An intelligent analysis method and system for construction deviation of an aluminum alloy structure

Abstract

The application relates to the technical field of building engineering information, and discloses an aluminum alloy structure construction deviation intelligent analysis method and system. The method comprises the following steps: acquiring structure design parameter sets of an aluminum alloy target component, establishing a node freedom degree matrix and a component coordinate reference under a standard temperature; extracting a component length, combining material parameters, node constraints and temperature differences to calculate a thermal deviation vector; obtaining a measured deviation vector according to measured coordinates and theoretical coordinates of measuring points, performing confidence weighted filtering after deducting the thermal deviation vector, and generating a net assembly deviation matrix; constructing a node constraint propagation model based on the node freedom degree matrix and topological mapping, inputting the net assembly deviation matrix into the model, and inversely calculating the probability of multiple error sources through Bayesian inference; and generating a component-level correction priority list by comprehensively considering error probability, adjustable margin, rework cost and a structure risk threshold value, and updating the error probability by using retest data in a closed loop. The application improves the controllability and accuracy of aluminum alloy structure construction quality.

Description

Technical Field

[0001] This application relates to the field of information technology in building engineering, and in particular to an intelligent analysis method and system for construction deviations of aluminum alloy structures. Background Technology

[0002] Large-scale aluminum alloy structures are increasingly used in large-span buildings such as stadiums, airport terminals, and convention centers due to their advantages of being lightweight, high-strength, and corrosion-resistant. However, the coefficient of linear expansion of aluminum alloy is approximately twice that of ordinary steel, making it extremely sensitive to temperature changes. In actual construction, factors such as ambient temperature fluctuations and solar radiation can cause significant thermal expansion and contraction deformation of components. The positional deviation data obtained from on-site measurements are often a complex result of the coupling of actual installation errors, thermal deformation, and measurement noise, making it difficult to directly use to judge construction quality.

[0003] Traditional deviation analysis methods typically compare measured coordinates directly with design coordinates, ignoring the interference of thermal deformation caused by temperature. This makes it difficult for construction workers to distinguish between "true installation misalignment" and "normal thermal expansion and contraction," often leading to a situation where adjustments only exacerbate the misalignment: unnecessary cutting of components that have elongated due to high-temperature expansion, or forcibly pulling on nodes with increased gaps due to low-temperature contraction, not only causes rework costs to skyrocket but may also damage the initial stress state of the structure, introducing new safety hazards. Furthermore, existing deviation analyses mostly remain at the level of detecting deviations, lacking the ability to trace the root cause. When a node shifts, it is difficult for on-site personnel to accurately determine whether it is caused by hole machining misalignment, support elevation error, bolt connection gap, or uneven preload, resulting in a lack of targeted corrective measures.

[0004] Therefore, how to isolate temperature interference from mixed field measurement data, restore the true assembly deviation, and deduce the probability distribution of the underlying error sources, and then formulate a scientific and economical correction priority, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the aforementioned technical issues, this application provides an intelligent analysis method and system for construction deviations in aluminum alloy structures, which aims to improve the controllability, accuracy, and intelligence level of aluminum alloy structure construction quality.

[0006] Firstly, this application provides an intelligent analysis method for construction deviations in aluminum alloy structures, the method comprising: Step S1: Obtain the structural design parameter set of the aluminum alloy target component, and establish the nodal degree of freedom matrix and component coordinate reference at standard temperature based on the structural design parameter set. The component coordinate reference includes the theoretical coordinates of each preset measuring point on each target component. Step S2: Extract the component length from the set of structural design parameters. Based on the aluminum alloy material parameters, component length, node degree of freedom matrix, and temperature difference between the component surface temperature and the design temperature in the set of structural design parameters, calculate the theoretical thermal expansion component of each measuring point on the target component and generate a thermal deviation vector. Step S3: Determine the measured deviation vector of each measuring point based on the measured coordinates and the theoretical coordinates of each measuring point, and obtain the initial assembly deviation vector by combining the thermal deviation vector. Then, according to the confidence level of each measuring point, perform weighted filtering on the initial assembly deviation vector to generate the net assembly deviation matrix. Step S4: Construct a node constraint propagation model based on the topological mapping relationship between the node degree of freedom matrix and the component end deviation. Input the net assembly deviation matrix into the node constraint propagation model and obtain the probability of various construction error sources through Bayesian inference inversion. Step S5: Based on the probability of error sources of various construction types, combined with adjustable margin, rework cost and structural risk threshold, generate a component-level correction priority list, obtain the corrected retest data, and update the error probability of various construction types based on the retest data.

[0007] Secondly, this application provides an intelligent analysis system for construction deviations of aluminum alloy structures, the system comprising: The acquisition module is used to acquire the structural design parameter set of the aluminum alloy target component, and to establish the nodal degree of freedom matrix and component coordinate reference at standard temperature based on the structural design parameter set. The component coordinate reference includes the theoretical coordinates of each preset measuring point on each target component. The calculation module is used to extract the component length from the structural design parameter set, and calculate the theoretical thermal expansion component of each measuring point on the target component based on the aluminum alloy material parameters, component length, node degree of freedom matrix and temperature difference between the component surface temperature and the design temperature in the structural design parameter set, and generate a thermal deviation vector. The generation module is used to determine the measured deviation vector of each measuring point based on the measured coordinates and the theoretical coordinates of each measuring point, and to obtain the initial assembly deviation vector by combining the thermal deviation vector. The initial assembly deviation vector is then weighted and filtered according to the confidence level of each measuring point to generate the net assembly deviation matrix. The inversion module is used to construct a node constraint propagation model based on the topological mapping relationship between the node degree of freedom matrix and the component end deviation. The net assembly deviation matrix is ​​input into the node constraint propagation model, and the probability of various construction error sources is obtained through Bayesian inference inversion. The calibration module is used to generate a component-level calibration priority list based on the probability of error sources of various construction types, combined with adjustable margin, rework cost and structural risk threshold, obtain the calibrated retest data, and update the error probability of various construction types based on the retest data.

[0008] Compared with the prior art, the beneficial effects of this application are at least as follows: First, by establishing the component coordinate reference and node degree-of-freedom matrix at standard temperature, and accurately calculating the thermal deviation vector based on the metal linear expansion coefficient, measured temperature difference, and node constraints, the thermal deviation vector is deducted from the measured total deviation. This achieves precise decoupling between temperature interference and actual construction assembly error in aluminum alloy structure construction deviation, and solves the deviation distortion problem caused by the inability of traditional methods to distinguish between normal thermal deformation and actual misalignment.

[0009] Secondly, by introducing measurement point confidence levels and applying graph Fourier domain weighted smoothing filtering to the initial assembly deviation vector, high-frequency measurement noise is effectively filtered out while retaining the true local defects at high-confidence measurement points, significantly improving the purity and reliability of the net assembly deviation matrix. Thirdly, a node constraint propagation model is constructed based on the node degree-of-freedom matrix and the topological mapping relationship of error sources. By combining Bayesian inference with Markov chain Monte Carlo sampling, the posterior probability distribution of hidden error sources such as hole offset, connection gap, support elevation, and uneven preload is inversely derived from apparent deviations, overcoming the technical bottleneck of traditional methods' inability to trace the root cause of errors.

[0010] Finally, by comprehensively considering the probability of error sources, structural stress increment, adjustable margin, and rework cost, a correction urgency index is calculated using normalized weighting, and a component-level correction priority list is generated. Combined with retest data, a closed-loop dynamic update is achieved, ensuring that correction resources are prioritized for key components with the highest risk, easiest adjustment, and lowest cost. This significantly reduces rework costs and construction blindness while ensuring structural safety, and improves the intelligent control level of aluminum alloy structure construction quality. Attached Figure Description

[0011] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a flowchart of an intelligent analysis method for construction deviations of aluminum alloy structures in an embodiment of this application; Figure 2 This is a schematic diagram of the design grid and node connection relationship of the aluminum alloy structure in the embodiments of this application. Figure 3 This is a schematic diagram of an intelligent analysis system for construction deviations of aluminum alloy structures according to an embodiment of this application. Detailed Implementation

[0013] This application provides an intelligent analysis method and system for construction deviations in aluminum alloy structures. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0014] For ease of understanding, the specific process of the embodiments of this application is described below. Please refer to [link / reference]. Figure 1 The intelligent analysis method for construction deviation of aluminum alloy structures in this application includes: Step S1: Obtain the structural design parameter set of the aluminum alloy target component, and establish the nodal degree of freedom matrix and component coordinate reference at standard temperature based on the structural design parameter set. The component coordinate reference contains the theoretical coordinates of each preset measuring point on each target component.

[0015] Step S1 includes: obtaining the structural design parameter set of the aluminum alloy target component, including the design grid, aluminum alloy material parameters, component cross-section, node constraints, and design temperature; determining the physical property characteristics of the target component based on the aluminum alloy material parameters and component cross-section; determining the translational and rotational degrees of freedom of each connection node based on the node constraints, and constructing the node degree of freedom matrix; using the design temperature as the reference temperature, calculating the theoretical spatial coordinates of each target component at the reference temperature based on the design grid and physical property characteristics, and generating the component coordinate reference.

[0016] Specifically, the structural design parameter set of the target aluminum alloy component is obtained. This parameter set is the starting point for the entire deviation analysis and includes the design grid, aluminum alloy material parameters, component cross-sections, node constraints, and design temperature. The design grid defines the macroscopic geometric skeleton of the building in three-dimensional space, consisting of a series of intersecting principal and secondary axes, forming a rigorous coordinate reference system. The aluminum alloy material parameters include elastic modulus, shear modulus, Poisson's ratio, and coefficient of linear expansion, which reflect the inherent physical response of the material under stress or heat. The component cross-section data describes the geometric shape and dimensional characteristics of the aluminum alloy member, such as I-shaped, box-shaped, or circular tube cross-sections. Based on these cross-sectional dimensions, the cross-sectional area, moment of inertia, and other geometric and mechanical properties can be further calculated. Node constraints reflect the connection method between components or between components and supports, specifically indicating the restriction or release state of each connection point in the six degrees of freedom in space, namely three translational degrees of freedom and three rotational degrees of freedom; design temperature refers to the ambient temperature of the structure under the assumed closed or reference state during the design stage, usually taken as the local annual average temperature or the value recommended by the construction code, and all subsequent calculations of thermal deformation are based on this temperature as the zero point reference.

[0017] After obtaining the aforementioned set of structural design parameters, the physical properties of the target component are first determined based on the aluminum alloy material parameters and the component's cross-section. Specifically, the axial compressive stiffness of the component is obtained by multiplying the elastic modulus by the cross-sectional area, and the bending stiffness is obtained by multiplying the elastic modulus by the moment of inertia of the cross-section. These physical properties characterize the component's ability to resist deformation under external loads or temperature changes, and they will be encapsulated as stiffness parameters required for subsequent finite element analysis. Next, the translational and rotational degrees of freedom of each connection node are determined based on the nodal constraints, and a nodal degree of freedom matrix is ​​constructed. For fully fixed rigid nodes, all three translational and three rotational degrees of freedom are restricted; for hinged nodes, only the translational degrees of freedom are restricted while the rotational degrees of freedom are released; for sliding supports, the translational degrees of freedom are released on specific coordinate axes according to the installation direction. By digitally encoding the constraint states of all nodes, a sparse nodal degree of freedom matrix is ​​formed. This matrix reflects the boundary support conditions and internal force transmission paths of the entire structure and is the core mathematical tool for subsequently assembling the overall stiffness matrix and solving the displacement field. Finally, using the design temperature as the reference temperature, the theoretical spatial coordinates of each target component at the reference temperature are calculated based on the design grid and physical property characteristics, generating a component coordinate reference. Specifically, the topological intersections of the design grid are used as initial node coordinates. Utilizing the determined physical property characteristics and node degree-of-freedom matrix, the finite element equilibrium equations are solved under the condition of considering only the structure's self-weight and zero temperature difference, obtaining the elastic deflection of each node under its own weight. The initial coordinates of the design grid are then superimposed with this elastic deflection to obtain the theoretical spatial coordinates of each target component at the reference temperature. This component coordinate reference also pre-sets the theoretical coordinates of each measuring point on each target component. These measuring points are key feature points selected on the component surface during subsequent field measurements, such as component ends, node centers, or mid-span points. The theoretical position of each measuring point in the global coordinate system is accurately recorded as an absolute reference for subsequent deviation comparisons.

[0018] The component coordinate datum and nodal degree-of-freedom matrix established through step S1 provide a digital reference system unaffected by ambient temperature for the entire intelligent analysis of construction deviations. This solves the fundamental problem of traditional methods failing to effectively isolate thermal deformation due to the lack of a unified datum, making it possible to subsequently subtract thermal deviations from field measurement data and invert error sources. By establishing an ideal model under standard temperature, the datum drift caused by the temperature difference between the design temperature and the construction environment is eliminated, ensuring that all subsequent deviation comparisons are performed under the same thermodynamic state, thereby significantly improving the accuracy and reliability of error identification.

[0019] Step S2: Extract the component length from the structural design parameter set. Based on the aluminum alloy material parameters, component length, node degree of freedom matrix, and temperature difference between the component surface temperature and the design temperature in the structural design parameter set, calculate the theoretical thermal expansion components of each measuring point on the target component and generate a thermal deviation vector.

[0020] Step S2 includes: calculating the temperature difference of the target component based on the surface temperature of the target component and the design temperature; calculating the theoretical thermal expansion component of the target component under unconstrained state based on the metal linear expansion coefficient, component length and temperature difference in the aluminum alloy material parameters; using the theoretical thermal expansion component as the initial thermal strain load, and combining the constraint boundary conditions defined by the nodal degree of freedom matrix, obtaining the thermal expansion component of each measuring point in three orthogonal directions by solving the structural thermoelastic equilibrium equation, and combining the thermal expansion components of all measuring points in a preset order to generate a thermal deviation vector.

[0021] Specifically, after establishing the component coordinate reference and nodal degree-of-freedom matrix at standard temperature, the aluminum alloy components will undergo significant thermal expansion and contraction deformation due to the fact that the actual temperature at the construction site often deviates from the design temperature. Directly comparing the measured coordinates with the theoretical coordinates will introduce a large amount of thermal interference. Therefore, it is necessary to first quantify the thermal deformation component caused by temperature so that it can be subtracted from the total deviation to restore the true assembly deviation.

[0022] First, the stress-free geometric length of each target component at the design temperature is extracted from the structural design parameter set or design grid. This length reflects the theoretical original size of the component without considering any external loads or temperature changes, and is the basis for subsequent calculations of thermal expansion and contraction. Simultaneously, the surface temperature of the component measured at the construction site and the design temperature are obtained to calculate the temperature difference of the target component. In actual construction, due to factors such as solar radiation and shading, the temperature of different parts of the same component may differ. Therefore, temperature measurement points are usually arranged on both the sun-facing and shaded sides of the target component. The surface temperatures of multiple measurement points are weighted and averaged to obtain the average body temperature of the component. The design temperature is then subtracted from the average body temperature to obtain the temperature difference of the target component. A positive temperature difference indicates thermal expansion of the component, while a negative value indicates contraction upon cooling.

[0023] Then, based on the pre-stored linear expansion coefficient of the aluminum alloy material parameters, the previously extracted component length, and the calculated temperature difference, the theoretical thermal expansion and contraction of the target component under unconstrained conditions is calculated. The linear expansion coefficient is an inherent physical constant of aluminum alloy materials, reflecting the elongation or shortening ratio of a unit length of material per unit temperature change. According to the law of linear thermal expansion, multiplying the linear expansion coefficient, component length, and temperature difference yields the absolute expansion and contraction length of the component when it is completely free and unconstrained at both ends. This theoretical thermal expansion and contraction represents the amount of natural deformation that the component "attempts" to occur under the current temperature difference. However, because the two ends of the component in the actual structure are constrained by adjacent components or supports, this free deformation cannot be fully realized, but is instead transformed into internal forces and coordinated deformation of the structure.

[0024] Therefore, it is necessary to consider the above-mentioned theoretical free thermal expansion and contraction. As the initial thermal strain load, combined with the constraint boundary conditions defined by the nodal degree-of-freedom matrix, the thermal expansion and contraction components in three orthogonal directions at each measuring point are obtained by solving the thermoelastic equilibrium equation of the structure. Specifically, the theoretical free thermal expansion and contraction of each component is first converted into an equivalent nodal force vector. For example, for a rod element, the equivalent nodal force can be expressed as: ,in, Let A be the elastic modulus, A be the cross-sectional area, and n be the cosine vector of the axial direction of the component, which includes three components. , , , representing the direction cosine values ​​of the angles between the component's axial direction (length direction) and the positive directions of the three coordinate axes X, Y, and Z of the global coordinate system, respectively. Let K be the length of the component; then, based on the physical properties of each component, such as axial stiffness, bending stiffness, and node topology, i.e., the connection relationship between each component and its two end nodes, the overall stiffness matrix K is assembled.

[0025] Before assembly, the constraint state of each node in space (whether the translational and rotational degrees of freedom are restricted) is determined based on the nodal degree-of-freedom matrix. Assembly begins at the component level: for each component in the structure, its cross-sectional area A, elastic modulus E, and moment of inertia are first determined. and length An element stiffness matrix is ​​established in the component's local coordinate system. This matrix describes the force and displacement relationships at the two end nodes of the component in six degrees of freedom, specifically composed of axial linear stiffness. Lateral translational stiffness Rotational stiffness coefficient It is composed of various elements. Subsequently, the axial direction cosine vector of this component is used. , A coordinate transformation matrix T is constructed to convert the element stiffness matrix in the local coordinate system to the element stiffness matrix in the global coordinate system. After calculating the global element stiffness matrix of all components, the stiffness coefficients of each component in the element stiffness matrix of the global coordinate system corresponding to the degrees of freedom are superimposed onto the corresponding positions of the global stiffness matrix K according to the node topology (i.e., the node numbers connected to both ends of each component) and the node degree of freedom numbers, forming a complete global stiffness matrix. This assembly process essentially accumulates the stiffness contributions of all structural components according to the degree of freedom correspondence, ultimately obtaining a sparse matrix K describing the force and displacement relationship of the entire structure in the global coordinate system, with a dimension of N×N, where N is the total number of degrees of freedom of all nodes in the structure. After the structural assembly of the global stiffness matrix K is completed, zero displacement constraints (i.e., constraint boundary conditions) are applied to the specified degree of freedom directions according to the node degree of freedom matrix. Subsequently, the linear equation system is solved. This yields the thermal displacement of each node in the X, Y, and Z directions in the global coordinate system. Finally, the corresponding displacement component of the preset measuring point on each target component is extracted from the thermal displacement of each node, which is the thermal expansion and contraction component of that measuring point.

[0026] For example, when an aluminum alloy beam fixed at both ends is heated, although its free thermal expansion and contraction... The thermal displacement is large, but due to the constraints at both ends, the actual calculated thermal displacement may be close to zero in the axial direction, while a bending bulge will occur in the vertical direction. This bending component is accurately calculated and distributed to the vertical degrees of freedom of each measuring point on the beam. For each pre-set measuring point on each target component, the thermal expansion and contraction components in three directions are recorded according to their relative positions on the component, such as ends, mid-span, quarter points, etc. Finally, the thermal expansion and contraction components of all measuring points are combined in a preset order, such as ascending order of node numbers, to generate a high-dimensional thermal deviation vector. This thermal deviation vector fully describes the spatial displacement of each measuring point in the entire aluminum alloy structure due to temperature effects under the current temperature difference field and node constraints.

[0027] By integrating the thermophysical properties of aluminum alloy materials with structural topological constraints through the above steps, the technical challenge of distinguishing between "normal thermal deformation" and "actual installation error" using traditional methods is solved. This allows subsequent error source inversion to be based on pure assembly deviation data, significantly improving the accuracy of error location. Furthermore, by introducing constraint mapping of the nodal degree-of-freedom matrix into the calculation of thermal expansion, misjudgments of deviations caused by simply using free thermal expansion are avoided, ensuring the physical authenticity and engineering practicality of deviation analysis in thermo-mechanical coupling scenarios.

[0028] Step S3: Determine the measured deviation vector of each measuring point based on the measured coordinates and theoretical coordinates of each measuring point, and obtain the initial assembly deviation vector by combining it with the thermal deviation vector. Then, according to the confidence level of each measuring point, perform weighted filtering on the initial assembly deviation vector to generate the net assembly deviation matrix.

[0029] Step S3 includes: calculating the coordinate difference between the measured coordinates and the component coordinate reference, generating a measured deviation vector, performing vector subtraction on the measured deviation vector and the thermal deviation vector to obtain an initial assembly deviation vector; determining whether the confidence level of each measuring point is greater than a preset confidence level threshold, and if so, using the measuring point confidence level as a weighting coefficient to perform weighted smoothing filtering on the initial assembly deviation vector to filter out high-frequency measurement noise data; and constructing a net assembly deviation matrix based on the filtered initial assembly deviation vector.

[0030] Specifically, after calculating the thermal deviation vector, the results of error inversion using the initial assembly deviation are unstable because the field measurement data still contains random noise from the measuring instruments themselves and high-frequency interference from environmental disturbances. Therefore, it is necessary to perform weighted filtering on the data based on the confidence level of each measuring point to extract the net assembly deviation that reflects the actual construction quality.

[0031] First, the measured coordinates of each measuring point at the construction site are obtained using a total station or 3D laser scanner. Since the measuring instrument is set up and oriented differently each time, these measured coordinates are usually located in a local station coordinate system with the instrument center as the origin. However, the component coordinate reference established in step S1 is located in the global absolute coordinate system of the entire building project. The origins and coordinate axes of the two systems are different, making direct coordinate difference calculation impossible. To solve this problem, several high-precision common control points need to be pre-established at the construction site. These control points are typically selected in locations free from construction interference and structurally stable, such as forced-alignment observation piers fixed to independent concrete foundations or permanent prism targets. The theoretical coordinates of each common control point in the global coordinate system are precisely known beforehand through a high-level control network. Simultaneously with each measurement of the component's measured coordinates, the same total station is used to measure the measured coordinates of these common control points in the current local station coordinate system. Since the same common control point has a set of coordinate values ​​in both the global coordinate system and the local station coordinate system, for three or more non-collinear common control points, a coordinate transformation mathematical model can be established, including three translation parameters, three rotation parameters, and one scale parameter. Solving this transformation model using the least squares method yields the coordinate transformation matrix from the local station coordinate system to the global coordinate system. Subsequently, this coordinate transformation matrix is ​​applied to the measured coordinates of all components, accurately mapping the measured coordinates of each measuring point from the local station coordinate system to the same global absolute coordinate system as the component coordinate datum, thus ensuring the consistency of the spatial datum for subsequent coordinate difference calculations. After coordinate unification, for each measuring point, the difference between its measured coordinates and the theoretical coordinates of the corresponding measuring point in the component coordinate datum is calculated, yielding the coordinate differences in the X, Y, and Z directions. These differences from all measuring points are arranged in a preset order to generate a measured deviation vector, which comprehensively reflects the total displacement under the combined effects of temperature deformation, construction errors, and measurement noise.

[0032] Next, a vector subtraction operation is performed between the measured deviation vector and the thermal deviation vector generated in step S2. Since the two vectors have the same dimension and the same measurement point arrangement order, the elements at corresponding positions are directly subtracted, and the thermal deformation component caused by temperature can be deducted from the total deviation to obtain the initial assembly deviation vector. This initial assembly deviation vector has been stripped of temperature interference and mainly retains the construction assembly error and measurement noise.

[0033] However, the initial assembly deviation vector still contains high-frequency random noise introduced during the measurement process. This noise may originate from factors such as instrument vibration, atmospheric refraction changes, and uneven quality of reflective patches. To filter out this noise, a weighted filtering method is needed using the confidence level of each measurement point. The confidence level of a measurement point is a value between 0 and 1, which comprehensively reflects the reliability of the measurement. It is determined by a weighted calculation of multiple factors. The closer the measurement distance, the more positive the laser incident angle, and the higher the nominal accuracy of the instrument, the higher the confidence level. Regarding environmental stability, the confidence level decreases accordingly when the real-time wind speed exceeds five meters per second, the temperature change rate exceeds two degrees Celsius per hour, or the ground vibration amplitude exceeds 0.1 millimeters per second squared. Each factor is mapped to a score between zero and one, and the comprehensive confidence level is obtained by multiplication or weighted harmonic averaging. The system pre-sets a confidence threshold. For each measurement point, it checks whether its confidence level is greater than the threshold. Only measurement point data with a confidence level greater than the threshold is considered to be of acceptable quality and can participate in subsequent filtering processing. Measurement points with a confidence level lower than the threshold are either discarded or supplemented by interpolation using neighboring high-confidence measurement points. For measurement points with acceptable confidence levels, their confidence levels are used as weighting coefficients to perform weighted smoothing filtering on the initial assembly deviation vector.

[0034] The weighted smoothing filtering process for the initial assembly deviation vector includes: obtaining the high-frequency and low-frequency components in the initial assembly deviation vector; attenuating the high-frequency components according to the weighting coefficients; retaining the low-frequency components; obtaining the smoothed assembly deviation vector after filtering out high-frequency measurement noise data; and constructing the net assembly deviation matrix based on the smoothed assembly deviation vector.

[0035] Specifically, the core of weighted smoothing filtering lies in utilizing graph signal processing theory to treat the gradually changing low-frequency components in the initial assembly deviation vector as the true assembly deviation trend, while treating the drastically jumping high-frequency components as measurement noise. First, a graph structure is constructed based on the design grid and node connection relationships of the aluminum alloy structure, such as... Figure 2The diagram shows the design grid and node connection relationships of an aluminum alloy structure. The nodes correspond to various measuring points in the structure, and the edges correspond to the connections between these measuring points. Specifically, if two measuring points are located on the same component and are adjacent along the component's axis (e.g., end and mid-span, mid-span and another end), then there is an edge between them. Measuring points on different components are not directly connected to each other to avoid introducing non-physical coupling relationships. The Laplace matrix is ​​calculated based on this diagram. The Laplace matrix is ​​a square matrix with dimensions equal to the total number of measuring points. Diagonal elements represent the number of measuring points connected to that point, while off-diagonal elements are negative when there is a connection between corresponding measuring points, and zero otherwise. Eigenvalue decomposition of the Laplace matrix yields a series of eigenvalues ​​and corresponding eigenvectors. Smaller eigenvalues ​​correspond to low-frequency basis functions in the spatial domain, representing smooth and continuous spatial deformation modes throughout the structure; larger eigenvalues ​​correspond to high-frequency basis functions, representing locally abrupt deformation modes.

[0036] Next, the initial assembly deviation vector is considered as a signal defined on the nodes of the graph structure. This signal is projected onto the orthogonal basis formed by the aforementioned eigenvectors using a graph Fourier transform, yielding a set of frequency domain coefficients. Based on a preset spatial cutoff frequency, which can be determined according to the average distance between adjacent measuring points in the structure (e.g., 0.5 times the reciprocal of the average distance), the frequency domain coefficients are divided into low-frequency and high-frequency components. The low-frequency component corresponds to the low-frequency component in the initial assembly deviation vector, reflecting the continuous deformation trend caused by systematic construction errors such as overall support settlement and excessively long component fabrication. The high-frequency component corresponds to the high-frequency component in the initial assembly deviation vector, mainly reflecting random noise during the measurement process and isolated outliers.

[0037] In the filtering process, high-frequency components need to be adaptively attenuated based on the confidence level of each measurement point. The confidence level of a measurement point is a pre-calculated dimensionless value between 0 and 1. It is already a normalized index and can therefore be directly used as a weighting coefficient without further normalization. Specifically, a weighting coefficient is constructed for each measurement point, and this coefficient is equal to the confidence level of that measurement point. For each frequency component in the high-frequency components, its corresponding frequency domain coefficient is multiplied by an attenuation factor, which is related to the weighting coefficients of each measurement point involved in that frequency component. One implementation method is to calculate the weighted average weight of the components of the corresponding eigenvector at each measurement point for each high-frequency component, and then multiply this weighted average weight by a preset attenuation intensity coefficient to control the degree of attenuation. For measurement points with higher confidence levels, the attenuation factor is closer to 1, meaning that high-frequency components are preserved to a greater extent. This prevents genuine severe local defects at high-confidence measurement points, such as severe bending of a member causing node misalignment, from being mistakenly identified as noise and filtered out. For measurement points with lower confidence levels, but still above the threshold, the corresponding attenuation factor is smaller, and high-frequency components are significantly attenuated or even reduced to zero, thus effectively suppressing noise. Low-frequency components are completely preserved without any attenuation processing to ensure that the overall deformation trend of the structure is not affected.

[0038] After weighted attenuation of the high-frequency components, the retained low-frequency components are added to the attenuated high-frequency components to obtain the processed frequency domain coefficients. Then, a graphical inverse Fourier transform is performed to remap the frequency domain coefficients back to three-dimensional physical space, resulting in a smoothed assembly deviation vector. In this smoothed assembly deviation vector, high-frequency jumps caused by measurement random noise are effectively suppressed, while the continuous deformation trend caused by actual construction assembly errors is fully preserved. Finally, the smoothed assembly deviation vector is rearranged into a matrix according to node number and degree of freedom direction. For measurement points that are rejected due to confidence levels below a threshold, the smoothed deviations of surrounding high-confidence measurement points are used to fill in the gaps using a spatial interpolation algorithm, ensuring the integrity of the matrix and thus constructing the net assembly deviation matrix.

[0039] The above steps address the problems of traditional filtering methods in structural deviation analysis, such as their inability to distinguish between real local defects and measurement noise, and their tendency to over-smooth, leading to the loss of key information. By introducing a weighted attenuation mechanism based on measurement point confidence, high-frequency random noise is effectively suppressed while retaining real anomalies at high-confidence measurement points. This ensures that the net assembly deviation matrix is ​​both clean and accurate, providing high-quality data input for the subsequent accurate inversion of the node constraint propagation model. This process significantly improves the sensitivity of deviation analysis in identifying real construction defects, avoiding the omission of key error sources due to over-filtering.

[0040] Step S4: Construct a node constraint propagation model based on the topological mapping relationship between the node degree of freedom matrix and the component end deviation. Input the net assembly deviation matrix into the node constraint propagation model and obtain the probability of various construction error sources through Bayesian inference inversion.

[0041] The step S4 of constructing the node constraint propagation model includes: obtaining an error source space containing hole offset, connection gap, support elevation, and uneven preload; establishing a positive mapping rule from error source parameters to component end displacement or rotation based on the topological mapping relationship between each error source in the error source space and the component end deviation; assembling an overall stiffness matrix containing all component stiffness and node constraints based on the node degree of freedom matrix, and constructing the node constraint propagation model in combination with the positive mapping rule.

[0042] Specifically, following the net assembly deviation matrix, this matrix only shows the spatial deviation values ​​of each measuring point after eliminating temperature interference and measurement noise, but it cannot directly answer "what caused these deviations". In order to trace the root cause of underlying construction errors from apparent deviations, it is necessary to construct a physical model that can simulate how various error sources are transmitted through the structural stiffness network and ultimately manifest as nodal deviations, and use this model for reverse reasoning.

[0043] First, we obtain the error source space, including hole position offset, connection gap, support elevation, and uneven preload. These four types of error sources are the most common factors in aluminum alloy structure construction and have the most significant impact on the structure's geometry. Hole position offset refers to the deviation of the bolt hole center coordinates on the end plate of the component from the design position. It usually originates from the positioning error during the CNC machining stage in the factory, and its value range is generally set within ±3 mm. Connection gap refers to the clearance between the bolt hole diameter and the bolt shank diameter. During on-site assembly, it will be compacted by its own weight or preload, causing slight slippage or rotation of the node. Its value range is usually 0 to 2 mm. Support elevation error is a common problem at the interface between civil construction and the superstructure. Due to the limited accuracy of concrete pouring, the actual elevation of the embedded parts deviates from the design value. Its value range can be set to ±20 mm depending on the accuracy of civil construction. Uneven preload refers to the failure of high-strength bolts to reach the design torque or the inconsistent preload of each bolt during tightening, resulting in unexpected slight rotation of the node when subjected to bending moment. By setting a reasonable range of values ​​and a discretization step size for each error source, a multidimensional error source parameter space is formed, in which each coordinate point represents a set of possible error combination hypotheses.

[0044] Next, based on the topological mapping relationship between each error source in the error source space and the deviation at the component end, a positive mapping rule from error source parameters to component end displacement or rotation is established. Specifically, the physical equivalent of hole offset is the initial translational displacement at the component end connecting plate. For example, when the bolt hole shifts to the left by two millimeters, it is equivalent to the component end being forcibly pushed to the left by two millimeters. The physical equivalent of connection gap is the nonlinear spring element in the node stiffness matrix. When the shear force on the node is less than the friction force, the spring stiffness is infinite; when it exceeds the friction force, the stiffness drops sharply, allowing slippage. The support elevation error is directly used as the forced displacement input of the boundary node into the model. For example, if the actual elevation of the support is ten millimeters lower than the design value, a downward ten-millimeter displacement constraint is applied to the support node. Uneven preload is physically equivalent to the additional rotation of the node under bending moment. When the difference in bolt preload on both sides of the node is large, the node will rotate slightly in a certain direction after being subjected to force. These mapping rules transform the abstract error source parameters into equivalent loads or boundary conditions that can be directly applied in finite element analysis.

[0045] Then, based on the nodal degree-of-freedom matrix constructed in step S1, an overall stiffness matrix containing the stiffness of all components and nodal constraints is assembled. The nodal degree-of-freedom matrix records the constraint state of each node in three translational degrees of freedom and three rotational degrees of freedom. Combined with the physical property characteristics of each component determined in step S1, such as axial stiffness and bending stiffness, the element stiffness matrix of each component is integrated into the global coordinate system using the finite element method, forming a sparse matrix describing the stiffness characteristics of the entire structure. Taking the equivalent load or boundary conditions obtained by the aforementioned forward mapping rule as input, and combining it with the overall stiffness matrix, a forward mapping function is established from the error source parameters to the theoretical deviation vectors of each node in the entire structure. The core of this function is to solve the structural mechanical equilibrium equations: given a set of error source parameters, they are converted into equivalent nodal forces or forced displacements, and then the theoretical displacements of each node in the X, Y, and Z directions are calculated by inverting or iteratively solving the overall stiffness matrix. This forward mapping function is the nodal constraint propagation model. The input of this model is any set of parameter vectors in the error source space, and the output is the theoretical deviation vector of each node.

[0046] The node constraint propagation model constructed through the above steps realizes the forward physical simulation from the underlying construction error to the apparent node deviation. This model solves the technical problem that traditional methods cannot quantitatively describe the error propagation path and superposition effect, and provides the core engine for calculating the likelihood function for subsequent Bayesian inference.

[0047] In step S4, obtaining the probabilities of various construction error sources through Bayesian inference inversion includes: extracting the net assembly deviation matrix as the observed variable; obtaining the theoretical deviation vector of each node of the target component calculated by the node constraint propagation model for a set of assumed error source parameters; obtaining the prior probability distribution of each error source in the error source space; constructing a likelihood function to evaluate the degree of agreement between the theoretical deviation vector and the observed variable; substituting the prior probability distribution, likelihood function, and observed variable into the Bayesian formula to calculate the posterior probability distribution of each error source; using the Markov chain Monte Carlo sampling algorithm to iteratively sample in the error source space, collecting sampled samples, statistically analyzing the sampled samples, and extracting the marginal posterior probability density curve of each error source as the posterior probability distribution.

[0048] Specifically, after constructing the node constraint propagation model, although the model can forward calculate the theoretical deviation of the structure based on any set of error source parameters, the actual values ​​of the error sources in actual construction are unknown, while the net assembly deviation matrix is ​​known observation data. Therefore, it is necessary to use Bayesian inference methods to inversely deduce the probability distribution of each error source from the observation data, that is, to infer the most likely state of the underlying construction error given the known apparent deviation of the structure.

[0049] First, the net assembly deviation matrix is ​​extracted as an observation variable, denoted as . Each element in this matrix represents the pure construction assembly deviation of the corresponding measuring point after excluding temperature interference and measurement noise, and is flattened into a one-dimensional column vector. The forward calculation of the node constraint propagation model takes a set of assumed error source parameters θ as input. The assumed error source parameters refer to a specific numerical combination set for various construction errors that may exist in the structure. The forward calculation process of the node constraint propagation model is as follows: For a given set of assumed error source parameters θ, firstly, according to the mapping rule established in step S4 from the error source to the displacement or rotation of the component end, each component in θ is converted into the corresponding equivalent load or boundary condition. Specifically, the hole offset is converted into the initial translational displacement of the component end node, the support elevation is converted into the forced displacement of the boundary node, the connection gap is converted into the initial nonlinear spring gap at the node, and the uneven preload is converted into the additional rotation of the node under bending moment. Then, these equivalent boundary conditions are applied to the overall finite element model, which includes the nodal degree-of-freedom matrix constructed in step S1 and the overall stiffness matrix K assembled from the physical properties of each component, such as axial stiffness and bending stiffness.

[0050] Solve the structural mechanics equilibrium equations Where K is the global stiffness matrix, This refers to the theoretical displacement vectors of all nodes in the directions of the three translational degrees of freedom and the three rotational degrees of freedom. It is the equivalent nodal force vector obtained by transforming the assumed error source parameter vector θ through topological mapping rules. The error source parameter vector θ includes construction type errors such as hole position offset, connection gap, support elevation, and uneven preload. Different types of error sources are converted into corresponding loads or boundary conditions according to their physical mechanisms: For hole offset, it manifests as the center of the bolt hole on the connecting plate at the end of the component deviating from the design position. Its physical equivalent is the initial translational displacement of the node at the end of the component, i.e., the forced displacement boundary condition. The penalty function method is used to introduce the global equation. Specifically, a large number much larger than the order of magnitude of other stiffness elements of the structure is superimposed on the diagonal element of the global stiffness matrix corresponding to the direction of the corresponding degree of freedom as a penalty. (Usually, the element with the highest stiffness in the structure is taken as...) to (times), and simultaneously fill in the corresponding position of the load vector. The product of this forced displacement value forces the displacement on that degree of freedom to approximate the specified offset; for example, hole offset (+0.5mm): this error manifests as the center of the bolt hole in the connecting plate deviating from the design position, physically equivalent to the forced displacement boundary condition at node 2, and the penalty function method is used to introduce the global equation. The penalty function method adds a large number (penalty number) to the diagonal elements of the global stiffness matrix. The product of this large number and the forced displacement value is then filled into the corresponding load term to approximate the application of the displacement constraint. In this calculation example, the boundary condition, after being processed by the penalty function method, is equivalent to the additional nodal force of node 2 in the X direction + 1.20 × N.

[0051] For support elevation errors, which manifest as a deviation between the actual and design elevations of the support node, this is converted into a forced displacement boundary condition for the support node in the vertical direction and applied using the penalty function method in the same way. For connection gaps, which manifest as clearances between bolts and hole walls, these are compacted by self-weight or preload during on-site assembly, causing slight slippage at the node. Physically, this is equivalent to the release of the initial displacement along the component's axial direction at the node. The magnitude of the equivalent node force equals the gap amount multiplied by the component's axial stiffness, and its direction is along the component's axial direction. This is mapped to the equivalent node force components in the global coordinate system via a direction cosine vector. For uneven preload, which manifests as inconsistent preload of high-strength bolts causing unexpected additional rotation at the node when subjected to bending moments, this is physically equivalent to additional bending moments or equivalent node forces at the node. This is addressed by converting the preload difference into node moment and distributing it to each degree of freedom. The equivalent node forces of the above components are then superimposed and assembled according to node number and degree of freedom direction to form an overall equivalent node force vector. Then, the theoretical deviation vector of each node is obtained by solving the system of linear equations. This forward calculation process simulates the nodal spatial deviation of the entire aluminum alloy structure after being transmitted through a stiffness network under the influence of assumed error sources. Since the stiffness matrix K and constraints in the model are pre-established based on structural design parameters, for different sets of θ, only reassembly is required. By solving the system of linear equations, the theoretical deviation vector corresponding to each error source can be obtained efficiently.

[0052] Next, the prior probability distribution of each error source in the error source space is obtained. This prior probability distribution is a preliminary judgment of the possible values ​​of various error sources based on historical experience or industry statistical data before any field measurement is performed. For example, for hole position offset, since the machining error of CNC machine tools usually follows a normal distribution, its prior probability can be set to a Gaussian distribution with a mean of zero and a standard deviation of 0.5 mm. This means that the probability of perfectly accurate hole position is the highest, and the probability decreases exponentially as the deviation increases. For support elevation error, if the civil engineering acceptance report shows that the concrete pouring quality in a certain area is good, its prior probability can be set to a normal distribution with a mean of zero and a small standard deviation; if the quality is poor, a wider uniform distribution can be used. For connection gap, since the clearance range allowed by design specifications is limited, a uniform distribution can be used as the prior. For uneven preload, an exponential distribution can be used, biased towards small errors.

[0053] Then, a likelihood function is constructed to evaluate the degree of agreement between the theoretical deviation vector and the observed variables. The likelihood function describes the theoretical deviation calculated by the nodal constraint propagation model under the assumption that a set of error source parameters θ holds. Net assembly deviation as observed By comparing the two, the closer they are, the more likely the error source parameter θ is to be the true cause of the current measured deviation. In engineering practice, the measurement data of each measuring point inevitably contains random measurement errors. These errors usually follow a normal distribution, so a likelihood function is constructed based on a multivariate Gaussian distribution. The exponential term of the likelihood function actually represents the Mahalanobis distance between the theoretical deviation and the observed deviation after covariance weighting. The smaller this distance, the larger the likelihood function value, that is, the more reasonable the set of hypotheses. The covariance matrix Σ is constructed diagonally weighted according to the confidence level of each measuring point. Σ is a diagonal matrix. The higher the confidence level of each measuring point, the smaller its corresponding covariance (variance), which means that the observation value of that measuring point is more reliable and occupies a larger weight in the likelihood function; conversely, the lower the confidence level of the measuring point, the larger its covariance, and the smaller its weight in the likelihood function, thereby effectively suppressing the adverse effects of low-quality measurement data on the error source inversion results. The normalization constant in the likelihood function is only used to ensure that the probability density function integrals to 1 over the entire space, thus ensuring mathematical completeness. However, its specific value is only used as a scaling factor in subsequent Bayesian inference calculations and does not affect the comparison of the relative probabilities of error sources.

[0054] The prior probability distribution, likelihood function, and observed variables Substituting these values ​​into Bayes' theorem allows us to calculate the posterior probability distribution of each error source. Bayes' theorem expresses the probability distribution of an error source given a known observation bias. Under the given conditions, the conditional probability density of the error source parameter θ is calculated as follows: multiply the prior probability by the likelihood function, and then divide by the evidence factor p. The prior probability distribution is set according to the engineering statistical characteristics of each error source. For example, borehole offset usually follows a normal distribution centered at zero, while support elevation error is set according to the accuracy of civil construction, reflecting the existing knowledge of the possible value ranges of various error sources before conducting on-site measurements; the likelihood function is used to measure the degree of matching between the model output and the observed data at a given θ; the evidence factor p( As a normalization constant, the posterior probability is theoretically obtained by integrating the product of the likelihood function and the prior probability over all possible values ​​of the error source parameter. This integral is a constant scalar whose sole purpose is to mathematically guarantee that the integral of the posterior probability over the entire parameter space is 1, thus satisfying the normalization requirement of the probability density function. In practical engineering calculations, directly solving this integral is often difficult because the error source parameter space is often high-dimensional and the integrand is complex. Since the Markov chain Monte Carlo sampling algorithm only requires comparing the relative magnitudes of the posterior probabilities under different hypotheses to complete sampling, the evidence factor, as a constant independent of specific parameter values, is reduced and canceled out during the comparison process; therefore, its specific value does not need to be explicitly calculated. In other words, although the evidence factor is theoretically defined through integration, in practice, it is completely unnecessary to solve this integral; the unnormalized posterior probability can directly drive subsequent sampling and inference. The above Bayesian inference framework is built upon the publicly available Bayes formula in probability theory, and those skilled in the art can directly implement it based on this well-known formula combined with the aforementioned prior distribution and likelihood function. This Bayesian inference framework enables inverse probabilistic inference from apparent biases to hidden error sources, quantifying the probability of occurrence of each type of construction error and providing a probabilistic basis for subsequent correction priority decisions.

[0055] Since the error source space typically has high dimensionality (e.g., multiple nodes each have multiple error sources), and the node constraint propagation model is nonlinear, the posterior probability distribution cannot be directly solved analytically. Therefore, a Markov Chain Monte Carlo Sampling (MCMC) algorithm is used to iteratively sample the error source space. The MCMC algorithm generates a large number of samples conforming to the posterior distribution by randomly walking through the error source space. Specifically, it first uses the prior probability distribution as the initial proposal distribution and randomly generates a set of initial error source parameter samples. Then, in each iteration, a candidate sample θ is generated according to a preset proposal distribution, such as a normal distribution centered on the current sample. The ratio of the likelihood function value L(θ) of the candidate sample to the likelihood function value L(θ_t) of the current sample is calculated, and then multiplied by the ratio of the prior probability of the candidate sample to that of the current sample to obtain the acceptance probability. This acceptance probability is a value between 0 and 1, representing the likelihood that the candidate sample should be adopted.

[0056] When the proposal distribution adopts a normal distribution centered on the current sample, due to symmetry, the ratio of the proposal distribution is exactly one, and this ratio is directly used as the acceptance probability. Subsequently, a random number is randomly generated from a uniform distribution between 0 and 1, and this random number is compared with the acceptance probability: if the random number is less than or equal to the acceptance probability, the candidate sample is accepted as the sample for the next time step; otherwise, if the random number is greater than the acceptance probability, the candidate sample is rejected, and the current sample is retained as the sample for the next time step. This random acceptance mechanism allows the Markov chain to explore the error source space in a probabilistic manner, allowing movement towards directions with higher likelihood values ​​while also having a certain probability of accepting candidate samples with lower likelihood values, thus avoiding getting trapped in local optima. This iterative process is repeated until the accumulated sample size reaches a preset threshold, while monitoring the traversal mean or the latent scaling factor to determine whether the posterior probability distribution has converged. The samples from the initial stage are discarded, a process called the aging period, and the subsequent effective sample set is retained.

[0057] Finally, statistical analysis is performed on the effective sample set. For example, a histogram is plotted for each error source, and kernel density estimation is used to extract the marginal posterior probability density curve of each error source, which serves as the posterior probability distribution of that error source. Using the Markov chain Monte Carlo sampling algorithm described above, approximate samples of the posterior distribution can be efficiently obtained in a high-dimensional nonlinear space, providing a reliable numerical basis for the accurate quantification of error source probabilities.

[0058] Step S5: Based on the probability of error sources for various construction types, combined with adjustable margin, rework cost and structural risk threshold, generate a component-level correction priority list, obtain the corrected retest data, and update the error probabilities of various construction types based on the retest data.

[0059] In step S5, generating a component-level correction priority list includes: determining whether the probability of each error source exceeds a preset probability threshold; when the probability of an error source exceeds the probability threshold, identifying the corresponding error source as the dominant error source; calculating the structural stress increment caused by the dominant error source in the target component, and determining whether the structural stress increment exceeds a preset structural risk threshold; when the structural stress increment exceeds the structural risk threshold, calculating the correction urgency index of each component based on the structural stress increment, adjustable margin, and rework cost; and generating a component-level correction priority list by arranging the components in descending order of correction urgency index.

[0060] Specifically, after completing the inversion of error source probabilities, the marginal posterior probability density curve for each error source has been obtained. For each error source, a negligible allowable deviation range is pre-defined according to engineering specifications. For example, the allowable range for hole position offset is ±0.5 mm, and the allowable range for support elevation error is ±5 mm. The integral value of the posterior probability density curve outside this allowable range is taken as the existence probability of the error source. When this existence probability exceeds a preset probability threshold, the error source is determined to be the dominant error source; otherwise, the evidence is considered insufficient and it is not included in the correction consideration. The probability threshold is usually set between 60% and 80%, specifically determined based on engineering quality requirements.

[0061] Next, the structural stress increment induced by each dominant error source in the target component is calculated, including: applying the deviation value of the dominant error source as a forced displacement boundary condition to the overall finite element model of the aluminum alloy structure; applying a combination of structural self-weight, design wind load, and temperature load in the overall finite element model, performing nonlinear statics solution to obtain the stress field of the structure before and after the introduction of the dominant error source; extracting the equivalent stress difference of each component and each node before and after the introduction of the dominant error source, and taking the maximum value among all differences as the structural stress increment.

[0062] Specifically, when calculating the structural stress increment induced in the target component by each dominant error source, the most probable deviation value is first extracted from the posterior probability density curve of that error source; that is, the peak position of the curve is taken as the optimal estimate of that error source. For example, for hole offset, this value represents the specific number of millimeters and direction in which the bolt hole center deviates from the design position; for support elevation error, this value represents the difference between the actual elevation and the design elevation of the support; for connection gap, this value represents the clearance between the bolt and the hole wall; for uneven preload, this value represents the equivalent additional rotation angle of the node under bending moment. Then, the deviation values ​​of the dominant error sources are converted into boundary conditions that can be applied by finite element analysis. Specifically, for support elevation error, its value is directly applied as a forced displacement boundary condition to the vertical degree of freedom of the corresponding support node; for hole offset, it is equivalent to the initial translational displacement of the node at the end of the component; for connection gap, it is equivalent to the initial relative displacement of the node in a specific direction; for uneven preload, it is equivalent to the additional angular displacement of the node under bending moment.

[0063] These forced displacement values ​​are directly input into the overall finite element model of the aluminum alloy structure constructed in step S4, serving as known displacement boundary conditions for the solution. Forced displacement boundary conditions refer to specifying fixed values ​​for the degrees of freedom displacement of certain nodes in finite element analysis, independent of load variation, to simulate inelastic deformation caused by manufacturing or installation errors. Simultaneously with applying forced displacement, the model also needs to be loaded with the structure's self-weight, design wind load, wind load corresponding to the local basic wind pressure, and temperature loads considering extreme conditions, such as the thermal effect caused by the difference between the closing temperature and the extreme operating temperature. This combination of three loads constitutes the most unfavorable stress condition. Subsequently, a nonlinear statics solver is invoked to solve the overall finite element model.

[0064] Because aluminum alloys exhibit nonlinear constitutive relations after stress exceeds the proportional limit, and forced displacement may alter local contact states, a nonlinear solution method must be employed, iterating incrementally until the convergence criterion is met. After solving, the full-field stress distribution of the structure before and after the introduction of the dominant error source is obtained. The stress field is typically output as equivalent stress, which comprehensively considers the combined effects of the three principal stresses, accurately determining whether the material has entered a yield state. Then, for each component and node, the difference in equivalent stress before and after the introduction of the dominant error source is calculated, representing the stress increment caused by that error source. The maximum stress increment among all components and nodes is taken as the structural stress increment caused by the dominant error source. This maximum value usually appears at connection nodes or member ends near the error source region, representing the most unfavorable impact of the error on structural safety. Through these calculations, the actual hazard to structural safety from each dominant error source can be quantitatively determined, providing a mechanical basis for whether to initiate correction and the urgency of correction.

[0065] The calculated structural stress increment is compared with a preset structural risk threshold, which is determined based on the yield strength, critical stability stress, and safety factor specified in national standards for aluminum alloy materials. For example, it can be taken as 80% of the material's yield strength. If the structural stress increment exceeds this threshold, it indicates that the error has severely eroded the structure's safety reserve and must be corrected in subsequent construction. If it does not exceed the threshold, it indicates that the structure's own redundancy is sufficient to absorb the disturbance, and it can be temporarily left untreated. For components where the dominant error sources of structural stress increment exceed the threshold, a correction urgency index for each component needs to be further calculated. This index comprehensively considers three factors: structural stress increment, adjustable margin, and rework cost. Since these three factors have different dimensions, they must be normalized to unify their dimensions. Specifically, the structural stress increment of each component is divided by the maximum allowable stress increment for that type of component, such as the material's yield strength minus the design stress, to obtain a normalized risk coefficient. This coefficient is between 0 and 1, with a higher value indicating a higher risk.

[0066] The adjustable margin, such as the adjustment range of the elongated hole reserved in the design of the connection node or the thickness of the shim that can be increased or decreased, is divided by the maximum design adjustment range of this type of node. The maximum design adjustment range refers to the maximum geometric adjustment range allowed for this type of node according to the design drawings or construction specifications. For example, the major axis length of the elongated hole minus the bolt diameter or the maximum allowable thickness of the shim. This yields a normalized construction convenience coefficient, the larger of which indicates easier adjustment. The rework cost, including the comprehensive cost of dismantling, reinstallation, material loss, and schedule delays, is divided by the total project budget or a baseline cost to obtain a normalized economic penalty coefficient. The larger this coefficient indicates higher correction costs. In actual engineering, the correction urgency index should be positively correlated with the risk coefficient, positively correlated with the construction convenience coefficient, and negatively correlated with the economic penalty coefficient. Therefore, a weighted summation method is used to calculate the correction urgency index: the risk coefficient and the construction convenience coefficient are given positive weights, and the economic penalty coefficient is given a negative weight. The specific values ​​of the three weights can be determined using either the Analytic Hierarchy Process (AHP) or the entropy weight method. Taking the AHP as an example: First, engineering experts compare the importance of the three factors pairwise, constructing a 3×3 judgment matrix. For example, risk is slightly more important than convenience (scale 2), risk is significantly more important than economic cost (scale 5), and convenience is equally important than economic cost (scale 1). Then, the eigenvector corresponding to the largest eigenvalue of this matrix is ​​calculated and normalized to obtain the initial weights. The final weights are then determined after a consistency test (CR < 0.1). If the entropy weight method is used, the weights are calculated based on the degree of variation of the three indicators in historical data from multiple completed projects. For example, the weight of the risk coefficient is 0.5, the weight of the construction convenience coefficient is 0.2, and the weight of the economic penalty coefficient is -0.3. The correction urgency index for each component is the sum of each normalized parameter multiplied by its corresponding weight.

[0067] After calculating the urgency index of all components to be corrected, they are sorted in descending order of the index. Components at the top of the list have the highest safety risk, are easiest to adjust, and have the lowest correction cost, and should be prioritized for on-site work. Finally, a component-level correction priority list is generated, which should include the component number, spatial location, dominant error source type, and suggested correction actions. After completing one round of correction work, on-site personnel carry out correction construction according to the priority list. Subsequently, the corrected re-measurement data is obtained, i.e., coordinate measurements and temperature data are collected again for the corrected components. The re-measurement data is merged with the original measurement condition data package, and steps S2 to S4 are re-executed, i.e., the thermal deviation vector, net assembly deviation matrix, and probabilities of various error sources are recalculated, and the previous data is updated with the recalculated results. Through this closed-loop update mechanism, the correction effect can be evaluated to see if it meets expectations, and the generation of the next round of correction priority list can be triggered if necessary, until the net assembly deviation of all components meets the design specifications.

[0068] Through step S5 above, the transformation from error source probability to scientific correction decision-making is achieved, solving the problems of blind adjustments, high costs, and easy introduction of new risks in traditional construction. By comprehensively considering structural safety risks, construction feasibility, and economic costs, and using a normalized weighted method to calculate the correction urgency index, the optimal allocation of correction resources is ensured, enabling limited construction resources to address the most urgent and economical correction tasks first. Through closed-loop updates of retest data, the digital model and the physical entity are kept synchronized, achieving dynamic control and continuous optimization of the construction process. The technical solution disclosed in this application can fundamentally improve the intelligent analysis level of construction deviations in aluminum alloy structures, significantly reducing rework costs and structural safety risks.

[0069] The above describes an intelligent analysis method for construction deviations of aluminum alloy structures in the embodiments of this application. The following describes an intelligent analysis system for construction deviations of aluminum alloy structures in the embodiments of this application. Please refer to [link / reference]. Figure 3 An intelligent analysis system for construction deviations of aluminum alloy structures, as described in this application embodiment, includes: The acquisition module is used to acquire the structural design parameter set of the aluminum alloy target component, and to establish the nodal degree of freedom matrix and component coordinate reference at standard temperature based on the structural design parameter set. The component coordinate reference contains the theoretical coordinates of each preset measuring point on each target component. The calculation module is used to extract the component length from the structural design parameter set, and calculate the theoretical thermal expansion components of each measuring point on the target component based on the aluminum alloy material parameters, component length, nodal degree of freedom matrix and temperature difference between the component surface temperature and the design temperature in the structural design parameter set, and generate thermal deviation vector. The generation module is used to determine the measured deviation vector of each measuring point based on the measured coordinates and theoretical coordinates of each measuring point, and to obtain the initial assembly deviation vector by combining it with the thermal deviation vector. The initial assembly deviation vector is then weighted and filtered according to the confidence level of each measuring point to generate the net assembly deviation matrix. The inversion module is used to construct a node constraint propagation model based on the topological mapping relationship between the node degree of freedom matrix and the component end deviation. The net assembly deviation matrix is ​​input into the node constraint propagation model, and the probability of error sources of various construction types is obtained through Bayesian inference inversion. The calibration module is used to generate a component-level calibration priority list based on the probability of error sources of various construction types, combined with adjustable margin, rework cost and structural risk threshold, obtain the calibrated retest data, and update the error probability of various construction types based on the retest data.

[0070] In summary, this application eliminates the influence of ambient temperature on the reference drift of deviation analysis by establishing a component coordinate reference and nodal degree-of-freedom matrix at standard temperature. Furthermore, based on aluminum alloy material parameters, measured temperature, and nodal constraints, it accurately calculates the thermal deviation vector and isolates it from the measured total deviation. Combined with weighted filtering based on measurement point confidence, a pure net assembly deviation matrix is ​​obtained. On this basis, a nodal constraint propagation model is constructed based on the nodal degree-of-freedom matrix and the topological mapping relationship of error sources. This transforms the deviation inversion problem into Bayesian inference. A Markov chain Monte Carlo sampling algorithm is used to efficiently sample in the error source space, quantifying the posterior probabilities of various construction errors such as hole offset, connection gap, support elevation, and uneven preload. Finally, by comprehensively considering error probabilities, structural stress increments, adjustable margins, and rework costs, a correction urgency index is calculated using normalized weighted calculation, generating a scientifically sound and economical component-level correction priority list. This list is then dynamically updated in a closed loop using remeasurement data. This application achieves precise decoupling between temperature interference and actual construction deviations, breaking through the limitations of traditional methods that cannot trace the root cause of errors. Under the premise of ensuring structural safety, it significantly reduces rework costs and improves the level of intelligence and quality control efficiency in aluminum alloy structure construction.

[0071] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0072] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0073] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for intelligent analysis of construction deviations in aluminum alloy structures, characterized in that, The method includes: Step S1: Obtain the structural design parameter set of the aluminum alloy target component, and establish the nodal degree of freedom matrix and component coordinate reference at standard temperature based on the structural design parameter set. The component coordinate reference includes the theoretical coordinates of each preset measuring point on each target component. Step S2: Extract the component length from the structural design parameter set. Based on the aluminum alloy material parameters, component length, nodal degree-of-freedom matrix, and temperature difference between the component surface temperature and the design temperature in the structural design parameter set, calculate the theoretical thermal expansion components at each measuring point on the target component and generate a thermal deviation vector. Step S2 includes: calculating the temperature difference between the target component and the design temperature based on the component surface temperature; calculating the theoretical thermal expansion components of the target component under unconstrained state based on the metal linear expansion coefficient in the aluminum alloy material parameters, the component length, and the temperature difference; using the theoretical thermal expansion components as the initial thermal strain load, and combining the constraint boundary conditions defined by the nodal degree-of-freedom matrix, solving the structural thermoelastic equilibrium equation to obtain the thermal expansion components of each measuring point in three orthogonal directions, and combining the thermal expansion components of all measuring points in a preset order to generate a thermal deviation vector. Step S3: Determine the measured deviation vector of each measuring point based on the measured coordinates and the theoretical coordinates of each measuring point, and obtain the initial assembly deviation vector by combining the thermal deviation vector. Then, according to the confidence level of each measuring point, perform weighted filtering on the initial assembly deviation vector to generate the net assembly deviation matrix. Step S4: Construct a node constraint propagation model based on the topological mapping relationship between the node degree-of-freedom matrix and the component end deviations. Constructing the node constraint propagation model includes: obtaining an error source space containing hole offsets, connection gaps, support elevations, and uneven preload; establishing a forward mapping rule from error source parameters to component end displacements or rotations based on the topological mapping relationship between each error source in the error source space and the component end deviations; assembling an overall stiffness matrix containing all component stiffness and node constraints based on the node degree-of-freedom matrix; constructing the node constraint propagation model by combining the forward mapping rule; inputting the net assembly deviation matrix into the node constraint propagation model; and obtaining the probabilities of various construction type error sources through Bayesian inference inversion. Step S5: Based on the probability of error sources of various construction types, combined with adjustable margin, rework cost and structural risk threshold, generate a component-level correction priority list, obtain the corrected retest data, and update the error probability of various construction types based on the retest data.

2. The intelligent analysis method for construction deviations of aluminum alloy structures according to claim 1, characterized in that, Step S1 includes: Obtain the structural design parameter set of the target aluminum alloy component, including design grid, aluminum alloy material parameters, component cross-section, node constraints, and design temperature; The physical properties of the target component are determined based on the aluminum alloy material parameters and the component cross-section. Based on the node constraints, determine the translational and rotational degrees of freedom of each connected node, and construct the node degree of freedom matrix; Using the design temperature as the reference temperature, the theoretical spatial coordinates of each target component at the reference temperature are calculated based on the design grid and the physical property characteristics, thus generating the component coordinate reference.

3. The intelligent analysis method for construction deviations of aluminum alloy structures according to claim 1, characterized in that, Step S3 includes: Calculate the coordinate difference between the measured coordinates and the component coordinate reference, generate the measured deviation vector, and perform vector subtraction operation between the measured deviation vector and the thermal deviation vector to obtain the initial assembly deviation vector; Determine whether the confidence level of each measurement point is greater than a preset confidence level threshold. If it is, use the confidence level of the measurement point as a weighting coefficient to perform weighted smoothing filtering on the initial assembly deviation vector to filter out high-frequency measurement noise data. The net assembly deviation matrix is ​​constructed based on the filtered initial assembly deviation vector.

4. The intelligent analysis method for construction deviations of aluminum alloy structures according to claim 3, characterized in that, The initial assembly deviation vector is subjected to weighted smoothing filtering, including: Obtain the high-frequency and low-frequency components in the initial assembly deviation vector, attenuate the high-frequency components according to the weighting coefficient, retain the low-frequency components, obtain the smoothed assembly deviation vector after filtering out high-frequency measurement noise data, and construct the net assembly deviation matrix based on the smoothed assembly deviation vector.

5. The intelligent analysis method for construction deviations of aluminum alloy structures according to claim 1, characterized in that, In step S4, the probabilities of various construction error sources obtained through Bayesian inference inversion include: Extract the net assembly deviation matrix as the observation variable, and obtain the theoretical deviation vector of each node of the target component calculated in the forward direction by the node constraint propagation model for a set of assumed error source parameters; Obtain the prior probability distribution of each error source in the error source space, and construct a likelihood function to evaluate the degree of agreement between the theoretical bias vector and the observed variable; Substitute the prior probability distribution, the likelihood function, and the observed variables into Bayes' formula to calculate the posterior probability distribution of each error source; The Markov chain Monte Carlo sampling algorithm is used to iteratively sample in the error source space, collect the sampled samples, perform statistics on the sampled samples, and extract the marginal posterior probability density curve of each error source as the posterior probability distribution.

6. The intelligent analysis method for construction deviations of aluminum alloy structures according to claim 1, characterized in that, Step S5 generates a list of component-level correction priorities, including: Determine whether the probability of each error source exceeds a preset probability threshold. When the probability of an error source exceeds the probability threshold, the corresponding error source is identified as the dominant error source. Calculate the structural stress increment caused by the dominant error source in the target component, and determine whether the structural stress increment exceeds the preset structural risk threshold. When the structural stress increment exceeds the structural risk threshold, calculate the correction urgency index for each component based on the structural stress increment, adjustable margin, and rework cost. A list of component-level correction priorities is generated by arranging the correction urgency indices from high to low.

7. The intelligent analysis method for construction deviations of aluminum alloy structures according to claim 6, characterized in that, Calculating the structural stress increment induced in the target component by the dominant error source includes: The deviation values ​​of the dominant error sources are used as forced displacement boundary conditions and applied to the overall finite element model of the aluminum alloy structure. In the overall finite element model, a combination of structural self-weight, design wind load, and temperature load is applied, and nonlinear statics solution is performed to obtain the stress field of the structure before and after the introduction of the dominant error source. Extract the equivalent stress difference of each component and each node before and after the introduction of the dominant error source, and take the maximum value among all differences as the structural stress increment.

8. An intelligent analysis system for construction deviations of aluminum alloy structures, used to implement the intelligent analysis method for construction deviations of aluminum alloy structures as described in any one of claims 1-7, characterized in that, The system includes: The acquisition module is used to acquire the structural design parameter set of the aluminum alloy target component, and to establish the nodal degree of freedom matrix and component coordinate reference at standard temperature based on the structural design parameter set. The component coordinate reference includes the theoretical coordinates of each preset measuring point on each target component. The calculation module is used to extract the component length from the structural design parameter set, and calculate the theoretical thermal expansion components of each measuring point on the target component based on the aluminum alloy material parameters, component length, nodal degree of freedom matrix, and temperature difference between the component surface temperature and the design temperature in the structural design parameter set, and generate a thermal deviation vector. Step S2 includes: calculating the temperature difference of the target component based on the component surface temperature and the design temperature; calculating the theoretical thermal expansion components of the target component under unconstrained state based on the metal linear expansion coefficient in the aluminum alloy material parameters, the component length, and the temperature difference; using the theoretical thermal expansion components as the initial thermal strain load, and combining the constraint boundary conditions defined by the nodal degree of freedom matrix, solving the structural thermoelastic equilibrium equation to obtain the thermal expansion components of each measuring point in three orthogonal directions, and combining the thermal expansion components of all measuring points in a preset order to generate a thermal deviation vector. The generation module is used to determine the measured deviation vector of each measuring point based on the measured coordinates and the theoretical coordinates of each measuring point, and to obtain the initial assembly deviation vector by combining the thermal deviation vector. The initial assembly deviation vector is then weighted and filtered according to the confidence level of each measuring point to generate the net assembly deviation matrix. The inversion module is used to construct a node constraint propagation model based on the topological mapping relationship between the node degree-of-freedom matrix and the component end deviations. Constructing the node constraint propagation model includes: obtaining an error source space containing hole offsets, connection gaps, support elevations, and uneven preload; establishing a forward mapping rule from error source parameters to component end displacements or rotations based on the topological mapping relationship between each error source in the error source space and the component end deviations; assembling an overall stiffness matrix containing all component stiffness and node constraints based on the node degree-of-freedom matrix; constructing the node constraint propagation model by combining the forward mapping rule; inputting the net assembly deviation matrix into the node constraint propagation model; and obtaining the probabilities of various construction type error sources through Bayesian inference inversion. The calibration module is used to generate a component-level calibration priority list based on the probability of error sources of various construction types, combined with adjustable margin, rework cost and structural risk threshold, obtain the calibrated retest data, and update the error probability of various construction types based on the retest data.

Images