Pump station structure performance simulation analysis method and system based on digital twinning

Abstract

The application provides a pump station structure performance simulation analysis method and system based on digital twinning, and relates to the technical field of water conservancy engineering structure health monitoring and simulation analysis. The method comprises the following steps: according to an initial design finite element model of a pump station to be monitored, a soil-structure interaction virtual entity reflecting the dynamic interaction mechanism of the pump station bottom plate, side wall and saturated soft soil foundation is constructed, and a reference digital twin is obtained, which contains initial spring-damping boundary constraints and preset visible nodes. The application can effectively represent the stiffness time-varying degradation effect caused by the accumulation of pore water pressure of the saturated soft soil foundation, and reasonably solve the problems of fatigue damage evaluation inaccuracy and lack of closed-loop iterative early warning capability.

Description

Technical Field

[0001] This invention relates to the field of structural health monitoring and simulation analysis technology for hydraulic engineering, and in particular to a method and system for simulation analysis of pump station structural performance based on digital twins. Background Technology

[0002] As a key node in regional water conservancy projects, pumping stations are subjected to alternating hydraulic vibrations caused by frequent start-ups and shutdowns of generating units. In soft soil areas, the pumping station's base slab and sidewalls are supported on saturated soft soil foundations. Under cyclic dynamic loads, the foundation soil exhibits rheological characteristics that gradually degrade in stiffness due to the accumulation of pore water pressure. This long-term evolution of the soil-structure interaction boundary directly threatens the service safety and remaining lifespan of the overall pumping station structure.

[0003] Current simulation analyses of pump station structural performance in soft soil areas generally employ a single static model built from initial parameters obtained during the geological exploration phase, using distributed spring-damping units to simulate the boundary constraints of the foundation on the structure. Once established, the spring stiffness and damping coefficients of this type of model are fixed and cannot be dynamically updated as service progresses, leading to the following drawbacks: In scenarios of continuously deteriorating foundation conditions, it is difficult to characterize the time-varying stiffness degradation effect caused by the accumulation of pore water pressure in saturated soft soil. Consequently, in long-term service simulations spanning flood season drainage and daily water transfer, the boundary constraint state of the simulation model gradually deviates from the actual mechanical behavior of the foundation. The structural fatigue damage increment calculated based on this model is at risk of inaccuracy and fails to reflect the actual damage accumulation process of the pump station structure under the combined effects of alternating hydraulic vibration and degraded foundation. Furthermore, because the simulation model lacks the ability to dynamically track and update the boundary degradation state, existing analysis methods mainly complete static verification during the design phase or assessment at specific time points, failing to meet the need for early warning of dangerous phases in long-term performance tracking of pump stations and iterative optimization during service. Summary of the Invention

[0004] This invention provides a method and system for simulating and analyzing the structural performance of pumping stations based on digital twins. It can effectively characterize the time-varying stiffness degradation effect caused by the accumulation of pore water pressure in saturated soft soil, and reasonably solve the problems of inaccurate fatigue damage assessment and lack of closed-loop iterative early warning capability.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows: Firstly, a method for simulating and analyzing the structural performance of pumping stations based on digital twins, the method comprising: Based on the initial design finite element model of the pumping station to be monitored, a virtual entity of soil-structure interaction reflecting the dynamic interaction mechanism between the pumping station's bottom plate, side walls, and saturated soft soil foundation is constructed, resulting in a reference digital twin containing initial spring-damped boundary constraints and preset visible nodes. Based on the benchmark digital twin, the measured vibration response data of the pump station unit under alternating start-stop conditions are continuously accessed. The time-frequency domain features associated with the cumulative effect of pore water pressure in the foundation are extracted from the measured vibration response data to obtain the real-time response feature set. Based on the real-time response feature set, the soil stiffness parameters characterized by the spring-damped boundary constraint are identified, the nonlinear degradation trajectory of the soil stiffness parameters with the number of load cycles is determined, and the time-varying stiffness degradation function of the soil is obtained. The spring-damped boundary constraints in the reference digital twin are dynamically corrected based on the soil time-varying stiffness degradation function to obtain a time-varying boundary twin model with degraded boundary state. The fatigue damage increment of the pump station structure under the combined action of alternating hydraulic vibration and degraded foundation is calculated based on the time-varying boundary twin model; the increment is mapped to the preset visible nodes of the time-varying boundary twin model to obtain the time-varying twin simulation results. Based on the time-varying twin simulation results and the preset long-term performance limit state, dangerous phases are marked for structural parts that exceed the preset safety threshold. Dangerous phase marks are added to the time-varying boundary twin model to obtain a time-varying twin model with marked dangerous phases. The degenerate boundary state in the time-varying twin model with marked dangerous phases is used as the initial spring-damped boundary constraint for the next round of real-time simulation analysis and fed back to the reference digital twin.

[0006] Secondly, a pump station structural performance simulation and analysis system based on digital twins includes: The benchmark twin construction module is used to construct a soil-structure interaction virtual entity that reflects the dynamic interaction mechanism between the pump station's base plate, side walls, and saturated soft soil foundation based on the initial design finite element model of the pump station to be monitored, and obtain a benchmark digital twin containing initial spring-damped boundary constraints and preset visible nodes. The response feature extraction module is used to continuously access the measured vibration response data of the pump station unit under the start-stop alternating operating conditions based on the reference digital twin, and extract the time-frequency domain features associated with the cumulative effect of pore water pressure in the foundation from the measured vibration response data to obtain a real-time response feature set. The stiffness degradation identification module is used to identify the soil stiffness parameters characterized by the spring-damped boundary constraint based on the real-time response feature set, determine the nonlinear degradation trajectory of the soil stiffness parameters with the number of load cycles, and obtain the time-varying stiffness degradation function of the soil. The time-varying boundary correction module is used to dynamically correct the spring-damped boundary constraints in the reference digital twin based on the soil time-varying stiffness degradation function, so as to obtain a time-varying boundary twin model with degraded boundary state. The fatigue damage mapping module is used to calculate the fatigue damage increment of the pump station structure under the combined action of alternating hydraulic vibration and degraded foundation according to the time-varying boundary twin model, and map the fatigue damage increment to the preset visible node in the time-varying boundary twin model to obtain the time-varying twin simulation result. The dangerous phase identification and feedback module is used to identify dangerous phases of structural parts that exceed a preset safety threshold based on the time-varying twin simulation results and the preset long-term performance limit state. The dangerous phase identification is attached to the time-varying boundary twin model to obtain a time-varying twin model with dangerous phases. The degenerate boundary state in the time-varying twin model with dangerous phases is used as the initial spring-damped boundary constraint for the next round of real-time simulation analysis and fed back to the reference digital twin in the reference twin construction module to form a closed-loop iteration.

[0007] The above-described solution of the present invention has at least the following beneficial effects: By continuously accessing measured vibration response data under alternating start-up and shutdown conditions of pump station units, time-frequency domain features related to the cumulative effect of pore water pressure in the foundation are extracted and a real-time response feature set is constructed. This achieves real-time perception of changes in the stiffness state of saturated soft soil foundations, providing a data foundation for dynamically identifying the degradation patterns of the foundation.

[0008] By identifying the soil stiffness parameters characterized by spring-damped boundary constraints through real-time response feature sets, and determining their nonlinear degradation trajectory with the number of load cycles, dynamic correction of boundary constraint parameters in the benchmark digital twin is performed, and a time-varying boundary twin model is established. This achieves the effect of simulating boundary conditions that continuously approximate the real mechanical behavior of the foundation as the service process progresses, overcoming the technical defect that static models are difficult to characterize the time-varying degradation of stiffness.

[0009] The fatigue damage increment of the pump station structure under the combined action of alternating hydraulic vibration and degraded foundation is calculated by using a time-varying boundary twin model. The fatigue damage increment is then mapped to visible nodes to identify dangerous phases at locations exceeding the safety threshold. The degraded boundary state is fed back to the next round of simulation to form a closed-loop iteration, achieving the effect of accurate tracking of the structural damage accumulation process and early warning of dangerous phases, thus realizing the continuous evaluation of the long-term service performance of the pump station. Attached Figure Description

[0010] Figure 1 This is a flowchart illustrating the pump station structural performance simulation analysis method based on digital twin provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of a pump station structural performance simulation and analysis system based on digital twins provided in an embodiment of the present invention; Figure 3This is a schematic diagram of a baseline digital twin soil-structure interaction virtual entity provided in an embodiment of the present invention; Figure 4 This is a fitting diagram of the nonlinear degradation trajectory of soil stiffness with the number of cycles provided in an embodiment of the present invention. Detailed Implementation

[0011] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0012] like Figure 1 As shown, embodiments of the present invention propose a method for simulation analysis of pump station structural performance based on digital twins, the method comprising the following steps: Step 1: Based on the initial design finite element model of the pumping station to be monitored, construct a virtual entity of soil-structure interaction that reflects the dynamic interaction mechanism between the pumping station's bottom plate, side walls, and saturated soft soil foundation, and obtain a reference digital twin containing initial spring-damped boundary constraints and preset visible nodes. Step 2: Based on the benchmark digital twin, continuously access the measured vibration response data of the pump station unit under alternating start-stop conditions, extract the time-frequency domain features associated with the cumulative effect of pore water pressure in the foundation from the measured vibration response data, and obtain the real-time response feature set; Step 3: Identify the soil stiffness parameters characterized by the spring-damped boundary constraint based on the real-time response feature set, determine the nonlinear degradation trajectory of the soil stiffness parameters with the number of load cycles, and obtain the time-varying stiffness degradation function of the soil. Step 4: Dynamically correct the spring-damped boundary constraints in the reference digital twin based on the soil time-varying stiffness degradation function to obtain a time-varying boundary twin model with degraded boundary state. Step 5: Calculate the fatigue damage increment of the pump station structure under the combined action of alternating hydraulic vibration and degraded foundation according to the time-varying boundary twin model; map the increment to the preset visible nodes of the time-varying boundary twin model to obtain the time-varying twin simulation results; Step 6: Based on the time-varying twin simulation results and the preset long-term performance limit state, identify the dangerous phases of structural parts that exceed the preset safety threshold, and add the dangerous phase identification to the time-varying boundary twin model to obtain the time-varying twin model with the dangerous phase; use the degenerate boundary state in the time-varying twin model with the dangerous phase as the initial spring-damped boundary constraint for the next round of real-time simulation analysis, and feed it back to the reference digital twin.

[0013] In this embodiment of the invention, by constructing a benchmark digital twin that reflects the dynamic interaction mechanism between the pump station base plate, side walls and saturated soft soil foundation, continuously accessing measured vibration response data under start-stop alternating working conditions and extracting time-frequency domain features associated with the cumulative effect of pore water pressure, the nonlinear degradation trajectory of soil stiffness parameters with the number of load cycles is identified, thereby achieving the effect of accurately capturing the time-varying degradation law of stiffness under long-term service conditions of saturated soft soil foundation.

[0014] By dynamically correcting the spring-damped boundary constraints in the benchmark digital twin using the soil time-varying stiffness degradation function, a time-varying boundary twin model with degraded boundary states is established. This model is used to calculate the fatigue damage increment of the pump station structure under the combined action of alternating hydraulic excitation and degraded foundation. This achieves the effect that the simulated boundary conditions continuously approach the actual mechanical behavior of the foundation as the service progresses, overcoming the problem that traditional static boundary models cannot reflect the time-varying stiffness degradation, leading to inaccurate damage assessment.

[0015] By mapping the fatigue damage increment to the visible nodes of the time-varying boundary twin model and generating time-varying twin simulation results, and combining the preset long-term performance limit state to identify the dangerous phase of structural parts that exceed the safety threshold, and using the degraded boundary state as the initial spring-damped boundary constraint for the next round of simulation analysis for closed-loop feedback, the effect of continuously tracking the damage accumulation process of the pump station structure and providing early warning of dangerous phases is achieved.

[0016] In a preferred embodiment of the present invention, step 1 above may include: Step 1.1: Based on the initial design finite element model, extract the soil-structure contact interface between the pump station's base slab, side walls, and saturated soft soil foundation. Specifically, the initial design finite element model is a three-dimensional numerical model established based on the construction details and foundation treatment scheme of the pump station to be monitored. In this model, the pump station's base slab and side walls are discretized using three-dimensional solid elements or thick shell elements. Solid elements realistically reproduce the three-dimensional geometry and internal stress distribution of the base slab and side walls by dividing them into hexahedral or tetrahedral meshes, while thick shell elements accurately capture bending deformation and transverse shear effects by using multi-point integration in the wall thickness direction; saturated soft soil... Based on the layered profiles given in the geological exploration report during the construction period, the foundation was discretized layer by layer from top to bottom using three-dimensional solid units. Each layer was assigned physical and mechanical parameters such as density, elastic modulus, and initial void ratio. In terms of physical contact, the bottom surface of the pump station base plate and the top surface of the foundation form a normal bearing contact, that is, the base plate transmits the self-weight of the superstructure and the water load to the foundation along the normal direction of the contact surface, and the foundation reacts to the base plate in the form of compressive stress. The outer side of the side wall and the side of the foundation form a normal contact and tangential friction contact, that is, the side wall is in close contact with the side of the foundation under the action of horizontal earth pressure, and the contact surface can transmit normal compressive force and tangential friction force.

[0017] Based on this physical contact relationship, specifically, the spatial three-dimensional coordinates of all nodes on the bottom surface of the pump station base plate and all nodes on the top surface of the foundation are traversed. The Euclidean distance between each pair of nodes is calculated. Node pairs with a distance less than a preset tolerance are identified as nodes sharing spatial coordinates and included in the base plate contact node set. This preset tolerance is determined based on the minimum feature size of the finite element mesh, and is exemplarily taken as 0.01 of the mesh feature size. When the mesh feature size is between 0.1 meters and 1 meter, the tolerance value ranges from 1 × 10⁻⁶. -3 meters to 1×10 -2 Meanwhile, the node connection topology of the outer sidewall unit is traversed, and nodes that share adjacency lines with the foundation sidewall unit on the boundary are extracted and included in the sidewall contact node set. Furthermore, the connectivity of the bottom plate contact node set and the sidewall contact node set along the boundary of the contact area is checked, and isolated discontinuities caused by local mesh discontinuities are eliminated, finally obtaining a continuous and complete soil-structure contact interface. This interface completely covers all force transmission areas where dynamic interaction may occur between the pump station structure and the saturated soft soil foundation, providing a precise spatial positioning carrier for the subsequent layout of distributed dynamic boundary constraints.

[0018] Step 1.2: Based on the soil-structure contact interface, configure distributed spring-damping elements on the initial design finite element model to form a virtual boundary characterizing the dynamic impedance characteristics of the saturated soft soil foundation to the pump station structure. Specifically, this includes: at each node on the extracted soil-structure contact interface, calculating the unit normal vector and two mutually orthogonal unit tangent vectors at that point based on the shape function of the contact interface element to which the node belongs, thereby constructing a local coordinate system for the node; specifically, let the parameter coordinates of the node on the contact interface element be... and The total number of unit nodes is Unit 1 The global coordinates of each node are The shape function is The global coordinates at that node are then determined by the following interpolation formula:

[0019] The range of summation is to Based on the partial derivatives of the shape function with respect to the parametric coordinates, the covariant basis vector at that node is calculated. and The calculation formula is:

[0020]

[0021] in and For global coordinates to the parameter coordinates ( or The partial derivatives of ). It is the first Node-shaped functions For parametric coordinates The partial derivative (scalar); It is the first Node-shaped functions For parametric coordinates The partial derivative (scalar); the normal vector at that node. The cross product of covariant basis vectors After determining that its magnitude is normalized, the unit normal vector is obtained. ,in Representing vectors The Euclidean modulus; based on this, construct the first unit tangent vector. If the unit normal vector of absolute value of components Less than Then take the auxiliary vector. Otherwise take , Depend on and The cross product is determined by normalization, i.e. The second unit tangent vector Depend on and cross product Sure.

[0022] At this point, a system was established at this node. , , An orthogonal right-handed local coordinate system is constructed; within this local coordinate system, a two-node connection element of length 0 is constructed, with one node bound to the contact interface node on the pump station structure side, and the other node serving as a constraint reference point on the foundation side; in the normal direction... A spring unit and a damping unit are connected in parallel in the first tangential direction. A spring unit and a damping unit are connected in parallel in the second tangential direction. A spring unit and a damping unit are connected in parallel. The normal spring-damping unit is used to simulate the vertical support impedance and energy dissipation when the saturated soft soil foundation causes relative displacement of the pump station base plate and side wall along the normal direction of the contact surface. The vertical support impedance refers to the elastic restoring force generated when the foundation undergoes normal compression or tensile deformation. The energy dissipation refers to the physical mechanism by which vibration energy is dissipated in the form of heat due to the internal friction between soil particles and the radiation of stress waves to the infinite domain.

[0023] Tangential spring-damping elements are used to simulate the constraint impedance and frictional energy dissipation of the structure when the foundation causes tangential slip displacement along the contact surface. The constraint impedance refers to the elastic resistance provided by the foundation during shear deformation, and the frictional energy dissipation refers to the physical mechanism by which the contact interface consumes vibration energy due to soil shear hysteresis and interface frictional slip. Correspondingly, the normal spring-damping elements at all contact interface nodes and the two tangential spring-damping elements together form a dense network of spring-damping elements on the soil-structure contact interface. This element network transforms the dynamic impedance effect of the infinite domain saturated soft soil foundation on the pump station structure, which originally required large-scale discrete modeling of the foundation domain, into a lumped parameter mechanical boundary acting on the structural contact surface, thereby forming a virtual boundary characterizing the dynamic impedance characteristics of the foundation.

[0024] Step 1.3: Based on the pre-acquired foundation survey data during the pump station construction and initial operation periods, assign initial stiffness coefficients and initial damping coefficients to the distributed spring-damping unit to obtain the initial spring-damping boundary constraints. Specifically, this includes: the dynamic shear modulus of the saturated soft soil layer obtained from the geological exploration report during the pump station construction period. Poisson's ratio ,density Shear wave velocity Longitudinal wave velocity In-situ dynamic parameters, combined with the modal damping ratio of the structure-foundation system identified by micro-amplitude environmental vibration tests during the initial trial operation of the pump station. For each node on the soil-structure contact interface Contact influence area allocated and equivalent radius To determine, where the equivalent radius , Take the weighted average of the areas of all contact interface units adjacent to the node.

[0025] Based on this, the initial stiffness coefficient and initial damping coefficient are calculated node by node using the impedance function formula based on the elastic half-space theory, where the normal spring stiffness coefficient... In the formula Let be the initial dynamic shear modulus of the saturated soft soil at this node location. Poisson's ratio of the saturated soft soil at this node location; tangential spring stiffness coefficient. According to the formula Calculate the normal damping coefficient. The relationship between longitudinal wave impedance is expressed by the formula calculate, This represents the density of the saturated soft soil at that node location. The longitudinal wave velocity of the saturated soft soil at this node location; the tangential damping coefficient. Based on the shear wave impedance relationship according to the formula calculate, This represents the shear wave velocity of the saturated soft soil at this node location.

[0026] Furthermore, the measured modal damping ratio of the structure-foundation system identified by micro-amplitude environmental vibration testing was used. For all the normal damping coefficients calculated above and tangential damping coefficient Perform overall calibration adjustment, that is, adjust the coefficients. Multiply each damping coefficient by Make corrections so that the system damping ratio calculated by the corrected finite element model is consistent with... Consistent, the corrected normal damping coefficient is taken as The corrected tangential damping coefficient is taken as The corrected damping coefficient is used as the final initial damping coefficient; correspondingly, the calculated initial stiffness coefficient is used... and The stiffness property parameters of the normal and tangential spring elements at the corresponding nodes are assigned respectively, and the corrected initial damping coefficients are used. and Damping attribute parameters are assigned to the normal and tangential damping elements at the corresponding nodes, respectively. Thus, the mechanical parameters of each distributed spring-damped element are directly calibrated from the in-situ dynamic test data of the foundation, forming the initial spring-damped boundary constraint that is equivalently transformed from the initial physical state of the foundation. The stiffness of all nodes at the contact interface is summed to obtain the total equivalent stiffness characterizing the overall support effect of the foundation, in units of MN / m. This is used as a physical benchmark for subsequent degradation function inversion, ensuring that the virtual boundary can accurately reproduce the initial dynamic impedance and energy dissipation characteristics of the saturated soft soil foundation on the pump station structure at the start of the simulation analysis.

[0027] Step 1.4: Based on the initial spring-damped boundary constraints and the virtual boundary, construct a soil-structure interaction virtual entity that reflects the dynamic interaction mechanism between the pump station floor, side walls, and saturated soft soil foundation. Use this soil-structure interaction virtual entity as the reference digital twin. Specifically, this includes: rigidly binding the virtual boundary with the initial spring-damped boundary constraint configuration to the grid nodes of the pump station floor and side walls; specifically, applying coupling constraints of all six degrees of freedom, including displacement synchronization conditions of three translational degrees of freedom and three rotational degrees of freedom, to the structural side node of each spring-damped unit in the virtual boundary and the corresponding pump station structural contact interface node, so that the two maintain consistent displacement throughout the dynamic response process, thereby realizing the direct transmission of the normal and tangential restoring forces and dissipative forces provided by the virtual boundary to the pump station structure.

[0028] Based on this, the selection of preset visible nodes is carried out. Specifically, based on the virtual boundary and the initial spring-damped boundary constraints, combined with the geometric features and stress characteristics of the pump station structure, typical areas of stress concentration and damage sensitivity in the pump station structure are identified; the corner area connecting the pump station base plate and the side wall is selected, which is subjected to combined bending and shearing under alternating hydraulic excitation and is a high-incidence area for fatigue crack initiation; the central area of ​​the base plate is selected, which has a large span and significant bending normal stress under the action of vertical foundation reaction force; the middle area of ​​the side wall is selected, which is subjected to the combined action of horizontal earth pressure and hydraulic lateral pressure; the junction area between the channel wall and the main structure is selected, which has abrupt geometric cross-section and high stress concentration coefficient; in each of the above areas, the corresponding finite element surface nodes are extracted and marked as preset visible nodes, and each preset visible node is assigned a region category identifier and three-dimensional spatial coordinates; the preset visible nodes constitute the node set for fatigue damage state bearing and visualization in subsequent steps, that is, the target nodes mapped in the above-mentioned steps.

[0029] At this point, the spring restoring force and damping dissipation force applied by the virtual boundary directly become the mechanical boundary conditions for the analysis of free and forced vibration of the pump station structure. This fully characterizes the supporting constraint effect and vibration energy absorption and dissipation effect of the saturated soft soil foundation on the pump station structure, replacing the foundation dynamic effects that require discretization of a large-scale foundation area in traditional methods. Based on this, the initial material properties, element types, and geometric dimensions of the pump station base plate and sidewalls are kept unchanged, and the pump station structure model with this complete mechanical boundary condition is defined as a virtual entity of soil-structure interaction. In the initial undamaged state, this virtual entity can fully reflect the joint bearing and deformation coordination mechanism of the pump station base plate, sidewalls, and saturated soft soil foundation under dynamic loads. It is used as a reference digital twin for subsequent continuous access to the vibration response data monitored in real time during pump station operation, and for time-varying degradation analysis of foundation stiffness and service performance evaluation.

[0030] like Figure 3As shown, this virtual entity completely replicates the actual spatial topology of a certain region's drainage pumping station. The main body of the pumping station is discretized using C35 reinforced concrete solid elements. The base plate has a planar dimension of 10m × 5m and a thickness of 0.5m, while the side walls are 4m high and 0.3m thick. The minimum feature size of the mesh is 0.5m, consistent with the discretization rules of the initial design finite element model in step 1.1. The figure uses a right-handed Cartesian rectangular coordinate system. The horizontal axis represents the horizontal length of the pumping station along the water flow direction, the vertical axis represents the horizontal width perpendicular to the water flow direction, and the vertical axis represents the vertical height. The origin of the coordinate system is set at the intersection of the lower left corner of the pumping station's base plate and the foundation contact surface, consistent with the global reference system of the node local coordinate system in step 1.2. The red dots in the figure represent the base plate-side wall contact interface nodes, totaling 1929 (1247 for the base plate and 682 for the side walls). (This is just an example, not all are shown). Each node has an orthogonal local coordinate system constructed as described in step 1.2, and a distributed spring-damping unit (blue line segment) is configured. The normal spring stiffness coefficient is calculated as 12.6 to 31.4 MN / m, the tangential stiffness is 0.6 times that of the normal, and the damping coefficient, after correction by micro-amplitude vibration test, has an error of less than 5% compared with the measured modal damping ratio of 3.2%. The yellow highlighted nodes are the 86 preset visible nodes selected in step 1.4 (not all are shown in the figure), which are concentrated in the stress concentration areas such as the four corners of the bottom plate, the mid-span of the bottom plate, the middle of the side wall, and the flow channel junction area. Each node is bound to the area category identifier and three-dimensional coordinates. The fatigue damage increment in the subsequent step 5 will be directly mapped to this type of node for visualization, intuitively presenting the boundary equivalence logic of soil-structure dynamic interaction.

[0031] In a preferred embodiment of the present invention, step 2 above may include: Step 2.1: Based on the soil-structure interaction virtual entity in the reference digital twin that reflects the dynamic interaction mechanism between the pump station base plate, side walls, and saturated soft soil foundation, determine the deployment location of vibration measuring points associated with the cumulative effect of foundation pore water pressure, and obtain the measuring point deployment information. Specifically, this includes: based on the established soil-structure interaction virtual entity in the reference digital twin, this virtual entity, through a dense network of spring-damping elements on the base plate contact interface and side wall contact interface, fully reflects the dynamic interaction mechanism of normal support, tangential constraint, and energy dissipation between the pump station base plate, side walls, and saturated soft soil foundation. Specifically, under the long-term start-stop alternating load of the unit, the excess pore water pressure inside the saturated soft soil foundation will gradually accumulate with the increase of the number of load cycles, leading to a decrease in the effective stress of the foundation and a degradation of local stiffness. This degradation effect will first be reflected in the normal spring stiffness coefficient in the spring-damping element network. The larger edge areas of the base plate and the bottom corner areas of the side walls exhibited abnormal changes in vibration response; based on this, modal participation factor analysis and frequency response function sensitivity calculation were performed using a benchmark digital twin to identify structural surface nodes with the highest time-varying sensitivity to foundation stiffness degradation.

[0032] The specific process of modality participation factor analysis involves solving for the eigenvalues ​​of the baseline digital twin and extracting the prior art. The structure mode; for typical large pump station units, the rated speed range is approximately 60 to 300 revolutions per minute, corresponding to a rotational frequency of 1 Hz to 5 Hz. The blade passing frequency is the rotational frequency multiplied by the number of blades, with a common number of blades being 3 to 7. Therefore, the upper limit of the excitation frequency band is usually no more than 100 Hz. The value covers more than three times the frequency band of the pump station unit's start-up and shutdown excitation, which corresponds to approximately 300Hz. The value of should be between 50 and 100 to ensure complete coverage of the excitation frequency band and provide sufficient modal information for sensitivity analysis.

[0033] For the First mode, where Its modal participation factor In the formula, It is a dimensionless scalar. For the first The mode shape vectors of the first mode. Represents the column vector The row vector obtained after performing the transpose operation; The mass matrix of the baseline digital twin, The unit rigid body displacement vector represents the excitation distribution direction of the unit's unbalanced excitation force in the three orthogonal directions of the pump station: axial, lateral, and vertical. Modal participation factor The larger the absolute value of , the stronger the . The more significant the response contribution of each modal in the corresponding excitation direction, the better; furthermore, from the mode shape vectors of each modality... Extract the modal displacement components of all surface nodes of the pump station structure under each mode, and correlate the nodal modal displacement amplitudes with the corresponding modal participation factors. The absolute values ​​are weighted and summed according to modal order to obtain the comprehensive modal sensitivity index of each structural surface node under unit start-up and shutdown excitation; the weighted summation is calculated as follows for nodes numbered... The structural surface nodes, where , The total number of surface nodes of the pump station structure in the baseline digital twin is determined from the mode shape vector. Extracting nodes Modal displacement components in the three orthogonal directions of the pumping station: axial, lateral, and vertical. 、 and These three components are all scalars with the dimension of length, and nodes are calculated from these three components. In the Modal displacement amplitude under first mode :

[0034] Further, the absolute values ​​of the participation factors of each mode are... With nodes mode displacement amplitude Multiply, then divide From 1 to Summing yields the nodes. Comprehensive modal sensitivity index The formula is:

[0035] It is a scalar quantity with the dimension of length, comprehensively reflecting the nodes. in front The cumulative contribution of the modal displacement amplitude under each modal mode after weighting by the participation factor of each modality. The larger the node, the greater the variation in its vibration response with the degradation of the foundation spring stiffness.

[0036] The specific process for calculating the frequency response function sensitivity is as follows: In the reference digital twin, select any normal spring stiffness coefficient in the spring-damping element network on the contact interface. As a perturbation parameter, calculate any surface node on the pump station structure. At any excitation frequency acceleration frequency response function under The sensitivity to this stiffness coefficient, i.e., the amplitude of the frequency response function. right partial derivatives This sensitivity reflects the sensitivity when contacting interface nodes. When the normal spring stiffness changes slightly, the structural surface nodes The acceleration response amplitude at frequency Rate of change at the point; traverse all contact interface nodes and all candidate structure surface nodes The absolute values ​​of the sensitivity of the computer group at the power frequency and its five harmonics are summed to obtain the frequency response sensitivity index of each candidate structural surface node to the overall stiffness degradation of the foundation. The higher the frequency response sensitivity index of the node, the more sensitive its acceleration response is to the spring stiffness degradation caused by the accumulation of pore water pressure in the foundation.

[0037] Based on this, the intersection of structural surface nodes ranking in the top 20% for both comprehensive modal sensitivity and frequency response sensitivity is defined as the set of highly sensitive nodes. This 20% threshold is determined according to the Pareto principle commonly used in engineering, which states that approximately 20% of critical locations often contribute about 80% of the structural response sensitivity difference. In practical implementation, this threshold can be adjusted between 15% and 25% based on the feasible upper limit of the total number of structural nodes and the number of measuring points installed in the pump station: when the total number of candidate surface nodes is large, the percentage threshold can be appropriately lowered to improve screening accuracy; when the number of measuring points installed is limited by engineering conditions, the percentage threshold can be appropriately increased. The value ensures that a sufficient number of candidate measuring points are selected. The set of highly sensitive nodes is spatially concentrated at the four corners of the pump station base plate, the midpoint of the long side of the base plate, the intersection line of the side wall and the base plate, and the free edge of the top of the side wall. These locations are the areas where the structural dynamic response produces the greatest difference when the normal spring constraint stiffness changes in the spring-damping unit network. Furthermore, the highly sensitive structural surface nodes identified above are determined as the deployment locations of vibration measuring points associated with the cumulative effect of pore water pressure in the foundation. Each measuring point is assigned three-dimensional spatial coordinates, the name of the structural component to which it belongs, and the corresponding contact interface area number to form complete measuring point deployment information.

[0038] Step 2.2: Based on the measured point deployment information, continuously input the measured vibration response data of the pump station unit under alternating start-stop conditions. Specifically, this includes: installing triaxial acceleration sensors at the four corners of the pump station base plate, the midpoint of the long side of the base plate, the intersection line between the side wall and the base plate, and the free edge of the top of the side wall, according to the obtained measured point deployment information. The range of the sensors is selected based on twice the estimated maximum vibration acceleration peak value at the rated speed of the pump station unit. The estimated maximum vibration acceleration peak value can be obtained by performing harmonic response analysis on the reference digital twin established in Step 1 under rated conditions. Its typical value for large pump station structures is generally between 0.5 times the gravitational acceleration and 2 times the gravitational acceleration, approximately 5. Up to 20 Therefore, the sensor range should be selected as 10. Up to 40 That is, approximately 1 to 4 times the acceleration due to gravity.

[0039] The upper limit of the sensor's frequency response range should not be less than five times the passing frequency of the pump station blades. For pump station units with 3 to 7 blades and a rated speed of 60 to 300 revolutions per minute, the passing frequency range of the blades is approximately 3Hz to 35Hz. Five times this range is 15Hz to 175Hz. Therefore, the upper limit of the sensor's frequency response range should preferably not be less than 200Hz. In actual selection, standard industrial accelerometers with frequencies between 500Hz and 1000Hz are often chosen. Specifically, during the complete start-stop alternating cycle of the pump station unit, which includes start-up, stable operation, shutdown, and stop, the data... The acquisition system continuously acquires acceleration time history signals at each measuring point along the three orthogonal directions of the pump station: axial, lateral, and vertical, at a preset fixed sampling frequency. The sampling frequency is set to be at least 10 times the highest frequency component to be analyzed, which is 5 times the blade passing frequency, with an upper limit of no more than 200Hz. Therefore, the sampling frequency should not be lower than 2000Hz. In engineering implementation, a standard sampling frequency between 2000Hz and 5000Hz is usually used to ensure that the Nyquist frequency in the time-frequency analysis is much higher than the highest frequency component of the signal to be analyzed, thus avoiding spectral aliasing.

[0040] Meanwhile, through the timing synchronization module of the data acquisition system, the cycle number of each start-stop cycle, the start-up time when the unit speed rises from 0 to the rated speed, the start and end times of stable operation when the speed stabilizes within the rated range, the shutdown time when the speed drops from the rated to zero, and the duration of the rest interval when the unit is completely stationary are recorded sequentially. Based on this, the data integrity of the acquired acceleration time history signals is verified in real time during transmission, and abnormal data segments caused by poor sensor contact or transmission errors are eliminated. The acceleration time history signals of all channels are aligned and packaged with the corresponding operating condition timing parameters according to a unified time reference to form a time-consistent and traceable measured vibration response data stream. This data stream completely covers the entire time history of the structural vibration response of the pump station unit during multiple start-stop alternating operating condition cycles, as the pore water pressure of the saturated soft soil foundation gradually accumulates.

[0041] Step 2.3: Based on the measured vibration response data, extract the time-frequency domain features associated with the cumulative effect of pore water pressure in the foundation to obtain an initial time-frequency domain feature set. Specifically, this includes: based on the measured vibration response data, extracting vibration acceleration time history segments for the stable operating period of each start-stop cycle. The start and end times of the stable operating period are determined according to the stable operation start and end times recorded in Step 2.2. The extracted time history segments have a uniform length and cover no less than 80% of the total stable operating time of the unit; each extracted segment is sequentially processed to remove DC components and eliminate polynomial trend terms to eliminate sensor zero drift and The impact of low-frequency environmental interference on time-frequency analysis results; short-time Fourier transform is performed on each preprocessed acceleration time history segment: Hanning window is selected as the window function, and the window length is determined according to the desired frequency resolution. The frequency resolution corresponding to the window length is not higher than 1 / 20 of the unit's power frequency (for a unit with a power frequency of 5Hz, the corresponding window length is not less than 4 seconds; for a unit with a power frequency of 1Hz, the corresponding window length is not less than 20 seconds). The actual window length should be 4 to 20 seconds to keep the frequency resolution within the range of 0.05Hz to 0.25Hz; the overlap rate between adjacent windows is 50% to 75% to ensure the continuity of the time spectrum on the time axis. A Fast Fourier Transform is performed on the signal segment captured in each window, and the spectral amplitudes of each window are arranged chronologically to obtain a complete time-frequency spectrum of each measuring point in each start-stop cycle. From the time-frequency spectrum, the vibration amplitudes corresponding to the power frequency and its five harmonics of the pump station unit are extracted window by window along the time axis to form the power frequency amplitude sequence and harmonic amplitude sequence. At the same time, the ratio of energy in each frequency band to the total energy of the entire frequency band is calculated to obtain the frequency band energy ratio sequence. Then, the trajectory of the instantaneous frequency change of the power frequency component in the time-frequency spectrum is traced to obtain the instantaneous frequency change sequence over time.

[0042] Simultaneously, Hilbert-Huang transform is performed on the same preprocessed acceleration time history segment: first, empirical mode decomposition is performed, and by repeatedly screening the upper and lower envelopes of the signal, the original acceleration signal is adaptively decomposed layer by layer into several intrinsic mode function components arranged from high frequency to low frequency; each intrinsic mode function component must simultaneously satisfy two conditions: over the entire signal time length, the sum of the number of local maxima and minima is equal to or differs by at most one zero-crossing point; at any point in time of the signal, the instantaneous mean of the upper envelope constructed from local maxima and the lower envelope constructed from local minima at that point in time is 0; the screening process continues until the residual component becomes a monotonic function or the total number of extreme points of the residual component is less than two, making it impossible to continue constructing upper and lower envelopes.

[0043] Hilbert transforms were performed on each intrinsic mode function component to construct an analytical signal from which instantaneous amplitude and frequency were extracted. Among all intrinsic mode function components, low-frequency modulation characteristic components were identified, whose instantaneous frequencies were concentrated below half the power frequency and whose instantaneous amplitudes showed a slow increasing trend with the number of start-stop cycles. The identification conditions were: instantaneous frequency not greater than half the power frequency (not greater than 2.5Hz for 5Hz units and not greater than 0.5Hz for 1Hz units), and the correlation coefficient between the amplitude sequence and the number of cycles greater than 0.7, with the sign of the amplitude increment remaining non-negative for at least three consecutive cycles. The evolution law of this low-frequency modulation characteristic component is similar to that of saturated soft soil. The physical processes of the cumulative effect of pore water pressure in the foundation are highly matched on the time scale. The frequency domain features extracted by the short-time Fourier transform and the time-frequency domain features extracted by the Hilbert-Huang transform are fused: the power frequency amplitude sequence, the amplitude sequences of each order harmonic, the energy ratio sequence of each frequency band, the power frequency instantaneous frequency fluctuation sequence, and the instantaneous amplitude mean sequence and instantaneous frequency mean sequence of the low-frequency modulation characteristic component within the same start-stop cycle at the same measuring point are multidimensionally spliced ​​according to the measuring point number and the cycle number to form a multidimensional feature vector. The multidimensional feature vectors of all measuring points under all start-stop cycles together constitute the initial time-frequency domain feature set associated with the cumulative effect of pore water pressure in the foundation.

[0044] Step 2.4, based on the initial time-frequency domain feature set, construct the real-time response feature set, specifically including: based on the initial time-frequency domain feature set, arranging all multi-dimensional feature vectors in ascending order of start-stop loop number to form a feature matrix with the loop number as the row index and the feature dimension as the column index; then performing sliding window statistical analysis on each feature dimension in the feature matrix: setting a sliding window with a width covering 5 to 10 start-stop loops (10 loops when the total number of start-stop loops is large, and 5 loops when it is small), and opening the window from the first loop position. The window is moved backward by one cycle step each time along the direction of the number of cycles. Regarding the trend of mean change, the arithmetic mean of all values ​​of the feature dimension within the window is calculated after each move to obtain the trend sequence of mean change with the number of cycles. If the trend sequence of mean shows a monotonically increasing or monotonically decreasing trend, it is determined that the feature dimension has a trend correlation with the cumulative effect of foundation pore water pressure. Regarding the evolution of variance, the same sliding window is used. After each move, the variance of all values ​​of the feature dimension within the window is calculated. The gradual increase in variance indicates that the fluctuation amplitude of the feature dimension intensifies with foundation degradation. Regarding the incremental change rate between adjacent cycles, for the value of this feature dimension in two adjacent start-stop cycles, the increment is obtained by subtracting the value of the previous cycle from the value of the later cycle, and then the incremental change rate is obtained by dividing the increment by the value of the previous cycle. After calculating for each cycle, an incremental change rate sequence is formed. The gradual increase in the absolute value indicates that the sensitivity of this feature dimension to foundation degradation is increasing. The above mean trend sequence, variance evolution sequence, and incremental change rate sequence are used as extended feature dimensions and supplemented into the feature matrix according to their original feature dimensions and cycle numbers.

[0045] Correlation was calculated pairwise between each feature dimension in the expanded feature matrix. The correlation coefficient was obtained by taking the ratio of the product of the covariance and the standard deviation of each feature dimension over the entire cycle number. For feature dimension groups with an absolute correlation coefficient higher than 0.90, the feature dimension more directly related to the physical mechanism of the cumulative effect of foundation pore water pressure was retained, and the other redundant feature dimension was removed. Feature dimensions with no obvious trend of change in value over the entire cycle number and whose variance was always lower than the noise level were identified as outliers and removed. The noise level was calibrated by calculating the root mean square value of the acceleration signal at each measuring point during the shutdown and static period of the pump station unit. Feature dimensions with variance lower than this noise level were considered to be mainly composed of measurement noise rather than caused by structural response evolution, and were therefore removed. The final retained feature dimensions were all core feature dimensions that had a monotonic correlation with the cumulative effect of foundation pore water pressure and were statistically significant.

[0046] Furthermore, the feature matrix, after being filtered and retaining the core feature dimensions and supplementing the extended feature dimensions, is incrementally updated according to the real-time access sequence of the measured vibration response data. This update is performed every time a complete start-stop cycle of data is added; the incremental update uses a sliding window recursive algorithm: assuming the current accumulated... The data for each start / stop cycle is set, and the sliding window width is set to... The second loop ( Take values ​​from 10 to 20, with a typical value of 15); when adding the first... After the first iteration of the loop, extract the data for the second iteration. The feature vectors of each measurement point in the next iteration, and the most recent The eigenvectors of each iteration together constitute the eigenmatrix of the current window. For each feature dimension within the current window, its statistics are updated recursively: for the mean, the mean of the current iteration is obtained by adding the difference between the latest iteration value and the out-of-queue value to the mean of the previous iteration, divided by the window width, without needing to iterate through all the data within the window and sum again; for the variance, the variance of the previous iteration is recursively corrected by combining the latest in-queue value and the oldest out-of-queue value, avoiding repeated calculation of the sum of squared deviations of all data within the window; higher-order statistics such as kurtosis can also be updated efficiently using a similar recursive format; for the closed 1st to 1st... The statistical results from the previous historical window are retained and do not need to be recalculated; when the cumulative number of cycles... Exceeding the preset window width After that, the oldest Data in the secondary window is automatically dequeued, maintaining a constant window width. Each incremental update of this algorithm only needs to process the area covered by the current window. The second loop takes data, while a full recalculation requires iterating through all the data. The data in the next loop, when Much larger The efficiency is significantly improved, thereby ensuring the computational efficiency and timeliness of the real-time response feature set during long-term operation.

[0047] After the incremental update is completed, the feature dimensions retained in the current window are re-performed with correlation screening and outlier removal. Degenerate feature dimensions that no longer have significant correlation are removed, and newly emerging sensitive features (i.e., dimensions that were previously removed because their correlation coefficients did not reach the threshold, but have exceeded the threshold after the incremental update) are added. This ensures that each feature data in the feature matrix reflects the latest evolution of the structural vibration response. The resulting real-time response feature set can characterize the evolution of the pump station's structural vibration response with the accumulation of pore water pressure in the saturated soft soil foundation in real time and continuously. This provides a continuous, traceable, and physically meaningful data-driven foundation for the inversion identification of the time-varying degraded spring stiffness coefficient of the foundation and the dynamic updating of the benchmark digital twin in subsequent steps.

[0048] In a preferred embodiment of the present invention, step 3 above may include: Step 3.1: Based on the real-time response feature set, identify the state value of the soil stiffness parameter characterized by the spring-damped boundary constraint under the current load cycle, obtain the current state value of the soil stiffness parameter, and simultaneously acquire the number of load cycles accumulated by the pump station unit under the start-stop alternating operating conditions. Specifically, this includes: extracting the core feature dimension with monotonically correlated with the cumulative effect of foundation pore water pressure from the constructed real-time response feature set, and modulating the instantaneous amplitude mean sequence of the feature component with low frequency. As the primary inversion index, its physical basis is as follows: In each cycle of the alternating start-up and shutdown loads of the pump station unit, the excess pore water pressure inside the saturated soft soil foundation monotonically accumulates with the increase of the number of cycles, leading to a decrease in effective stress between soil particles and a degradation of the soil skeleton stiffness. This degradation effect manifests in the vibration response as a slow increase in the amplitude of the low-frequency modulation component. There is a one-to-one monotonic mapping relationship between the soil stiffness degradation degree and the soil stiffness degradation degree.

[0049] The specific calculation process for inversion identification is as follows: using the normal spring stiffness coefficients assigned at all contact interface nodes in step 1.3... Based on this, take all the contact surfaces of the base plate and the side wall. The arithmetic mean of the normal spring stiffness coefficients of each spring-damped element is used as the initial equivalent normal stiffness, denoted as . The unit is ;in This represents the total number of nodes on the contact interface where spring-damped elements are arranged. For step 1.3 The initial normal spring stiffness coefficient at each contact interface node; simultaneously, the average instantaneous amplitude of the low-frequency modulation characteristic component corresponding to the first start-stop cycle in the real-time response feature set of step 2.4. As a reference amplitude, record the initial equivalent normal stiffness corresponding to this reference amplitude. The cycle number 1 indicates that the pump station unit has completed the first complete start-stop cycle since it was put into operation.

[0050] Furthermore, for any currently accumulated load cycle number Read the first from the real-time response feature set The mean instantaneous amplitude of the low-frequency modulation characteristic component corresponding to the next start-stop cycle The current equivalent normal stiffness is calculated based on the monotonic mapping relationship between the amplitude ratio and the stiffness degradation ratio, which is expressed by the following formula:

[0051] In the formula The current specific value of the accumulated start-stop load cycles of the pump station unit and ; For experience The current state value of the equivalent normal stiffness of the soil after one load cycle, in units of ; It is a stiffness-amplitude mapping exponent and a dimensionless positive number. Its value is determined by the plasticity index of the saturated soft soil at the pump station site. and sensitivity Determined through empirical relationships; among which It is a dimensionless scalar, with a typical value range of 10 to 30; Also a dimensionless scalar, its typical value range is from 2 to 8. and , The empirical relationship between them is This relationship was established based on multiple linear regression analysis of cyclic triaxial test data of saturated soft soil samples with different plasticity indices and sensitivities, with correlation coefficients... For the above typical saturated soft soil working conditions ( , ), The value ranges from 1.0 to 3.5. When site-specific calibration data is lacking during project implementation, the median value of 2.0 is used. , Typical saturated soft soil working conditions; thus obtained That is, the soil stiffness parameter under the current load cycle The current status value is recorded simultaneously, and the start / stop cycle number accumulated by the timing synchronization module of the data acquisition system is also recorded. This number represents the number of load cycles accumulated by the pump station unit under the alternating start / stop conditions. to Stored one-to-one according to the cyclic number.

[0052] Step 3.2: Based on the current state value of the soil stiffness parameter and the accumulated number of load cycles, determine the nonlinear degradation trajectory of the soil stiffness parameter with the number of load cycles, and obtain the soil stiffness degradation trajectory data. Specifically, this includes: using the successively identified number of load cycles... Current state value of equivalent normal stiffness of soil The corresponding data pairs are arranged in ascending order of the number of iterations, forming a sequence of... A stiffness degradation sequence consisting of discrete data pairs, wherein... This represents the total number of start-stop cycles completed up to the current moment. When this sequence is plotted on a Cartesian coordinate system with the number of load cycles as the horizontal axis and the equivalent normal stiffness as the vertical axis, the scatter distribution exhibits a typical nonlinear decay pattern: in the initial stage of cyclic loading, the equivalent normal stiffness decreases rapidly with the increase of the number of cycles, the rate of decrease gradually slows down, and eventually approaches a certain stable residual stiffness level. This decay pattern is highly consistent with the physical process of excess pore water pressure in saturated soft soil under cyclic loading, which first accumulates rapidly and then tends to stabilize.

[0053] To extract continuous degradation patterns from the discrete data, a nonlinear least squares fit is performed. The degradation model used for fitting is a three-parameter exponential decay function, the expression of which is:

[0054] In the formula, The number of load cycles and ,when This indicates the initial state of the pumping station before it is put into operation; For a three-parameter exponential decay function in The function value at that point represents the experience. Predicted equivalent normal stiffness of saturated soft soil foundation after multiple load cycles, in units of ; The initial stiffness parameter is undetermined, characterizing the original stiffness level of the saturated soft soil foundation without any cyclic load disturbance, and is determined during fitting. As its initial guess; The parameter for residual stiffness is undetermined. It represents the asymptotic lower limit when the number of cycles approaches infinity. Its physical meaning is the minimum stable stiffness retained by the soil skeleton after the excess pore water pressure has been fully accumulated. The degradation rate is an undetermined parameter and is a dimensionless positive number, directly controlling the function value from... Towards The rate of decay, The larger the value, the faster the stiffness drops as the number of cycles increases.

[0055] During the fitting process, the number of load cycles is used as the basis for calculation. Equivalent normal stiffness as the independent variable Using the three-parameter exponential decay function as the dependent variable and the above three-parameter exponential decay function as the target fitting function, the Levenberg-Marquardt iterative algorithm is used in the parameter space. The algorithm searches for the optimal parameter combination. Its optimization objective is to minimize the sum of squares of the differences between the predicted values ​​of the fitted function at each discrete data point and the inversion values ​​from step 3.1. It adaptively switches between gradient descent and Gauss-Newton steps using a damped least squares strategy. After iterative convergence, the optimal estimates of the three parameters are obtained, denoted as follows: , and The asterisk (*) is derived from the optimal estimate of the nonlinear least squares fitting. After the fitting converges, a smooth and continuous nonlinear degradation curve is obtained. This curve, along with all its corresponding discrete data pairs and the fitting residual statistics, constitutes the soil stiffness degradation trajectory data. It should be noted that the three-parameter exponential decay function used in step 3.2 has three parameters. , , During the fitting phase, the variables are in the form of undetermined variables; step 3.3 then converts the optimal estimated value obtained after the fitting converges. , and Substituting into the same function structure, it is finally determined as an analytical function that can be directly called. The two are completely identical in function form, the only difference being whether the parameter is an undetermined variable or a calibrated constant. That is, step 3.2 completes the parameter calibration based on the measured data, and step 3.3 solidifies the calibration result into a continuously degenerate function that can be called in engineering.

[0056] like Figure 4 As shown, this is a degradation trajectory diagram in a two-dimensional rectangular coordinate system. The horizontal axis represents the cumulative number of start-stop load cycles since the pump station unit was put into operation, and the vertical axis represents the equivalent normal stiffness value of the saturated soft soil foundation, in units of MN / m, consistent with the unit of total stiffness of the boundary constraints defined in step 1.3. The scatter points in the figure are the measured equivalent stiffness values ​​obtained by successive inversion in the first 30 start-stop cycles. This set of values ​​shows a strict negative correlation with the extracted low-frequency modulation amplitude, which conforms to the inverse proportional mapping logic of stiffness and amplitude in step 3.1. The exponent coefficient in this mapping relationship is 2.02, which is directly calculated from the empirical formula based on the plasticity index of 22 and sensitivity of 4.8 of the saturated soft soil at the site, and is reasonably matched with the soil parameters in the site geological survey report.

[0057] The solid line in the figure represents the three-parameter exponential fitting curve. The optimal initial stiffness value obtained by fitting is 418 MN / m, which has an error of only 1.5% compared with the initial equivalent stiffness of 412 MN / m calculated based on geological data in step 1.3. The optimal residual stiffness value obtained by fitting is 192 MN / m, the optimal degradation rate coefficient is 0.048, and the goodness of fit reaches 0.91, verifying the consistency between the degradation model and the field measured data. The dashed line in the figure represents the residual stiffness asymptote, indicating that after the excess pore water pressure in the foundation is fully accumulated, the soil skeleton will still retain 46% of the initial stiffness. This proportion is consistent with the results of indoor cyclic triaxial tests on similar silty saturated soft soils. This fitting curve transforms the discrete inversion data into a continuous and predictable time-varying stiffness degradation function, providing a quantitative basis for subsequent boundary dynamic correction.

[0058] Step 3.3, based on the soil stiffness degradation trajectory data, determine the time-varying stiffness degradation function of the soil, specifically including: a three-parameter exponential degradation model and its optimal parameter estimates determined by nonlinear least squares fitting from the soil stiffness degradation trajectory data. , and The soil time-varying stiffness degradation function is determined as a function of the number of load cycles. A continuous analytic function with a unique independent variable has the following final expression:

[0059] in, To determine the optimal initial stiffness for fitting. The optimal residual stiffness determined by fitting is given in units of 1 / 2. ; To fit the determined optimal degradation rate coefficients, which are dimensionless positive numbers, the optimal estimates of the above three parameters are obtained by searching and converging in the parameter space using the Levenberg-Marquardt iterative algorithm in step 3.2. Their physical meaning and relationship with site soil parameters (plasticity index) are discussed. Sensitivity Permeability coefficient Drainage path length The correlation of ) has been fully explained in step 3.2, and will not be repeated here; the soil time-varying stiffness degradation function This provides a directly accessible quantitative basis for stiffness degradation in the dynamic correction of the benchmark digital twin spring-damping boundary constraints for subsequent steps.

[0060] In a preferred embodiment of the present invention, step 4 above may include: Step 4.1: Based on the soil time-varying stiffness degradation function and the number of load cycles accumulated by the pumping unit under start-stop alternating conditions, determine the soil stiffness degradation value corresponding to the current load cycle. Update the initial stiffness coefficient of the distributed spring-damping unit based on the soil stiffness degradation value to obtain the degradation boundary constraint parameters. Specifically, this includes: reading the number of accumulated start-stop load cycles of the pumping unit synchronously recorded and stored in step 3.1. This value represents the total number of complete alternating load cycles experienced by the saturated soft soil foundation since the pump station was put into operation; based on the soil time-varying stiffness degradation function. ,Will Substituting this into the function as the independent variable, we obtain the function in The fitted smooth value at point is denoted as . ; This value is determined by the final expression. exist The function value at this point is obtained; here, the superscript tilde indicates that this value comes from the smoothed value obtained by fitting the continuous degenerate function in step 3.3, to distinguish it from the discrete inversion value obtained directly by the amplitude ratio mapping formula in step 3.1. Both refer to experience. The equivalent normal stiffness of the soil after multiple load cycles is very similar in value, but the source paths are different: To ensure the fitted function takes smooth values, This is the direct inversion result of discrete data points.

[0061] This degradation value reflects the degree of macroscopic stiffness reduction in the foundation soil skeleton under the current cumulative cyclic load due to the accumulation of excess pore water pressure and the decrease in effective stress; further, the above fitted smooth values ​​are then... With a determined initial equivalent normal stiffness Comparison will be performed by The stiffness degradation scaling factor was calculated. It is a dimensionless scalar, and its physical meaning is the experience of The retention ratio of the equivalent normal stiffness of the foundation relative to the initial state after each cycle is determined by... It can be known Based on this, the stiffness degradation scaling factor is... Apply them one by one to all the contact interfaces of the reference digital twin in step 1.3. The initial normal spring stiffness coefficient of a distributed spring-damped element.

[0062] The specific update method is as follows: for the first... For each contact interface node, the updated normal spring stiffness coefficient is given by the formula. Obtain; among them Assign the first in step 1.3 The initial normal spring stiffness coefficient of each node, For this node experience Degraded normal spring stiffness coefficient after one load cycle, in units of This update method ensures that the stiffness degradation of all boundary constraint elements maintains a proportional relationship with the degradation of the equivalent normal stiffness in spatial distribution. Furthermore, for the stiffness coefficient of the tangential spring paired with the normal spring in the distributed spring-damped element, the same stiffness degradation scaling factor as the normal spring is used. Perform simultaneous reduction.

[0063] Correspondingly, for the damping coefficient in the distributed spring-damped unit, based on the initial damping coefficient assigned in step 1.3, a scaling factor identical to the one used for stiffness coefficient degradation is applied. The physical basis for synchronous adjustment lies in the fact that the damping energy dissipation of saturated soft soil foundations is mainly contributed by the hysteresis effect of the soil skeleton. This hysteresis effect originates from the frictional slippage and irreversible deformation at the contact points between soil particles. This mechanism weakens synchronously with the degradation of soil stiffness, i.e., the decrease in contact stiffness between soil particles and the reduction in the number of contact points. Existing cyclic triaxial test studies have shown that the damping ratio and shear modulus ratio of saturated soft soil under cyclic loading are approximately inversely proportional. Degradation accompanied by damping ratio The damping ratio changes approximately proportionally; therefore, within the range of engineering accuracy, reducing the damping coefficient using the same scaling factor as the stiffness degradation is an acceptable simplification. For higher accuracy, an independent damping ratio degradation curve can be established based on cyclic single-shear or cyclic triaxial tests of saturated soft soil samples from the site. In this step, the damping coefficient is independently updated using this curve instead of the stiffness degradation scaling factor, ensuring that the damping characteristics of the boundary constraint are coordinated and matched with the degraded stiffness characteristics. The updated normal spring stiffness coefficients, tangential spring stiffness coefficients, and damping coefficients together constitute the degraded boundary constraint parameters, which comprehensively describe the parameters under the current load cycle. Degradation boundary conditions of the dynamic impedance characteristics of pump station structures on subsaturated soft soil foundation.

[0064] Step 4.2: Based on the degenerate boundary constraint parameters, replace the stiffness coefficients corresponding to the initial spring-damped boundary constraints in the reference digital twin to obtain a time-varying boundary twin model with degenerate boundary states. Specifically, this includes: based on the degenerate boundary constraint parameters, locating the element attribute domains corresponding to all distributed spring-damped elements in the finite element model data structure of the reference digital twin. This finite element model data structure of the reference digital twin adopts a node-element topology organization method, where the contact interface node sets related to the foundation dynamic boundary have been classified and identified in the structured mesh of Step 1.2 according to the base plate contact interface and the side wall contact interface; in the element attribute management module of the finite element model data structure, using the node numbers of all contact interfaces defined in Step 1.2 as indexes, traversing the spring-damped element attribute records attached to each node, all nodes can be matched one by one. The domain of element attributes corresponding to each distributed spring-damped element; this domain records all parameter values ​​of the initial spring-damped boundary constraints assigned in step 1.3, including the initial normal spring stiffness coefficient, initial tangential spring stiffness coefficient, and initial damping coefficient of each node.

[0065] Furthermore, the updated normal spring stiffness coefficient in the degenerate boundary constraint parameters is... Replace the corresponding initial normal spring stiffness coefficient node by node This involves updating the normal constraint stiffness of each contact interface node from its initial value to a reduced value reflecting the current foundation degradation state. Correspondingly, the updated tangential spring stiffness coefficients in the degradation boundary constraint parameters replace the corresponding initial tangential spring stiffness coefficients node by node, completing the synchronous update of the tangential constraint stiffness. Based on this, the updated damping coefficients in the degradation boundary constraint parameters replace the corresponding initial damping coefficients node by node, completing the synchronous update of the boundary damping characteristics. After the replacement operation is completed, all nodes in the reference digital twin will be updated. The stiffness and damping coefficients of all distributed spring-damped boundary constraint elements have been updated to match the current load cycle. The corresponding degenerate state values ​​result in a time-varying boundary twin model with degenerate boundary states.

[0066] At this point, the variable boundary twin model fully inherits all the features of the benchmark digital twin in terms of geometric structure and material properties, including the spatial geometric model of the pump station base plate-side wall-saturated soft soil foundation constructed in step 1.1, the structured mesh partitioning scheme divided in step 1.2, and the constitutive parameters of the reinforced concrete structure material of the pump station assigned in step 1.2, thereby maintaining the consistency of the finite element model of the main structure of the pump station. At the boundary constraint level, the spring-damped boundary constraints of the variable boundary twin model are different from the initial steady boundaries assigned based on the original geological survey data in the benchmark digital twin. Instead, they are dynamically updated by the soil time-varying stiffness degradation function obtained from the measured vibration response in step 4.1. Specifically, the stiffness parameters of the boundary constraints are transformed from initial values ​​to current degradation values. Essentially, the long-term stiffness degradation information of the foundation extracted and quantified from the real-time operation vibration response of the pump station unit in steps 2 and 3 is injected into the boundary conditions of the benchmark digital twin in the form of calculable and updatable parameters. At this point, the variable boundary twin model can accurately characterize the dynamic interaction boundary state between the pump station floor, side walls, and saturated soft soil foundation at the current foundation degradation stage. This lays a numerical foundation reflecting the true degradation process of the foundation for subsequent structural dynamic response prediction and pump station operation status assessment based on this model.

[0067] In a preferred embodiment of the present invention, step 5 above may include: Step 5.1: Based on the time-varying boundary twin model with degenerate boundary state, using the degenerate boundary constraint parameters as the boundary conditions for structural dynamic time history analysis, apply alternating hydraulic excitation loads corresponding to the pump station unit under start-up and shutdown alternating operating conditions to obtain the structural stress-strain response. Specifically, the time-varying boundary twin model with degenerate boundary state fully inherits all the characteristics of the baseline digital twin constructed in Step 1 in terms of geometric structure, material properties, and mesh generation. Its boundary constraint parameters have been updated from those in Step 4.1 based on the foundation stiffness degradation function to those of the current load cycle. The degenerate boundary constraint parameters are as follows; specifically, the degenerate boundary constraint parameters include all... The updated normal spring stiffness coefficient, tangential spring stiffness coefficient, and damping coefficient of each distributed spring-damping element fully describe the degraded boundary conditions of the dynamic impedance characteristics of the pump station structure on the saturated soft soil foundation under the current load cycle. The above-mentioned degraded boundary constraint parameters are used as the boundary conditions for the structural dynamic time history analysis, replacing the initial steady spring-damping boundary constraints in the reference digital twin.

[0068] Based on this, the alternating hydraulic excitation load corresponding to the alternating start-up and shutdown conditions of the pump station unit is determined. This alternating hydraulic excitation load originates from the transient water pressure pulsation caused by the rapid change in water flow state during the start-up and shutdown of the pump station unit. Its load characteristics are consistent with the alternating start-up and shutdown conditions corresponding to the real-time operating vibration response of the pump station unit collected in step 2. This load is applied to the hydraulic excitation area such as the flow channel wall and the inner surface of the bottom plate of the pump station structure in the form of time-history surface pressure. The amplitude, frequency and duration of the load time history are determined by the operating parameters such as the rated head, rated speed and start-up and shutdown transition time of the pump station unit.

[0069] Furthermore, within the finite element solution framework of the time-varying boundary twin model, solution control parameters for dynamic time history analysis are defined. These parameters include the time step, total analysis duration, and damping model parameters. The time step should not exceed a certain percentage of the period corresponding to the dominant frequency of the alternating hydraulic excitation load. To ensure solution accuracy, the total analysis duration covers a single complete start-up and shutdown transition process of the pumping station unit. Based on the above settings, the direct integration method is used to solve the structural dynamic time history of the time-varying boundary twin model, calculating the dynamic response of the pumping station structure under the combined action of degraded foundation boundary constraints and alternating hydraulic excitation loads. After the solution is completed, the stress time history and strain time history of each finite element node of the pumping station structure are extracted from the numerical results to form the structural stress-strain response. This structural stress-strain response fully records the changes in stress and strain components at various locations of the pumping station base plate, side walls, and other structural components over time during a single start-up and shutdown cycle, providing load effect input for subsequent fatigue damage assessment.

[0070] Step 5.2: Based on the structural stress-strain response and a preset fatigue damage accumulation criterion, calculate the fatigue damage increment of the pump station structure under the combined action of alternating hydraulic vibration and degraded foundation. Specifically, this includes: reading the structural stress-strain response, which contains the stress time history and strain time history of each finite element node of the pump station structure within a single complete start-stop cycle; extracting the equivalent stress amplitude from the stress time history of each node as a load effect characteristic quantity for fatigue damage assessment; specifically, for the first... Each node is used to identify the maximum stress peak value from its stress-time history curve. and minimum stress valley value Then the equivalent stress amplitude at that node For nodes under multiaxial stress, the stress components at each time step are first synthesized into a von Mises equivalent stress time history. Then, the equivalent stress amplitude is extracted from it. Based on the extracted equivalent stress amplitude... Based on the preset fatigue damage accumulation criterion, the fatigue damage increment of the node in this start-stop cycle is calculated.

[0071] The preset fatigue damage accumulation criterion adopts the Miner-Palmgren linear accumulation criterion, which defines the damage increment generated in each load cycle as the share of fatigue life consumed in that cycle; for the th Each node, and its fatigue damage increment during this start-stop cycle. In the formula, For the first Each node at the current equivalent stress amplitude The fatigue life corresponding to the action, that is, the equivalent stress amplitude at that node. The total number of cycles that the load can withstand before fatigue failure. The fatigue SN curve of the pump station structural materials was used to determine the relationship between the equivalent stress amplitude and fatigue life, which describes the power function relationship between the equivalent stress amplitude and fatigue life. The calculation formula is: ;in The fatigue strength coefficient of the pump station structural materials (dimensions are the same as stress, units are...). ), The fatigue strength index (a dimensionless positive number) is obtained through experimental calibration from the fatigue strength SN curve of the reinforced concrete material of the pump station; it is then substituted into the fatigue damage increment calculation formula. , obtained the The fatigue damage increment of each node during this start-stop cycle This formula satisfies the basic laws of fatigue mechanics, namely the equivalent stress amplitude. The larger the size, the longer the fatigue life. The shorter the cycle, the greater the increase in fatigue damage per cycle. The larger.

[0072] Following the above calculation method, the fatigue damage increment of each finite element node of the pump station structure during this start-stop cycle is calculated one by one, obtaining the result compared with the current load cycle. The corresponding distribution of fatigue damage increments across the entire structure reflects the single-cycle fatigue damage of the pump station structure under the current degraded foundation boundary conditions caused by alternating hydraulic excitation loads. The magnitude of this damage is influenced by both the amplitude of the hydraulic excitation load and the degree of foundation stiffness degradation. The degradation of foundation stiffness affects the dynamic response characteristics and stress distribution of the pump station structure by changing the structural boundary constraints, thereby indirectly altering the equivalent stress amplitude and corresponding fatigue damage increments at each node.

[0073] Step 5.3: Map the fatigue damage increment to the preset visible nodes in the time-varying boundary twin model to update the damage state of the preset visible nodes and obtain the time-varying twin simulation results. Specifically, this includes: reading the fatigue damage increment distribution of the entire structure, and simultaneously reading the preset set of visible nodes in the time-varying boundary twin model generated in step 4.2. The preset set of visible nodes is a subset of key monitoring nodes pre-selected and identified based on the structured mesh partitioning results in step 1.2 when constructing the baseline digital twin in step 1.4. These visible nodes are distributed in stress concentration or damage-sensitive locations such as the corner area connecting the pump station base plate and side wall, the central area of ​​the base plate, the central area of ​​the side wall, and the junction area between the flow channel wall and the main structure, and are used to bear and display the fatigue damage state information of the pump station structure.

[0074] The calculated fatigue damage increments of each finite element node are mapped one by one to the pre-defined visible nodes in the time-varying boundary twin model according to the node number correspondence. For finite element nodes that do not belong to the set of visible nodes, their fatigue damage increments are retained as intermediate calculation data but do not participate in the visualization update of the damage state. After mapping, the current fatigue damage increment received by each visible node is... Historical fatigue damage accumulated at this node Accumulate the results to obtain the number of times the node has experienced the current load cycle. Cumulative fatigue damage Its update formula is In the formula The specific meaning is the first The visible nodes in the previous load cycle Historical cumulative fatigue damage value at the end of the period As a dimensionless scalar, when the cumulative fatigue damage value of a certain visible node... The first time the preset fatigue damage threshold is reached or exceeded When the node is in a critical state of fatigue damage, it is determined that the node has entered the critical state of fatigue damage; the preset critical threshold for fatigue damage is... According to the Miner-Palmgren linear accumulation criterion, the value is set to 1.0, indicating that the node location has reached the critical condition for fatigue failure.

[0075] The updated cumulative fatigue damage values ​​of all visible nodes and the fatigue damage critical state determination results of each node are integrated with the geometric structure and boundary constraint parameters of the time-varying boundary twin model generated in step 4.2 to jointly constitute the time-varying twin simulation result; this time-varying twin simulation result includes the results under the current load cycle. The complete state information of the pump station structure is obtained by integrating three levels of state information: first, the inherent geometric structure and material properties of the baseline digital twin in step 1; second, the foundation stiffness degradation state characterized by the degradation boundary constraint parameters updated according to the soil time-varying stiffness degradation function in step 4.1; and third, the structural fatigue damage state characterized by the cumulative fatigue damage distribution of visible nodes updated in this step. Thus, the time-varying twin simulation results realize a holographic mapping of the multi-dimensional deterioration state of the pump station structure under the combined action of alternating hydraulic excitation load and degraded foundation boundary conditions, providing a numerical twin basis for pump station operation and maintenance decisions that reflects the dual evolution process of foundation degradation and structural damage.

[0076] In a preferred embodiment of the present invention, step 6 above may include: Step 6.1: Based on the time-varying twin simulation results, extract the damage state of each part of the pump station structure to obtain structural damage distribution data. Specifically, this includes: the time-varying twin simulation results fully encompass the damage distribution data under the current load cycle. The three-layer state information of the pump station structure includes the inherent geometric structure and material properties of the baseline digital twin, the foundation stiffness degradation state information represented by the degradation boundary constraint parameters in step 4.1, and the cumulative fatigue damage distribution information after updating all visible nodes in step 5.3. The cumulative fatigue damage dataset of all visible nodes is located from the time-varying twin simulation results. This dataset records the cumulative fatigue damage of each visible node after experiencing the current load cycle. Cumulative fatigue damage value Based on the geometrical distribution of visible nodes in the pump station structure space, the cumulative fatigue damage values ​​are calculated. The data is mapped back to the actual structural parts of the pump station, such as the base plate, side walls, flow channel walls, and connecting corners.

[0077] Specifically, based on the classification labels of the structural parts to which visible nodes belong during the structured mesh generation in step 1.2, the cumulative fatigue damage value of each visible node is assigned to its corresponding structural part category, thus forming damage data groupings based on structural parts as the organizational dimension. For each structural part, the maximum value, average value, and damage distribution gradient of the cumulative fatigue damage are extracted from all visible nodes it contains, serving as the damage state characterization of that structural part. Furthermore, the damage state characterization of each structural part, along with the cumulative fatigue damage values ​​and spatial coordinates of each visible node, are comprehensively processed to generate structural damage distribution data. This structural damage distribution data fully presents the pump station structure under the current load cycle in a spatially distributed form. The degree of fatigue damage in each part provides a quantitative basis for subsequent determination of the limit state and identification of dangerous parts.

[0078] Step 6.2: Based on the structural damage distribution data and the preset long-term performance limit state, determine whether each structural component exceeds the preset safety threshold, and identify the target structural component that exceeds the preset safety threshold. Specifically, this includes: the structural damage distribution data includes the maximum and average cumulative fatigue damage values ​​and damage distribution gradients of each structural component of the pump station structure, as well as the cumulative fatigue damage values ​​and spatial coordinates of each visible node; retrieving the preset long-term performance limit state judgment standard, which is a performance control index system pre-established based on the pump station reinforced concrete structure design code and structural durability assessment criteria. This system defines the upper limit of fatigue damage allowed for various structural components of the pump station structure during long-term service; specifically, for the corner area connecting the pump station base plate and sidewalls, the preset safety threshold is taken as the critical fatigue damage threshold. The reduction value The reduction factor The value ranges from 0.60 to 0.85, determined based on the safety level of the pump station structure. For pump station structures with a safety level of I (design service life of 100 years, serious consequences of failure), the value is taken as follows: For pump station structures with a safety level of II (design service life of 50 years), take... For pump station structures with a safety level of III (design service life of 30 years), take... For the interface area between the flow channel wall and the main structure, the preset safety threshold is taken as the critical threshold for fatigue damage. It is 0.7 to 0.9 times, and the specific value is determined based on the dominant frequency of the water flow pulsation pressure in the flow channel. With the first natural frequency of the structure The degree of proximity is determined when When there is a risk of resonance, a lower reduction factor of 0.70 is used; when or When the frequency domain is fully separated, a higher reduction factor of 0.90 is used, and other operating conditions are determined by linear interpolation.

[0079] Based on the aforementioned preset safety thresholds, the maximum cumulative fatigue damage value of each structural component obtained in step 6.1 is compared with the corresponding preset safety threshold. When the maximum cumulative fatigue damage value of a structural component is greater than or equal to the preset safety threshold, the structural component is determined to be outside the safe range. All structural components determined to be outside the safe range are summarized and identified as target structural components. The target structural component represents the component that is outside the safe range during the current load cycle. Weak areas in the pump station structure that are close to or have reached the long-term performance limit are the key focus of subsequent dangerous phase identification and operation and maintenance intervention. At the same time, structural parts that have not exceeded the maximum value of cumulative fatigue damage but are close to the preset safety threshold are marked as warning structural parts and their current damage margin information is recorded separately. The warning structural parts and their damage margin information will be output together with the target structural parts information, providing a graded structural safety assessment conclusion for pump station operation and management.

[0080] Step 6.3: Based on the target structural parts and the loading phase of the alternating hydraulic excitation load corresponding to the time-varying boundary twin model when calculating the fatigue damage increment, the target structural parts are marked with dangerous phases to obtain a time-varying twin model that marks the dangerous phases. Specifically, this includes: the target structural parts list listing the parts that are in the current load cycle. The cumulative fatigue damage of all structural parts that have exceeded the preset safety threshold and their corresponding cumulative fatigue damage values ​​are recorded. At the same time, the complete time history curve of the alternating hydraulic excitation load applied during the structural dynamic time history analysis in step 5.1 is read. This time history curve records the complete time history of the alternating hydraulic excitation load from the initial zero value to the peak value and then back to zero during a single start-up and shutdown transition of the pump station unit. Its load characteristics strictly correspond to the three-stage transition process of the pump station unit: the start-up acceleration stage, the rated operation stage, and the shutdown deceleration stage.

[0081] The loading phase distribution of the load is identified from the complete time history curve of the alternating hydraulic excitation load. The loading phase refers to the time stage position of the alternating hydraulic excitation load in a single start-stop cycle, specifically divided into three characteristic phase intervals: the start-up impact phase, the rated alternating phase, and the shutdown backflow phase. The start-up impact phase corresponds to the transition process of the pump station unit from a static state to the rated speed. During this stage, the water flow suddenly accelerates from a static state, generating a positive water hammer impact load. The rated alternating phase corresponds to the stable operation process of the pump station unit at the rated speed. During this stage, the water flow pressure pulsates periodically with the blade passage frequency as the main frequency. The shutdown backflow phase corresponds to the transition process of the pump station unit decelerating from the rated speed to a static state. During this stage, the water flow generates a reverse backflow pressure load due to inertia.

[0082] Further, by tracing back from the structural stress-strain response in step 5.1, the stress time histories of each visible node corresponding to the target structural part in a single start-stop cycle are located. The moment when the maximum stress peak occurs for each target structural part is extracted from the stress time histories, and this moment is mapped onto the time axis of the alternating hydraulic excitation load time history curve to determine the load loading phase borne by each target structural part in this start-stop cycle. For a certain target structural part, if its maximum stress peak time falls within the start-up impact phase interval, the dangerous phase of the target structural part is marked as the start-up impact phase; if its maximum stress peak time falls within the rated alternating phase interval, the dangerous phase is marked as the rated alternating phase; if its maximum stress peak time falls within the shutdown backflow phase interval, the dangerous phase is marked as the shutdown backflow phase.

[0083] According to the above mapping rules, each target structural part identified in step 6.2 is marked with a dangerous phase. The dangerous phase mark is attached to the corresponding visible node in the time-varying boundary twin model in the form of attribute tags, so that each marked visible node carries not only the cumulative fatigue damage state information, but also the load dangerous phase information that caused its damage to exceed the limit. After completing the dangerous phase marking of all target structural parts, the marking information is integrated with the time-varying twin simulation results generated in step 5.3 to obtain the time-varying twin model with marked dangerous phases. This time-varying twin model with marked dangerous phases adds a fourth layer of information on the basis of the original three layers of state information, namely the correlation mapping relationship between the target structural parts and their corresponding load dangerous phases. Thus, the model can not only answer the question of which parts of the pump station structure have exceeded the safety threshold, but also further reveal under which characteristic phase of the start-stop alternating load these target structural parts are subjected to the most unfavorable load effect, thereby providing a precise phase dimension intervention basis for the optimization of pump station operation control strategy.

[0084] Step 6.4: The degenerate boundary states in the time-varying twin model identifying the dangerous phase are used as the initial spring-damped boundary constraints for the next round of real-time simulation analysis and fed back to the reference digital twin to form a closed-loop iteration. Specifically, this includes: the time-varying twin model identifying the dangerous phase fully carries the current load cycle. The system collects all state information, including the geometry and material properties of the baseline digital twin, the foundation stiffness degradation state represented by the degradation boundary constraint parameters, the cumulative fatigue damage distribution of all visible nodes, and the critical phase markers for the target structure. Degradation boundary constraint parameters are extracted from the time-varying twin model that identifies the critical phases. These parameters, calculated in step 4.1, contain all... A distributed spring-damping unit in the current load cycle The updated normal spring stiffness coefficient, updated tangential spring stiffness coefficient, and updated damping coefficient provide a complete description of the boundary constraint state during the current stage of foundation stiffness degradation.

[0085] The aforementioned degenerate boundary constraint parameters are encapsulated as the initial spring-damped boundary constraints for the next round of real-time simulation analysis, where the next round of real-time simulation analysis refers to the analysis for subsequent load cycles. The complete simulation analysis process executed from step 2 to step 6; the current load cycle The degraded boundary constraint parameters are used as initial values ​​and passed to the baseline digital twin in step 4.1 of the next round of simulation analysis. This ensures that the starting point of the boundary constraints in the next round of simulation analysis is no longer based on the initial steady spring-damped boundary constraints assigned by the original geological data in step 1.3, but on the degraded boundary state updated in this cycle. At the same time, the cumulative fatigue damage state information and the dangerous phase identification information of the target structural parts in this round are also passed to the next round of simulation analysis as historical cumulative fatigue damage in step 5.3 of the next round. The initial values ​​and key areas of focus for structural safety monitoring are referenced.

[0086] The aforementioned process of transmitting degradation boundary states, cumulative fatigue damage states, and dangerous phase identification information constitutes the process from the current load cycle. To the next load cycle The complete state information closed-loop feedback link; this closed-loop feedback link injects the foundation stiffness degradation effect and structural fatigue damage effect generated by each start-up and shutdown cycle of the pump station unit into the boundary conditions and structural state parameters of the reference digital twin in a recursive and cumulative manner. Thus, the complete simulation analysis process consisting of steps 2 to 6 forms a self-updating and recursively evolving closed-loop iterative mechanism.

[0087] This closed-loop iterative mechanism ensures that the initial spring-damped boundary constraints of the baseline digital twin are continuously updated as the actual operation of the pumping station progresses, so that the starting point of each simulation analysis always reflects the true deterioration state of the foundation and structure after the end of the previous operating cycle. Ultimately, this closed-loop iterative mechanism realizes the synchronous tracking and dynamic mapping of the time-varying stiffness degradation of the foundation and the progressive fatigue damage of the structure in the digital twin of the pumping station structure throughout its entire operating life cycle, providing a digital twin support platform with physical consistency and recursive continuity for long-term decision-making in pumping station operation and maintenance.

[0088] like Figure 2 As shown, embodiments of the present invention also provide a pump station structural performance simulation and analysis system based on digital twins, including: The benchmark twin construction module is used to construct a soil-structure interaction virtual entity that reflects the dynamic interaction mechanism between the pump station's base plate, side walls, and saturated soft soil foundation based on the initial design finite element model of the pump station to be monitored, and obtain a benchmark digital twin containing initial spring-damped boundary constraints and preset visible nodes. The response feature extraction module is used to continuously access the measured vibration response data of the pump station unit under the start-stop alternating operating conditions based on the reference digital twin, and extract the time-frequency domain features associated with the cumulative effect of pore water pressure in the foundation from the measured vibration response data to obtain a real-time response feature set. The stiffness degradation identification module is used to identify the soil stiffness parameters characterized by the spring-damped boundary constraint based on the real-time response feature set, determine the nonlinear degradation trajectory of the soil stiffness parameters with the number of load cycles, and obtain the time-varying stiffness degradation function of the soil. The time-varying boundary correction module is used to dynamically correct the spring-damped boundary constraints in the reference digital twin based on the soil time-varying stiffness degradation function, so as to obtain a time-varying boundary twin model with degraded boundary state. The fatigue damage mapping module is used to calculate the fatigue damage increment of the pump station structure under the combined action of alternating hydraulic vibration and degraded foundation according to the time-varying boundary twin model, and map the fatigue damage increment to the preset visible node in the time-varying boundary twin model to obtain the time-varying twin simulation result. The dangerous phase identification and feedback module is used to identify dangerous phases of structural parts that exceed a preset safety threshold based on the time-varying twin simulation results and the preset long-term performance limit state. The dangerous phase identification is attached to the time-varying boundary twin model to obtain a time-varying twin model with dangerous phases. The degenerate boundary state in the time-varying twin model with dangerous phases is used as the initial spring-damped boundary constraint for the next round of real-time simulation analysis and fed back to the reference digital twin in the reference twin construction module to form a closed-loop iteration.

[0089] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0090] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for simulation analysis of pump station structural performance based on digital twins, characterized in that, The method includes: Based on the initial design finite element model of the pumping station to be monitored, a virtual entity reflecting the dynamic interaction mechanism between the pumping station's base slab, side walls, and saturated soft soil foundation is constructed. This results in a baseline digital twin containing initial spring-damped boundary constraints and preset visible nodes, including: Based on the initial design finite element model, the soil-structure contact interface between the pump station base slab, side walls and saturated soft soil foundation is extracted; Based on the soil-structure contact interface, distributed spring-damping elements are configured on the initial design finite element model to form a virtual boundary characterizing the dynamic impedance characteristics of the saturated soft soil foundation to the pump station structure. Based on the foundation survey data obtained in advance during the construction and initial operation phases of the pumping station, the distributed spring-damping unit is assigned an initial stiffness coefficient and an initial damping coefficient to obtain the initial spring-damping boundary constraints. Based on the initial spring-damped boundary constraints and the virtual boundary, finite element surface nodes at the corners connecting the base plate and side walls, the center of the base plate, the middle of the side walls, and the junction area of ​​the flow channel walls on the pump station structure are selected as preset visible nodes to construct a soil-structure interaction virtual entity that reflects the dynamic interaction mechanism between the pump station base plate, side walls, and saturated soft soil foundation; this soil-structure interaction virtual entity is used as the reference digital twin; wherein the preset visible nodes are used to bear and display the fatigue damage state information of the pump station structure; Based on the benchmark digital twin, the measured vibration response data of the pump station unit under alternating start-stop conditions are continuously accessed. Time-frequency domain features related to the cumulative effect of pore water pressure in the foundation are extracted from the measured vibration response data to obtain a real-time response feature set. Based on the real-time response feature set, the soil stiffness parameters characterized by the spring-damped boundary constraint are identified, the nonlinear degradation trajectory of the soil stiffness parameters with the number of load cycles is determined, and the time-varying stiffness degradation function of the soil is obtained. The spring-damped boundary constraints in the reference digital twin are dynamically corrected based on the soil time-varying stiffness degradation function to obtain a time-varying boundary twin model with degraded boundary state. The fatigue damage increment of the pump station structure under the combined action of alternating hydraulic vibration and degraded foundation is calculated based on the time-varying boundary twin model; the increment is mapped to the preset visible nodes of the time-varying boundary twin model to obtain the time-varying twin simulation results. Based on the time-varying twin simulation results and the preset long-term performance limit state, dangerous phases are marked for structural parts that exceed the preset safety threshold. Dangerous phase marks are added to the time-varying boundary twin model to obtain a time-varying twin model with marked dangerous phases. The degenerate boundary state in the time-varying twin model with marked dangerous phases is used as the initial spring-damped boundary constraint for the next round of real-time simulation analysis and fed back to the reference digital twin.

2. The method for simulation analysis of pump station structural performance based on digital twins according to claim 1, characterized in that, Based on the baseline digital twin, the measured vibration response data of the pump station unit under alternating start-up and shutdown conditions are continuously input. Time-frequency domain features related to the cumulative effect of pore water pressure in the foundation are extracted from the measured vibration response data to obtain a real-time response feature set, including: Based on the soil-structure interaction virtual entity in the reference digital twin that reflects the dynamic interaction mechanism between the pump station base plate, side walls and saturated soft soil foundation, the deployment location of vibration measuring points associated with the cumulative effect of foundation pore water pressure is determined, and the measuring point deployment information is obtained. Based on the aforementioned measurement point deployment information, the measured vibration response data of the pump station unit under alternating start-stop conditions are continuously collected. Based on the measured vibration response data, time-frequency domain features associated with the cumulative effect of pore water pressure in the foundation are extracted to obtain an initial time-frequency domain feature set; Based on the initial time-frequency domain feature set, the real-time response feature set is constructed.

3. The method for simulation analysis of pump station structural performance based on digital twins according to claim 2, characterized in that, Based on the real-time response feature set, the soil stiffness parameters characterized by the spring-damped boundary constraints are identified, and the nonlinear degradation trajectory of the soil stiffness parameters with the number of load cycles is determined, resulting in the time-varying stiffness degradation function of the soil, including: Based on the real-time response feature set, identify the state value of the soil stiffness parameter characterized by the spring-damped boundary constraint under the current load cycle, obtain the current state value of the soil stiffness parameter, and simultaneously obtain the number of load cycles accumulated by the pump station unit under the start-stop alternating working conditions. Based on the current state value of the soil stiffness parameter and the accumulated number of load cycles, the nonlinear degradation trajectory of the soil stiffness parameter with the number of load cycles is determined, and the soil stiffness degradation trajectory data is obtained. Based on the soil stiffness degradation trajectory data, the time-varying stiffness degradation function of the soil is determined.

4. The method for simulation analysis of pump station structural performance based on digital twins according to claim 3, characterized in that, Based on the dynamic correction of the spring-damped boundary constraints in the reference digital twin using the time-varying stiffness degradation function of the soil, a time-varying boundary twin model with degraded boundary states is obtained, including: Based on the soil time-varying stiffness degradation function and the number of load cycles accumulated by the pump station unit under start-stop alternating conditions, the soil stiffness degradation value corresponding to the current load cycle is determined, and the initial stiffness coefficient of the distributed spring-damping unit is updated based on the soil stiffness degradation value to obtain the degradation boundary constraint parameters. Based on the degenerate boundary constraint parameters, the stiffness coefficients corresponding to the initial spring-damped boundary constraints in the baseline digital twin are replaced to obtain a time-varying boundary twin model with degenerate boundary states.

5. The method for simulation analysis of pump station structural performance based on digital twins according to claim 4, characterized in that, The fatigue damage increment of the pump station structure under the combined action of alternating hydraulic excitation and degraded foundation was calculated based on the time-varying boundary twin model. Mapping the increment to the preset visible node of the time-varying boundary twin model yields the time-varying twin simulation results, including: Based on the time-varying boundary twin model with degenerate boundary state, the degenerate boundary constraint parameters are used as the boundary conditions for structural dynamic time history analysis. Alternating hydraulic excitation loads corresponding to the pump station unit in start-up and shutdown alternating conditions are applied to obtain the structural stress-strain response. Based on the stress-strain response of the structure and combined with the preset fatigue damage accumulation criterion, the fatigue damage increment of the pump station structure under the combined action of alternating hydraulic vibration and degraded foundation is calculated. The fatigue damage increment is mapped to a preset visible node of the time-varying boundary twin model to update the damage state of the preset visible node and obtain the time-varying twin simulation result.

6. The method for simulation analysis of pump station structural performance based on digital twins according to claim 5, characterized in that, Based on the time-varying twin simulation results and the preset long-term performance limit state, dangerous phases are marked for structural parts that exceed the preset safety threshold, and time-varying twin models with marked dangerous phases are obtained. The degenerate boundary states in the time-varying twin model that identify the dangerous phase are used as the initial spring-damped boundary constraints for the next round of real-time simulation analysis, including: Based on the time-varying twin simulation results, the damage state of each part of the pump station structure is extracted to obtain structural damage distribution data; Based on structural damage distribution data and preset long-term performance limit states, determine whether each structural part exceeds the preset safety threshold, and identify the target structural parts that exceed the preset safety threshold. Based on the target structural part, and combined with the loading phase of the alternating hydraulic excitation load corresponding to the time-varying boundary twin model when calculating the fatigue damage increment, the target structural part is marked with a dangerous phase, and a time-varying twin model with a dangerous phase is obtained. The degenerate boundary states in the time-varying twin model that identify the dangerous phase are used as the initial spring-damped boundary constraints for the next round of real-time simulation analysis and fed back to the reference digital twin to form a closed-loop iteration.

7. A pump station structural performance simulation and analysis system based on digital twins, wherein the system implements the method as described in any one of claims 1 to 6, characterized in that, include: The benchmark twin construction module is used to construct a soil-structure interaction virtual entity that reflects the dynamic interaction mechanism between the pump station's base plate, side walls, and saturated soft soil foundation based on the initial design finite element model of the pump station to be monitored, and obtain a benchmark digital twin containing initial spring-damped boundary constraints and preset visible nodes. The response feature extraction module is used to continuously access the measured vibration response data of the pump station unit under the start-stop alternating operating conditions based on the reference digital twin, and extract the time-frequency domain features associated with the cumulative effect of pore water pressure in the foundation from the measured vibration response data to obtain a real-time response feature set. The stiffness degradation identification module is used to identify the soil stiffness parameters characterized by the spring-damped boundary constraint based on the real-time response feature set, determine the nonlinear degradation trajectory of the soil stiffness parameters with the number of load cycles, and obtain the time-varying stiffness degradation function of the soil. The time-varying boundary correction module is used to dynamically correct the spring-damped boundary constraints in the reference digital twin based on the soil time-varying stiffness degradation function, so as to obtain a time-varying boundary twin model with degraded boundary state. The fatigue damage mapping module is used to calculate the fatigue damage increment of the pump station structure under the combined action of alternating hydraulic excitation and degraded foundation according to the time-varying boundary twin model, and map the fatigue damage increment to the preset visible nodes of the time-varying boundary twin model to obtain the time-varying twin simulation results. The dangerous phase identification and feedback module is used to identify dangerous phases of structural parts that exceed a preset safety threshold based on the time-varying twin simulation results and the preset long-term performance limit state. The dangerous phase identification is attached to the time-varying boundary twin model to obtain a time-varying twin model with dangerous phases. The degenerate boundary state in the time-varying twin model with dangerous phases is used as the initial spring-damped boundary constraint for the next round of real-time simulation analysis and fed back to the reference digital twin in the reference twin construction module to form a closed-loop iteration.

Images