BIM-based collaborative control system for foundation pit support dismantling and stress conversion

Abstract

The application belongs to the technical field of engineering monitoring, and provides a foundation pit inner support dismantling and stress conversion collaborative control system based on BIM, which comprises an initial digital twin construction module: a BIM model of a foundation pit inner support system and a finite element analysis model associated therewith are acquired, and support stress data obtained through actual monitoring, material mechanics parameters and boundary constraint conditions are introduced to form an initial digital twin. Through a complete closed loop scheme, after each cutting step, the subsequent operation can be reevaluated and replanned according to the field measured data, when the actual measured stress does not match the expected value, instead of falling into a dilemma of continuing to execute the risk accumulation and suspending the construction to ensure the construction period, the deviation is automatically and scientifically dealt with through the ways of correcting the model parameters and re-planning the cutting sequence, so that the construction safety is ensured and unnecessary construction period delay is reduced.

Description

Technical Field

[0001] This invention belongs to the field of engineering monitoring technology, specifically a BIM-based collaborative control system for the removal of internal foundation pit supports and the transformation of stress. Background Technology

[0002] In the construction of deep foundation pit projects, temporary internal support systems are usually set up inside the pit to resist the pressure of the soil outside the pit. After the foundation pit bottom slab and basement floor slab are completed, the temporary internal support needs to be gradually removed so that the permanent structure can gradually bear the soil load. This process is called force transfer. At present, BIM technology has been widely used in foundation pit projects. For example, BIM models can be used to split the support, divide the dividing points, and determine the dismantling length, thereby generating a static dismantling plan. However, in actual engineering applications, although the BIM model specifies the order of dismantling supports in different areas, when workers cut a support beam, it is not clear in the planning scheme which section of the beam to start first and how to control the cutting depth of each cut. Taking a support beam that needs to be cut into three sections as an example, whether to cut the left end first, the middle end first, or the right end first, different choices will lead to completely different stress paths. After the first section is cut, the remaining part may instantly become a cantilever beam, and the bending moment and deformation state are completely different from the complete beam. Due to the lack of clear regulations on the cutting order within the beam section, there have been accidents where the remaining beam section suddenly collapsed or even the entire support lost control and fell, exposing the safety hazards caused by the hidden steps. With the application of monitoring technology, it has been further discovered that even if the cutting sequence is specified in advance, the actual stress response often deviates significantly from the BIM expectation. This is because the BIM model is based on idealized design parameters, while the actual elastic modulus of the concrete on site and the actual stiffness of the joints after long-term compression often differ significantly from the design values. Furthermore, the supports may develop microcracks and loosening of joints during service, which cannot be predicted in static planning. When cutting according to the preset plan, if the measured stress at a certain step does not match the expectation, the site will be caught in a dilemma: continuing to carry out the operation may lead to the accumulation of risks, while suspending construction may jeopardize the construction period. To address this, the present invention provides a BIM-based collaborative control system for the removal of internal foundation pit supports and the transfer of stress. Summary of the Invention

[0003] In order to overcome the shortcomings of the prior art, at least one technical problem raised in the background art is solved.

[0004] The technical solution adopted by this invention to solve its technical problem is: a BIM-based collaborative control system for the removal of internal supports and the transfer of stress in foundation pits, comprising: Initial digital twin construction module: Obtain the BIM model of the support system in the foundation pit and the associated finite element analysis model, and import the support stress data, material mechanical parameters and boundary constraints obtained from actual monitoring to form the initial digital twin; Static dismantling path generation module: Based on the initial digital twin, determine the number of segments to be cut for each support beam, and preliminarily plan the dismantling sequence of all segments to generate a static optimal dismantling path diagram; Segmented dismantling execution module: Using the static optimal dismantling path map as the initial reference trajectory, during dismantling, it plans subsequent cutting instructions based on the actual stress state of the current support and surrounding structure, combined with the preset rolling time domain window length; Real-time sensing and updating module: During the cutting process, it collects multi-source sensing data of the entire demolition process in real time, compares it with the expected response of the planned cutting command, outputs residual values, and corrects uncertain parameters in the initial digital twin, so that the digital twin gradually approaches the actual physical state. Rolling Replanning Module: After completing the current cutting instruction, the module re-simulates and calculates the force response of the remaining cutting segments under different cutting sequences based on the corrected digital twin, and selects the cutting sequence that minimizes the peak impact force and meets the safety constraints as the subsequent execution plan.

[0005] The beneficial effects of this invention are as follows: This invention determines the optimal number of segments to be cut for each support beam using a multi-objective optimization algorithm and the Pareto front screening method, and automatically selects the optimal disassembly sequence that meets safety constraints using a graph search algorithm, thus fundamentally solving the sorting decision problem. This invention dynamically adjusts the length of the rolling time domain window based on the real-time risk index. When construction is safe, it plans the next three cutting operations each time to improve efficiency. When the risk increases, it automatically shortens the window to plan only the next step to avoid risk accumulation. This reduces the dilemma of either being too conservative and delaying the construction period or being too reckless and accumulating risks. This invention utilizes the residual between the measured data and the expected value after each cutting step to correct the key uncertain parameters in the digital twin online, so that the model gradually approaches the real physical state on site as construction progresses, and has the ability to cope with on-site uncertainties such as the discreteness of concrete materials, the degradation of joint stiffness, and the accumulation of construction errors. This invention, through a complete closed-loop scheme, can re-evaluate and re-plan subsequent operations based on on-site measured data after each cutting step. When the measured force at a certain step does not match the expectation, it no longer falls into the dilemma of risk accumulation if it continues to execute, or risk of jeopardizing the construction period if it suspends construction. Instead, it automatically and scientifically addresses the deviation by correcting the model parameters and replanning the cutting sequence, which not only ensures construction safety but also reduces unnecessary delays in the construction period. Attached Figure Description

[0006] The invention will now be further described with reference to the accompanying drawings.

[0007] Figure 1 This is a system block diagram of an embodiment of the present invention; Figure 2 This is a flowchart of the steps in an embodiment of the present invention. Detailed Implementation

[0008] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0009] Example Please see Figure 1 and Figure 2 As shown in the embodiment of the present invention, the BIM-based collaborative control system for the removal of internal foundation pit supports and the transfer of stress includes: Initial digital twin construction module: acquire the BIM model of the support system in the foundation pit and the associated finite element analysis model, and import the support stress data, material mechanical parameters and boundary constraints obtained from actual monitoring to jointly form the initial digital twin; The specific process is as follows: extract relevant information about the support system within the foundation pit based on the existing BIM model of the current engineering project; Among them, the relevant information of the internal support system of the foundation pit includes at least the geometric dimensions, spatial location and material properties of permanent structures such as internal support beams, walers, columns and basement floor slabs that have been constructed. The BIM model records the design dimensions, material specifications, spatial relationships, and connection node types of each component, such as fixed or hinged connections. It should be noted that if a complete BIM model has not yet been established for the current project, there are two alternative solutions: A 3D laser scanner was used to scan the on-site support system, and a model was built from the point cloud data. Manually model the structure based on the construction drawings in commonly used BIM software such as Revit or Tekla. The resulting BIM model needs to be able to clearly distinguish each supporting beam, each node, and its connection relationship with the surrounding structure. The process of associating the obtained BIM model with the finite element analysis model is as follows: Export the BIM model as an IFC or STEP format file, and then import it into Abaqus, MIDAS FEA, or ANSYS finite element analysis software. In finite element software, components such as supporting beams, walers, and floor slabs are meshed; Among them, the size of the beam element can be controlled between 0.2 meters and 0.5 meters to meet the calculation accuracy requirements. However, the support ends and the locations where the cutting points are planned will be the parts where the stress changes most drastically. Therefore, the mesh needs to be refined to below 0.1 meters to accurately capture the stress distribution. Assign initial parameters to each component based on the design designation in the BIM model; The system continuously collects real-time data based on a pre-deployed sensor network and sends it to the industrial control computer. Specifically, the axial force gauges installed at the ends or mid-span of the supports provide the current axial pressure value of each support, the test results of the test blocks cured under the same conditions provide a reference value for the actual compressive strength of the concrete, and the displacement gauges and inclinometers installed at the ends of the supports record the actual rotation angle and relative displacement at the nodes, which are used to infer the current constraint stiffness of the nodes. The collected real-time data is integrated with the design parameters of the BIM model, and the measured values ​​are used to replace the original design values ​​to obtain the initial digital twin. For example, taking a C30 concrete support beam in the foundation pit of a subway station as an example, the design axial force of the support beam is 2000 kN, but the actual measured reading of the axial force gauge is 1850 kN. The designed elastic modulus was 30 GPa, while the measured value of the test block under the same conditions was 28.5 GPa. The design assumed that both ends were ideally fixed, but the displacement gauge data showed that there was a relative slip of about 0.3 mm at the ends, indicating that the node stiffness had degraded. The three measured values ​​were automatically replaced with the original design values ​​and filled into the corresponding parameter fields of the finite element analysis model. The purpose of this module is to organically integrate the geometric information of the BIM model, the mechanical framework of the finite element model, and the on-site measured data to construct a digital twin that is close to the actual site conditions in its initial state. Although it is not completely accurate, because the uncertain parameters in the model need to be further corrected online during the subsequent demolition process, it has reduced the most basic "design-actual" deviation, so that the subsequent static demolition path planning is based on reliable data.

[0010] Static dismantling path generation module: Based on the initial digital twin, determine the number of segments to be cut for each support beam, and preliminarily plan the dismantling sequence of all segments to generate a static optimal dismantling path diagram; The specific process is as follows: based on the initial digital twin, read the finite element model after calibration using measured data; Each support beam that needs to be removed is handled separately; Preferably, taking a concrete support beam fixed at both ends as an example, the corresponding length, cross-sectional dimensions, current axial force value and concrete elastic modulus are read, and the stress response when the concrete support beam is cut into different numbers of segments is simulated in the finite element model; Specifically, different schemes such as dividing into 1 segment, 2 segments, 3 segments, and 4 segments were simulated respectively; Calculate the peak impact force at the moment of cutting and the maximum cantilever length of the remaining beam segment under each scheme; A multi-objective optimization algorithm is used to generate the optimal number of segment cuts, with the objectives of minimizing peak impact force and minimizing the number of cuts. The process is as follows: For each support beam to be dismantled, the number of segments to be cut is used as an optimization variable, which is usually taken as 1 to 6 segments. Set two mutually constraining optimization objectives: Minimize peak impact force, that is, minimize the maximum impact force generated when the remaining beam segment suddenly breaks during the cutting process; Minimize the number of cuts, that is, reduce the total number of cuts as much as possible to improve construction efficiency; Since there is an inherent contradiction between minimizing peak impact force and minimizing cutting technology, the more segments there are, the smoother the energy release during each segment cutting, and the smaller the peak impact force, but the number of cuttings increases and the construction period is extended. Conversely, the fewer segments there are, the higher the efficiency, but the greater the risk. The purpose of optimization is to find the optimal balance between the two. In the finite element model, each segmentation scheme is dynamically simulated. Taking a concrete support beam with two fixed ends, a length of L, and an axial pressure of N as an example, it is uniformly divided into n segments. In the simulation, each segment was cut off in the order of "first the two ends, then the middle", and the release of strain energy inside the remaining beam segment was calculated at each cut. Specifically, when a cutter cuts through a beam section, the elastic strain energy that was originally constrained within the corresponding beam segment will be released instantaneously and converted into the kinetic and vibrational energy of the remaining beam segment. Through explicit finite element dynamic analysis, the relative velocity of the nodes on both sides of the cut is captured at the moment of cutting, and then the peak impact force generated at this time is calculated. The peak impact force is calculated based on the momentum theorem and the law of conservation of energy. According to the momentum theorem, the product of the nodal mass and the relative velocity is equal to the calculated impact force. Since the actual concrete material is not perfectly elastic, there is material damping and frictional energy loss during the cutting process. Therefore, the calculated impact force needs to be multiplied by a correction factor to obtain a peak impact force that is closer to the actual value. The correction factor is usually taken as 0.6-0.8. Impact force monitoring points were defined in the finite element model, located in the middle of the uncut segment adjacent to the cutting point. When the cutting was completed, the maximum stress change rate recorded by the impact force monitoring points was positively correlated with the peak impact force. Through statistical analysis of a large number of simulation results, an empirical relationship between the number of segments n and the peak impact force was established: as the number of segments n increases from 1 to 3, the peak impact force decreases most significantly, usually by more than 50%. When n exceeds 3, the marginal benefit of further increasing the number of segments to reduce the peak impact force decreases significantly, and each additional segment can only reduce the peak impact force by 5%-10%. Set safety constraints, including: cantilever length constraints, lifting capacity constraints, and workspace constraints; Cantilever length constraint: When a beam segment is cut off, the length of the longest cantilever portion of the remaining beam segment shall not exceed 6 times the height of the supporting beam section (refer to the limit for cantilever members in the Code for Design of Concrete Structures), otherwise the remaining beam segment may be bent and fail due to its own weight or residual stress. Lifting capacity constraints: The weight of each cut concrete block must not exceed the rated lifting capacity of the on-site lifting equipment; Working space constraints: The length of the cutting section should also take into account the working space requirements of the hydraulic shears or wire saw cutting equipment on site. Generally, the length of a single cut should not be less than 0.5 meters or more than 3 meters. The two optimization objectives and constraints mentioned above are input into a multi-objective optimization algorithm, and the Pareto front screening method is used, specifically as follows: For each support beam to be dismantled, all possible number of segments and possible segment positions under each number of segments are combined as candidate solutions. Different combinations of segment positions were evaluated separately, and each candidate scheme corresponded to a set of two target values: peak impact force and number of cuts. Place all candidate solutions in a two-dimensional coordinate system with peak impact force as the horizontal axis and the number of cuts as the vertical axis. For any two solutions A and B, determine their merits according to the following rules: If the peak impact force of scheme A is less than or equal to the peak impact force of scheme B, and the number of cuts of scheme A is less than or equal to the number of cuts of scheme B, and at least one of the indicators is strictly less than, that is, scheme A is no worse than scheme B in both indicators, and at least one is better, then scheme A is considered to dominate scheme B. A dominated solution implies that there is another solution that is better or equal in both safety and efficiency, so the dominated solution can be removed from the candidate set. After filtering by dominance relationships, all the remaining non-dominant schemes constitute the Pareto front. The remaining options are characterized by the fact that none of them is superior to the other option on the Pareto front in both objectives. That is, if you want to reduce the peak impact, you have to accept more cuts, and conversely, if you want to reduce the number of cuts, you have to accept a higher peak impact. Each option on the Pareto front is mathematically optimal, but represents a different risk-efficiency trade-off. After obtaining the Pareto front, instead of directly selecting a solution, we recommend an intermediate solution on the front as the initial suggestion based on pre-set default preferences (usually prioritizing safety while taking efficiency into account). All options on the Pareto front are presented in a comparative chart, with each option's peak impact force, estimated construction period, and impact risk level on the floor below marked. Based on the actual site conditions (such as the current construction period pressure, whether construction is carried out at night, and whether the floor below has been fully cured), the recommended option is selected and confirmed, or other options are manually switched to, and the optimal number of segments is output. The role of Pareto frontier screening is to transform the segmented decision-making process, which originally required repeated calculations and relied on personal experience, into a visualized and comparable multi-option selection process. Instead of guessing how many segments are appropriate, it allows for a direct view of the risk level and efficiency cost corresponding to different numbers of segments, thus enabling more rational decisions. After determining the optimal number of segments for each concrete support beam, arrange all segments in a safe order. The stress transformation process under different disassembly sequences is simulated sequentially in the initial digital twin. After each simulated cutting of a segment, the stress distribution of the remaining structure is recalculated. Ensure that the maximum stress of the remaining structure after each cutting step does not exceed the allowable value of the material strength; A graph search algorithm is used to find a path that satisfies the safety constraints from all possible permutations. If multiple safe paths exist, the one with the shortest total time is selected as the recommended solution. The process is as follows: All segments to be cut are considered as a group of nodes. Each segment has two states: not yet removed or already removed. The entire removal process starts from the initial state where all segments are not yet removed and ends in the final state where all segments are already removed. Each time a cut is performed, exactly one cut segment changes from never being removed to being removed. Thus, the removal order problem is transformed into finding a path in the state space from the initial state to the final state. Starting from the current state, enumerate all possible next cutting operations, simulate the cutting in the initial digital twin, and check whether the maximum stress of the remaining structure after cutting exceeds the allowable value of the material strength. If it does not exceed the allowable value, the cutting operation is feasible and the corresponding next state is added to the search queue. If it exceeds the allowable value, the cutting operation is not feasible, that is, the branch is pruned in the current state. Starting from the initial state, the algorithm expands outward layer by layer. For each state in the current layer, a feasible edge screening operation is performed. Since the time taken to cut each segment is roughly the same, safety constraints are prioritized. A breadth-first search strategy is adopted, advancing layer by layer according to the number of cutting steps. When the terminal state is reached for the first time, the path found is the solution with the fewest cutting steps that satisfies the safety constraints. If there are multiple paths with the same number of cutting steps, the total time of each path is further compared (considering the different time taken per segment due to the difference in hoisting distance for different cutting segments), and the path with the shortest total time is selected as the final recommended solution. After the search is complete, the optimal path found will be output as a static optimal demolition path map.

[0011] The purpose of this module is twofold: First, through multi-objective optimization and Pareto front screening, it determines the optimal number of segments to be cut for each support beam, solving the decision-making problem of whether to cut in one go or one cut at a time, and how many cuts are most appropriate. Second, through a graph search algorithm, it automatically selects the optimal disassembly sequence that meets safety constraints, solving the problem of deciding which segment to disassemble first and which segment to disassemble later, and optimizing the shortcomings of only specifying the order of regions without specifying the cutting order within beam segments. Furthermore, although the static optimal demolition path map generated by this module has undergone sufficient calculation and optimization, it is not a fixed construction command. Instead, it provides a mechanically verified initial reference trajectory for subsequent modules. In the actual construction process, the segmented demolition execution module will take the path map as the starting point and dynamically adjust the subsequent cutting sequence based on real-time monitoring data. This two-stage design of static planning followed by dynamic correction has both theoretical calculations as a basis and retains the flexibility to deal with uncertainties on site, achieving a more reasonable balance between safety and efficiency.

[0012] Segmented dismantling execution module: Based on the static optimal dismantling path map as the initial reference trajectory, during the dismantling process, subsequent cutting instructions are planned according to the actual stress state of the current support and surrounding structure, combined with the preset rolling time domain window length. The specific process is as follows: before the actual cutting begins, the length of the rolling time domain window is determined based on the real-time risk index of the current construction site; Among them, the length of the rolling time domain window refers to how many steps forward to look at each planning step, whether to plan only the next step, or to plan the next two or three steps. The calculation process for the real-time risk index is as follows: Read the current axial force value measured by the axial force gauge at each support end, compare it with the axial force value at the previous moment (usually the previous 5 minutes), calculate the axial force change rate of each support, that is, the amount of axial force change per unit time, and take the maximum change rate among all supports as the current axial force change rate index. Read the displacement gauge readings installed on the foundation pit cap beam and the surrounding ground surface, and take the average value of the three points with the largest cumulative displacement value among all monitoring points as the current displacement status index; Calculate the ratio of the number of remaining uncut segments to the total number of cut segments, then calculate the ratio of the measured total axial force of the current remaining support to the initial total axial force. Multiply the two ratios to obtain the redundancy coefficient. Subtract 1 from the redundancy coefficient to obtain the bearing capacity loss index. The redundancy coefficient reflects how much load-bearing capacity the remaining support system still retains. The higher the ratio, the greater the safety reserve of the support system, which can withstand a certain amount of disturbance. The lower the ratio, the less the number of remaining supports, and the more significant the impact of each subsequent cutting step on the overall stress pattern, requiring greater caution. The current axial force change rate index is compared with the preset change rate warning value to obtain the normalized current axial force change rate index. The current displacement state index is compared with the preset displacement state warning value to obtain the displacement state index before normalization. The maximum value among the normalized current axial force change rate index, the normalized displacement state index, and the bearing capacity loss index is extracted as the real-time risk index. The reason for extracting the maximum value is that if the normalized risk value of any one of the three dimensions is high, the comprehensive risk index will be correspondingly high, which follows the barrel effect principle in engineering safety monitoring, where overall safety depends on the weakest link. The closer the real-time risk index is to 0, the safer the current construction status is; the closer the real-time risk index is to 1, the closer the current construction status is to a warning, and more caution is needed. Convert the real-time risk index into a rolling time-domain window length N; When the real-time risk index is in the low range, it indicates that the current monitoring indicators of all dimensions are far away from the warning threshold, the construction status is safe, and the rolling time domain window length is 3, that is, each time the subsequent three steps of the cutting operation are planned, it is appropriate to look forward a few more steps, reduce the frequency of interaction between the system and the on-site equipment, and improve construction efficiency. When the real-time risk index is in the median range, it indicates that at least one dimension is in a middle state that needs attention. The rolling time window length is 2, which maintains a trade-off between efficiency and risk. When the real-time risk index is in a high range, it indicates that at least one dimension has approached or reached the warning threshold, and the construction risk is high. The rolling time window length is set to 1, that is, only the next operation is planned each time, and long-term planning is abandoned. The "take it one step at a time" strategy is adopted. Each cut must be re-evaluated based on the latest monitoring data before it can continue, so as to avoid the accumulation of risks to the greatest extent. The division of the aforementioned low, medium, and high intervals is mainly based on the monitoring specifications for foundation pit engineering, such as the "Technical Standard for Monitoring Foundation Pit Engineering." The national standard clearly states that foundation pit engineering monitoring and early warning should simultaneously use both cumulative change and rate of change indicators for control. Specifically, when the monitoring indicator is far below the warning threshold, it is considered a normal state, corresponding to the low interval in this embodiment; when the monitoring indicator reaches or exceeds 60%–80% of the warning value, it is considered an abnormal state, requiring attention and strengthened monitoring, corresponding to the medium interval in this embodiment; and when the monitoring indicator reaches or exceeds the warning threshold, it is considered a dangerous state, requiring immediate emergency measures, corresponding to the high interval in this embodiment. Using the static optimal demolition path diagram as the initial reference trajectory, and combining it with the actual stress state in the current digital twin, the specific cutting instructions for the next N steps are planned as follows: The cutting segment suggested in the current step of the static optimal demolition path diagram is used as the initial candidate. The cutting operation is simulated in the digital twin. The maximum stress of the remaining structure after cutting is calculated to see if it exceeds the allowable value of the material strength. If it does not exceed the allowable value, the current cutting segment is included in the instruction sequence as the first step operation in the current window. If it exceeds the allowable value, it means that due to deviations in the previous construction or actual degradation of the support system, the current cutting segment is no longer safe in the current state. The suggestion in the static path diagram is abandoned. Instead, the current digital twin is used as the benchmark to perform single-step cutting simulations on all remaining uncut segments. The peak impact force and the maximum stress of the remaining structure under each cutting operation are calculated. The cutting segment with the smallest peak impact force and that meets all safety constraints is selected as the actual operation object for the first step in the current window. If N is greater than 1, after completing the selection in the first step, the selected segment in the first step is marked as cut off in the digital twin, the state is updated, and the second step is planned in the same way based on the updated state. First, try to refer to the next step suggestion corresponding to the static path diagram. If it is not feasible, choose the optimal solution from the remaining segments. And so on, until the entire N-step cutting instruction sequence is planned. If N equals 1, only the next step is planned. Once the planning is completed, the cutting command sequence of the current window will be sent to the on-site cutting equipment. The cutting command shall include at least the support beam number to be cut and the segment number to be cut in each step. It is important to note that before executing the cutting command, the real-time spatial position of the cutting equipment is obtained through the UWB positioning tag to determine whether it is within the allowable working area corresponding to the current segment to be cut. If the equipment position is correct, the equipment is unlocked and cutting is allowed. If the equipment position deviates, the hydraulic system power source is locked and a position deviation warning is issued. The operation can continue after the operator moves the equipment to the correct position.

[0013] The purpose of this module is to transform the static optimal demolition path map from a rigid command into a flexible reference. It neither completely relies on static planning nor completely abandons static planning. Instead, it uses the static path map as the starting point for each decision, and verifies and corrects it in combination with the current actual situation. After making a certain number of cuts within the current window, it pauses and re-evaluates, so that each subsequent operation is based on the latest monitoring data and the corrected model, achieving the collaborative control goal of demolishing while monitoring and adjusting simultaneously.

[0014] Real-time sensing and updating module: During the cutting process, multi-source sensing data of the entire demolition process is collected in real time and compared with the expected response of the planned cutting command. The residual value is output and the uncertain parameters in the initial digital twin are corrected based on the residual value, so that the digital twin gradually approaches the actual physical state. Among them, the uncertain parameters include at least: the actual elastic modulus of concrete and the actual stiffness coefficient of the support node; The specific process is as follows: During each step of the cutting command execution, multi-source sensing data of the entire demolition process is collected in real time. The multi-source sensing data includes at least: Axial force gauges installed at the ends or mid-span of the supports continuously record the axial pressure value of each support at the current moment; Strain gauges were installed in the mid-span of the support and near the cut to monitor strain changes on the concrete surface; Displacement gauges and inclinometers installed at the connection points between the support end and the waler measure the relative slippage and rotation changes at the nodes; Accelerometers record the vibration response of the remaining beam segment during the cutting process; The collected real-time sensing data is compared with the expected force response item by item; The specific comparison method is as follows: Before performing a cutting operation on a certain segment, the expected axial force, expected strain and expected displacement values ​​at each monitoring point are calculated in advance based on the current state of the digital twin. When the actual cutting is completed and the on-site sensor data is stable, the actual readings of each measuring point are read, and the actual values ​​are subtracted from the expected values ​​to obtain the residual value, which is the deviation between the actual response and the theoretical expectation. If the residual value is within the preset allowable range, it indicates that the current digital twin's description of the support system is still relatively accurate; If the residual value exceeds the preset allowable range, it means that some parameters in the digital twin no longer match the actual physical state and need to be corrected. The preset allowable range is based on the warning value requirements for the rate of change and cumulative change of monitoring data in the "Technical Standard for Monitoring of Building Foundation Pit Engineering". It is combined with the design bearing capacity and construction stage division of the support system of this project and is set separately according to the ultimate limit state of bearing capacity and the serviceability limit state. The parameter identification method is used to correct uncertain parameters in the digital twin online. The specific method is as follows: The calculated residual value, along with the concrete elastic modulus and nodal stiffness coefficients in the current digital twin, is input into the state estimator built based on the Kalman filter principle. The state estimator pre-defines the state vector and observation vector: The state vector consists of uncertain parameters that need to be identified online, namely the actual elastic modulus of concrete for each support beam and the actual stiffness coefficient for each support node. The observation vector is a physical quantity that can be directly or indirectly measured by a sensor, namely the measured displacement, strain, and end axial force of each monitoring point near the cut. The state estimator internally maintains a current optimal parameter estimate. When a new residual value is obtained, the state estimator determines which parameters need to be adjusted based on the physical location and dimensions corresponding to the residual. Specifically, the state estimator associates each component in the residual vector with each parameter in the state vector: the displacement residual mainly reflects the deviation of the elastic modulus parameter, because the elastic modulus of concrete directly determines the deformation of the member after being subjected to force; the strain residual reflects the changes in both the elastic modulus and the load transfer path; and the axial force residual mainly reflects the deviation of the nodal stiffness parameter, because changes in the nodal constraint stiffness will change the internal force distribution of the support system. The state estimator calculates the gain matrix based on the correlation. Specifically, the state estimator internally maintains an error covariance matrix. The number of rows and columns of the error covariance matrix is ​​the same as the number of parameters to be identified. The elements on the diagonal reflect the uncertainty of the current estimated value of each parameter. The greater the uncertainty, the more the parameter needs to be corrected. When new residual data is obtained, the state estimator combines the current error covariance matrix with the observation matrix to calculate an intermediate matrix. The intermediate matrix is ​​multiplied by the inverse of the observation noise covariance matrix to obtain the gain matrix. The rows of the gain matrix correspond to the parameters to be corrected (i.e., the elastic modulus of each support beam and the stiffness coefficient of each node), and the columns correspond to the observed residual data (i.e., the displacement residual, strain residual, and axial force residual of each monitoring point). The value of each element represents the proportion of the corresponding residual that should be allocated to the corresponding parameter. For example, when the displacement residual at a certain monitoring point is positive (i.e., the measured displacement is greater than the expected displacement) and the strain residual near the monitoring point is negative (i.e., the measured strain is less than the expected strain), the state estimator will identify this combination pattern as a characteristic of low elastic modulus. At this time, the state estimator will use 60% of the current displacement residual value as a correction amount and add it to the current value of the concrete elastic modulus parameter of the corresponding support beam to adjust it upward. This setting is based on engineering experience and reflects the sensitivity of displacement residual to elastic modulus correction. In actual use, it can be adjusted within the range of 50% to 70% according to the noise level of the on-site monitoring data. When both displacement and strain residuals are positive, the state estimator identifies this combination as a possible composite feature of nodal stiffness degradation. If the ratio of displacement residual to strain residual is greater than 2, it indicates that the deformation deviation is much greater than the strain deviation, suggesting that the main source of the deviation is nodal stiffness degradation rather than elastic modulus deviation. The state estimator uses all of this set of residual values ​​to correct the stiffness coefficient of adjacent support nodes, adjusting it downwards. This threshold is determined based on the theoretical ratio of the sensitivity of the mid-span measuring point to the end measuring point of the support beam. If the difference between displacement residual and strain residual is not greater than 2, it indicates that the two deviations are similar in degree. The state estimator allocates 50% of the residual value to the concrete elastic modulus parameter of the corresponding support beam and the remaining 50% to the stiffness coefficient of the adjacent support node, with both parameters corrected synchronously. After completing the gain calculation, the state estimator multiplies the gain matrix with the current residual vector to obtain the change in each parameter to be corrected. The change in the parameter to be corrected is superimposed on the current parameter value in the state vector to obtain the updated optimal parameter estimate. Through this recursive method, after each cut, the key uncertain parameters are corrected using measured data, so that the parameters in the digital twin gradually approach the real physical state of the scene. The correction process is executed automatically once after each cutting operation, throughout the entire demolition construction. As the demolition work progresses, the accuracy of the digital twin's prediction of the stress response of the on-site support system will gradually improve, thus providing a more reliable model basis for subsequent rolling replanning modules.

[0015] The role of this module is to serve as a crucial bridge between static planning and dynamic collaborative control. By utilizing the residual between measured data and expected values, it reverse-corrects key uncertain parameters in the digital twin, enabling the model to gradually approximate the actual physical state on site. Through continuous parameter correction, it ensures that the digital twin can accurately reflect the actual stress state on site at each stage of demolition construction, thereby enabling the subsequent calculation of a truly instructive cutting sequence. It possesses the technical capability to cope with uncertainties on site, such as the discreteness of concrete materials, the degradation of joint stiffness, and the accumulation of construction errors.

[0016] Rolling Replanning Module: After completing the current cutting instruction, the module re-simulates and calculates the force response of the remaining cutting segments under different cutting sequences based on the corrected digital twin, and selects the cutting sequence that minimizes the peak impact force and meets the safety constraints as the subsequent execution plan. The specific process is as follows: Obtain the corrected digital twin. At this time, the digital twin has updated the key parameters of concrete elastic modulus and joint stiffness coefficient, which are closer to the actual physical state on site than the initial state. Read all remaining uncut segments, and perform a new force analysis and sequence optimization on these remaining segments in the corrected digital twin. The specific steps are as follows: The remaining segments are grouped into a new candidate set, and the stress state at the current moment is used as the initial condition to re-simulate and calculate all segments in the candidate set. It is important to note that although both this module and the segmented dismantling execution module are responsible for selecting the next segment to be cut, they rely on different model states: the segmented dismantling execution module plans based on the current digital twin (which has not yet absorbed the measured data of this cut) before each cut execution, aiming to generate instructions within the current window under the guidance of the static path diagram, which is equivalent to making local decisions at the execution layer. This module, on the other hand, replans based on the digital twin (which has absorbed the measured data of this cut) after the cut is completed, based on the real-time perception and update module's correction, aiming to re-examine and optimize the global order of the remaining segments, provided that the model has been updated. The two have different functions and complement each other: the former is responsible for how to cut at the present moment, and the latter is responsible for how to adjust next. Compared with the static demolition path generation module, this module no longer enumerates all possible segment combinations and all permutations from scratch, because the number of segments for each beam has been determined in the static planning stage. This module only needs to redetermine the execution order of the remaining cutting segments based on the existing segmentation scheme. Using the current stress state as the initial condition, each remaining uncut segment is taken as a candidate operation for the next step. The cutting is simulated one by one in the corrected digital twin. For each candidate operation, the peak impact force after the cutting is completed and the maximum stress of the remaining structure are calculated, and it is checked whether the safety constraints are met. The safety constraints are consistent with those set in the static demolition path generation module, including cantilever length constraints, hoisting capacity constraints, and workspace constraints. After completing the simulation calculations for all candidate operations, the cutting segment with the smallest peak impact force and that satisfies all safety constraints is selected as the actual operation target for the next step. If the length N of the rolling time-domain window is greater than 1, after performing the above selection process, the selected cutting segments are marked as cut off in the digital twin, the model state is updated, and based on the updated state, candidate operation simulation and peak impact force comparison are performed again from the remaining cutting segments to select the second step operation. This process continues until all N cutting steps are planned to form a complete subsequent execution plan. If N equals 1, the system only plans the next operation and does not arrange subsequent steps in advance.

[0017] The purpose of this module is to immediately apply the updated parameters to the subsequent re-optimization of the cutting sequence, determine the safest next step in the current state, and enable rapid response and timely adjustment of construction strategies after each model correction, reducing the passive situation caused by adhering to static planning schemes when site conditions change.

[0018] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A BIM-based collaborative control system for the removal of internal supports and stress transfer in foundation pits, characterized in that: include: Initial digital twin construction module: acquire the BIM model of the support system in the foundation pit and the associated finite element analysis model, and import the support stress data, material mechanical parameters and boundary constraints obtained from actual monitoring to form the initial digital twin; Static dismantling path generation module: Based on the initial digital twin, determine the number of segments to be cut for each support beam, and preliminarily plan the dismantling sequence of all segments to generate a static optimal dismantling path diagram; Segmented dismantling execution module: Using the static optimal dismantling path map as the initial reference trajectory, during dismantling, it plans subsequent cutting instructions based on the actual stress state of the current support and surrounding structure, combined with the preset rolling time domain window length; The preset scrolling time domain window length is determined based on the real-time risk index; The real-time risk index is obtained by extracting the maximum value from the normalized current axial force change rate index, the normalized displacement state index, and the bearing capacity loss index. When the real-time risk index is in the low range, the rolling time domain window length is 3; when the real-time risk index is in the middle range, the rolling time domain window length is 2; when the real-time risk index is in the high range, the rolling time domain window length is 1. The calculation process for the current axial force change rate index, displacement state index, and bearing capacity loss index is as follows: Read the current axial force value at each support end, compare it with the axial force value at the previous moment, calculate the axial force change rate of each support, and take the maximum change rate among all supports as the current axial force change rate index; Read the displacement gauge readings and take the average of the three points with the largest cumulative displacement values ​​among all monitoring points as the current displacement status index; Calculate the ratio of the number of remaining uncut segments to the total number of cut segments, then calculate the ratio of the measured total axial force of the current remaining support to the initial total axial force. Multiply the two ratios to obtain the redundancy coefficient. Subtract 1 from the redundancy coefficient to obtain the bearing capacity loss index. Real-time sensing and updating module: During the cutting process, it collects multi-source sensing data of the entire demolition process in real time, compares it with the expected response of the planned cutting command, outputs residual values, and corrects uncertain parameters in the initial digital twin, so that the digital twin gradually approaches the actual physical state. Among them, the uncertain parameters include: the actual elastic modulus of concrete and the actual stiffness coefficient of the support node; Rolling Replanning Module: After completing the current cutting instruction, the module re-simulates and calculates the force response of the remaining cutting segments under different cutting sequences based on the corrected digital twin, and selects the cutting sequence that minimizes the peak impact force and meets the safety constraints as the subsequent execution plan.

2. The BIM-based collaborative control system for the removal of internal foundation pit supports and the transfer of stress as described in claim 1, characterized in that: The static demolition path generation module includes: With the optimization objectives of minimizing peak impact force and minimizing the number of cuts, and combining cantilever length constraints, hoisting capacity constraints, and working space constraints, a multi-objective optimization algorithm is used to generate the optimal number of segment cuts for each support beam. After determining the optimal number of segments to be cut for each support beam, the stress transformation process under different dismantling sequences is simulated sequentially in the initial digital twin. A graph search algorithm is used to find the cutting sequence that satisfies the safety constraints and has the shortest total time, and the static optimal dismantling path diagram is output.

3. The BIM-based collaborative control system for the removal of internal foundation pit supports and the transfer of stress as described in claim 2, characterized in that: The multi-objective optimization algorithm employs the Pareto front selection method, specifically: All possible number of segments and possible segment positions under each number of segments are combined as candidate schemes, and each candidate scheme corresponds to a set of peak impact force and number of segments. For any two candidate solutions, if one solution is not inferior to the other solution in both objective values ​​and is superior in at least one objective value, then the solution is considered to dominate the other solution, and the dominated solution is removed from the candidate set. After removing all dominated solutions from the candidate solutions, all remaining non-dominated solutions constitute the Pareto front. All solutions on the Pareto front are displayed in a comparative chart, and the optimal number of segment cuts is output.

4. The BIM-based collaborative control system for the removal of internal foundation pit supports and the transfer of stress as described in claim 2, characterized in that: The graph search algorithm is as follows: Using the cut segments as nodes and the states of not demolished and demolished, search from the initial state of all not demolished to the final state of all demolished. Starting from the current state, enumerate all possible next cutting operations, simulate the cutting in the initial digital twin, and expand the branch when the maximum stress of the remaining structure after the simulated cutting does not exceed the allowable value. The process proceeds layer by layer according to the number of cutting steps. When the termination state is reached for the first time, the path found is the solution with the fewest cutting steps that satisfies the safety constraints. If there are multiple paths with the same number of cutting steps, the one with the shortest total time is selected as the final recommended solution.

5. The BIM-based collaborative control system for the removal of internal foundation pit supports and the transfer of stress as described in claim 1, characterized in that: The real-time sensing and updating module includes: Before performing the cutting segment removal operation, the expected axial force, expected strain and expected displacement values ​​of each monitoring point are pre-calculated based on the current state of the digital twin. After the cutting is completed, the actual readings of each measuring point are read, and the residual value is obtained by subtracting the expected value from the actual value. When the residual value exceeds the preset allowable range, the residual value, along with the concrete elastic modulus and the nodal stiffness coefficient, is input into the state estimator built based on the Kalman filter principle for online correction. The state vector of the state estimator consists of the actual elastic modulus of concrete for each support beam and the actual stiffness coefficient for each support node. The observation vector consists of the measured displacement, strain, and end axial force. The state estimator calculates the gain matrix based on the relationship between the residual vector and the parameters in the state vector. The gain matrix is ​​multiplied by the current residual vector to obtain the change of each parameter to be corrected, and then superimposed on the current parameter value to complete the correction.

6. The BIM-based collaborative control system for the removal of internal foundation pit supports and the transfer of stress as described in claim 5, characterized in that: The state estimator maintains an error covariance matrix internally, with the diagonal elements reflecting the degree of uncertainty of the current estimated values ​​of each parameter. When new residual data is acquired, the current error covariance matrix is ​​combined with the observation matrix to calculate the intermediate matrix, and then the intermediate matrix is ​​multiplied by the inverse of the observation noise covariance matrix to obtain the gain matrix. The rows of the gain matrix correspond to the parameters to be corrected, and the columns correspond to the observed residual data.

7. The BIM-based collaborative control system for the removal of internal foundation pit supports and the transfer of stress as described in claim 1, characterized in that: The rolling replanning module includes: Obtain the corrected digital twin, read all the remaining uncut segments to form a candidate set, use the current stress state as the initial condition, simulate the cutting in the corrected digital twin one by one, calculate the peak impact force after cutting and the maximum stress of the remaining structure for each candidate operation, and check whether the safety constraints are met. Select the cutting segment with the minimum peak impact force that satisfies all safety constraints as the actual operation object for the next step; if the length N of the rolling time domain window is greater than 1, mark the selected cutting segment as cut off and update the model state, repeat the above process to select subsequent operations, until the cutting sequence of all N steps is planned. If N equals 1, then only the next operation is planned.

Images