A method and system for detecting settlement of a grain silo during construction

Abstract

The application discloses a kind of grain shallow round warehouse in construction process settlement detection method and system, it is related to the technical field of settlement detection, the method includes monitoring the construction process of grain shallow round warehouse, deformation data and initial settlement data are fused, and settlement characterization data are generated in the fusion process;The settlement characterization data are input into the preset digital twin space, and the settlement evolution is carried out in the digital twin space, the accuracy of the multidimensional settlement displacement field of grain shallow round warehouse is improved.The hoop stress distribution of bin wall is combined with the used time and strength data of bin wall to determine the damage evolution trend of bin wall;Based on the identification of the damage evolution trend of bin wall, the corresponding crack characteristics are predicted, and the corresponding settlement early warning event is determined in combination with the construction history of grain shallow round warehouse, according to the position of settlement early warning area, settlement rate and preset knowledge graph Generation and settlement decision content associated with the next construction process, the accuracy of settlement decision content is improved.

Description

Technical Field

[0001] This invention relates to the technical field of settlement detection, and in particular to a method and system for settling detection during the construction of shallow circular grain silos. Background Technology

[0002] As a crucial infrastructure for modern grain and oil storage, shallow circular grain silos are massive in size, extremely heavy, and subject to high loads. During construction, the uneven bearing capacity of the foundation and the complexity of dynamic loads can easily lead to uneven settlement of the foundation. Uneven settlement of the shallow circular grain silos directly causes additional circumferential tensile stress and localized stress concentration in the silo walls, potentially leading to wall cracking or even structural instability. Therefore, settlement monitoring and safety decision-making for shallow circular grain silos during construction are of paramount importance.

[0003] During the construction of shallow circular grain silos, existing monitoring methods typically use raw observations or simply filtered data as input for settlement field analysis. This lacks effective separation of dynamic loads from construction machinery and material loads, as well as decoupling of dynamic and static loads. Consequently, deformation data contains a large amount of data that does not reflect the true structural response. Existing technologies cannot obtain deformation data that reflects the true deformation of the shallow circular grain silos, resulting in low accuracy of settlement characterization data fusion. This affects the accuracy of the multi-dimensional settlement displacement field of the shallow circular grain silos, leading to lower accuracy in settlement decision-making. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art. This invention provides a method and system for detecting settlement during the construction of shallow circular grain silos.

[0005] This invention provides a method for settling detection during the construction of shallow circular grain silos, comprising:

[0006] The initial readings of the multi-source sensor array of the shallow grain silo are collected and used as the initial settlement data; the construction process of the shallow grain silo is monitored, the current readings of the multi-source sensor array are obtained, and the material load and construction machinery dynamic load of the current construction process are extracted to determine the deformation data of the shallow grain silo.

[0007] The deformation data and the initial settlement data are fused together to generate settlement characterization data during the fusion process. The settlement characterization data is then input into a preset digital twin space, and settlement evolution is performed in the digital twin space to determine the multi-dimensional settlement displacement field of the shallow grain silo.

[0008] Based on the identification of the multidimensional settlement displacement field, the circumferential curvature change of the silo wall is determined. The circumferential stress distribution of the silo wall is inverted based on the curvature change. The damage evolution trend of the silo wall is determined by combining the service time and strength data of the silo wall.

[0009] Based on the identification of the damage evolution trend of the silo wall, the corresponding crack characteristics are predicted, and the corresponding settlement early warning events are determined in combination with the construction history of the shallow circular grain silo, and the settlement early warning area is marked; based on the location of the settlement early warning area, the settlement rate, and the preset knowledge graph, settlement decision content associated with the next construction procedure is generated.

[0010] This invention provides a settlement detection system for shallow circular grain silos during construction. This system is applied to the aforementioned settlement detection method for shallow circular grain silos during construction. The settlement detection system includes:

[0011] The monitoring module is used to collect the initial readings of the multi-source sensor array of the shallow grain silo and use them as initial settlement data; monitor the construction process of the shallow grain silo, obtain the current readings of the multi-source sensor array, extract the material load and construction machinery dynamic load of the current construction process, and thus determine the deformation data of the shallow grain silo.

[0012] The settlement evolution module is used to fuse the deformation data with the initial settlement data and generate settlement characterization data during the fusion process; the settlement characterization data is input into a preset digital twin space and settlement evolution is performed in the digital twin space to determine the multi-dimensional settlement displacement field of the shallow grain silo.

[0013] The damage evolution trend module is used to determine the circumferential curvature change of the silo wall based on the identification of the multi-dimensional settlement displacement field, invert the circumferential stress distribution of the silo wall based on the curvature change, and determine the damage evolution trend of the silo wall in combination with the silo wall's service time and strength data.

[0014] The settlement decision module is used to predict the corresponding crack features based on the identification of the damage evolution trend of the silo wall, and to determine the corresponding settlement early warning events in combination with the construction history of the shallow circular grain silo, and to mark the settlement early warning area; and to generate settlement decision content associated with the next construction procedure based on the location of the settlement early warning area, the settlement rate and the preset knowledge graph.

[0015] Compared with the prior art, the beneficial effects of the present invention are:

[0016] (1) Collect the initial readings of the multi-source sensor array of the shallow grain silo and use them as the initial settlement data; monitor the construction process of the shallow grain silo, obtain the current readings of the multi-source sensor array, extract the material load and construction machinery dynamic load of the current construction process, and thus determine the deformation data of the shallow grain silo; fuse the deformation data with the initial settlement data, and generate settlement characterization data during the fusion process; input the settlement characterization data into the preset digital twin space, and perform settlement evolution in the digital twin space, thereby determining the multi-dimensional settlement displacement field of the shallow grain silo. The deformation data of the shallow grain silo is introduced, and the settlement characterization data is further controlled, which improves the accuracy of the multi-dimensional settlement displacement field of the shallow grain silo.

[0017] (2) Based on the identification of the multi-dimensional settlement displacement field, the curvature change of the silo wall in the circumferential direction is determined. The circumferential stress distribution of the silo wall is inverted based on the curvature change. The damage evolution trend of the silo wall is determined by combining the service time and strength data of the silo wall. Based on the identification of the damage evolution trend of the silo wall, the corresponding crack characteristics are predicted. The corresponding settlement warning events are determined by combining the construction history of the shallow circular grain silo, and the settlement warning area is marked. Based on the location of the settlement warning area, the settlement rate and the preset knowledge graph, the settlement decision content associated with the next construction procedure is generated. The damage evolution trend of the silo wall is further controlled. The settlement warning area and knowledge graph are fully considered, and the accuracy of the settlement decision content is improved. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a method for detecting settlement during the construction of a shallow circular grain silo, as described in an embodiment of the present invention.

[0019] Figure 2 This is a schematic diagram of the structural composition of a settlement detection system for a shallow circular grain silo during construction, according to an embodiment of the present invention.

[0020] Explanation of reference numerals in the attached figures:

[0021] 21. Monitoring module; 22. Settlement evolution module; 23. Damage evolution trend module; 24. Settlement decision module. Detailed Implementation

[0022] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0023] Please see Figure 1 A method for detecting settlement during the construction of a shallow circular grain silo, applicable to settlement detection scenarios; the method for detecting settlement during the construction of a shallow circular grain silo includes:

[0024] Step S11: Collect the initial readings of the multi-source sensor array of the shallow grain silo and use them as the initial settlement data; monitor the construction process of the shallow grain silo, obtain the current readings of the multi-source sensor array, extract the material load and construction machinery dynamic load of the current construction process, and thus determine the deformation data of the shallow grain silo.

[0025] Step S12: The deformation data and the initial settlement data are fused together to generate settlement characterization data during the fusion process; the settlement characterization data is input into a preset digital twin space, and settlement evolution is performed in the digital twin space to determine the multi-dimensional settlement displacement field of the shallow grain silo.

[0026] Step S13: Based on the identification of the multi-dimensional settlement displacement field, determine the circumferential curvature change of the silo wall, invert the circumferential stress distribution of the silo wall according to the curvature change, and determine the damage evolution trend of the silo wall by combining the service time and strength data of the silo wall.

[0027] Step S14: Based on the identification of the damage evolution trend of the silo wall, predict the corresponding crack characteristics, and determine the corresponding settlement warning events in combination with the construction history of the shallow circular grain silo, and mark the settlement warning area; generate settlement decision content associated with the next construction procedure based on the location of the settlement warning area, the settlement rate and the preset knowledge graph.

[0028] In step S11, the specific steps are as follows:

[0029] S111: A multi-source sensor array is deployed on the foundation and walls of the shallow circular grain silo to synchronously collect the initial readings of the multi-source sensor array, and the initial readings are aligned and calibrated in a spatiotemporal coordinate system to serve as initial settlement data.

[0030] S112: Dynamically monitor the construction process of shallow circular grain silos, collect the spatial trajectory of construction machinery in real time, and determine the dynamic load of construction machinery in the current construction process by combining the corresponding vibration frequency, while extracting the corresponding material load.

[0031] S113: The dynamic load of the construction machinery and the material load are used as boundary conditions and applied to the current readings of the multi-source sensor array to decouple the dynamic and static loads and perform spatiotemporal interpolation, eliminate noise drift caused by ambient temperature and construction vibration, and thus determine the deformation data reflecting the shallow circular grain silo.

[0032] In the embodiments of this application, a multi-source sensor array is deployed on the foundation and walls of the shallow circular grain silo, and the initial readings of the multi-source sensor array are collected synchronously. The initial readings are then aligned and calibrated in a spatiotemporal coordinate system to serve as initial settlement data, thus introducing the initial readings.

[0033] At this point, based on the force transmission path and easily deformable area distribution of the shallow circular grain silo, various types of sensor arrays are deployed on the foundation and walls of the silo. Multiple static levels are evenly spaced along the circumferential direction on the bottom surface of the circular foundation beam of the silo to capture the absolute elevation changes and uneven settlement trends of the foundation. Vibrating wire earth pressure cells are radially deployed in the soil layer below the foundation slab to sense the distribution of base reaction force. Meanwhile, distributed fiber optic strain sensors are layered along the height direction on the inner and outer surfaces of the silo wall to obtain the circumferential and vertical strain fields of the silo wall. Inertial measurement units (IMUs) are fixedly deployed at the characteristic nodes of the ring beam on the outer side of the silo wall to monitor the macroscopic tilt and attitude changes of the silo wall. This forms a three-dimensional multi-source sensor array covering the "foundation-soil-silo wall".

[0034] During the static curing period after the foundation construction of the shallow circular grain silo is completed and before the upper silo wall structure is loaded, the multi-source sensor array is triggered to synchronously acquire initial readings. To ensure the time synchronization of multi-source heterogeneous data, a data acquisition gateway supporting Precise Time Protocol (PTP) is used to perform unified clocking and synchronous sampling of the liquid level electrical signal of the hydrostatic level, the frequency signal of the earth pressure cell, the Brillouin frequency shift signal of the distributed optical fiber, and the acceleration and angular velocity signals of the IMU.

[0035] Meanwhile, a steady-state determination mechanism is introduced, which involves continuously collecting sensor data for a preset duration, such as 24 consecutive hours, and calculating the variance of each sensor reading. When the variance remains within a preset minimum fluctuation threshold range, the structure is determined to be in a quasi-steady state with minimal environmental interference. At this point, the output signals of each sensor are captured and converted into physical quantities to form the original initial reading set.

[0036] The initial readings must be collected when the structure is in a quasi-stable state. This quasi-stable state is defined as follows: within 12 hours of continuous monitoring, the difference between adjacent readings of each hydrostatic level does not exceed 0.05 mm, and there is no large construction machinery operating outside the warehouse wall, and the temperature change does not exceed ±1 degree Celsius per hour.

[0037] Steady-state determination is achieved by automatically calculating the standard deviation of the readings of each sensor over the past 12 hours. When the standard deviation of all sensors is lower than 0.5% of their respective ranges, the system determines that a steady state has been reached and automatically triggers the initial reading acquisition. The initial reading acquisition period is 10 consecutive minutes, with the sampling frequency uniformly set to 5 Hz. The arithmetic mean of the readings of each sensor during this period is used as the initial settlement data.

[0038] The absolute three-dimensional spatial coordinates of each sensor installation point were measured using a total station and real-time dynamic differential positioning technology. A local spatial rectangular coordinate system was established with the geometric center of the bottom plate of the shallow grain silo as the origin, the north direction as the X-axis, the east direction as the Y-axis, and the vertical upward as the Z-axis. The spatial coordinates of all sensors were uniformly transformed into this coordinate system to achieve the alignment and calibration of the initial settlement data in the spatial dimension. In the time dimension, the initial acquisition trigger time was taken as time zero, and all subsequent sensor data carried a precise timestamp relative to this time zero to ensure that the initial settlement data had a complete spatiotemporal reference.

[0039] A local three-dimensional spatial coordinate system was constructed with the vertical line of the geometric center of the shallow circular grain silo as the Z-axis, the north direction as the X-axis, and the east direction as the Y-axis. The initial acquisition time was set as the zero-time coordinate system to establish a global spatiotemporal reference. The absolute three-dimensional spatial coordinates of each static level, earth pressure cell, and IMU installation node were measured using a total station and RTK equipment. The topological path coordinates of the distributed optical fiber laid on the silo wall surface were also recorded. The spatial position parameters of the above sensors were uniformly converted and mapped to the global three-dimensional spatial coordinate system, and each sensor was given a unique spatial coordinate identifier, realizing the precise positioning of physical sensors in virtual space.

[0040] The original initial reading set is associated with the sensor spatial coordinate identifier and spatiotemporal alignment calibration is performed. For the spatial dimension, based on the global spatial coordinates of each sensor, a three-dimensional kriging interpolation algorithm is used to spatially extend the initial readings of discrete sensors, generating a continuous initial distribution field of physical quantities covering the entire foundation and warehouse wall.

[0041] For the time dimension, the initial acquisition time of each sensor is aligned to the zero-time coordinate system to eliminate data misalignment caused by acquisition time difference. The multi-dimensional data array, which includes absolute elevation, base reaction force, strain distribution and tilt attitude and has precise spatiotemporal labels after alignment and calibration, is fused and determined as the initial settlement data, which serves as the zero-point benchmark for settlement evolution in the subsequent construction process.

[0042] Specifically, taking a shallow circular grain silo in a grain depot expansion project as an example, the shallow circular grain silo is designed with a diameter of 30 meters and a wall height of 15 meters. It adopts the slipform construction process: the construction workers install static level instruments at "36 measuring points every 10 degrees at the bottom of the ring foundation beam of the shallow circular grain silo", and 27 earth pressure cells are embedded in three concentric circles radially below the base plate. As the slipform construction reaches the top elevation of the silo wall, four distributed optical fiber sensing cables are spirally laid from bottom to top along the four directions of 0 degrees, 90 degrees, 180 degrees and 270 degrees on the outer surface of the silo wall. Two high-precision IMU sensors are added at the intersection of the ring beam in the 0 degree and 180 degree directions to form a multi-source sensor array dedicated to the shallow circular grain silo.

[0043] After the foundation and slipform construction of the shallow circular grain silo were completed and the concrete reached its initial setting strength, the on-site data acquisition host started synchronous data acquisition at night when external construction machinery was stopped and the temperature was stable, such as at 2:00 AM. The acquisition gateway synchronously acquired the liquid level height of 36 hydrostatic instruments, the frequency modulus of 27 earth pressure cells, the Brillouin frequency shift distribution of 4 optical fibers, and the Euler angles of 2 IMUs at a frequency of 10Hz. The system monitored that the fluctuation of the readings at each measuring point was less than 0.1mm equivalent settlement for 4 consecutive hours, and determined that this was a steady state. Therefore, the data at the current moment was extracted as the original initial reading.

[0044] With the geometric center of the bottom plate of the shallow circular grain silo as the "origin: X=0, Y=0, Z=0", the spatial coordinates of the static level at the foundation beam in the "0-degree direction: due north" are accurately measured using a total station, and the spatial coordinates of the IMU at the silo wall ring beam in the 180-degree direction are "X=-15m, Y=0, Z=15m". The spiral cabling path of the distributed optical fiber is discretized into spatial coordinate strings with an interval of 0.5m, and all measuring points and paths are uniformly mapped to the three-dimensional local coordinate system of the shallow circular grain silo.

[0045] The system binds the 36 hydrostatic level heights captured at 2:00 AM to the aforementioned three-dimensional coordinates. Using the Kriging interpolation algorithm, it calculates the continuous three-dimensional initial elevation surface of the foundation plate within a 30-meter diameter range of the entire shallow grain silo based on these 36 elevation data. Simultaneously, it aligns the fiber optic strain, earth pressure, and IMU attitude data to the same acquisition time point, generating a comprehensive data matrix that includes the initial base reaction field, initial strain field of the silo wall, and initial spatial attitude of the shallow grain silo. This matrix serves as the initial settlement data benchmark for comparing settlement evolution during subsequent construction processes such as slipform lifting and material loading of the shallow grain silo.

[0046] Furthermore, the construction process of the shallow circular grain silo is dynamically monitored, the spatial trajectory of the construction machinery is collected in real time, and the dynamic load of the construction machinery for the current construction procedure is determined by combining the corresponding vibration frequency. At the same time, the corresponding material load is extracted, and the material load is fully considered.

[0047] At this time, multiple base station RTK positioning receivers and panoramic vision monitoring arrays are deployed around the construction area of ​​the shallow grain silo to perform real-time dynamic tracking of construction machinery operating around the silo and on the silo wall surface. RTK rover stations and inertial navigation modules are installed at the working end and key nodes of the chassis of the construction machinery, such as crawler cranes, concrete pump trucks, and material transport vehicles, to continuously acquire the three-dimensional coordinate sequence of mechanical feature points in the global spatiotemporal coordinate system at a preset high-frequency sampling rate, such as 10Hz. At the same time, the edge computing nodes of the panoramic vision array are used to perform real-time target detection and contour extraction on the video stream. The spatial position of the machinery identified by vision is fused with the RTK / inertial navigation data by Kalman filtering to generate a high-precision, low-latency three-dimensional spatial motion trajectory curve of the construction machinery.

[0048] Based on the acquisition of the three-dimensional spatial trajectory of the construction machinery, the operating speed and acceleration vectors of the machinery's working end are extracted. High-frequency acceleration sensors are simultaneously deployed on the chassis and boom of the construction machinery to collect vibration signals during operation in real time and perform Fast Fourier Transform (FFT) to extract the dominant vibration frequency and amplitude. Combining the machinery's own mass parameters and the acceleration of the working end, the quasi-static inertial force generated by macroscopic movement and lifting actions is calculated. At the same time, the vibration amplitude and dominant frequency are substituted into the preset mechanical dynamic response model, and a dynamic impact coefficient is introduced to calculate the dynamic load effect caused by the mechanical excitation force. Finally, the quasi-static inertial force and the dynamic load effect are vector-superimposed to determine the dynamic load vector of the construction machinery for the current construction process, which includes the spatial position, load magnitude, and frequency characteristics.

[0049] By integrating the Building Information Model (BIM) and material scheduling system at the construction site, the types, total mass, and designed stacking areas of materials used in the current construction process are obtained. Real-time point cloud data is collected from the surrounding area and the floor of the shallow grain silo using a 3D laser scanner or UAV oblique photography. The actual 3D envelope of the material stacking is extracted through point cloud denoising and semantic segmentation. This envelope is then converted into a voxel model in the global spatial coordinate system. Combined with measured or empirical bulk density parameters of the materials, the total mass of the materials is spatially integrated based on the voxel volume, thereby extracting a continuously distributed 3D material storage area precisely correlated with its spatial location.

[0050] Specifically, four panoramic high-definition cameras were deployed at a high point 50 meters from the center of the shallow grain silo, and an RTK base station was set up on site. At this time, a 56-meter boom pump truck was parked at the shallow grain silo in the "0 to 90 degree orientation: due north to due east" to pour concrete. RTK mobile stations were installed at the end of the boom and the chassis of the pump truck. During the pouring operation, the system continuously collected the spatial coordinates of the end of the pump truck boom at a frequency of 10Hz, and combined the panoramic vision to identify the swing trajectory of the pump truck boom, generating a precise three-dimensional spatial trajectory curve of the end of the pump truck boom reciprocating above and outside the silo wall.

[0051] High-frequency acceleration sensors deployed on the outriggers of the pump truck chassis capture vibration signals in real time caused by intermittent impacts from the pumping hydraulic system. After FFT frequency domain conversion, the dominant vibration frequency under the current operating condition is identified as 3.5Hz, and the vibration acceleration amplitude is 0.25g. The system calls the pump truck parameter library, knowing that the total weight of the pump truck boom and body is 40 tons, and calculates the "amplitude acceleration: 0.1m / s²" based on the spatial trajectory of the boom end. 2 The quasi-static overturning moment caused by boom movement was calculated; based on the dominant frequency and amplitude, a dynamic impact coefficient of 1.2 was introduced to calculate the vertical and horizontal dynamic load components brought about by 3.5Hz excitation. Finally, the construction machinery dynamic load vector applied by the pump truck to the foundation of the shallow circular grain silo was determined by superposition, including horizontal thrust and pulsating vertical pressure and their spatial points of action.

[0052] At this time, the slipform operation platform of the shallow circular grain silo is piled with steel bars and supporting rods to be poured, and sand and gravel aggregates are temporarily piled up outside the silo. The system reads the BIM construction progress module and learns that the "5th formwork of the silo wall: elevation 6m-9m" is currently being poured, requiring approximately 150 cubic meters of C40 concrete. At the same time, the drone performs oblique photography of the area around the shallow circular grain silo, generates point clouds, and then segments out the three-dimensional envelope surface of the sand and gravel aggregate pile piled on the northeast side of the silo. The system voxelsizes the envelope surface of the aggregate pile, based on the "medium sand bulk density: 1.5t / m³". 3 "The weight represented by each voxel is calculated, and the distribution position and self-weight of the steel bars on the platform are extracted based on the BIM coordinates of the slipform platform. Finally, the material loading field, which includes the ground loading outside the warehouse and the dynamic loading on the warehouse, is extracted and accurately mapped to the spatial position of the shallow circular grain warehouse, providing accurate static boundary input for subsequent decoupling of dynamic and static loads."

[0053] Therefore, the dynamic load of construction machinery and the load of material accumulation are used as boundary conditions and applied to the current readings of the multi-source sensor array to decouple the dynamic and static loads and perform spatiotemporal interpolation. This eliminates noise drift caused by ambient temperature and construction vibration, thereby determining the deformation data of the shallow circular grain silo that reflects the true deformation of the grain silo, thus introducing deformation data.

[0054] At this point, the dynamic load vector of the construction machinery and the material loading field are used as external excitation boundary conditions and mapped to the spatial nodes where the multi-source sensor array is located, respectively. The theoretical sensing response characteristics caused solely by the current construction load are calculated. For the theoretical sensing response characteristics, under given structural geometry and material parameters, using only the material loading and dynamic load of the construction machinery in the current construction process as external excitations, the theoretical response vector set at each measurement point of the multi-source sensor array is calculated through a physical information neural network surrogate model. The theoretical sensing response characteristics include theoretical displacement, strain, tilt, and acceleration. The theoretical sensing response characteristics are then... The theoretical sensing response characteristics are algebraically superimposed with the current raw readings acquired in real time by the multi-source sensor array to generate a mixed-state initial observation dataset containing real deformation, temperature drift, construction vibration noise, and load coupling response. For the mixed-state initial observation dataset, at the same time, the current raw readings acquired in real time by the multi-source sensor array are algebraically superimposed with the above-mentioned "theoretical sensing response characteristics" to form a mixed observation vector set containing multiple components such as real deformation response, temperature drift, construction vibration noise, and load coupling response. This set serves as the initial dataset for subsequent dynamic and static load decoupling, noise removal, and settlement characterization data extraction.

[0055] For the mixed-state initial observation dataset, the empirical wavelet transform algorithm is used to decompose it into different frequency bands. Based on the difference in frequency band distribution between the "dominant vibration frequency: high frequency band" of construction machinery dynamic load and the "foundation consolidation settlement: low frequency band" of material load and "foundation consolidation settlement", an adaptive bandpass filter bank is constructed. The high frequency cutoff frequency of the filter bank is set to be higher than the material load response frequency and covers the mechanical dynamic load frequency band, thereby extracting and separating the high frequency transient elastic response component caused by construction machinery dynamic load. The remaining low frequency component is regarded as the coupled response of static load and long-term settlement, realizing the physical decoupling of dynamic and static loads at the sensor reading level.

[0056] For each empirical wavelet component obtained from empirical wavelet decomposition, its nominal vibration frequency band is first determined according to the type of construction machinery. For example, for vibratory rollers, their operating frequency is usually in the 15–35Hz range; for concrete pump trucks, their pumping excitation and boom vibration frequency is usually in the 1–30Hz range; while the ground vibration frequency caused by heavy vehicles is mostly concentrated in the 10–50Hz range. Based on this, a frequency higher than the preset high-frequency cutoff frequency f is selected. high,min Empirical wavelet components (e.g., 5Hz or 10Hz) are labeled as high-frequency candidate frequency bands. Subsequently, the power spectral density of each candidate component is calculated to identify whether there are significant spectral peaks located within the typical excitation frequency band of the construction machinery. At the same time, based on the operation log of the construction machinery, the energy ratio of the candidate components during the machine's operation period to the machine's downtime period is calculated. When this ratio is greater than a set threshold (e.g., 3–5 times), the frequency band is determined to be strongly correlated with the dynamic load of the construction machinery.

[0057] Furthermore, by comparing the energy distribution of this component at different spatial locations, frequency bands that have no obvious attenuation characteristics in space or are unrelated to the mechanical location are eliminated. Finally, empirical wavelet components that meet the requirements of frequency band range, significant spectral peaks, and strong synchronization with the mechanical operating conditions are identified as high-frequency transient elastic response components caused by the dynamic load of construction machinery, while the remaining components below the low-frequency cutoff frequency are regarded as the coupled response of static load and long-term settlement.

[0058] Based on dynamic and static decoupling, for the slow-varying drift noise caused by ambient temperature, the temperature sensor readings built into the multi-source sensor array or the pure temperature-sensitive channels in the distributed optical fiber are extracted. According to the sensor's thermal expansion coefficient and temperature-strain calibration model, the temperature self-compensation correction at each spatial node is calculated and subtracted from the low-frequency response component. Simultaneously, for the low-frequency construction vibration noise remaining in the decoupling low-frequency signal, such as low-frequency foundation micro-vibrations caused by heavy vehicle movement, an independent component analysis algorithm is used, combined with the spatial topology of the multi-source sensor array, to separate and eliminate spatially globally correlated but structurally different noise. The residual vibration noise component of the actual settlement mode is used to obtain the true physical deformation response. The true physical deformation response refers to the substantial changes in the geometric shape and spatial position of the shallow grain silo structure and foundation soil caused by quasi-static loads such as material loading, structural self-weight, and foundation consolidation settlement during the construction of the shallow grain silo. This response is the set of effective settlement displacement and strain fields that purely reflect the mechanical behavior of the structure after eliminating interference components such as high-frequency transient elastic vibration caused by construction machinery dynamic load, thermal expansion and contraction drift caused by environmental temperature changes, and low-frequency foundation micro-vibration noise caused by construction vibration.

[0059] To address the real physical deformation response of discrete spatial nodes, a spatiotemporal variogram function is constructed by introducing a time dimension, using a spatiotemporal coordinate system as the reference. Taking into account the elevation sensitivity of the hydrostatic level, the strain continuity of the distributed optical fiber, and the attitude tilt characteristics of the IMU, a spatiotemporal co-Kriging interpolation algorithm is adopted. Using the continuous strain data of the distributed optical fiber as a covariate, the settlement and displacement of discrete points are jointly estimated and extended in a spatiotemporal manner. In the time dimension, the phase misalignment caused by the small deviation of the sensor sampling clock is eliminated, and in the spatial dimension, the blind spots of the sensor deployment are filled. Finally, a spatiotemporally continuous three-dimensional multidimensional deformation data field reflecting the real deformation of the shallow circular grain silo is reconstructed.

[0060] In the context of settlement monitoring during the construction of shallow circular grain silos, a three-dimensional multidimensional deformation data field refers to a set of structural deformation states with spatiotemporally continuous distribution characteristics obtained by spatiotemporally reconstructing and extending the real physical deformation responses of discrete nodes within the walls, foundation, and subgrade soil of a shallow circular grain silo under a unified three-dimensional spatial coordinate system and time axis. This data field covers the entire three-dimensional geometric domain of the shallow circular grain silo in the spatial dimension, uses the time axis of construction process as an index in the temporal dimension, and simultaneously includes heterogeneous but mechanically coupled deformation observations such as settlement displacement components, circumferential and radial strain components, and attitude tilt components in the "multidimensional" dimension. It fully represents the real deformation state of the shallow circular grain silo in a digital twin space in a unified spatiotemporal function form.

[0061] Specifically, the system uses the pump truck's "dynamic load vector: 3.5Hz pulsating load" and the sand and gravel pile outside the silo and the steel reinforcement material pile on top of the silo as boundary conditions. It inputs the numerical transfer function matrix of the shallow circular grain silo and calculates the theoretical strain and settlement response of the sensors in the silo wall and foundation area at 0 to 90 degrees under the above load. At this time, the system collects the current raw readings of the static level, fiber optic cable and IMU in the area in real time. It finds that the reading of the static level of the foundation at 0 degrees fluctuates drastically and the strain value of the fiber optic cable outside the silo wall is also significantly larger. The system superimposes the theoretical response and the raw readings to generate a mixed initial observation dataset that includes the effects of pump truck vibration, material pressure and ambient temperature.

[0062] For the mixed-state data from the 0-degree azimuth foundation static level and the fiber optic cable of the silo wall, the system initiates empirical wavelet transform for frequency domain decomposition. Since the main frequency of the pump truck dynamic load is 3.5Hz, while the material loading and the actual settlement of the shallow grain silo are slow-developing low-frequency processes with frequencies usually below 0.01Hz, the system constructs an adaptive filter bank with a cutoff frequency of 0.1Hz. Through filtering, the 3.5Hz frequency and the accompanying high-frequency elastic vibration response are successfully separated from the original readings. The remaining extremely low-frequency signal represents the coupled response of the static material loading and the actual settlement of the foundation, thus achieving dynamic and static decoupling between the pump truck dynamic load and the material static load.

[0063] The system detected a slow upward trend in the low-frequency strain component of the optical fiber at 0 degrees azimuth during the afternoon. By extracting temperature channel data from the fiber, it was found that the temperature of the warehouse wall in this area, exposed to direct sunlight, increased by 8°C. Based on the thermal expansion coefficient of concrete (approximately 1.0 × 10⁻⁶), this is consistent with the observation that the temperature of the warehouse wall in this area increased by approximately 8°C. -5The temperature strain compensation calculated from " / ℃" is approximately 80 microstrains. The system accurately deducts this temperature drift from the low-frequency component. At the same time, for the "low-frequency foundation micro-vibration: frequency about 0.5Hz" caused by heavy trucks entering and leaving the construction site, which causes synchronous small fluctuations in all static levels, the system uses an independent component analysis algorithm to identify this global noise mode that is highly correlated at each measuring point in space and removes it from the signal, ensuring that the remaining signal only reflects the true stress and deformation of the shallow circular grain silo structure.

[0064] The system reconstructs the discrete points of the actual deformation after purification. For example, the static level of the foundation at 0 degrees measured a settlement of 0.5 mm, and at 90 degrees measured a settlement of 0.1 mm, while the fiber optic cable between 0 and 90 degrees of the silo wall measured a continuous circumferential strain distribution. Using fiber optic strain as a covariate, the system uses a spatiotemporal co-Kriging interpolation algorithm to calculate the continuous settlement curve of the foundation in the area between 0 and 90 degrees where no static level is installed. It also synchronizes the sampling time difference of each sensor and finally outputs a spatiotemporally continuous three-dimensional deformation data field that reflects the actual deformation, covering the entire foundation and walls of the shallow circular grain silo. This provides a clean input source for the subsequent settlement evolution in the digital twin space.

[0065] In step S12, the specific steps are as follows:

[0066] S121: Input the deformation data and the initial settlement data into a preset estimation framework, and perform multi-scale data fusion in the estimation framework to generate corresponding settlement characterization data, the settlement characterization data including vertical displacement and tilt vector;

[0067] S122: Input the settlement characterization data into a preset digital twin space to drive the synchronous deformation of the multi-dimensional coupled grid in the digital twin space. Further, use the settlement characterization data as a boundary constraint condition. The boundary constraint condition is applied to the neural network corresponding to the digital twin space to trigger the settlement evolution of the settlement characterization data, and then output the corresponding multi-dimensional settlement displacement field. The multi-dimensional settlement displacement field includes the overall three-dimensional spatial displacement of the shallow grain silo, the differential settlement of the foundation, and the uneven tilt state of the silo wall.

[0068] In the embodiments of this application, deformation data and initial settlement data are input into a preset estimation framework, and multi-scale data fusion is performed in the estimation framework to generate corresponding settlement characterization data. The settlement characterization data includes vertical displacement and tilt vector, thus introducing settlement characterization data.

[0069] At this point, a state-space estimation framework centered on the overall settlement state of the shallow grain silo is constructed. This estimation framework includes state equations and observation equations. Deformation data reflecting the actual deformation of the shallow grain silo is used as the observation input vector, initial settlement data is used as the initial prior benchmark of the state vector, and the temporal characteristics of material loading and mechanical dynamic load during construction are used as the system control input vector. Process noise covariance and observation noise covariance matrices are introduced to establish a dynamic recursive relationship of multi-source heterogeneous data in the time dimension.

[0070] The core of this estimation framework consists of state equations and observation equations. The state equation is defined as follows: the current state vector equals the previous state vector plus the product of the system control input vector and the sampling time interval, plus the process noise vector. The observation equation is defined as follows: the current observation vector equals the current state vector plus the observation noise vector. The state vector is a five-dimensional column vector, containing the vertical displacement of the center point of the shallow circular grain silo floor, the tilt angles around the X and Y axes of the local coordinate system, and the radial displacement and circumferential strain of a reference point in the middle of the silo wall. The observation vector is a four-dimensional column vector, containing the vertical displacement of a point on the edge of the floor measured by a hydrostatic level, the tilt angles around the X and Y axes measured by an inertial measurement unit, and the circumferential strain in the middle of the silo wall measured by a distributed fiber optic sensor.

[0071] The system control input vector is a three-dimensional column vector, which sequentially includes the total material load applied to the shallow grain silo in the current construction process, the maximum vertical dynamic load amplitude of the construction machinery, and the working azimuth angle of the construction machinery. Both the process noise covariance matrix and the observation noise covariance matrix are diagonal matrices. Their diagonal elements are determined based on the variance values ​​measured by each sensor in the static calibration test and the empirical estimates, respectively. For example, the process noise variance of the hydrostatic level is taken as 0.01 mm², and the observation noise variance is taken as 0.04 mm². These matrices remain unchanged in each filter update, thereby achieving a weighted fusion of measured values ​​and prior estimates.

[0072] Within the estimation framework, multi-scale decomposition and filtering recursion are performed on the state vector. The "overall rigid body motion trend of the shallow grain silo, including overall vertical translation and macroscopic rotation around the horizontal axis" is defined as the macroscopic scale state, and the elastic deflection of the local structure and the micro-strain of the material are defined as the microscopic scale state. By constructing a block state transition matrix, the optimal estimate of the macroscopic scale state and the correction value of the microscopic scale state at the current moment are calculated using the prediction and update steps in Kalman filtering. During the update process, high-gain weights are assigned based on the high confidence of the hydrostatic level and IMU for the macroscopic rigid body displacement, and corresponding local correction weights are assigned based on the high sensitivity of the distributed optical fiber for the micro-strain. This achieves the decoupling and optimal fusion of physical responses at different spatial scales within a unified framework.

[0073] Within the estimation framework, the state vector is divided into two sub-blocks: macro-scale state and micro-scale state. Correspondingly, the state transition matrix is ​​written in a block-based form, called the block state transition matrix. Its mathematical meaning is: in a unified state-space model, the block matrix describes the dynamic relationship between macro-scale and micro-scale states during the time update process, including their respective self-evolution laws and their mutual coupling effects.

[0074] In constructing the block state transition matrix, the state vector of the shallow circular grain silo-foundation system is decomposed into macroscopic state (x_macro) and microscopic state (x_micro) according to multiple scales. The macroscopic state describes the overall rigid body motion trend, including overall vertical translation and macroscopic rotation about the horizontal axis; the microscopic state describes the elastic deflection and micro-strain of the local structure, such as radial displacement and strain at key nodes. In the program implementation, the state vector is arranged in the form x = [x_macro; x_micro], where the first n_macro components correspond to the macroscopic state and the last n_micro components correspond to the microscopic state.

[0075] For macroscopic states, a rigid body dynamics model is established based on overall settlement monitoring data (hydrostatic level, IMU, etc.) and the overall structural kinematics. For example, low-order differential equations or autoregressive models are used to describe the evolution of overall vertical displacement and macroscopic rotation angles, yielding the macroscopic state transition matrix F_macro. For microscopic states, a dynamics model of local deflection and strain evolution is established based on finite element models or elasticity relationships. For example, structural dynamic equations or strain transfer relationships are used to describe the evolution of the microscopic state, yielding the microscopic state transition matrix F_micro. These two local models are typically identified or modeled independently at their respective time scales.

[0076] Based on structural mechanics principles, this study determines how macroscopic rigid body motion drives local microscopic deformation, and how local microscopic strain reacts to the overall rigid body motion. For example, overall tilting significantly alters the local stress field and strain distribution of the silo wall, while local damage and strain concentration affect the overall rotational stiffness through stiffness degradation. Linearizing this coupling relationship yields two coupling block matrices: the influence matrix F_macro_to_micro of the macroscopic state on the microscopic state, and the influence matrix F_micro_to_macro of the microscopic state on the macroscopic state. In many simplified engineering scenarios, the macroscopic and microscopic time scales differ significantly; a slow-fast two-scale approximation can be used, neglecting the instantaneous feedback from the microscopic to the macroscopic. In this case, F_micro_to_macro can be approximated as zero or small.

[0077] The four blocks are assembled into a unified block state transition matrix F using a matrix concatenation method, in the form of:

[0078] F=[F_macro,F_macro_to_micro;F_micro_to_macro,F_micro];

[0079] The top-left block, F_macro, describes the contribution of the macroscopic state to its own evolution; the bottom-right block, F_micro, describes the contribution of the microscopic state to its own evolution; the top-right block, F_macro_to_micro, describes how the macroscopic state drives changes in the microscopic state; and the bottom-left block, F_micro_to_macro, describes how the microscopic state, in turn, modifies the macroscopic state. In the specific implementation, based on the physical mechanism and identification results, it is determined whether each block is zero, whether it is a constant matrix, and whether it needs to be updated over time or under operating conditions.

[0080] In the prediction step, the state estimate from the previous time step is recursively extrapolated to the current time step using a block state transition matrix F, from both macroscopic and microscopic scales, to obtain the predicted state value and the predicted covariance. In the update step, a measurement matrix is ​​constructed based on the monitoring characteristics of different sensors: for sensors such as hydrostatic levels and IMUs that are mainly sensitive to macroscopic rigid body displacement, the corresponding measurement matrix mainly takes the x_macro component; for sensors such as distributed optical fibers that are sensitive to microscopic strain, the measurement matrix mainly takes the x_micro component. Then, based on the covariance prediction and measurement noise statistics, the Kalman gain is calculated, and different degrees of correction weights are applied to the macroscopic and microscopic states according to high confidence or high sensitivity, respectively, to achieve decoupling and optimal fusion of macroscopic and microscopic responses within a unified framework.

[0081] Based on the topological and geometric characteristics of the shallow circular grain silo, the center point of the silo's bottom plate is defined as the reference origin. Based on the optimal estimate of the macroscopic state, the translation component of the reference origin in the Z-axis direction of the global spatial coordinate system is extracted and determined as the overall vertical displacement of the shallow circular grain silo. At the same time, the rotation components around the X-axis and Y-axis of the global spatial coordinate system in the macroscopic state are extracted. Combined with the geometric radius of the shallow circular grain silo, the spatial eccentric displacement of the top of the silo wall relative to the bottom plate is calculated using the rigid body kinematics rotation transformation matrix. This eccentric displacement is then transformed into an inclination vector containing the inclination direction and inclination angle.

[0082] The vertical displacement and tilt vector determined by the macro-scale state estimate are algebraically superimposed and reconstructed with the local strain flexural field determined by the micro-scale state correction. Using the three-dimensional geometric topology of the shallow grain silo, the superimposed physical quantities are mapped onto the surface and basic grid nodes of the shallow grain silo, generating a multi-dimensional data set that simultaneously contains overall settlement trend characteristics and local micro-deformation characteristics. This data set is defined as settlement characterization data, which represents the spatiotemporal distribution of the vertical displacement field and tilt vector field in a structured matrix form.

[0083] Specifically, the system constructs a state-space estimation framework for shallow circular grain silos; it sets the "initial settlement data, such as the zero-point elevation of the foundation and the zero-strain state of the silo wall calibrated at 2:00 AM" as the initial a priori benchmark for the state vector; as the slipform construction progresses, the system uses the deformation data of the foundation settlement of 0.5mm-0.1mm in the 0-90 degree azimuth calculated by S113 and the circumferential strain data of the silo wall as the observation input vector, and the current steel reinforcement load and pump truck dynamic load on the silo as the control input vector, inputting them into the estimation framework to establish a dynamic recursive model of the shallow circular grain silo from the initial zero time to the current slipform construction time.

[0084] The system performs multi-scale decoupling and fusion of deformation data within the estimation framework. For shallow circular grain silos, the overall subsidence and tilt of the entire 30-meter diameter silo are defined as macro-scale states, while the elastic deformation of the silo wall caused by sliding mode lifting is defined as micro-scale states. In the Kalman filter update, since the 0.5mm subsidence measured by the 0-degree azimuth foundation static level mainly reflects macro-rigid displacement, the system assigns it a high gain weight. Meanwhile, the micro-strain measured by the distributed optical fiber of the silo wall from 0 to 90 degrees mainly reflects local structural deformation, and the system assigns it a local correction weight. This separates the overall rigid subsidence component and the local micro-strain component of the shallow circular grain silo under the current load.

[0085] The system uses the geometric center of the 30-meter bottom plate of the shallow circular grain silo as the reference origin and extracts the vertical displacement and tilt vector based on the optimal estimate of the macroscopic state. The filtering results show that the center of the bottom plate has a translation of 0.3 mm in the global Z-axis direction, which is determined to be the overall vertical displacement. At the same time, the filtering identifies a tiny rotation of 0.001 degrees around the X-axis of the silo body. Combined with the 15-meter wall height of the shallow circular grain silo, and based on the rigid body kinematic rotation transformation, the top of the silo wall has an eccentric displacement of 0.26 mm in the Y-direction relative to the bottom plate. This generates a tilt vector pointing in the positive Y-direction, which includes a tilt angle of 0.001 degrees and the eccentric direction.

[0086] The system algebraically superimposes and geometrically maps the macroscopic overall "vertical displacement: 0.3mm" and "tilt vector: Y-axis eccentricity 0.26mm" with the local elastic deflection strain of the silo wall at the microscopic scale from 0 degrees to 90 degrees. The reconstructed results show that the actual vertical displacement of the bottom plate of the shallow circular grain silo at the 0-degree azimuth is 0.5mm. The 0.3mm overall settlement superimposed with the local tilt and deflection mapping amount, which includes a matrix set containing the overall spatial attitude and local refined deformation, serves as the settlement characterization data of the shallow circular grain silo under the current slipform process.

[0087] Furthermore, the settlement characterization data is input into a preset digital twin space to drive the synchronous deformation of the multi-dimensional coupled grid in the digital twin space. The settlement characterization data is further used as a boundary constraint condition, which is applied to the neural network corresponding to the digital twin space to trigger the settlement evolution of the settlement characterization data, thereby outputting the corresponding multi-dimensional settlement displacement field. This multi-dimensional settlement displacement field includes the overall three-dimensional spatial displacement of the shallow circular grain silo, the differential settlement of the foundation, and the uneven tilt state of the silo wall, fully considering the multi-dimensional settlement displacement field.

[0088] At this point, in the preset digital twin space, a multi-dimensional coupled mesh model that is completely consistent with the geometric topology of the physical shallow circular grain silo is pre-constructed. This model includes a foundation soil mesh, a foundation plate mesh, and a silo shell mesh, and the mesh nodes have mechanical coupling correlation attributes. The "settlement characterization data: including vertical displacement and tilt vector" is accurately projected onto the corresponding boundary nodes of the multi-dimensional coupled mesh through a spatial coordinate mapping matrix. The vertical displacement and tilt vector are used as driving commands to drive the coupled mesh in the digital twin space to undergo geometric deformation updates, so that the appearance of the digital twin model and the macroscopic real-time deformation of the physical shallow circular grain silo are kept in spatiotemporal synchronization. Among them, the spatial coordinate mapping matrix refers to the matrix expression of the linear or piecewise linear coordinate transformation relationship established between the preset overall three-dimensional global coordinate system (such as the geodetic coordinate system or engineering coordinate system) and the local digital twin grid coordinate system. This matrix maps the three-dimensional physical coordinate vector of any point on the physical grain shallow circular silo and its foundation into the local coordinates or node number index of the corresponding node in the multi-dimensional coupled grid, thereby realizing the accurate spatial positioning of settlement characterization data and the application of boundary conditions.

[0089] A pre-trained Physical Information Neural Network (PINN) is deployed within a digital twin space. The loss function of this PINN incorporates the structural equilibrium equations of the shallow circular grain silo, the constitutive relations of the soil, and the slip criteria of the contact surface. The settlement characterization data is transformed into boundary constraints for the neural network input layer. Vertical displacement is transformed into forced displacement constraints at the bottom boundary of the computational domain, and tilt vectors are transformed into moment equivalent load constraints at the side boundaries of the computational domain. When the boundary constraints are applied to the neural network, its forward inference and backward gradient optimization processes are triggered. Under the constraint of satisfying the physical conservation laws, the network adaptively infers the mechanical response and displacement distribution of the "uncovered area: i.e., the blind zone of physical sensor deployment" of the boundary constraints, thereby realizing the evolution of settlement characterization data from discrete observation points to a continuous physical field across the entire field.

[0090] The physical information neural network employs a fully connected feedforward neural network structure, comprising one input layer, four hidden layers, and one output layer. The input layer contains three neurons, receiving the circumferential angular coordinates, radial coordinates, and vertical elevation coordinates of a spatial point in the local cylindrical coordinate system. Each hidden layer contains sixty-four artificial neurons, all using the hyperbolic tangent function as the activation function. This function ensures that the output value is continuously differentiable between -1 and +1, facilitating the embedding of physical constraints. The output layer contains three neurons, outputting the displacement components of the spatial point along the X, Y, and Z directions in the global coordinate system. The network has approximately 80,000 trainable parameters, capable of adequately fitting the spatial distribution of the settling field in a shallow grain silo.

[0091] The loss function of the physical information neural network consists of a weighted sum of three parts: a data loss term, a physical loss term, and a boundary loss term. The data loss term is calculated by summing the squares of the differences between the network's predicted displacement values ​​and the measured displacement values ​​at known spatial points in the training dataset, and then averaging the sums. This ensures that the network output matches the measured settlement data. The physical loss term is calculated by using automatic differentiation to calculate the partial derivatives of the displacement with respect to spatial coordinates based on the network's predicted displacement field, thereby obtaining the strain and stress components. These components are then substituted into the equilibrium differential equations and linear elastic constitutive equations of the shallow circular grain silo thin-shell structure. The squares of the differences between the left and right sides of the equations are calculated, integrated within the computational domain, and then averaged. This ensures that the network output satisfies the laws of conservation of mechanics. The boundary loss term is calculated by extracting the displacement values ​​at the boundary nodes corresponding to the preset boundary conditions based on the network's predicted displacement field, summing the squares of the differences between these values ​​and the given displacement values ​​under the boundary conditions, and then averaging the sums. This ensures that the network output satisfies the displacement boundary constraints. The weighting coefficients for the three items mentioned above are 1.0, 0.1, and 0.5, respectively. This combination of weighting coefficients was determined through multiple sets of preliminary experiments and can fully satisfy the constraints of the physical equations while ensuring the accuracy of data fitting.

[0092] The training dataset for the physical information neural network is entirely derived from the settlement characterization data generated in step S121 and the measured displacement data obtained after driving the digital twin space in step S122. Specifically, one thousand spatial sampling points are selected on the surface and inside the foundation of the shallow grain silo. Two hundred of these points are the actual sensor placement locations, and the remaining eight hundred points are encrypted points obtained through spatiotemporal co-kriging interpolation. Each sampling point contains a five-dimensional vector, namely, spatial circumferential angular coordinates, radial coordinates, vertical elevation coordinates, and the corresponding measured X-direction and Y-direction displacements. This five-dimensional vector constitutes a training sample. All one thousand training samples form the complete training dataset. It should be noted that this neural network does not require additional external labeled data; all training samples come from the measured or interpolated settlement data output from the preceding steps of this system, thus constituting self-supervised training.

[0093] The full computational domain node displacement tensor is extracted from the physical information neural network under boundary constraints. The full computational domain node displacement tensor refers to the three-dimensional tensor formed by organizing and storing the displacement vectors at all grid nodes in an orderly manner within the overall computational domain of the shallow circular grain silo-foundation system output by the physical information neural network, using grid node numbers, spatial coordinate components, and time steps as indices. This tensor covers all structural areas such as the foundation soil, foundation plate, and silo shell in space, and corresponds to each load step or time discrete point under boundary constraints in time. It is used to characterize the continuous or discrete deformation state of the shallow circular grain silo-foundation system in the computational domain. Based on the transformation relationship between the local coordinate system and the global coordinate system, the displacement tensor is decomposed into translational components along the X, Y, and Z axes and rotational components around the coordinate axes. By calculating the Z-direction displacement difference of each node of the foundation plate relative to the initial reference plane, the differential settlement matrix of the foundation is extracted. By calculating the radial offset and elevation change of the top feature node of the silo wall relative to the bottom node, the uneven tilt distribution map of the silo wall is extracted. The global translational components, the differential settlement matrix of the foundation, and the uneven tilt map of the silo wall are tensor-stitched together to generate a structured multidimensional settlement displacement field.

[0094] The differential settlement matrix refers to the numerical matrix obtained by organizing the vertical displacement difference of each node relative to the initial reference plane in matrix form on the design reference plane of the foundation slab, using the coordinates of the grid nodes as indexes. This matrix represents the non-uniform distribution of vertical settlement at different locations of the foundation slab and is a numerical expression of differential settlement on a discrete grid. It is used to evaluate the overall tilt, local depression, and uneven compression characteristics of the foundation.

[0095] The multi-dimensional settlement displacement field is fed back into the digital twin space, and the node attributes of the multi-dimensional coupled mesh are re-assigned and physically rendered. Based on the magnitude and gradient of the differential settlement and uneven tilt of the silo wall, the mesh cells are given differentiated color-level mapping and vector arrow labels. Thus, the dynamic cloud map of the overall three-dimensional spatial displacement of the shallow circular grain silo, the three-dimensional contour surface of the foundation settlement, and the vector field of the silo wall tilt are intuitively presented in the digital twin space, completing the dimensional reconstruction and digital twin mapping from one-dimensional settlement characterization data to a three-dimensional multi-dimensional settlement displacement field.

[0096] Specifically, a high-precision, multi-dimensional coupled mesh of a shallow circular grain silo has been constructed within the digital twin space. The system projects the settlement characterization data, namely the overall vertical displacement of 0.3 mm and the tilt vector pointing in the positive Y direction, with a tilt angle of 0.001 degrees, onto the center of the bottom plate of the shallow circular grain silo and the boundary mesh nodes at the 0-degree azimuth through a spatial mapping matrix. Under this driving command, the mesh model of the shallow circular grain silo in the digital twin space undergoes synchronous deformation. The basic mesh at the 0-degree azimuth sinks significantly, while the mesh at the 90-degree azimuth rises slightly, intuitively reproducing the macroscopic posture of the physical shallow circular grain silo tilting towards the 0-degree azimuth.

[0097] A physical information neural network customized for shallow circular grain silos was deployed within the digital twin space. The system applied 0.3 mm vertical displacement and 0.001 degree tilt vector as boundary constraints to the input of the neural network. The 0.3 mm vertical displacement served as a forced displacement of the bottom boundary, and the 0.001 degree tilt was converted into an equivalent torque applied to the side boundary. Under the constraints of the embedded Winkler foundation model and the mechanical equations of the thin shell structure, the network began to deduce and automatically calculate the displacement response of the foundation in the 180-degree to 270-degree blind zone of the shallow circular grain silo within a 30-meter diameter area where no sensors were deployed, as well as the redistribution of internal forces and deformations of the 15-meter-high silo wall structure, realizing the evolution from local observation data to the physical field of the entire silo.

[0098] The system extracts and solves the displacement tensor of the entire computational domain of the shallow circular grain silo from the neural network inference output. Decomposition along the Z-axis shows that the maximum settlement of the bottom plate at 0 degrees is 0.52 mm, while the settlement at 180 degrees is only 0.08 mm. The difference of 0.44 mm between the two constitutes the basic differential settlement matrix of the shallow circular grain silo. The solution along the height of the silo wall shows that the top of the silo at 0 degrees has a radial displacement of 0.35 mm outward, while the displacement at 90 degrees is minimal. This distribution feature constitutes the uneven tilt state of the silo wall. Combining the overall translation in the X and Y directions, a multi-dimensional settlement displacement field covering all the above physical information is finally generated.

[0099] The system writes back the calculated multi-dimensional settlement displacement field to the digital twin space. In the rendering engine, the foundation plate of the shallow grain silo is rendered as a thermal cloud map transitioning from "red: 0.52mm large settlement" to "blue: 0.08mm small settlement" based on the differentiated settlement amount. The surface of the silo wall is attached with an array of green vector arrows indicating the direction and magnitude of the offset according to the uneven tilt state. Through this field update, engineers can holographically see the overall three-dimensional spatial displacement of the shallow grain silo, the foundation settlement depression area, and the most unfavorable tilt position of the silo wall in the digital twin space, providing a complete three-dimensional field input for subsequent curvature calculation and damage inversion.

[0100] In step S13, the specific steps are as follows:

[0101] S131: Dynamically identify the multi-dimensional settlement displacement field, extract the radial displacement distribution of the silo wall during the identification process, and combine the cubic spline interpolation mechanism to trigger the smooth fitting of the radial displacement fitting surface of the silo wall, thereby calculating the circumferential curvature change of the silo wall.

[0102] S132: Invert the curvature change and determine the circumferential stress distribution of the silo wall during the inversion process; obtain the usage time and strength data of the silo wall, and perform multiple calculations in combination with the circumferential stress distribution of the silo wall to determine the damage evolution trend of the silo wall during the calculation process. The damage evolution trend of the silo wall presents the evolution of the silo wall from an intact state to a local failure state.

[0103] In the embodiments of this application, the multi-dimensional settlement displacement field is dynamically identified, and the radial displacement distribution of the silo wall is extracted during the identification process. The radial displacement fitting surface of the silo wall is smoothed by combining the cubic spline interpolation mechanism, thereby calculating the curvature change of the silo wall in the circumferential direction, thus introducing the curvature change.

[0104] At this point, a spatiotemporal sliding window that rolls with the construction process time step is constructed and applied to the multi-dimensional settlement displacement field. Within each time step, by extracting the node displacement increments within the window coverage area, the dynamic evolution trajectory of the overall three-dimensional spatial displacement of the shallow circular grain silo, the differential settlement of the foundation, and the uneven tilt of the silo wall is identified. Simultaneously, a displacement increment threshold judgment mechanism is introduced. When the change in displacement field variables in adjacent time steps exceeds a preset threshold, high-frequency identification and capture of the displacement field are triggered, thereby accurately locking the key construction moment when the settlement displacement field undergoes a significant change in the time domain.

[0105] For the identified multi-dimensional settlement displacement field, a local three-dimensional cylindrical coordinate system {r,θ,z} is established with the vertical line of the center of the shallow circular grain silo as the Z-axis and the center of the silo bottom as the origin. Using coordinate rotation and projection transformation matrices, the three-dimensional spatial displacement vector of the silo wall in the global Cartesian coordinate system is projected onto this cylindrical coordinate system. The displacement along the cylindrical coordinate tangent and the vertical displacement along the Z-axis are separated, and only the displacement component ur(θ,z) along the radial coordinate axis is extracted. ur is resampled and spatially meshed according to the circumferential angle θ and the vertical elevation z, thereby extracting the radial displacement distribution matrix that characterizes the magnitude of radial deformation of the silo wall surface along the circumferential and height directions.

[0106] To address the discreteness and high-frequency measurement noise in the extracted radial displacement distribution matrix caused by sensor spacing, a bicubic spline interpolation mechanism is introduced. Using the circumferential angle θ and vertical elevation z as independent variables and the radial displacement ur as the dependent variable, a bicubic polynomial interpolation function is constructed within a grid interval formed by four adjacent discrete measurement points. This interpolation function is set to satisfy continuity constraints of first-order partial derivatives and second-order mixed partial derivatives at the grid nodes. By solving the tridiagonal equations with the aforementioned continuity constraints, the interpolation coefficients in the entire domain are obtained, thereby triggering the smooth fitting of the silo wall radial displacement fitting surface and generating a radial displacement fitting surface u~r(θ,z) that eliminates local sawtooth fluctuations and possesses second-order continuous smoothness.

[0107] Based on the smoothed radial displacement fitting surface u~r(θ,z), according to the definition of surface curvature in differential geometry, the theoretical initial radius R of the shallow circular grain silo is introduced; the circumferential curvature change Δκ of the silo wall is defined as the difference between the circumferential curvature of the deformed silo wall and the "initial circular curvature: 1 / R". By calculating the second-order partial derivative of the radial displacement fitting surface along the circumferential angle θ, and combining the principle of local geometric approximation of the surface, and further combining the circumferential curvature change, the distribution field of the circumferential curvature change of the entire surface of the silo wall is solved node by node and output. At this point, the principle of local geometric approximation of a surface refers to the optimal second-order approximation of the shape of the surface near any point on a smooth surface by establishing a local orthogonal coordinate system and Taylor expansion, using a quadratic surface (elliptic parabola, hyperbolic parabola, or parabolic cylinder). This allows the local curvature of the surface to be completely characterized by the principal curvature and its direction. This principle ensures that near a point, a "closest parabola" can be used to replace the original surface to study its curvature and shape, and is the theoretical basis for curvature calculation and local geometric analysis of surfaces.

[0108] A cylindrical coordinate system is established with the geometric center of the bottom plate of the shallow circular grain silo as the origin, the vertically upward direction as the positive Z-axis, and the 0-degree circumferential angle starting line from the center of the silo bottom pointing due north. Under this coordinate system, the position of any point on the silo wall surface is uniquely determined by three coordinates: the circumferential angle (ranging from 0 to 360 degrees), the radial distance (constantly 15 meters for a shallow circular grain silo with a radius of 15 meters), and the vertical elevation (ranging from 0 meters at the top of the bottom plate to 15 meters at the top of the silo wall).

[0109] For any point on the surface of the warehouse wall, its displacement components in the X and Y directions in the global Cartesian coordinate system are projected onto the radial direction in the cylindrical coordinate system through coordinate rotation transformation. The specific transformation formula is: radial displacement equals the X-direction displacement multiplied by the cosine of the circumferential angle plus the Y-direction displacement multiplied by the sine of the circumferential angle. For a measuring point at a circumferential angle of 0 degrees and a vertical elevation of 5 meters, if the measured X-direction displacement is 0.35 mm outward and the Y-direction displacement is 0 mm, then the radial displacement is 0.35 mm; a positive value indicates that the warehouse wall bulges outward. Following the above method, all sensor deployment points and interpolated densification points are calculated one by one to obtain a discrete set of radial displacement sample points. Each sample point contains three pieces of information: circumferential angle, vertical elevation, and radial displacement value.

[0110] The circumferential angle, ranging from 0 to 360 degrees, is divided into 180 intervals of two degrees each. The vertical elevation, ranging from 0 to 15 meters, is divided into 75 intervals of 0.2 meters each, resulting in a total of 13,500 grid nodes (180 x 75). For each grid node, a bicubic spline interpolation function is used to calculate its radial displacement. Within each small grid bounded by four adjacent sample points, this interpolation function is a cubic polynomial with respect to both the circumferential angle and the vertical elevation, containing a total of sixteen undetermined coefficients. The determination of these coefficients requires that the function values, first-order partial derivatives, and second-order mixed partial derivatives of the interpolation surface at the sample points be continuous.

[0111] In the specific solution, firstly, cubic spline interpolation is performed on the sample points at each elevation position along the circumferential angle direction to obtain the first-order partial derivative values ​​along the circumferential direction at each sample point. Then, cubic spline interpolation is performed on the sample points at each angle position along the vertical elevation direction to obtain the first-order partial derivative values ​​along the vertical direction at each sample point. Finally, using these partial derivative values ​​and the function values ​​at the sample points, a strip-shaped sparse linear equation system is solved to obtain sixteen coefficients within each small grid. This linear equation system is solved using the chasing method, which is fast and numerically stable. After the solution is completed, the circumferential angle and vertical elevation of any grid node are substituted into the interpolation polynomial of the corresponding small grid to obtain the radial displacement fitting value of that node. The resulting bicubic spline interpolation surface has a second-order continuous derivative, eliminating measurement noise and sawtooth fluctuations in the original discrete data. The circumferential curvature of the surface is defined as the second-order partial derivative of the radial displacement with respect to the circumferential angle divided by the square of the initial radius of the shallow grain silo.

[0112] Specifically, the system constructs a sliding window with a time step of 1 hour to continuously scan the multi-dimensional settlement displacement field of the shallow grain silo. When the pump truck moves from the 0-degree azimuth to the 45-degree azimuth to continue pouring at a certain moment, the system identifies that the incremental change in the displacement field at the 45-degree azimuth exceeds the preset threshold of 0.05mm, and then triggers the dynamic recognition mechanism to capture the instantaneous snapshot of the displacement field of the foundation and silo wall of the shallow grain silo caused by the construction action.

[0113] The system establishes a cylindrical coordinate system with the center of the shallow circular grain silo as the origin, and projects the Cartesian displacement vectors within the range of 0 degrees to 360 degrees and 0 meters to 15 meters in elevation. After stripping away the vertical compression displacement of the silo wall, the radial displacement component ur is extracted separately. For example, at 0 degrees in elevation and 5 meters in elevation, the radial outward displacement is extracted as 0.35 mm, and at the same elevation at 180 degrees in elevation, the radial inward displacement is extracted as -0.10 mm. These radial displacement data arranged by angle and elevation constitute the radial displacement distribution matrix of the shallow circular grain silo at this time.

[0114] Because the fiber optic cable and the static level are "discretely distributed" on the surface of the shallow grain silo, such as a feature point every 10 degrees, direct calculation of curvature is easily affected by the amplification of discrete errors. The system starts a bicubic spline interpolation mechanism. In a two-dimensional spatial grid of 0 to 360 degrees and 0 to 15 meters, with discrete displacements at 10-degree intervals and 1-meter elevation intervals as anchor points, a bicubic polynomial that satisfies the continuity of the first and second derivatives is constructed. Through smooth fitting, a continuous radial displacement fitting surface u~r(θ,z) covering the entire 30-meter diameter outer wall of the shallow grain silo is generated, which effectively eliminates surface burrs caused by local measurement jitter.

[0115] The system calculates the circumferential curvature change based on the smooth surface. For the 0-degree azimuth elevation of 5 meters, the radial displacement u~r=0.35 of the fitted surface is obtained. At the same time, the system calculates the second partial derivative of the point along the circumferential direction θ. Since the 0-degree azimuth is a local extreme value region of radial displacement, its second partial derivative is negative, such as -0.28mm. Thus, the circumferential curvature change at this point is obtained. This change quantitatively characterizes the degree of local outward deformation caused by uneven settlement at the 0-degree azimuth of the shallow circular grain silo, providing accurate geometric parameter input for subsequent inversion of the circumferential stress distribution.

[0116] Furthermore, the curvature change was inverted, and the circumferential stress distribution of the silo wall was determined during the inversion process. The usage time and strength data of the silo wall were obtained, and multiple calculations were performed in combination with the circumferential stress distribution of the silo wall to determine the damage evolution trend of the silo wall during the calculation process. The damage evolution trend of the silo wall shows the evolution of the silo wall from an intact state to a local failure state, and the evolution content is fully considered.

[0117] At this point, based on the theory of thin-shell structure mechanics, a physical mapping relationship between the curvature change of the silo wall and the internal forces of the structure is established. The distribution field of the circumferential curvature change of the silo wall, Δκ(θ,z), is taken as input. Combined with the equivalent bending stiffness D of the silo wall cross section, it is determined by the silo wall thickness and the elastic modulus of the material. Based on the principle of calculating bending normal stress in mechanics of materials, the additional circumferential bending moment Mθ(θ,z) = D multiplied by Δκ(θ,z) caused by uneven settlement is inverted. Furthermore, based on the plane section assumption, the additional circumferential bending moment is linearly decomposed along the silo wall thickness direction. Combined with the circumferential tension of the membrane derived from the radial displacement of the silo wall, the maximum bending stress at the edge of the section and the membrane stress are vector-superimposed to determine the circumferential stress distribution field σθ(θ,z) of the silo wall containing tension and compression partitioning characteristics.

[0118] By connecting to the database of shallow circular grain silos, the design strength grade, mix proportion, and pouring records of the silo wall concrete were obtained, and the time t used from the completion of pouring to the current moment was extracted. A time-varying strength model of concrete based on maturity theory was introduced to consider the nonlinear growth law of compressive strength fc(t) and tensile strength ft(t) over time caused by early hydration reaction of concrete, as well as the influence of creep relaxation effect on long-term stress redistribution, and dynamically calculate the actual bearing capacity parameters of the silo wall material under the current construction process. At the same time, the yield strength and distribution parameters of the circumferential reinforcement configured in the silo wall were retrieved to construct a multiphase material strength dataset of the silo wall that reflects the current material age characteristics.

[0119] Using the circumferential stress distribution field σθ(θ,z) as the applied load and the time-varying intensity dataset as the bearing capacity boundary, a multi-field coupled multi-operational architecture is constructed by introducing the theory of continuum damage mechanics. Based on the stress-intensity interference model, the stress ratio at the current moment, i.e., the ratio of applied stress to time-varying intensity, is calculated. The stress ratio is substituted into the preset damage evolution equation to calculate the elastic modulus degradation coefficient and the proportion of microcrack propagation area under the current load step. The degraded elastic modulus is then substituted back into the inversion model for iterative correction until the stress field and damage field reach self-equilibrium, thereby outputting the damage variable Dmg(θ,z) characterizing the degree of material performance degradation during the calculation process. Here, the multi-operational architecture refers to a hierarchical, iterative computational framework constructed in the digital twin space to solve multi-field coupled, nonlinear damage evolution problems.

[0120] Using the time axis of construction procedures as the horizontal axis and the damage variable Dmg as the vertical axis, the time-series tracking and interpolation reconstruction of multiple calculation results throughout the entire life cycle of the shallow circular grain silo wall are performed. Multi-level damage failure criteria are set, where Dmg∈[0,0.1) is defined as the intact state, Dmg∈[0.1,0.6) is defined as the microcrack initiation and propagation state, Dmg∈[0.6,0.9) is defined as the macrocrack penetration state, and Dmg≥0.9 is defined as the local failure state of local crushing or tensile fracture. By extracting the damage variable trajectory of key nodes of the silo wall with the used time t, a damage evolution trend map showing the dynamic evolution of the silo wall from the intact state to the local failure state is generated.

[0121] Specifically, the system inverts the circumferential curvature change Δκ at an elevation of 5 meters in the 0-degree azimuth position; the designed thickness of the shallow circular grain silo wall is 0.25 meters, and the current elastic modulus of concrete is 30 GPa, from which the equivalent bending stiffness D is calculated; based on this stiffness and curvature change, the additional circumferential bending moment Mθ caused by local outward deformation in the 0-degree azimuth position is inverted; since the 0-degree azimuth position protrudes outward, the outer surface of the silo wall is under tension and the inner surface is under compression. The system superimposes the bending stress caused by the bending moment with the circumferential tensile stress of the silo wall membrane, determining that the outer surface in the 0-degree azimuth position bears a circumferential tensile stress of 2.5 MPa, forming a circumferential stress distribution field on the entire silo surface.

[0122] The system obtained the usage time t of the 0-degree orientation silo wall of the shallow circular grain silo from the BIM construction log, which is 7 days, i.e., 1 day after slipforming was removed. Due to the extremely short concrete age, the system calculated the actual tensile strength of the current C40 concrete based on the time-varying strength model to be only 1.2 MPa, far lower than the 28-day standard value of 2.4 MPa, and the actual compressive strength to be 15 MPa. This data indicates that the shallow circular grain silo is currently in the early fragile stage with the lowest material strength, providing stringent boundary conditions for subsequent damage calculations.

[0123] The system performs multiple calculations on the circumferential tensile stress of 2.5 MPa on the outer surface at 0 degrees and the current tensile strength of 1.2 MPa. The stress ratio is as high as 2.08, far exceeding the material's bearing capacity limit, triggering the Mazars damage model calculation. The calculation shows that the elastic modulus degradation rate of this region exceeds 60% under the current load step, and the damage variable Dmg reaches 0.65. The system resubmits the degraded elastic modulus into the inversion model for iteration. Due to local stiffness softening, the stress in adjacent regions is redistributed. When the calculation converges, the system outputs a damage variable field that includes the obviously softened region at 0 degrees.

[0124] The system tracks the time-series damage evolution trend of the outer surface nodes at the 0-degree azimuth of the shallow circular grain silo. The trend graph shows that during the first 0-5 days of use, the slipform construction period, the structure was under pressure but did not exceed the limits, and the damage variable Dmg was close to 0, indicating an intact state. On the 6th day, when the slipform was lifted, the local stress surged, and Dmg jumped to 0.15, entering the microcrack initiation state. On the 7th day, under the dynamic load of the pump truck, Dmg climbed to 0.65, and a macroscopic crack penetration state appeared. It is predicted that if the construction sequence is not adjusted, with the subsequent increase in load, Dmg will exceed 0.9 within a few hours, showing a local failure state of concrete tensile cracking. Thus, the entire evolution process of the shallow circular grain silo from intact to local failure is presented intuitively and quantitatively.

[0125] In step S14, the specific steps are as follows:

[0126] S141: Dynamically identify the damage evolution trend, and determine the local stress concentration and damage accumulation area of ​​the warehouse wall during the identification process. Combine the fracture mechanics criteria to predict the corresponding crack characteristics, which show the crack initiation location, propagation direction and crack width.

[0127] S142: Map the crack features to the construction history of the shallow grain silo, assess the impact of uneven settlement on structural connectivity and integrity, thereby triggering the corresponding settlement early warning event, and marking the corresponding settlement early warning area in the digital twin space.

[0128] S143: Based on the detection of the settlement warning area, determine the three-dimensional coordinate position, current settlement rate and historical cumulative settlement of the settlement warning area, and then input the preset knowledge graph of the construction safety of shallow grain silos, and generate settlement decision content strongly associated with the next construction procedure through graph neural network reasoning and semantic matching. The settlement decision content includes material loading path change plan, construction machinery travel avoidance route and foundation dynamic reinforcement strategy.

[0129] In the embodiments of this application, the damage evolution trend is dynamically identified, and the local stress concentration and damage accumulation area of ​​the warehouse wall are determined during the identification process. The corresponding crack characteristics are predicted by combining fracture mechanics criteria. The crack characteristics show the crack initiation location, propagation direction and crack width, thus introducing crack characteristics.

[0130] At this point, using the timeline of construction sequence progression as a dynamic index, the damage evolution trend is continuously tracked and dynamically identified. A damage rate field is constructed by calculating the incremental gradient of the damage variable Dmg within adjacent time steps. Specifically, during the damage evolution process indexed by the construction sequence progression timeline, the spatiotemporal field organized by the spatial point x, representing the gradient of the damage variable D(x,t) with respect to time or the increment of adjacent time steps, is called the damage rate field. A dynamic adaptive threshold determination mechanism is introduced: when the incremental damage gradient of a spatial node at the current time step exceeds a preset damage acceleration evolution threshold, the region is marked as an active evolution zone. Furthermore, the absolute value level of the damage variable within the active evolution zone is extracted. When the absolute value exceeds a critical "damage threshold," such as Dmg ≥ 0.5, the region is defined as a local stress concentration and damage accumulation zone, thereby achieving dynamic framing of high-risk areas of the structure. Optionally, Dmg ≥ 0.5 is not limited here and is derived based on a threshold reference table.

[0131] For the identified local stress concentration and damage accumulation zone, the circumferential stress distribution field within this region is extracted. By solving the stress tensor eigenvalues ​​of this local stress field, the direction of the maximum principal stress and the principal stress trajectory in this region are determined. Combining fracture mechanics criteria, the spatial coordinate point where the maximum principal stress reaches the material's time-varying tensile strength is taken as the crack initiation location. Simultaneously, based on the local strain energy density factor and the maximum energy release rate criterion, the driving force direction for crack propagation is calculated, predicting that the crack will extend perpendicular to the "direction of maximum principal stress: i.e., along the principal compressive stress trajectory or shear band," thereby determining the crack propagation direction vector on the bin wall surface.

[0132] In the framework of linear elastic fracture mechanics, fracture mechanics criteria refer to a class of mechanical judgments used to determine whether a crack has started, in which direction it propagates, and whether the propagation has become unstable.

[0133] The local strain energy density factor is a direction-dependent scalar function S(θ) constructed near the crack tip or stress concentration point, using the stress intensity factor as a variable. It characterizes the "energy-driven intensity" of crack propagation in different directions; its minimum direction corresponds to the predicted crack propagation direction. Crack initiation is determined when Smin reaches the material's critical value. In numerical implementation, the discretized S(θ) field can be obtained by calculating the strain energy density distribution in the local region and integrating or fitting it along different directions. This field is used to predict the crack propagation direction vector on the bin wall surface.

[0134] The maximum energy release rate criterion is an energy-based fracture mechanics criterion for determining the direction of crack propagation and crack initiation: First, it is assumed that the crack will extend in the direction that maximizes the energy released per unit area of ​​crack propagation (energy release rate G); Second, when the maximum energy release rate Gmax in this direction reaches the fracture toughness Gc of the material, crack initiation and propagation are determined.

[0135] After determining the crack initiation location and propagation direction, a virtual crack model is introduced along the predicted propagation path. The crack propagation process is simulated as a fracture process zone of material strain softening. Based on the fracture energy Gf of concrete and the tensile strength softening curve, the crack opening displacement is calculated. The relative displacement on both sides of the crack is decomposed into elastic deformation and cracking deformation caused by material softening. By integrating the stress-crack opening displacement curve within the fracture process zone, a mechanical equilibrium equation for the crack tip opening displacement and crack propagation length is established. This equation is iteratively solved to deduce the macroscopic crack width extending along the propagation direction to the current stress level, achieving a three-dimensional quantitative characterization of the crack characteristics. The virtual crack model refers to simplifying the microcrack zone at the macroscopic crack tip in concrete fracture analysis into a virtual crack capable of transmitting cohesive stress, and defining a softening constitutive relationship between cohesive stress and crack opening displacement on it, using the fracture energy Gf as the basis for the calculation. F A nonlinear fracture mechanics model that serves as a material fracture parameter.

[0136] The spatial coordinates (x, y, z) of the crack initiation location, the angle vector d of the expansion direction, and the macroscopic crack width w are fused to generate a structured set of crack feature parameters. This set of parameters, in the form of a multi-dimensional matrix, accurately maps the spatial topology and mechanical opening state of the crack on the warehouse wall in the digital twin space, providing an intuitive physical criterion for the subsequent generation of settlement early warning events.

[0137] Specifically, the system dynamically identifies the damage evolution trend during the current slipform construction period of the shallow circular grain silo. Calculations show that in the area at 0 degrees azimuth and 5 meters elevation, the incremental gradient of the damage variable Dmg reached 0.15 / hour during the slipform lifting process in the last 2 hours, far exceeding the preset accelerated evolution threshold of 0.05 / hour. Moreover, the current absolute damage value Dmg in this area has reached 0.65, exceeding the critical threshold of 0.5. The system then clearly defines the outer surface area of ​​the silo wall at 0 degrees azimuth and 5 meters elevation as a local stress concentration and damage accumulation area, and issues a high-risk area lock signal.

[0138] The system extracts the stress field of the high-risk area at 0 degrees azimuth and solves for its tensor eigenvalues. It determines that the maximum principal stress in this area is a circumferential tensile stress of 2.5 MPa, with its direction of action along the circumferential tangent. Since the maximum principal stress exceeds the tensile strength of the concrete at the current age, the system predicts that the crack initiation location is at the point of maximum stress on the outer surface at 0 degrees azimuth, with coordinates: X=15m, Y=0, Z=5m, based on the maximum energy release rate criterion. It also predicts that the crack will extend in a direction perpendicular to the circumferential tensile stress, that is, along the vertical direction of the silo wall (Z-axis) from the crack initiation point upwards and downwards, thus determining the vertical extension direction vector of the crack.

[0139] The system introduces a virtual crack model along the predicted vertical expansion path. Considering that the concrete in the shallow circular grain silo area is only 7 days old, the system retrieves the "fracture energy parameters, such as Gf=80" and the bilinear softening constitutive curve of the early concrete. By integrating, the stress-opening displacement relationship in the fracture process zone is calculated. Under the current circumferential tensile stress of 2.5MPa, it is deduced that the crack has penetrated the softening zone and formed a macroscopic crack. The total length of the crack extending vertically is about 0.4 meters, and the maximum crack opening displacement at the crack initiation point reaches 0.25mm, thus completing the quantitative deduction of the crack width.

[0140] The system integrates the above simulation results to generate a set of current crack feature parameters for the shallow grain silo. This set of parameters clearly indicates that there is a macroscopic crack extending vertically at an elevation of 5 meters at 0 degrees, with a length of about 0.4 meters and a maximum opening width of 0.25 mm. These multi-dimensional crack features are synchronously mapped into the digital twin space, and the spatial location and opening shape of the crack are accurately marked with a red warning line on the three-dimensional model of the shallow grain silo. This provides conclusive physical evidence of damage for the next stage of generating settlement early warning events in conjunction with the construction process.

[0141] Furthermore, the crack features are mapped to the construction history of the shallow grain silo to assess the impact of uneven settlement on the structural connectivity and integrity, thereby triggering corresponding settlement early warning events and marking the corresponding settlement early warning areas in the digital twin space, fully considering the settlement early warning areas.

[0142] At this point, the characteristic parameter set of the crack is extracted, and the timestamp of the crack occurrence is used as an index to connect with the construction progress log and load history database of the shallow circular grain silo. The crack initiation location, expansion direction and width evolution path are spatiotemporally aligned with the spatial distribution of material loading, dynamic load trajectory of construction machinery and differential settlement development curve of the foundation in the historical construction process. Through correlation analysis, the dominant construction process that induces the crack feature and the corresponding extreme load time are determined, thereby establishing a causal mapping relationship between macroscopic cracks on the structural surface and specific disaster-causing loads in the construction process.

[0143] The predicted crack features were embedded into a three-dimensional finite element mesh model of the shallow circular grain silo wall. Contact surface elements or element birth and death techniques were used to simulate the local structural discontinuity caused by the cracks. Based on the defective model, the redistribution of the internal force transmission path of the silo wall under the continuous development of uneven settlement was calculated. The deflection and truncation of the principal stress traces were extracted, and it was assessed whether stress extreme value superposition zones were formed at both ends of the crack, thus causing further penetration of the crack. At the same time, by comparing the overall strain energy or equivalent bending stiffness of the cracked model and the intact model, the degree of weakening of the overall deformation resistance and structural connectivity of the shallow circular grain silo by local cracks was quantified, and the overall degradation factor was calculated.

[0144] Regarding the overall degradation factor, it is defined as a dimensionless index used to quantify the degree of weakening of the overall structural mechanical properties of shallow circular grain silos by local crack defects. It aims to reflect the ratio of stiffness degradation and interruption of internal force transmission path caused by damage to the structure.

[0145] The overall degradation factor is calculated using a comparative analysis method. First, in a finite element simulation environment, the overall elastic strain energy or equivalent bending stiffness of a healthy shallow circular grain silo model under uneven settlement load conditions is extracted as a benchmark parameter. Then, for a damaged model with embedded crack features, its current overall elastic strain energy or equivalent bending stiffness is calculated under identical boundary conditions and load conditions. Based on the variational principle of energy, there is a definite correspondence between changes in structural stiffness and changes in strain energy. Therefore, in the calculation process, the ratio of strain energy between the healthy model and the damaged model is used to characterize the degree of structural stiffness retention, or the relative increment of strain energy is used to reflect the degree of degradation of the overall structural performance. Finally, this relative change in physical quantity is converted into a numerical output of the overall degradation factor. The larger this value, the more significant the decrease in the overall structural deformation resistance caused by the local discontinuity due to cracks, and the more severe the weakening of the overall connectivity of the structure.

[0146] Based on the degree of structural connectivity weakening and the overall degradation factor, combined with the preset multi-level thresholds for the safe operation of shallow circular grain silos, a triggering logic for settlement early warning events is constructed. When a local crack only causes stress redistribution but does not cut off the main force transmission path, and the overall degradation factor is in the low-level range, a "local damage early warning" event is triggered. When the crack expands and causes a significant truncation of the force transmission path in the circumferential compression zone or tension zone, posing a risk of connection and penetration, and the overall degradation factor exceeds the medium threshold, a "structural overall damage early warning" event is triggered. When the predicted crack characteristics indicate that a penetrating macroscopic crack will form and cause local instability of the silo wall, a "critical overall failure early warning" event is triggered, and the type of early warning event, the disaster-causing process, and the risk level are written into the system message queue.

[0147] The characteristics of cracks that trigger settlement warning events and their impact range are projected into a digital twin space. Centered on the crack initiation coordinates, a three-dimensional spatial envelope is delineated as the settlement warning area based on the gradient extrapolation range of the crack propagation direction and stress redistribution. In the digital twin space, a high-contrast semi-transparent warning color block is used to cover and render the envelope, while a warning line marker with a pulse flashing effect is dynamically drawn along the crack propagation trajectory. The warning event level, dominant construction procedures, crack characteristic parameters, and overall degradation indicators are displayed in a floating window next to the warning area, realizing a visual mapping of abstract warning data to an intuitive spatial expression.

[0148] Specifically, the system backtracked the vertical crack (0.25mm wide, 0.4m long) predicted by S141 at the 0-degree azimuth. Through comparison, it was found that the crack initiation time coincided with the time when the 56-meter boom pump truck was pumping concrete at full load at the 0-degree azimuth. During this period, the dynamic load of the pump truck outriggers and the eccentric material load at the 0-degree azimuth of the silo wall were superimposed, resulting in a differential settlement of 0.44mm in the foundation at the 0-degree azimuth. The system thus established a mapping relationship and confirmed that the vertical crack was caused by uneven settlement of the foundation due to a sudden increase in local load at the 0-degree azimuth.

[0149] The system embedded the 0.25mm wide and 0.4m long vertical crack into the digital analysis model of the shallow circular grain silo. Calculations showed that since the shallow circular grain silo mainly relies on circumferential tension to resist the lateral pressure of the internal material, the vertical crack cut off the main path of circumferential tensile force transmission at the 0-degree azimuth, resulting in significant stress redistribution on both sides of the crack. The circumferential tensile force was forced to bypass the crack end, resulting in more than twice the stress concentration at the crack end. At the same time, the overall stiffness assessment showed that the overall bending stiffness of the model with the crack decreased by 8%, the structural connectivity was locally damaged at the 0-degree azimuth, and the overall degradation factor was calculated to be 0.65.

[0150] The system determines the degradation factor of 0.65 based on the early warning threshold of the shallow circular grain silo. Since 0.65 exceeds the overall damage threshold of 0.6 and the crack has cut off the circumferential force transmission path, the system determines that this state may cause the crack to penetrate the silo wall at any time during the subsequent slipform lifting, thereby triggering the "structural overall damage early warning" event. This early warning event is clearly marked as follows: the risk level is II, the disaster-causing process is the pump truck pumping at 0 degrees with full load and slipform lifting, and the core risks are the circumferential force transmission path cut-off and the crack penetration and expansion.

[0151] The system performs spatial marking within the digital twin model of the shallow circular grain silo. Using a 0-degree azimuth elevation of 5 meters as the geometric center, it extends outwards by 1.5 meters based on the stress redistribution range, generating a three-dimensional envelope covering "4 meters: circumferential" × "3 meters: vertical". In the digital twin space, this envelope is rendered as a semi-transparent orange-red warning block, and the predicted crack trajectory, 0.4 meters long, is dynamically marked with a pulsed, flashing dark red line. Next to the orange-red block, a pop-up window details the "Structural Overall Damage Warning," allowing on-site engineers to intuitively perceive the spatial location and degree of danger of potential hazards within the shallow circular grain silo, providing a solid spatial positioning basis for subsequent construction decisions.

[0152] Therefore, based on the detection of the settlement warning area, the three-dimensional coordinates, current settlement rate, and historical cumulative settlement of the settlement warning area are determined. Then, the preset knowledge graph of the construction safety of shallow grain silos is input, and through graph neural network reasoning and semantic matching, settlement decision content strongly associated with the next construction procedure is generated. The settlement decision content includes material loading path change plan, construction machinery travel avoidance route, and foundation dynamic reinforcement strategy, thus introducing settlement decision content.

[0153] At this point, for the settlement warning area marked in the digital twin space, the three-dimensional coordinate position of the geometric center of the area's envelope in the global spatiotemporal coordinate system is extracted; simultaneously, the displacement field data derived from the digital twin space is called, the derivative of the displacement of the area with respect to time within the current time window is calculated to determine the current settlement rate, and the displacement increment of the area from the initial settlement data to the current time is integrated to determine the historical cumulative settlement; the above three-dimensional coordinate position, current settlement rate and historical cumulative settlement are tensor-concatenated according to a preset data structure to construct a query vector characterizing the spatiotemporal evolution and mechanical state of the warning area.

[0154] The query vector is input into a preset knowledge graph for the construction safety of shallow grain silos. This knowledge graph uses construction procedures, geological conditions, structural responses, load types, and treatment strategies as node entities, and causal triggering, spatial constraints, and mechanical coupling relationships between entities as directed edges. Using the three-dimensional coordinates and settlement rate in the query vector, a dual matching is performed at the spatial index layer and the mechanical threshold layer of the knowledge graph to locate the set of entity nodes similar to the current warning state. Based on the graph walk algorithm, starting from the matched entity nodes, a local subgraph containing adjacent procedure nodes, related load nodes, and historical treatment strategy nodes is extracted to lock the topology associated with the current settlement situation.

[0155] The extracted local subgraphs are input into a pre-trained graph neural network (GNN) model. Under the message passing mechanism, node features are aggregated and iteratively updated along directed edges in the subgraph, so that the spatiotemporal state information of the warning area is propagated to adjacent construction process nodes and disposal strategy nodes. High-dimensional semantic matching is performed by calculating the cosine similarity between the warning feature vector and the embedded representation of each candidate strategy node. The set of strategy nodes with the highest similarity and confidence exceeding the preset threshold is selected, and the constraints and expected goals of activating the strategy are traced back along the directed edges, thereby inferring a candidate set of settlement decision content that is strongly correlated with the spatial location and load characteristics of the next construction process.

[0156] The candidate set of settlement decision content is physically constrained and logically reorganized; based on the three-dimensional coordinates of the warning area and the work scope of the next process, a material loading path change plan and construction machinery travel avoidance route are generated to avoid the settlement sensitive area; based on the foundation bearing capacity gap reflected by the current settlement rate and the historical cumulative settlement, a dynamic foundation reinforcement strategy with corresponding grouting materials and reinforcement depth is matched; the above plans, routes and strategies are formatted and encapsulated to output settlement decision content including spatial coordinate guidance, load threshold limit and execution time window, so as to realize closed-loop intervention for the next construction process.

[0157] For physical constraint verification, when screening and revising the candidate set of settlement decision content, each candidate scheme is placed in the real physical environment of the grain shallow circular silo-foundation system for "feasibility" and "safety" verification. The core is to ensure that the scheme does not overload, become unstable, or exceed the bearing capacity of materials and foundation at the physical level.

[0158] First, based on structural design data, geological survey reports, and monitoring data, a series of physical constraints are established in advance, including: allowable bearing capacity of the foundation, allowable settlement and tilt limits, material strength and deformation limits, additional load limits of construction machinery and slack loads on shallow grain silos, and geometric boundary conditions such as working space and passage height.

[0159] For each candidate material loading path change scheme, construction machinery travel avoidance route, or foundation reinforcement strategy, the system sequentially verifies: whether the foundation exceeds the allowable bearing capacity under the new load, whether it will cause the settlement rate or cumulative settlement of the already warned area to exceed the preset threshold, whether the machinery travel route conflicts with existing facilities or the space of the ongoing construction process, and whether the reinforcement depth and grouting pressure will cause additional deformation or secondary risks to the existing structure. The verification process uses the stress field, settlement field, and deformation field in the digital twin model for rapid simulation calculation. The candidate scheme is applied to the current structure-foundation system, and the response of key physical quantities is extracted and compared with preset thresholds. If the response value exceeds the threshold, the scheme is determined to violate physical constraints and is eliminated or the parameters are adjusted and re-verified until all candidate schemes meet the physical feasibility requirements.

[0160] For logical reorganization, under the premise of satisfying physical constraints, the preliminary screening of decision content candidate sets is reorganized and rearranged at the process level to eliminate spatiotemporal conflicts, optimize resource utilization and improve overall executability.

[0161] The system first extracts the logical dependencies between various processes based on the construction organization design, overall schedule, and current construction status, such as "foundation reinforcement of a certain section must be completed before subsequent silo wall loading" and "material loading must not overlap with critical machinery operations within the same time window." Then, it performs temporal and spatial conflict detection on candidate solutions to identify whether incompatible activities are scheduled within the same work period or spatial area, such as simultaneously scheduling heavy machinery passage and heavy material loading in a settlement-sensitive area.

[0162] For sets of conflicting solutions, the system reorganizes them by adjusting the sequence of procedures, staggering execution time windows, changing the layout of workspaces, or replacing them with lower-risk alternatives. This ensures that activities are staggered in time and separated in space, forming a logically feasible and non-conflicting decision sequence. During the reorganization process, the system also comprehensively evaluates and ranks candidate solutions based on objectives such as critical path, total project duration, and resource balance, prioritizing solutions that have minimal impact on the total project duration, smooth resource fluctuations, and effectively avoid settlement-sensitive areas. The solutions, routes, and strategies, after physical constraint verification and logical reorganization, are formatted and encapsulated, outputting settlement decision content including spatial coordinate guidance, load threshold limits, and execution time windows, thus achieving closed-loop intervention for the next construction procedure.

[0163] Specifically, for the 0-degree azimuth settlement warning area, the system extracts the three-dimensional coordinates of the geometric center of the area's envelope: X=15.5m, Y=0m, Z=5m; it calls the digital twin simulation data to calculate that the current settlement rate of the area is 0.15mm / h, and the historical cumulative settlement since the start of construction is 1.2mm; the system concatenates the above parameters according to the data structure [15.5,0,5,0.15,1.2] to construct a query vector representing the high-risk state at 0 degrees azimuth.

[0164] The system inputs the query vector into the knowledge graph of construction safety for shallow circular grain silos. The graph stores a large number of historical engineering cases and standards. Based on the features of "coordinates: 0 degrees orientation" and "high settlement rate: 0.15mm / h", the system quickly locates entity nodes such as "differential settlement of the foundation at 0 degrees orientation of the shallow circular grain silo" and "early tensile cracking of concrete". Through the graph walk algorithm, the system extracts local subgraphs containing nodes such as "next process: slipform lifting of silo walls", "related load: dynamic load of pump truck outriggers", and treatment strategies such as "adjusting the load position", "far-end avoidance of equipment", and "shallow grouting" adopted under similar historical conditions.

[0165] The system inputs this local subgraph into a graph neural network (GNN). During message transmission, the high settlement rate feature at 0 degrees propagates to the adjacent "pump truck outrigger dynamic load" and "slipform lifting" nodes, triggering an update of the risk weights for these nodes. Through semantic matching calculation, the GNN infers that the current settlement situation has the highest cosine similarity to the "dynamic load superposition induced structural instability" mode, activating three high-confidence strategy nodes: "dynamic load stripping," "load transfer," and "foundation bearing capacity reinforcement," forming a candidate set of decision content for the next slipform lifting process.

[0166] The system, considering the actual spatial constraints of the shallow circular grain silo, reorganizes the candidate set to generate specific decisions: for material stacking, it generates a material stacking path change plan, moving the steel bars originally planned to be stacked in the 0-degree orientation to the "180-degree orientation: safety zone"; for construction machinery, it generates a construction machinery travel avoidance route, planning to move the concrete pump truck to the 90-degree orientation for operation, and setting its outriggers to be at least 5 meters away from the outer wall of the shallow circular grain silo; for insufficient foundation bearing capacity, it generates a dynamic foundation reinforcement strategy, guiding the implementation of pressure grouting 2 meters outside the foundation in the 0-degree orientation, with a grouting depth of 3 meters below the foundation; the above decisions are simultaneously pushed to the construction control end, directly guiding the safe execution of the next process.

[0167] Please see Figure 2 The settlement detection system for the shallow circular grain silo during construction is applied to the aforementioned settlement detection method for the shallow circular grain silo during construction; the settlement detection system for the shallow circular grain silo during construction includes:

[0168] The monitoring module 21 is used to collect the initial readings of the multi-source sensor array of the shallow grain silo and use them as initial settlement data; monitor the construction process of the shallow grain silo, obtain the current readings of the multi-source sensor array, extract the material load and construction machinery dynamic load of the current construction process, and thus determine the deformation data of the shallow grain silo.

[0169] The settlement evolution module 22 is used to fuse the deformation data and the initial settlement data, and generate settlement characterization data during the fusion process; input the settlement characterization data into a preset digital twin space, and perform settlement evolution in the digital twin space, thereby determining the multi-dimensional settlement displacement field of the shallow grain silo.

[0170] The damage evolution trend module 23 is used to determine the circumferential curvature change of the silo wall based on the identification of the multi-dimensional settlement displacement field, invert the circumferential stress distribution of the silo wall based on the curvature change, and determine the damage evolution trend of the silo wall in combination with the service time and strength data of the silo wall.

[0171] Settlement decision module 24 is used to predict the corresponding crack features based on the identification of the damage evolution trend of the silo wall, and to determine the corresponding settlement early warning event in combination with the construction history of the shallow circular grain silo, and to mark the settlement early warning area; and to generate settlement decision content associated with the next construction procedure based on the location of the settlement early warning area, the settlement rate and the preset knowledge graph.

[0172] It should be noted that although multiple modules are mentioned in the detailed description above, this division is not mandatory; in fact, according to the embodiments of this disclosure, the features and functions of two or more modules or described above can be embodied in one module; conversely, the features and functions of one module described above can be further divided into multiple modules to be embodied.

[0173] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for detecting settlement during the construction of shallow circular grain silos, characterized in that, include: The initial readings of the multi-source sensor array in the shallow circular grain silo were collected and used as initial settlement data. The construction process of the shallow circular grain silo is monitored, the current readings of the multi-source sensor array are obtained, and the material load and dynamic load of the construction machinery in the current construction process are extracted to determine the deformation data of the shallow circular grain silo. The deformation data and the initial settlement data are fused together to generate settlement characterization data during the fusion process; The settlement characterization data is input into a preset digital twin space, and settlement evolution is performed in the digital twin space to determine the multi-dimensional settlement displacement field of the shallow grain silo. Based on the identification of the multidimensional settlement displacement field, the circumferential curvature change of the silo wall is determined. The circumferential stress distribution of the silo wall is inverted based on the curvature change. The damage evolution trend of the silo wall is determined by combining the service time and strength data of the silo wall. Based on the identification of the damage evolution trend of the silo wall, the corresponding crack characteristics are predicted, and the corresponding settlement early warning events are determined by combining the construction history of the shallow circular grain silo, and the settlement early warning area is marked. Based on the location of the settlement warning area, the settlement rate, and the preset knowledge graph, settlement decision content associated with the next construction procedure is generated.

2. The method for detecting settlement during the construction of shallow circular grain silos according to claim 1, characterized in that, The initial readings of the multi-source sensor array for collecting grain shallow circular silos are used as initial settlement data; the construction process of the grain shallow circular silos is monitored, the current readings of the multi-source sensor array are obtained, and the material load and dynamic load of the construction machinery in the current construction process are extracted to determine the deformation data of the grain shallow circular silos, including: A multi-source sensor array is deployed on the foundation and walls of a shallow circular grain silo. Initial readings of the multi-source sensor array are collected synchronously, and the initial readings are aligned and calibrated in a spatiotemporal coordinate system to serve as initial settlement data.

3. The method for detecting settlement during the construction of shallow circular grain silos according to claim 2, characterized in that, The method of collecting initial readings from the multi-source sensor array of the shallow grain silo and using them as initial settlement data; monitoring the construction process of the shallow grain silo, obtaining the current readings of the multi-source sensor array, extracting the material load and construction machinery dynamic load of the current construction process, thereby determining the deformation data of the shallow grain silo, also includes: The construction process of shallow circular grain silos is dynamically monitored, the spatial trajectory of construction machinery is collected in real time, and the dynamic load of construction machinery in the current construction process is determined by combining the corresponding vibration frequency, while the corresponding material load is extracted. The dynamic load of the construction machinery and the material load are used as boundary conditions and applied to the current readings of the multi-source sensor array to decouple the dynamic and static loads and perform spatiotemporal interpolation, thereby eliminating noise drift caused by ambient temperature and construction vibration, and thus determining the deformation data reflecting the shallow circular grain silo.

4. The method for detecting settlement during the construction of shallow circular grain silos according to claim 1, characterized in that, The process involves fusing the deformation data with the initial settlement data to generate settlement characterization data; inputting the settlement characterization data into a preset digital twin space and performing settlement evolution within the digital twin space to determine the multi-dimensional settlement displacement field of the shallow grain silo, including: The deformation data and the initial settlement data are input into a preset estimation framework, and multi-scale data fusion is performed in the estimation framework to generate corresponding settlement characterization data, which includes vertical displacement and tilt vector.

5. The method for detecting settlement during the construction of shallow circular grain silos according to claim 4, characterized in that, The process of fusing the deformation data with the initial settlement data to generate settlement characterization data during the fusion process, and inputting the settlement characterization data into a preset digital twin space to perform settlement evolution in the digital twin space, thereby determining the multi-dimensional settlement displacement field of the shallow grain silo, further includes: The settlement characterization data is input into a preset digital twin space to drive the synchronous deformation of the multi-dimensional coupled grid in the digital twin space. The settlement characterization data is then used as a boundary constraint condition, which is applied to the neural network corresponding to the digital twin space to trigger the settlement evolution of the settlement characterization data, thereby outputting a corresponding multi-dimensional settlement displacement field. The multi-dimensional settlement displacement field includes the overall three-dimensional spatial displacement of the shallow circular grain silo, the differential settlement of the foundation, and the uneven tilt state of the silo wall.

6. The method for detecting settlement during the construction of shallow circular grain silos according to claim 1, characterized in that, The determination of the circumferential curvature change of the silo wall based on the identification of the multi-dimensional settlement displacement field, the inversion of the circumferential stress distribution of the silo wall based on the curvature change, and the determination of the damage evolution trend of the silo wall in conjunction with the service time and strength data of the silo wall include: The multidimensional settlement displacement field is dynamically identified, and the radial displacement distribution of the silo wall is extracted during the identification process. The radial displacement fitting surface of the silo wall is then smoothed by combining a cubic spline interpolation mechanism, thereby calculating the circumferential curvature change of the silo wall.

7. The method for detecting settlement during the construction of shallow circular grain silos according to claim 6, characterized in that, The method of determining the circumferential curvature change of the silo wall based on the identification of the multi-dimensional settlement displacement field, retrieving the circumferential stress distribution of the silo wall based on the curvature change, and determining the damage evolution trend of the silo wall in conjunction with the service time and strength data of the silo wall, further includes: The curvature change is inverted, and the circumferential stress distribution of the silo wall is determined during the inversion process. The usage time and strength data of the silo wall are obtained, and multiple calculations are performed in combination with the circumferential stress distribution of the silo wall to determine the damage evolution trend of the silo wall during the calculation process. The damage evolution trend of the silo wall presents the evolution of the silo wall from an intact state to a local failure state.

8. The method for detecting settlement during the construction of shallow circular grain silos according to claim 1, characterized in that, The corresponding crack features are predicted based on the identification of the damage evolution trend of the silo wall, and the corresponding settlement early warning events are determined in combination with the construction history of the shallow circular grain silo, and the settlement early warning area is marked. Based on the location and settlement rate of the settlement warning area and a preset knowledge graph, settlement decision content associated with the next construction procedure is generated, including: The damage evolution trend is dynamically identified, and the local stress concentration and damage accumulation zone of the warehouse wall are determined during the identification process. The corresponding crack characteristics are predicted by combining fracture mechanics criteria. The crack characteristics show the crack initiation location, propagation direction and crack width. The crack features are mapped to the construction history of the shallow circular grain silo to assess the impact of uneven settlement on structural connectivity and integrity, thereby triggering corresponding settlement early warning events and marking the corresponding settlement early warning areas in the digital twin space.

9. The method for detecting settlement during the construction of shallow circular grain silos according to claim 8, characterized in that, The corresponding crack features are predicted based on the identification of the damage evolution trend of the silo wall, and the corresponding settlement early warning events are determined in combination with the construction history of the shallow circular grain silo, and the settlement early warning area is marked. Based on the location, settlement rate, and preset knowledge graph of the settlement early warning area, settlement decision content associated with the next construction procedure is generated, which also includes: Based on the detection of the settlement warning area, the three-dimensional coordinates, current settlement rate, and historical cumulative settlement of the settlement warning area are determined. Then, a preset knowledge graph of the construction safety of shallow grain silos is input, and settlement decision content strongly associated with the next construction procedure is generated through graph neural network reasoning and semantic matching. The settlement decision content includes material loading path change plan, construction machinery travel avoidance route, and foundation dynamic reinforcement strategy.

10. A settlement detection system for shallow circular grain silos during construction, characterized in that, The settlement detection system for shallow circular grain silos during construction is applied to the settlement detection method for shallow circular grain silos during construction as described in any one of claims 1-9. The settlement detection system for the shallow circular grain silo during construction includes: The monitoring module is used to collect the initial readings of the multi-source sensor array of the shallow grain silo and use them as initial settlement data; monitor the construction process of the shallow grain silo, obtain the current readings of the multi-source sensor array, extract the material load and construction machinery dynamic load of the current construction process, and thus determine the deformation data of the shallow grain silo. The settlement evolution module is used to fuse the deformation data with the initial settlement data and generate settlement characterization data during the fusion process; the settlement characterization data is input into a preset digital twin space and settlement evolution is performed in the digital twin space to determine the multi-dimensional settlement displacement field of the shallow grain silo. The damage evolution trend module is used to determine the circumferential curvature change of the silo wall based on the identification of the multi-dimensional settlement displacement field, invert the circumferential stress distribution of the silo wall based on the curvature change, and determine the damage evolution trend of the silo wall in combination with the silo wall's service time and strength data. The settlement decision module is used to predict the corresponding crack features based on the identification of the damage evolution trend of the silo wall, and to determine the corresponding settlement early warning events in combination with the construction history of the shallow circular grain silo, and to mark the settlement early warning area; and to generate settlement decision content associated with the next construction procedure based on the location of the settlement early warning area, the settlement rate and the preset knowledge graph.

Images