the support, etc., to monitor the magnitude and distribution of the stress on the support in real time. The data collected by these sensors is transmitted to the control system in real time. Once the stress approaches or exceeds the design allowable range, the system immediately issues an alarm to prompt the operator to take corresponding measures to avoid damage to the support due to overload. The overall balance module of the whole machine ensures the stability of the bridge girder erecting machine as a whole. By installing load cells on the front and rear outriggers respectively, the difference value of the support forces of the front and rear outriggers is measured to judge whether the weight distribution of the front and rear of the bridge girder erecting machine is balanced. At the same time, laser alignment devices are installed on both sides of the transverse movement track to monitor the symmetry of the transverse movement track, ensuring that the bridge girder erecting machine remains balanced during the transverse movement process and preventing safety accidents caused by skew.

[0037] During the installation process of the bridge girder erecting machine, displacement deviation, angle deviation and stress distribution parameters of each module are collected through the sensor network. For the main girder positioning module, the longitudinal displacement deviation is collected by a laser rangefinder every 10 seconds, and the deviation value of the main girder in the longitudinal direction relative to the ideal installation position is recorded; the transverse torsion angle is obtained by using biaxial inclinometers installed at both ends of the main girder, and this sensor can measure the change of the torsion angle of the main girder on the horizontal plane in real time; the track levelness is monitored by multiple levels installed on the track, with a level set every 5 meters to collect the height difference data of the track, and then the track levelness deviation is calculated.

[0038] Regarding the outrigger synchronization module, the hydraulic pressure difference value is collected in real time by pressure sensors installed on the hydraulic cylinders of each outrigger, and the data is collected once per second to detect the pressure difference between each outrigger in time; the telescopic speed synchronization rate is calculated by collecting the displacement data during the telescopic process of the outrigger through displacement sensors and combining the time interval, and it is recorded once every 20 seconds; the vertical inclination angle is continuously monitored by an inclinometer installed on the outrigger, and the data collection frequency is 1 time per second to ensure timely grasp of the outrigger inclination situation.

[0039] In the overall balance module of the whole machine, the difference value of the support forces of the front and rear outriggers is measured in real time by load cells installed at the bottoms of the front and rear outriggers, and the data is collected once every 15 seconds; the symmetry of the transverse movement track is monitored by a laser alignment device. The laser alignment device continuously emits laser signals, and the receiving end receives and analyzes the signals in real time to calculate the track symmetry deviation, and the data is recorded once every 30 seconds.

[0040] According to the historical fault database of this type of bridge-building machine, as well as design standards, construction specifications and other information, the critical threshold ranges of each module are extracted. For the main beam positioning module, the critical threshold of the longitudinal displacement deviation is set at ±50 mm. If it exceeds this range, it may cause the installation position of the box beam to deviate too much, affecting the stress of the bridge structure; the critical threshold of the lateral torsion angle is ±0.5°. Exceeding this angle will cause uneven stress on the main beam, posing a safety hazard; the critical threshold of the track horizontality is ±3 mm / m. Exceeding this range will affect the walking stability of the bridge-building machine.

[0041] In the outrigger synchronization module, the critical threshold of the hydraulic pressure difference is set to ±0.5MPa. Excessive pressure difference will cause uneven force on the outriggers and cause the bridge-building machine to tilt; the critical threshold of the telescopic speed synchronization rate is ±5%. Synchronization rate exceeding this range will lead to uncoordinated outrigger movements; the critical threshold of the vertical tilt angle is ±1°. Too large a tilt angle can easily cause the bridge-building machine to become unstable.

[0042] In the whole machine balance module, the critical threshold of the difference in the support force of the front and rear legs is set to ±10 tons. Exceeding this range will affect the overall balance of the bridge crane; the critical threshold of the symmetry of the transverse track is ±5 mm. Excessive symmetry deviation may cause the bridge crane to become stuck or even derail when it moves laterally. During the entire bridge crane installation process, the data of each module is collected in real time and transmitted to the control system. The control system compares and analyzes the collected data with the critical threshold set at the risk factor layer to provide basic data support for subsequent safety evaluations and ensure that the initial state of the bridge crane installation is safe and controllable.

[0043] Embodiment 2: Divide the multi-dimensional feature intervals of the dynamic parameter layer based on the fuzzy clustering algorithm, and set the fuzzy membership function according to the parameter distribution characteristics of the dynamic parameter layer. The parameters of the dynamic parameter layer have different distribution characteristics. Some parameters may exhibit a normal distribution, while others may be skewed. When setting the fuzzy membership function, these characteristics need to be fully considered. For the displacement deviation parameter, its change may be relatively continuous and have a certain fluctuation range, and a Gaussian-type fuzzy membership function can be used to describe the degree to which it belongs to a certain feature category within different value ranges. For the angle deviation parameter, since its value range is usually limited, a triangular or trapezoidal fuzzy membership function may be more appropriate. Calculate the similarity matrix between different parameters through an iterative optimization algorithm. The iterative optimization algorithm can adopt common genetic algorithms, particle swarm optimization algorithms, etc. Taking the genetic algorithm as an example, first randomly generate a set of initial solutions, which represent the similarity relationships between different parameters. Then, through operations such as selection, crossover, and mutation, continuously optimize these solutions so that the generated similarity matrix can more accurately reflect the internal connections between parameters. In the selection operation, according to the fitness values of the solutions, select the solutions with higher fitness to enter the next generation; the crossover operation exchanges some information between two solutions to generate new solutions; the mutation operation randomly changes some elements in the solutions with a certain probability to prevent the algorithm from falling into a local optimum.

[0044] Perform feature clustering grouping on the dynamic parameter layer according to the similarity matrix. After obtaining the similarity matrix, the parameters can be clustered based on the values in the matrix. A hierarchical clustering algorithm can be used. Starting from each parameter as a separate class, gradually merge the classes with high similarity until a certain clustering stop condition is met. The clustering stop condition can be that the number of classes reaches a preset value, or the similarity within the class reaches a certain threshold. In this way, the dynamic parameter layer is divided into different feature intervals, providing a more targeted data processing basis for subsequent safety evaluations.

[0045] In a large highway bridge construction project, a specific type of highway bridge erecting machine was used for box girder erection work. When conducting a safety evaluation of the initial installation state of the bridge erecting machine, it involved dividing the multi-dimensional feature intervals of the dynamic parameter layer based on the fuzzy clustering algorithm.

[0046] The dynamic parameter layer of this bridge erecting machine includes numerous parameters, such as the longitudinal displacement deviation, lateral torsion angle, and track levelness of the main girder positioning module, and the hydraulic pressure difference, telescopic speed synchronization rate, and vertical tilt angle of the outrigger synchronization module. The change laws of these parameters are different during the installation process of the bridge erecting machine. To achieve a more accurate safety evaluation, it is necessary to reasonably divide their feature intervals.

[0047] First, the fuzzy membership function is set according to the parameter distribution characteristics of the dynamic parameter layer. Taking the longitudinal displacement deviation as an example, through a large amount of collection and analysis of the longitudinal displacement deviation data in previous similar bridge erector installation projects, it is found that the data distribution is approximately normal. Therefore, a Gaussian fuzzy membership function is used to describe the degree to which the longitudinal displacement deviation belongs to different characteristic categories. Let the longitudinal displacement deviation be , and the Gaussian fuzzy membership function can be expressed as , where is the central value of the function, which is determined according to the mean of historical data; is the standard deviation, reflecting the degree of data dispersion. In practical applications, through the calculation of the longitudinal displacement deviation data collected in the early stage of this project, the mean value (unit: millimeter) and the standard deviation are obtained. In this way, when the collected longitudinal displacement deviation is millimeters, substituting it into the function can obtain the membership value belonging to a certain characteristic category, so as to judge the relative degree of the displacement deviation in the safety evaluation.

[0048] For the lateral torsion angle, since its value range is limited and the change is relatively gentle, it is more appropriate to use a triangular fuzzy membership function. Assume that the value range of the lateral torsion angle is , and according to engineering experience and safety standards, it is divided into three characteristic categories: safe range, warning range, and dangerous range. When the angle is in , it is the safe range, and the corresponding triangular fuzzy membership function takes values within this interval as ; in and is the warning range, and the function value linearly decreases from to ; in and is the dangerous range, and the function value is . When the actually collected lateral torsion angle is , through this function, the membership degree in the warning range can be determined, providing a basis for safety evaluation.

[0049] Then, the similarity matrix between different parameters is calculated through an iterative optimization algorithm. Here, the particle swarm optimization algorithm (PSO) is selected. The particle swarm optimization algorithm is an optimization algorithm that simulates the foraging behavior of bird flocks. Each particle represents a potential solution and continuously adjusts its position in the solution space to find the optimal solution. In this application scenario, each particle represents an assumption about the similarity relationship between different parameters.

[0050] When the algorithm is initialized, randomly generate A particle, and the position vector of each particle contains the similarity values of all parameter pairs for which similarity is to be calculated. For example, for the two parameters of longitudinal displacement deviation and hydraulic pressure difference, one element in the particle position vector represents the assumed similarity value between them. At the same time, an initial velocity is assigned to each particle.

[0051] During the iterative process, the particles adjust their velocities and positions based on their own historical optimal positions and the group historical optimal position. At each iteration, the objective function value under the similarity relationship represented by each particle is calculated. The objective function can be defined as the sum of the similarities of the data within the classes being maximized and the sum of the similarities of the data between the classes being minimized after clustering all the collected parameter data according to these similarity relationships. Through continuous iteration, the particles gradually approach the optimal solution. After a certain number of iterations, a relatively stable similarity matrix is obtained. For example, it can be seen from the final similarity matrix that the similarity between the longitudinal displacement deviation and the lateral torsion angle is , indicating that there is a certain correlation in the change trends of these two parameters; while the similarity between the longitudinal displacement deviation and the telescopic speed synchronization rate is , indicating that the correlation between them is relatively weak.

[0052] Finally, feature clustering and grouping are performed on the dynamic parameter layer according to the similarity matrix. The hierarchical clustering algorithm is used to achieve this goal. The hierarchical clustering algorithm is a clustering algorithm based on clusters, which is divided into two methods: agglomerative clustering (bottom-up) and divisive clustering (top-down). Here, agglomerative clustering is adopted.

[0053] At the beginning, each parameter is regarded as a separate class. According to the similarity matrix, the similarity between every two classes is calculated. For example, the similarity between the longitudinal displacement deviation and the lateral torsion angle is relatively high, and they are first merged into a new class. Then, the similarity between the new class and other classes is recalculated, and the classes with high similarity are continuously merged. During the merging process, the information of each merge is recorded to form a clustering dendrogram in a tree structure. When the number of classes reaches the preset value, such as preset to be divided into classes, the clustering process stops. Finally, the dynamic parameter layer is divided into feature intervals, namely the parameter class related to the main girder position (including longitudinal displacement deviation, lateral torsion angle, etc.), the leg hydraulic parameter class (including hydraulic pressure difference, etc.), the leg movement synchronization parameter class (including telescopic speed synchronization rate, etc.), the leg inclination parameter class (including vertical inclination angle, etc.), and other comprehensive parameter classes (including some parameters related to overall safety but with relatively weak correlation with the above categories). Through such feature clustering and grouping, the subsequent analysis and safety evaluation of the dynamic parameter layer are made more targeted and efficient, and can more accurately judge the safety of the initial installation state of the bridge erecting machine.

[0054] Example 3: The grey relational analysis method is used to calculate the relational degree coefficients of each parameter in the risk factor layer, and a benchmark reference sequence for each parameter in the risk factor layer is selected. The risk factor layer contains multiple parameters, and for each parameter, a benchmark reference sequence needs to be determined. This benchmark reference sequence can be an ideal value sequence determined according to historical experience, design standards, or safety specifications. For the stress distribution parameter of the main girder, its benchmark reference sequence can be a sequence composed of the theoretical stress values of each part of the main girder under ideal working conditions. Calculate the relational degree coefficients between each parameter sequence and the benchmark reference sequence. By comparing the differences between the actually collected parameter sequence and the benchmark reference sequence, the relational degree coefficients are calculated. The smaller the difference, the larger the relational degree coefficient, indicating that the parameter is closer to the ideal state and the safety risk is relatively low; on the contrary, the larger the difference, the smaller the relational degree coefficient and the higher the safety risk. When calculating the relational degree coefficients, factors such as the length of the parameter sequence and the trend of data changes need to be considered. Generate a relational degree coefficient matrix according to the mean value of the relational degree coefficients. The relational degree coefficients of each parameter are averaged to obtain a mean value representing the degree of association between the parameter and the benchmark reference sequence. Combining the mean values of the relational degree coefficients of all parameters forms a relational degree coefficient matrix. This matrix can intuitively reflect the degree of association between each parameter in the risk factor layer and the ideal state, providing an important basis for subsequent weight determination and safety evaluation.

[0055] Taking a highway bridge erecting machine used in a certain highway bridge construction project as an example, this bridge erecting machine is used to erect 30-meter prestressed concrete box girders, and its risk factors need to be evaluated in the initial installation state.

[0056] Among the numerous parameters in the risk factor layer, the stress of the main girder is selected as a key parameter for analysis. First, the benchmark reference sequence of the main girder stress needs to be determined. By referring to the design documents of this bridge erecting machine, the theoretical stress values at different positions of the main girder under standard working conditions, that is, when the bridge erecting machine is in a stable installation state and hoisting a box girder with the rated weight, are obtained. For example, at the mid-span position of the main girder, the standard stress value required by the design should be stably maintained at about 200 MPa during the installation process, and at the position near the support point, the stress value is about 120 MPa. Arrange these theoretical stress values in a certain order to form a benchmark reference sequence, assumed to be , where the ellipsis represents the theoretical stress values of other key positions of the main girder.

[0057] Calculate the relational degree coefficients between each parameter sequence and the benchmark reference sequence. During the installation of the bridge erecting machine, using stress sensors installed at key positions of the main girder, actual stress data is collected every certain time interval (such as 5 minutes) to form an actual stress sequence . For example, at a certain moment, the stress at the mid-span of the main girder is collected as 210 MPa, and the stress at the position near the support point is 130 MPa, obtaining the actual stress sequence at this time When calculating the correlation coefficient, the difference between the actual sequence and the corresponding point of the reference sequence needs to be considered. For the mid-span position of the main beam, the difference between the actual stress and the reference stress is , here It represents the difference between the first data in the first actual sequence and the corresponding data in the reference sequence. After calculating the difference of all corresponding points, find the maximum value among all the differences. and minimum value . Introducing the resolution factor (Generally, the value is between 0 and 1, here we take ), calculate the correlation coefficient of each corresponding point according to the correlation coefficient formula. Taking the mid-span position of the main beam as an example, the correlation coefficient , substitute the previously calculated values ​​and assume , ,but In the same way, the correlation coefficient between each point in the actual stress sequence and the corresponding point in the reference sequence is calculated to obtain the correlation coefficient sequence , where the ellipsis represents the correlation coefficient at other positions.

[0058] Generate a correlation coefficient matrix based on the mean of the correlation coefficient. Average the correlation coefficient of each actual stress sequence to obtain the correlation coefficient between the sequence and the reference sequence. For example, for the previous actual stress sequence , whose correlation coefficient sequence The average value of is 0.7, which is the correlation coefficient between the actual stress sequence and the reference sequence. In actual operation, multiple actual stress sequences at different times are collected, such as , Etc., and calculate their correlation coefficients with the reference sequence respectively. Arrange all these correlation coefficients in a certain order to form a correlation coefficient matrix. Assume that after multiple collections and calculations, the correlation coefficient matrix is ​​obtained. , each row in the matrix represents the correlation coefficient between an actual stress sequence and a reference sequence, and each column represents the correlation coefficient corresponding to the data collected at different times. Through this correlation coefficient matrix, we can intuitively understand the correlation between the actual stress of the main beam and the standard stress at different times. The closer the correlation coefficient is to 1, the closer the actual stress is to the stress under the standard working condition, and the safety risk is relatively low; conversely, the smaller the correlation coefficient is, the greater the deviation of the actual stress from the standard working condition is, and the higher the safety risk is.

[0059] Embodiment 4: Determine the objective weights of the parameters in the dynamic parameter layer through the entropy weight method, and perform probability density function conversion on the historical data set of the dynamic parameter layer. The dynamic parameter layer has accumulated a large amount of historical data, which reflects the parameter changes of the bridge erecting machine under different installation conditions.

[0060] Perform probability density function conversion on this historical data, and represent the distribution characteristics of the data in the form of probability. Methods such as kernel density estimation can be used to estimate the probability density function of each parameter. Through the probability density function, the probability of the parameter appearing in different value ranges can be understood. Calculate the information entropy value and the coefficient of variation of each parameter. According to the probability density function, calculate the information entropy value of each parameter. The information entropy value can measure the degree of uncertainty of the parameter data. The larger the information entropy value, the more dispersed the data of the parameter, and the higher its importance in safety evaluation may be.

[0061] Calculate the coefficient of variation, which can reflect the degree of difference between different parameters. Generate the objective weights of the dynamic parameter layer through the normalization of the coefficient of variation. Perform normalization processing on the coefficient of variation so that the sum of the weights of all parameters is 1. In this way, the objective weights of the parameters in the dynamic parameter layer are obtained. The objective weights reflect the relative importance of the parameters in safety evaluation and provide an objective basis for subsequent parameter fusion and safety evaluation.

[0062] In a bridge construction project of a newly built expressway, a specific type of highway bridge erecting machine is used. When conducting a safety evaluation of its initial installation state, it is necessary to determine the objective weights of the parameters in the dynamic parameter layer through the entropy weight method. The implementation process is described in detail below in combination with the actual situation.

[0063] The dynamic parameter layer of this bridge erecting machine involves many parameters, such as the longitudinal displacement deviation of the main girder positioning module, the hydraulic pressure difference of the leg synchronization module, etc. Before determining the objective weights, a large amount of historical data during the installation process of the bridge erecting machine has been accumulated. These data were collected when using the same type of bridge erecting machine in multiple previous similar bridge construction projects, covering the parameter changes under different construction environments and different operation processes. The data volume is rich and representative.

[0064] Process the historical data set of the dynamic parameter layer and convert it into a probability form to reflect the distribution characteristics of the parameters. Taking the longitudinal displacement deviation data as an example, thousands of groups of longitudinal displacement deviation data were collected in past construction records. Organize these data and count the frequencies of different displacement deviation values. For example, during the statistics, it was found that the displacement deviation appeared 200 times in the range of 10 - 20 mm and 150 times in the range of 20 - 30 mm, etc. By calculating the proportion of the frequency of data in each interval to the total data volume, approximately represent the probability of the longitudinal displacement deviation occurring in that interval. In this way, the historical data of the longitudinal displacement deviation is presented in the form of probability, enabling us to intuitively understand the likelihood of different displacement deviation values occurring in actual construction.

[0065] Calculate the information entropy value and coefficient of variation for each parameter. For the longitudinal displacement deviation, calculate the information entropy value based on the probability distribution obtained previously. The information entropy value can be understood as an index to measure the uncertainty or dispersion degree of the data. If the distribution of the longitudinal displacement deviation data is relatively concentrated, that is, most of the data are concentrated in a relatively small range, then its information entropy value is smaller, indicating that the change of this parameter is relatively stable; on the contrary, if the data distribution is relatively dispersed and distributed in a large range, then the information entropy value is larger, meaning that the uncertainty of this parameter is higher. During the calculation process, comprehensively consider factors such as the probability of each displacement deviation interval and the number of intervals to determine the information entropy value. At the same time, calculate the coefficient of variation, which is used to reflect the difference degree between different parameters. When comparing the two parameters of longitudinal displacement deviation and hydraulic pressure difference, determine their coefficient of variation by analyzing their respective probability distributions and the fluctuation range of the data, etc. The larger the coefficient of variation, the more obvious the difference between these two parameters in the safety evaluation and the different impacts on the evaluation results.

[0066] Generate the objective weights of the dynamic parameter layer through coefficient of variation normalization. Process the calculated coefficient of variation of all parameters to make them meet the requirements of weights, that is, the sum of the weights of all parameters is 1. In this process, according to the size of the coefficient of variation of each parameter, allocate according to a certain proportion to determine the objective weight of each parameter in the safety evaluation. For example, after calculation and normalization, the objective weight of the longitudinal displacement deviation is determined to be 0.3, and the objective weight of the hydraulic pressure difference is determined to be 0.25, etc. These objective weights reflect the relative importance of each parameter in the safety evaluation. The larger the objective weight, the more crucial the role of this parameter in evaluating the safety of the initial state of the bridge girder erection machine installation. In the subsequent safety evaluation process, reasonable weighted calculations can be carried out on different parameters based on these objective weights to more accurately evaluate the safety of the initial state of the bridge girder erection machine installation.

[0067] Example 5: The parameter fusion is carried out by combining the correlation coefficient and the entropy weight, and the comprehensive weight of the dynamic parameter layer is generated. The correlation coefficient is used as the subjective weight factor, and the entropy weight is used as the objective weight factor. The correlation coefficient reflects the degree of correlation between the parameter and the ideal state, embodying a certain subjective judgment factor, so it can be used as the subjective weight factor. The entropy weight is calculated based on the objective distribution characteristics of the data, representing the objective importance of the parameter, and is used as the objective weight factor. The Bayesian network model is used to perform probability fusion on the subjective and objective weights, and the comprehensive weight of the dynamic parameter layer is generated through the posterior probability distribution. The Bayesian network model is a graphical model based on probabilistic inference, which can well integrate information from different sources. In the present invention, the correlation coefficient and the entropy weight are used as input information to construct the Bayesian network model. Through the inference algorithm of the Bayesian network, the posterior probability distribution under the given subjective and objective weights is calculated. The comprehensive weight of the dynamic parameter layer is extracted from the posterior probability distribution. The comprehensive weight comprehensively considers subjective and objective factors, more comprehensively reflects the importance of the parameter in safety evaluation, and lays a solid foundation for accurately calculating the risk index and determining the safety level.

[0068] In a city viaduct construction project, a large highway bridge girder erecting machine was used. When conducting a safety evaluation of the initial installation state of the bridge girder erecting machine, the operation process of combining the correlation coefficient and the entropy weight to generate the comprehensive weight of the dynamic parameter layer in this embodiment is as follows.

[0069] The dynamic parameter layer of the bridge girder erecting machine includes key parameters such as the longitudinal displacement deviation and the lateral torsion angle of the main girder positioning module, and the hydraulic pressure difference of the leg synchronization module. In the previous work, the correlation coefficients of each parameter have been calculated through the grey correlation analysis method, and at the same time, the entropy weights of each parameter have been determined by using the entropy weight method.

[0070] Taking the longitudinal displacement deviation as an example, through grey correlation analysis, the correlation coefficient between it and the parameter sequence in the ideal state is 0.75. This means that at the subjective level, the longitudinal displacement deviation has a relatively high degree of correlation with the ideal state and has a certain importance in safety evaluation. The entropy weight of the longitudinal displacement deviation calculated by the entropy weight method is 0.3. The entropy weight is calculated based on the objective distribution characteristics of historical data, reflecting the objective importance of the longitudinal displacement deviation at the data level.

[0071] Taking the correlation coefficient as the subjective weight factor and the entropy weight as the objective weight factor, the Bayesian network model is used to perform probability fusion on the subjective and objective weights. The Bayesian network model is a directed acyclic graph, where the nodes represent random variables and the edges represent the dependence relationships between the variables. In this case, the correlation coefficient and the entropy weight of the longitudinal displacement deviation are used as input nodes, and the corresponding weight nodes of other parameters are used to construct the Bayesian network together.

[0072] When constructing the Bayesian network, based on the professional knowledge and experience in the field of bridge erecting machines, the conditional probability relationships between various nodes are determined. For example, it is known that there is a certain correlation between the longitudinal displacement deviation and the lateral torsion angle during the actual operation of the bridge erecting machine. When the longitudinal displacement deviation is large, the probability of the lateral torsion angle being abnormal also increases. Through a large amount of historical data and expert experience, the conditional probability table between them is determined.

[0073] After the Bayesian network is constructed, the Bayesian inference algorithm is used for calculation. The Bayesian inference algorithm is based on Bayes' theorem and calculates the posterior probability distribution according to the known conditional probabilities and evidence. In this example, through the inference of the Bayesian network, combined with information such as the correlation coefficient and entropy weight of the longitudinal displacement deviation, the posterior probability distribution of the longitudinal displacement deviation considering subjective and objective factors is calculated. The comprehensive weight of the longitudinal displacement deviation is extracted from this posterior probability distribution.

[0074] Suppose that after the calculation of the Bayesian network and the posterior probability analysis, the comprehensive weight of the longitudinal displacement deviation is 0.4. This comprehensive weight no longer depends solely on subjective judgment (correlation coefficient) or simply on the objective characteristics of the data (entropy weight), but integrates the information of both. Compared with using the correlation coefficient or entropy weight alone, the comprehensive weight more comprehensively reflects the actual importance of the longitudinal displacement deviation in safety assessment.

[0075] For other parameters such as the lateral torsion angle and the hydraulic pressure difference, in the same way, their respective correlation coefficients and entropy weights are input into the Bayesian network for probability fusion to obtain their comprehensive weights. In this way, comprehensive weights are generated for each parameter in the dynamic parameter layer. These comprehensive weights play a key role in subsequent calculations of the risk indices of each structural layer of the bridge erecting machine and the total risk score, and can more accurately determine the safety level of the initial installation state of the highway bridge erecting machine, providing strong support for ensuring the safe installation of the bridge erecting machine.

[0076] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or also includes elements inherent to such process, method, article or device.

[0077] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A safety evaluation method for the initial installation state of a highway bridge erecting machine, characterized in that, It includes the following steps: S1. Establish a multi-modal safety evaluation model for a highway bridge girder erecting machine, including a structure layer, a dynamic parameter layer, and a risk factor layer; S2. Divide the multi-dimensional feature intervals of the dynamic parameter layer based on the fuzzy clustering algorithm; S3. Calculate the correlation degree coefficients of each parameter in the risk factor layer using the grey relational analysis method; S4. Determine the objective weights of each parameter in the dynamic parameter layer through the entropy weight method to obtain the entropy weight; S5. Combine the correlation degree coefficients and the entropy weight to perform parameter fusion to generate the comprehensive weight of the dynamic parameter layer; S6. Collect multi-source sensing data during the installation process of the bridge girder erecting machine in real time and perform discretization processing; S7. Calculate the risk index layer by layer and determine the safety level of the initial state of the highway bridge girder erecting machine installation; The calculation of the risk index includes: Multiply the comprehensive weight of the dynamic parameter layer by the discretized data to generate the risk index of each structure layer; Generate the total risk score by superimposing the preset structure layer weight coefficient and the risk index, and divide the safety level based on the score interval.

2. The safety evaluation method for the initial installation state of a highway bridge erecting machine according to claim 1, characterized in that, In the step S1: The structure layer is divided into a main girder positioning module, a leg synchronization module, a support bearing module, and an overall balance module based on the physical topological relationship of the bridge girder erecting machine installation components; The dynamic parameter layer collects displacement deviation, angle deviation, and stress distribution parameters of each module through a sensor network; The risk factor layer extracts the critical threshold intervals of each module based on the historical failure database.

3. The safety evaluation method for the initial installation state of a highway bridge erecting machine according to claim 1, characterized in that, The steps of dividing the feature intervals based on the fuzzy clustering algorithm in the step S2 include: Set the fuzzy membership function according to the parameter distribution characteristics of the dynamic parameter layer; Calculate the similarity matrix between different parameters through the iterative optimization algorithm; Perform feature clustering grouping on the dynamic parameter layer according to the similarity matrix.

4. The safety assessment method for the initial installation state of a highway bridge erecting machine according to claim 1, characterized in that, The implementation steps of the grey relational analysis method in the step S3 include: Select the reference sequence for each parameter in the risk factor layer; Calculate the correlation coefficient between each parameter sequence and the reference sequence; Generate the correlation degree coefficient matrix according to the mean value of the correlation coefficients.

5. The safety evaluation method for the initial installation state of a highway bridge erecting machine according to claim 1, wherein The steps of determining the objective weight by the entropy weight method in the step S4 include: Perform probability density function conversion on the historical data set of the dynamic parameter layer, calculate the information entropy value and the difference coefficient of each parameter, and generate the objective weight of the dynamic parameter layer through the normalization of the difference coefficient.

6. The safety evaluation method for the initial installation state of a highway bridge erecting machine according to claim 1, characterized in that The steps of parameter fusion in the step S5 include: Use the correlation degree coefficient as the subjective weight factor and the entropy weight as the objective weight factor; Adopt the Bayesian network model to perform probability fusion on the subjective and objective weights, and generate the comprehensive weight of the dynamic parameter layer through the posterior probability distribution.

7. A safety evaluation method for the initial installation state of a highway bridge erecting machine according to claim 1, characterized in that, The discretization processing in the step S6 includes: Perform abnormal fluctuation detection and noise filtering on the multi-source sensing data, convert the data with different dimensions into a unified discrete interval through the range normalization method, and generate the discretized distribution map of the dynamic parameters according to the time series.

8. The safety evaluation method for the initial installation state of a highway bridge erector according to claim 2, wherein The dynamic parameter layer includes at least one of the following parameters: The longitudinal displacement deviation, lateral torsion angle, and track levelness of the main girder positioning module; The hydraulic pressure difference, telescopic speed synchronization rate, and vertical tilt angle of the leg synchronization module; The difference value of the front and rear leg support forces and the symmetry degree of the transverse movement track of the overall balance module.

9. A safety evaluation method for the initial installation state of a highway bridge erecting machine according to claim 7, characterized in that, The steps of generating the discretized distribution map include: The standardized sensing data is segmented into equal-length data segments according to a time window. The wavelet packet decomposition algorithm is used to extract the frequency-domain energy features of each data segment, and the Mahalanobis distance is used to calculate the feature difference matrix of data segments in different time windows.

Images