A real-time monitoring data processing system for stress of offshore steel platform structure

Abstract

The application relates to the technical field of aerial work equipment, and discloses a real-time monitoring data processing system for offshore steel platform structure stress, which comprises a data acquisition module, an equipment telemetry module and an edge computing node. The edge computing node is used for converting the main shaft rotating speed parameter into a center frequency variable of a band-pass filter, combining wave data alignment value to extract a high-frequency excitation response signal, and separating marine environmental noise and mechanical vibration; according to a two-dimensional coordinate interpolation result, a comprehensive additional static load stress parameter is generated, a static stress data of tide compensation is subtracted from the comprehensive additional static load stress parameter to generate a static load correction stress vector, local uneven additional static load bias is removed to restore stress distribution; a space adjacency matrix, a transfer response column vector and a transpose of the transfer response column vector are spliced and fused to obtain an extended dimension correlation matrix, a feature vector multidimensional space included angle is calculated to generate a structure instability risk index, and a structure instability state caused by bottom steel pipe pile bearing capacity degradation is determined.

Description

Technical Field

[0001] This invention relates to the field of high-altitude work equipment technology, specifically to a real-time stress monitoring and data processing system for offshore steel platform structures. Background Technology

[0002] Offshore steel platforms operating in the marine environment are subject to the combined effects of environmental and mechanical loads. They are also affected by high-frequency interference from ocean waves, gradual tidal changes, and the coupling interference of localized static bias and high-frequency excitation caused by the movement and operation of large construction machinery. Current stress measurement data processing methods do not separate the vibration response characteristics of mechanical operation from the background noise of the marine environment, and the high-frequency white noise generated by wave impacts interferes with the extraction of mechanical excitation.

[0003] Meanwhile, the positional changes of crawler cranes and rotary drilling rigs on the surface of offshore steel platforms can cause localized uneven additional static load offsets. Existing technologies do not remove these localized uneven additional static load offsets from the underlying data, making it impossible to restore the benchmark force distribution that truly reflects the global state of the offshore steel platform.

[0004] The existing stress monitoring data analysis model does not establish an evaluation matrix that characterizes the fusion state of static deformation characteristics and dynamic wave propagation characteristics, and does not combine the characteristic vector deflection angle for calculation and evaluation. As a result, the monitoring platform cannot accurately determine the structural instability state caused by the bearing capacity degradation of the bottom steel pipe pile.

[0005] Therefore, this invention proposes a real-time stress monitoring data processing system for offshore steel platform structures to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a real-time stress monitoring data processing system for offshore steel platform structures. This system solves the problem that the stress measurement data processing process fails to separate the vibration response characteristics of mechanical operation from the background noise of the marine environment, and fails to remove the local uneven additional static load offset caused by the positional changes of crawler cranes and rotary drilling rigs. This results in the monitoring platform being unable to determine the structural instability state caused by the degradation of the bearing capacity of the bottom steel pipe piles.

[0007] To achieve the above objectives, the present invention provides the following technical solution: This invention provides a real-time stress monitoring and data processing system for offshore steel platform structures, comprising: The data acquisition module is deployed at various nodes of the offshore steel platform. The data acquisition module is equipped with strain gauges and environmental sensors. The strain gauges are fixed to the surface of the main crossbeam, Bailey bridge sheet, and steel pipe pile of the offshore steel platform, respectively. The environmental sensors are set at the support nodes of the offshore steel platform. The strain gauges measure the stress time series signal on the surface of each component of the offshore steel platform, and the environmental sensors measure the sea wave data and tide data. The data acquisition module sends the stress time series signal, wave data, and tide data to the edge computing node. The equipment telemetry module is installed inside the 50t crawler crane and rotary drilling rig operating on the offshore steel platform. The equipment telemetry module is equipped with a real-time positioning antenna and a main control data interface. The real-time positioning antenna reads the two-dimensional coordinates of the 50t crawler crane and rotary drilling rig, and the main control data interface reads the spindle speed parameters of the rotary drilling rig. The equipment telemetry module sends the two-dimensional coordinates and spindle speed parameters to the edge computing node according to a preset time period. Edge computing nodes are located within the terminal box of the offshore steel platform. Each edge computing node contains a synchronization scheduling unit and an algorithm execution engine. The synchronization scheduling unit timestamps the received stress time series signals, wave data, tide data, two-dimensional coordinates, and spindle speed parameters. The algorithm execution engine performs narrowband filtering, static load offset correction, active excitation transmission response extraction, and spatial deflection early warning calculation based on the timestamp-aligned data. The edge computing node determines the settlement status of the offshore steel platform based on the output of the spatial deflection early warning calculation and outputs an early warning command.

[0008] The synchronization scheduling unit performs resampling and interpolation calculations on the received multi-source parameters to generate unified timestamp data frames. The algorithm execution engine sets the cutoff frequency parameter of the digital low-pass filter to extract basic static stress data from the stress time series signal, converts the tide level data into a global seawater buoyancy change reference drift, and performs a scalar multiplication vector operation on the global seawater buoyancy change reference drift and a preset node elevation compensation coefficient vector to generate a differentiated drift compensation vector.

[0009] The algorithm execution engine subtracts the differential drift compensation vector from the basic static stress data to complete low-frequency environmental stress compensation and obtain tidal level compensated static stress data. The engine calculates the true high-frequency cutting fundamental frequency based on the spindle speed parameters and the pre-stored rotary drill bit tooth quantity parameters. This true high-frequency cutting fundamental frequency is assigned to the center frequency variable of the bandpass filter. Wave data is used to align numerical calculations of wave height to adjust the frequency offset parameter of the bandpass filter. Frequency band filtering is performed on the stress time series signal, retaining waveform components and integrating them to generate a high-frequency excitation response signal, thus avoiding environmental interference caused by high-frequency white noise from ocean waves on mechanical excitation extraction.

[0010] The algorithm execution engine retrieves the pre-stored spatial grid static load offset mapping table based on the two-dimensional coordinate interpolation results, extracts the first and second additional static load stress parameters, performs vector addition operations, and merges them to generate a comprehensive additional static load stress parameter. The algorithm execution engine then subtracts the comprehensive additional static load stress parameter from the tidal level compensated static stress data, performs vector subtraction operations to generate a static load correction stress vector, and eliminates the localized uneven additional static load offsets caused by the positional changes of the crawler crane and rotary drilling rig, restoring the benchmark stress distribution that reflects the true global state of the offshore steel platform.

[0011] The algorithm execution engine sets the sampling time window length based on the center frequency variable and calculates the theoretical time delay of mechanical wave propagation according to the spatial topology model of the offshore steel platform. The engine then sets a time-sliding search window to perform dynamic advance compensation on the high-frequency excitation response signal of the steel pipe piles. Finally, the engine performs cross-correlation integration on the high-frequency excitation response signals around the rotary drilling rig and the steel pipe piles, extracting the maximum absolute value of the integral as the cross-correlation coefficient. Based on the cross-correlation coefficients of all steel pipe piles, the engine generates a transmission response column vector, which reflects the dynamic transmission damping characteristics of the offshore steel platform's underlying foundation.

[0012] The algorithm execution engine calculates edge connection weights according to the spatial structural connection relationships of the offshore steel platform, and transforms the static load modified stress vector into a spatial adjacency matrix. The engine then uses a block matrix combination algorithm to concatenate and fuse the spatial adjacency matrix, the dimension-aligned transferred response column vector, and the transpose of the dimension-aligned transferred response column vector to obtain an extended-dimensional correlation matrix. Finally, the engine performs singular value decomposition on the extended-dimensional correlation matrix to extract the first principal component eigenvector corresponding to the largest singular value.

[0013] The algorithm execution engine matches the target baseline feature vector from a pre-stored baseline feature vector matrix library and calculates the multidimensional spatial angle between the first principal component feature vector and the target baseline feature vector using the inverse cosine function. The engine then normalizes and compares this multidimensional spatial angle with a structural instability judgment threshold to generate a structural instability risk index. If the structural instability risk index is greater than 1, an anti-instability anomaly warning command is triggered. The system establishes an extended-dimensional correlation matrix representing the fusion state of static deformation characteristics and dynamic wave propagation characteristics, and extracts the degree of deflection of the first principal component feature vector to determine the structural instability state of the underlying steel pipe piles caused by bearing capacity degradation.

[0014] This invention provides a real-time stress monitoring and data processing system for offshore steel platform structures. It offers the following advantages: 1. This invention uses an algorithm execution engine to calculate and convert the spindle speed parameters into the center frequency variable of a bandpass filter. It then combines wave data alignment values ​​to adjust the frequency offset parameters, performs frequency band filtering on the stress time series signal, retains waveform components, and integrates them to generate a high-frequency excitation response signal. This avoids environmental interference caused by high-frequency white noise from ocean waves on mechanical excitation extraction, and achieves the separation and extraction of mechanical vibration response characteristics from marine environmental background noise.

[0015] 2. This invention uses an algorithm execution engine to retrieve the spatial grid static load offset mapping table based on the two-dimensional coordinate interpolation results, extracts the first and second additional static load stress parameters, performs vector addition to generate a comprehensive additional static load stress parameter, subtracts the comprehensive additional static load stress parameter from the tide level compensated static stress data to generate a static load correction stress vector, eliminates the local uneven additional static load offset caused by the position changes of crawler cranes and rotary drilling rigs, and restores the benchmark force distribution that reflects the true global state of the offshore steel platform.

[0016] 3. This invention uses an algorithm execution engine to concatenate and fuse the spatial adjacency matrix, the transitive response column vector, and the transpose of the transitive response column vector to obtain an extended-dimensional correlation matrix. The first principal component feature vector is extracted from the extended-dimensional correlation matrix, and the multi-dimensional spatial angle is calculated in combination with the target benchmark feature vector to generate a structural instability risk index. An evaluation matrix is ​​established to characterize the fusion state of static deformation characteristics and dynamic wave propagation characteristics, and the structural instability state caused by the degradation of bearing capacity of the bottom steel pipe pile is determined. Attached Figure Description

[0017] Figure 1 This is a system block diagram of the present invention.

[0018] Figure 2 This is a flowchart of the method of the present invention.

[0019] Figure 3 This is a flowchart of the multi-source heterogeneous data timestamp alignment operation of the present invention.

[0020] Figure 4 This is the logic diagram for the basic static stress separation and high-frequency excitation extraction of the present invention.

[0021] Figure 5 This is a schematic diagram illustrating the principle of spatial grid static load bias mapping and correction in this invention.

[0022] Figure 6 This is a diagram illustrating the frequency domain separation effect of the stress time series signal according to the present invention.

[0023] Figure 7 This is a calculation diagram for the cross-correlation coefficient compensation of the high-frequency excitation response signal of the present invention.

[0024] Figure 8 This is a comparison chart of the structural instability risk index early warning of the present invention. Detailed Implementation

[0025] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] See attached document Figure 1 This invention provides a real-time stress monitoring data processing system for offshore steel platform structures, comprising: The data acquisition module is deployed at various nodes of the offshore steel platform. The data acquisition module is equipped with strain gauges and environmental sensors. The strain gauges are fixed to the surface of the main crossbeam, Bailey bridge sheet, and steel pipe pile of the offshore steel platform, respectively. The environmental sensors are set at the support nodes of the offshore steel platform. The strain gauges measure the stress time series signal on the surface of each component of the offshore steel platform, and the environmental sensors measure the sea wave data and tide data. The data acquisition module sends the stress time series signal, wave data, and tide data to the edge computing node. The equipment telemetry module is installed inside the 50t crawler crane and rotary drilling rig operating on the offshore steel platform. The equipment telemetry module is equipped with a real-time positioning antenna and a main control data interface. The real-time positioning antenna reads the two-dimensional coordinates of the 50t crawler crane and rotary drilling rig, and the main control data interface reads the spindle speed parameters of the rotary drilling rig. The equipment telemetry module sends the two-dimensional coordinates and spindle speed parameters to the edge computing node according to a preset time period. Edge computing nodes are located within the terminal box of the offshore steel platform. Each edge computing node contains a synchronization scheduling unit and an algorithm execution engine. The synchronization scheduling unit timestamps the received stress time series signals, wave data, tide data, two-dimensional coordinates, and spindle speed parameters. The algorithm execution engine performs narrowband filtering, static load offset correction, active excitation transmission response extraction, and spatial deflection early warning calculation based on the timestamp-aligned data. The edge computing node determines the settlement status of the offshore steel platform based on the output of the spatial deflection early warning calculation and outputs an early warning command.

[0027] After the synchronization scheduling unit within the edge computing node completes timestamp alignment, it triggers the algorithm execution engine to execute the data processing workflow. The algorithm execution engine sets the center frequency of the bandpass filter according to the spindle speed parameters. The algorithm execution engine separates the stress time series signal into basic static stress data and high-frequency excitation response signal. The algorithm execution engine retrieves the additional static load stress parameter from the pre-stored spatial grid static load offset mapping table based on two-dimensional coordinates. The algorithm execution engine performs vector subtraction between the basic static stress data and the additional static load stress parameter to obtain the static load correction stress vector. The algorithm execution engine extracts the height of the main beam within a preset time window. The algorithm execution engine integrates the cross-correlation coefficients of the high-frequency excitation response signals of all steel pipe piles to generate a transfer response column vector. The algorithm execution engine maps the static load modified stress vector into a spatial adjacency matrix according to the spatial structural connection topology of the offshore steel platform. The algorithm execution engine injects the transfer response column vector into the spatial adjacency matrix to form an extended dimension correlation matrix. The algorithm execution engine performs singular value decomposition on the extended dimension correlation matrix to extract the first principal component eigenvector corresponding to the maximum singular value. The algorithm execution engine calculates the spatial angle between the first principal component eigenvector and the pre-stored reference eigenvector.

[0028] The algorithm execution engine compares the spatial angle with the judgment threshold. When the spatial angle is greater than the judgment threshold, the algorithm execution engine determines that the offshore steel platform has experienced structural deflection and settlement. After determining that the offshore steel platform has experienced structural deflection and settlement, the algorithm execution engine outputs an abnormal warning command.

[0029] See attached document Figure 2 This invention provides a method for real-time monitoring and data processing of stress in offshore steel platform structures, comprising the following steps: S10, the synchronization scheduling unit within the edge computing node receives stress time-series signals, wave data, and tide data sent by the data acquisition module. The synchronization scheduling unit also receives two-dimensional coordinates and spindle speed parameters sent by the equipment telemetry module. The synchronization scheduling unit performs timestamp alignment operations on the stress time-series signals, wave data, tide data, two-dimensional coordinates, and spindle speed parameters. The algorithm execution engine within the edge computing node calculates and sets the center frequency of the bandpass filter according to the spindle speed parameters and the pre-stored number of rotary drill bit teeth. The algorithm execution engine uses the bandpass filter with the set center frequency to perform frequency domain separation operations on the stress time-series signals. The algorithm execution engine separates the foundation static stress data from the stress time-series signals and performs low-frequency drift compensation operations on the foundation static stress data in combination with the tide data. At the same time, it uses wave data to adaptively and dynamically adjust the cutoff frequency boundary of the bandpass filter to obtain a high-frequency excitation response signal.

[0030] S20, the algorithm execution engine retrieves the spatial grid static load offset mapping table stored in the edge computing node based on the two-dimensional coordinates, extracts the additional static load stress parameter corresponding to the two-dimensional coordinates, and performs vector subtraction operation between the basic static stress data and the additional static load stress parameter to generate the static load correction stress vector.

[0031] S30, the algorithm execution engine extracts the high-frequency excitation response signals around the rotary drilling rig and the high-frequency excitation response signals of each steel pipe pile within the time window. The algorithm execution engine calculates the cross-correlation coefficient between the high-frequency excitation response signals around the rotary drilling rig and the high-frequency excitation response signals of each steel pipe pile. The algorithm execution engine generates a transitive response column vector based on the cross-correlation coefficients of all steel pipe piles. The algorithm execution engine transforms the static load modified stress vector into a spatial adjacency matrix according to the structural connection relationship of the offshore steel platform. The algorithm execution engine combines the transitive response column vectors and incorporates them into the spatial adjacency matrix to obtain an extended dimension correlation matrix.

[0032] S40, the algorithm execution engine performs singular value decomposition on the extended dimension correlation matrix. The algorithm execution engine extracts the first principal component eigenvector corresponding to the largest singular value from the result of the singular value decomposition operation. The algorithm execution engine calculates the spatial angle between the first principal component eigenvector and the pre-stored benchmark eigenvector. The algorithm execution engine compares the spatial angle with the judgment threshold. When the algorithm execution engine determines that the spatial angle is greater than the judgment threshold, it triggers an abnormal warning command.

[0033] The following section details the technical solution for real-time stress monitoring and data processing of offshore steel platform structures, outlining the technical steps involved.

[0034] See attached document Figure 3 The synchronization scheduling unit within the edge computing node performs timestamp alignment operations on stress time series signals, wave data, tide data, two-dimensional coordinates, and spindle speed parameters. The timestamp alignment operation for multi-source heterogeneous data consists of the following steps: S101, the synchronization scheduling unit receives stress time-series signals, wave data, and tide data sent by the data acquisition module. The synchronization scheduling unit also receives two-dimensional coordinates and spindle speed parameters sent by the equipment telemetry module. The stress time-series signals, wave data, tide data, two-dimensional coordinates, and spindle speed parameters all carry local transmission timestamps and have different initial sampling frequencies. The synchronization scheduling unit is internally divided into a high-frequency data buffer queue, a low-frequency data buffer queue, and a medium-frequency data buffer queue. The synchronization scheduling unit stores the stress time-series signals in the high-frequency data buffer queue. The synchronization scheduling unit stores the wave data and tide data in the low-frequency data buffer queue. The synchronization scheduling unit stores the two-dimensional coordinates and spindle speed parameters in the medium-frequency data buffer queue. Regarding the local timestamp generation method of the hardware layer of the data acquisition module and the equipment telemetry module, those skilled in the art can use satellite second pulse synchronization or network precise time protocols. Hardware time synchronization mechanisms are well-known technologies in this field and will not be elaborated upon here.

[0035] S102, the synchronization scheduling unit sets the global reference timestamp sequence. The time step of the global reference timestamp sequence is equal to the sampling period of the stress time series signal. The synchronization scheduling unit reads the target time node from the global reference timestamp sequence. The synchronization scheduling unit performs resampling interpolation calculations on the low-frequency data buffer queue and the medium-frequency data buffer queue according to the target time node.

[0036] S103, for wave data and tide data in the low-frequency data buffer queue, the synchronization scheduling unit retrieves two original sampled values ​​before and after the target time node, and uses a linear interpolation algorithm to calculate and generate wave data alignment values ​​and tide data alignment values ​​corresponding to the target time node.

[0037] S104, for the two-dimensional coordinates and spindle speed parameters in the intermediate frequency data buffer queue, the synchronization scheduling unit uses a zero-order hold interpolation algorithm to obtain the interpolation result. The synchronization scheduling unit reads the most recently occurring real telemetry value on the time axis at the target time node, assigns the real telemetry value to the current target time node, and maintains the real telemetry value until it receives the next batch of telemetry updates on the time axis. The zero-order hold interpolation algorithm can reconstruct the mechanical operating state of the 50t crawler crane and rotary drilling rig within the sampling interval.

[0038] S105, the synchronization scheduling unit extracts the stress time series signal, wave data alignment values, tide data alignment values, two-dimensional coordinate interpolation results, and spindle speed parameter interpolation results corresponding to the same target time node. The synchronization scheduling unit packages and encapsulates the extracted multi-source parameters into a unified timestamp data frame. The synchronization scheduling unit continuously transmits the unified timestamp data frame to the algorithm execution engine to perform frequency domain separation operation.

[0039] See attached document Figure 4 The algorithm execution engine extracts the static stress data of the foundation through a digital low-pass filter, and adaptively adjusts the band-pass filter parameters according to the excitation frequency of the equipment spindle to extract the high-frequency excitation response signal. The specific steps are as follows: S106, the algorithm execution engine continuously reads the unified timestamp data frame output by the synchronization scheduling unit. The algorithm execution engine parses and extracts the stress time series signal corresponding to each measuring point location from the unified timestamp data frame. The stress time series signal is a mixture of static strain components generated by the self-weight of the offshore steel platform, low-frequency strain components generated by the gradual change of tide level, random high-frequency components generated by wave impact, and high-frequency components of mechanical vibration generated by rotary drilling operations.

[0040] S107, the algorithm execution engine first reads the spindle speed parameters and the pre-stored rotary drill bit tooth quantity parameters from the unified timestamp data frame, calculates the current cutting fundamental frequency, and pre-calculates the lower cutoff frequency of the bandpass filter in conjunction with preset frequency offset parameters. Subsequently, the algorithm execution engine sets the cutoff frequency parameter of the digital low-pass filter, setting it to 1 Hz or one-third of the calculated lower cutoff frequency, taking the smaller of the two. This adaptive threshold setting can intercept random high-frequency components generated by wave impact and high-frequency components of mechanical vibration generated by rotary drilling operations, while avoiding overlap and interference with the bandpass filter's frequency band. For the underlying mathematical model of the digital low-pass filter, those skilled in the art can use finite impulse response (FIR) filters or infinite impulse response (IR) filters for code writing. The transfer function design and code writing of the digital filter are well-known techniques in this field and will not be elaborated upon here.

[0041] In S108, the algorithm execution engine inputs the stress time series signal into a digital low-pass filter with a pre-defined cutoff frequency parameter for calculation. The digital low-pass filter removes waveform data with frequencies higher than the cutoff frequency from the stress time series signal. The algorithm execution engine retains waveform data with frequencies less than or equal to the cutoff frequency. The algorithm execution engine then integrates the retained waveform data into basic static stress data. This basic static stress data reflects only the stress baseline state of the offshore steel platform. The algorithm execution engine stores the basic static stress data in the memory blocks of the edge computing nodes for subsequent computation.

[0042] The algorithm execution engine extracts the aligned values ​​of tide level data under the same target time node. The algorithm execution engine retrieves the preset tide level to strain linear conversion coefficient in the internal memory and converts the aligned values ​​of tide level data into the global seawater buoyancy change reference drift through scalar multiplication.

[0043] Subsequently, the algorithm execution engine retrieves the preset node elevation compensation coefficient vector. This vector is obtained by applying node buoyancy boundary conditions under a series of discrete drafts using a pre-established finite element fluid-structure interaction model of an offshore steel platform. The theoretical static strain increment of the surface of each strain gauge installation node is solved, and a one-dimensional array with pre-calibrated slope is extracted by polynomial fitting. Its value shows a non-linear increasing distribution characteristic as the node elevation decreases.

[0044] The algorithm execution engine performs a scalar multiplication vector operation on the global seawater buoyancy change benchmark drift amount and the preset node elevation compensation coefficient vector to generate differentiated drift compensation vectors for nodes at different heights. The algorithm execution engine subtracts the differentiated drift compensation vectors from the basic static stress data to complete the tidal environmental stress compensation that conforms to the physical spatial distribution state, and then overwrites and saves the compensated data as tidal level compensated static stress data.

[0045] The algorithm execution engine extracts the wave data alignment values ​​at the same target time node, calculates the wave height characteristics, and adaptively reduces the upper cutoff frequency of the bandpass filter when the sea surface waves surge, so as to avoid environmental interference caused by the high-frequency white noise of the waves hitting the mechanical excitation extraction.

[0046] S109, when configuring the bandpass filter to extract the high-frequency excitation response signal, the algorithm execution engine retrieves the spindle speed parameters parsed in the previous steps. The spindle speed parameters characterize the rotation state of the rotary drilling rig spindle. The algorithm execution engine synchronously retrieves the pre-stored rotary drilling bit tooth quantity parameters in the memory block. The algorithm execution engine divides the spindle speed parameters in revolutions per minute (RPM) by a constant 60 to convert them to Hertz (Hz) units, and then multiplies them by the rotary drilling bit tooth quantity parameters to calculate the true high-frequency cutting fundamental frequency generated when the rotary drilling rig teeth cut the seabed rock strata. The algorithm execution engine assigns the value of the true high-frequency cutting fundamental frequency to the center frequency variable of the bandpass filter.

[0047] S110, the algorithm execution engine calculates the upper and lower cutoff frequencies of the bandpass filter based on the center frequency variable. The algorithm execution engine adds a frequency offset parameter to the center frequency variable to generate the upper cutoff frequency. The algorithm execution engine subtracts the frequency offset parameter from the center frequency variable to generate the lower cutoff frequency. The algorithm execution engine retrieves wave data alignment values ​​at the same target time node, calculates the current wave height, and compares it with the preset wave height threshold.

[0048] When the wave height is less than or equal to the wave height threshold, the algorithm execution engine presets the value of the frequency offset parameter to a baseline value of 2 Hz. When the wave height is greater than the wave height threshold, the algorithm execution engine calculates a dynamic attenuation coefficient based on the proportion by which the wave height exceeds the threshold. The dynamic attenuation coefficient is used to proportionally reduce the value of the frequency offset parameter, and the reduced value of the frequency offset parameter is limited to not being lower than the preset minimum bandwidth operating limit threshold (e.g., 0.5 Hz). This adaptively narrows the bandwidth of the bandpass filter when the high-frequency harmonics of the waves surge, and ensures that the bandpass filter always has the tolerance to allow small fluctuations in the spindle speed, avoiding the filter bandwidth from returning to zero and causing non-convergence.

[0049] S111, after the algorithm execution engine completes the setting of the upper and lower cutoff frequencies, it turns on the bandpass filter. The algorithm execution engine inputs the stress time series signal in the unified timestamp data frame into the bandpass filter for filtering operations.

[0050] S112, the bandpass filter performs frequency band filtering on the stress time series signal. The bandpass filter blocks waveform components with frequency values ​​lower than the lower cutoff frequency, blocks waveform components with frequency values ​​higher than the upper cutoff frequency, and retains waveform components with frequency values ​​between the lower and upper cutoff frequencies.

[0051] S113, the algorithm execution engine integrates the waveform components retained by the bandpass filter to generate a high-frequency excitation response signal. The center frequency of the bandpass filter moves dynamically with the rotation state of the rotary drilling spindle and avoids random environmental interference caused by high-frequency harmonics of ocean waves. The algorithm execution engine stores the high-frequency excitation response signal in the memory block of the edge computing node for subsequent operation scheduling.

[0052] See attached document Figure 5 The algorithm execution engine uses a pre-built mapping table and real-time location coordinates to match topological weights to perform Hadamard product correction on the foundation static stress data. The specific steps are as follows: S201, the algorithm execution engine reads the structural dimension parameters of the offshore steel platform, establishes a two-dimensional coordinate system covering the surface of the offshore steel platform, and divides the surface of the offshore steel platform into an array of two-dimensional coordinate grids according to the set distance step size.

[0053] S202, the algorithm execution engine reads the weight parameters of the 50t crawler crane and the rotary drilling rig respectively. In the finite element simulation model, the algorithm execution engine introduces the constant of gravitational acceleration to convert the weight parameters of the two into virtual static loads with force dimensions. The algorithm execution engine applies the virtual static load of a single device to the center node of each two-dimensional coordinate grid. The algorithm execution engine extracts the simulated stress values ​​of each strain gauge installation node of the offshore steel platform when each two-dimensional coordinate grid is subjected to virtual static loads. For the establishment of the finite element model of the offshore steel platform and the simulation calculation of the nodal stress, those skilled in the art can use finite element analysis software to complete it. Structural finite element mechanical simulation is a well-known technology in this field and will not be described in detail here.

[0054] S203, the algorithm execution engine extracts the baseline simulated stress values ​​at each strain gauge installation node location without virtual static load. The engine then calculates the difference vector between the simulated stress value after adding the virtual static load and the baseline simulated stress value. The engine arranges and combines the difference vectors corresponding to a single two-dimensional coordinate grid according to the node order to generate an additional static load stress parameter. This additional static load stress parameter characterizes the static interference stress generated by the crawler crane at that grid location on each node. The dimension of the additional static load stress parameter is equal to the number of strain gauges deployed in the data acquisition module.

[0055] S204, the algorithm execution engine extracts the boundary coordinate parameters of the two-dimensional coordinate grid. The algorithm execution engine binds the boundary coordinate parameters with the additional static load stress parameters to generate mapping entries. The algorithm execution engine summarizes and compiles the mapping entries corresponding to all two-dimensional coordinate grids to generate a spatial grid static load offset mapping table. The algorithm execution engine writes the spatial grid static load offset mapping table into the internal memory of the edge computing node to provide real-time retrieval and calling services.

[0056] S205, the algorithm execution engine parses the unified timestamp data frame to extract two-dimensional coordinates. These two-dimensional coordinates reflect the real-time positions of the 50t crawler crane and rotary drilling rig on the offshore steel platform surface. The algorithm execution engine inputs these two-dimensional coordinates as search keywords into its internal memory.

[0057] In step S206, when the algorithm execution engine determines that the two-dimensional coordinates fall within the boundary coordinate parameter range of the target two-dimensional coordinate grid, it extracts the first additional static load stress parameter corresponding to the two-dimensional coordinates of the 50t crawler crane and the second additional static load stress parameter corresponding to the two-dimensional coordinates of the rotary drilling rig. The algorithm execution engine performs vector addition on the first and second additional static load stress parameters, merging them to generate a comprehensive additional static load stress parameter. The algorithm execution engine then loads the comprehensive additional static load stress parameter into the calculation block.

[0058] S207, the algorithm execution engine retrieves the tidal level compensated static stress data extracted through low-pass separation and tidal level environmental stress compensation operations. The algorithm execution engine subtracts the corresponding value from the comprehensive additional static load stress parameter from the numerical value in the tidal level compensated static stress data, performing a vector subtraction operation. The algorithm execution engine derives the static load correction stress vector based on the result of the vector subtraction operation. The formula is as follows: ; In the formula, Represents the static load corrected stress vector; Represents the static stress data of the foundation; This represents the combined additional static stress parameter, which includes the offset of the self-weight of the 50t crawler crane and the rotary drilling rig.

[0059] The movement of the S208, 50t crawler crane on the surface of the offshore steel platform alters the static boundary loads of local components and induces static stress baseline drift. The algorithm execution engine utilizes the additional static load stress parameter to perform spatial feature stripping on the foundation static stress data affected by local dynamic load disturbances. Through vector subtraction, the engine merges and eliminates localized, uneven additional static load biases caused by the positional changes of the crawler crane and rotary drilling rig, thereby restoring the baseline stress distribution that reflects the true global state of the offshore steel platform. This prevents local static load distortion from dominating subsequent global matrix dimensionality reduction feature extraction. The algorithm execution engine stores the static load correction stress vector in a data buffer, awaiting invocation in subsequent correlation matrix reconstruction operations.

[0060] The algorithm execution engine integrates dynamic cross-correlation coefficients and static structural topology to construct an extended-dimensional correlation matrix representing the static-dynamic coupling state. The specific steps are as follows: S301, the algorithm execution engine calculates and sets a dynamic time length containing at least 10 complete excitation mechanical wave cycles, based on the center frequency of the bandpass filter set according to the spindle speed parameters in the previous steps, as the current sampling time window length. The adaptively adjusted sampling time window length determines the sufficiency of energy samples for cross-correlation integration, thus avoiding integration distortion caused by excessive truncation of waveform characteristics at low rotary drilling rig speeds. The algorithm execution engine calculates the planar Euclidean distance between the two-dimensional coordinates of the rotary drilling rig and the three-dimensional spatial projection coordinates of each strain gauge installation node on the offshore steel platform. The algorithm execution engine determines strain gauge installation nodes whose planar Euclidean distance is less than a preset search radius threshold as target nodes. When no strain gauge installation node exists within the search radius threshold, the algorithm execution engine sorts them in ascending order according to the planar Euclidean distance, forcibly locking the nearest at least one strain gauge installation node as the target node, thereby accurately locking the strain gauges around the rotary drilling rig and avoiding an empty reference signal set. The algorithm execution engine retrieves the high-frequency excitation response signals generated by strain gauge measurements around the rotary drilling rig within the sampling time window length from the memory block. The algorithm execution engine synchronously retrieves the high-frequency excitation response signals generated by strain gauge measurements on the surface of each underwater steel pipe pile within the sampling time window.

[0061] S302, the vibrational mechanical waves generated by the rotary drilling operation are transmitted downwards to the steel pipe piles through the main crossbeam and Bailey panels. The propagation of these vibrational mechanical waves in the steel medium involves a time delay. The algorithm execution engine reads the spatial topology model of the offshore steel platform, assigns the actual geometric lengths of each physical component of the offshore steel platform as weights to the corresponding connected edges in the spatial topology model, and calls the graph theory shortest path algorithm based on these weights to calculate the shortest physical transmission path length along the actual steel structure connecting components between the strain gauges around the rotary drilling rig and the strain gauges on the surfaces of each underwater steel pipe pile.

[0062] The algorithm execution engine calculates the vibration mechanical wave propagating from the periphery of the rotary drilling rig to the first stage based on the shortest physical transmission path length and the sound velocity constant of the mechanical wave in the steel. The theoretical time delay of a steel pipe pile. For the lookup of the material's sound velocity constant and the calculation of the mechanical wave propagation delay, those skilled in the art can consult the Handbook of Acoustics and Elasticity to establish a calculation program. The calculation of the time delay of internal material fluctuations is a well-known technique in this field and will not be elaborated here.

[0063] S303. Due to the dispersion and multipath effects of vibrational mechanical waves propagating in the complex steel frame structure, the actual arrival time will fluctuate unpredictably near the theoretical time delay. The algorithm execution engine sets a time sliding search window centered on the theoretical time delay, and substitutes each dynamic offset parameter within the time sliding search window into the cross-correlation integral function to perform dynamic advance compensation on the time axis of the high-frequency excitation response signal of the steel pipe pile. The algorithm execution engine performs cross-correlation integral calculations on the high-frequency excitation response signals around the rotary drilling rig and the high-frequency excitation response signals of each steel pipe pile after dynamic advance compensation, and extracts the maximum absolute value of the integral within the time sliding search window as the cross-correlation coefficient of the corresponding node to absorb the waveform phase dispersion caused by the physical topology.

[0064] ; In the formula, Representing the The number of cross-relationships among the nodes of a steel pipe pile; Represents the length of the sampling time window; Represents the time variable of integration; This represents the high-frequency excitation response signal around the rotary drilling rig; Representing the High-frequency excitation response signal of a steel pipe pile; Represents theoretical time delay; Represents the dynamic offset parameter; This represents the boundary range of the time-based sliding search window.

[0065] S304, the cross-correlation coefficient characterizes the degree of energy attenuation of vibratory mechanical waves after transmission in the structural medium. When the bearing state of the weathered mudstone or silt layer in the bottom layer changes, the damping parameters of the entire offshore steel platform change accordingly, causing the cross-correlation coefficient values ​​to shift. The algorithm execution engine extracts the corresponding cross-correlation coefficients according to the structural numbering order of the steel pipe piles. The algorithm execution engine arranges all cross-correlation coefficients sequentially along the vertical dimension. The algorithm execution engine generates a transfer response column vector. The transfer response column vector reflects the dynamic transfer damping characteristics of the entire offshore steel platform's bottom foundation. The algorithm execution engine writes the transfer response column vector into a register, awaiting the call to the dimension expansion and reconstruction operation.

[0066] S305, the algorithm execution engine reads the structural drawing data of the offshore steel platform. The engine parses the spatial structural connections between the main crossbeams, Bailey bridge sections, and steel pipe piles. Based on the installation node positions on the surfaces of the main crossbeams, Bailey bridge sections, and steel pipe piles, the engine establishes a spatial topology graph model. The engine assigns a unique index number to each installation node in the spatial topology graph model. The establishment of the graph data model and the assignment of node numbers can be accomplished by those skilled in the art using graph theory-based algorithms; the establishment of the graph data structure is a well-known technique in this field and will not be elaborated upon here.

[0067] S306, the algorithm execution engine retrieves the static load correction stress vector stored in the data buffer. The static load correction stress vector is arranged in a one-dimensional sequence structure. The algorithm execution engine initializes and establishes a blank two-dimensional square matrix in the memory block. The number of rows and columns of the two-dimensional square matrix are equal to the total number of installed nodes. The algorithm execution engine sets the initial value of all elements in the two-dimensional square matrix to zero.

[0068] S307, the algorithm execution engine fills the two-dimensional matrix with numerical parameters based on the spatial structural connection relationships. When the algorithm execution engine determines that there is a direct structural connection between any two installation nodes, it extracts the two corresponding stress values ​​from the static load correction stress vector. The algorithm execution engine calculates the absolute values ​​of these two stress values, takes the arithmetic mean of the two absolute values, and adds this arithmetic mean to a preset structural baseline connectivity minimum constant to generate edge connection weights. The introduction of the structural baseline connectivity minimum constant aims to prevent edge connection weights from becoming zero due to local nodes being on a stress-neutral surface. Combined with the arithmetic mean calculation of absolute values, this avoids the illusion of broken connectivity in the algebraic matrix of the topological graph.

[0069] In a preferred embodiment, the range of the structural reference connectivity minimum constant is set to be much smaller than the value of the normal stress signal noise floor (e.g., 10). -6 (Micro-strain level) to ensure that it maintains matrix connectivity without causing algebraic interference to the principal energy features. The algorithm execution engine fills the edge connection weights into the matrix element positions where the corresponding row and column numbers intersect in the two-dimensional square matrix.

[0070] In S308, when the algorithm execution engine determines that there is no direct structural connection between any two installation nodes, it keeps the corresponding element value in the two-dimensional matrix zero. The engine then sequentially fills the main diagonal of the two-dimensional matrix with the absolute values ​​of the stress values ​​corresponding to each installation node in the static load correction stress vector. After extracting and filling all numerical parameters, the engine generates a spatial adjacency matrix. This matrix transforms the isolated node stress data into a two-dimensional algebraic feature representing the overall load distribution. The engine stores the spatial adjacency matrix in an internal cache, awaiting subsequent matrix assembly and dimensionality reduction calculations. Because the absolute value arithmetic mean operation is used, the generated spatial adjacency matrix is ​​a real symmetric matrix, ensuring the undirected nature of the graph data topology and the stability of subsequent dimension expansion and correlation matrix decomposition calculations.

[0071] In step S309, the algorithm execution engine retrieves the spatial adjacency matrix from its internal cache and simultaneously retrieves the transitive response column vector from its registers. The spatial adjacency matrix is ​​a two-dimensional square matrix of order N rows and N columns, where N represents the total number of installed nodes. The transitive response column vector is a vector structure of order K rows and 1 column, where K represents the number of underlying steel pipe piles. To meet the dimension alignment requirements of matrix block splicing, the algorithm execution engine performs a dimension alignment mapping operation. Based on the index number of the installed nodes, the algorithm execution engine fills the corresponding steel pipe pile node positions in a newly created alignment column vector of order N rows and 1 columns with the K cross-correlation coefficients from the transitive response column vector. The algorithm execution engine then fills all elements in the alignment column vector other than the steel pipe pile positions with 0, thus obtaining the dimension-aligned transitive response column vector. This zero-padding dimension alignment operation is equivalent in mathematical graph theory space to introducing a virtual global root node that is dynamically coupled only to the underlying steel pipe piles.

[0072] To eliminate the algebraic masking effect of physical dimension differences on the extraction of principal energy features, the algorithm execution engine extracts the global maximum absolute value from the spatial adjacency matrix and the dimension-aligned transferred response column vector before performing the following matrix concatenation. It then divides all stress weight elements and cross-correlation coefficients by their corresponding global maximum absolute values ​​for proportional scaling and normalization, mapping their values ​​to the same dimensionless range of [0,1] to ensure that the sparse topological disconnection features represented by the zero elements in the matrix are not destroyed.

[0073] S310, the algorithm execution engine allocates a matrix storage space in the memory block to store the extended dimension correlation matrix. The order of the extended dimension correlation matrix is ​​set to N+1 rows and N+1 columns. The algorithm execution engine uses a block matrix combination algorithm to combine and concatenate the spatial adjacency matrix, the dimension-aligned transitive response column vector, and the transpose matrix of the dimension-aligned transitive response column vector.

[0074] S311, the algorithm execution engine fills the spatial adjacency matrix into the upper-left sub-matrix region of the extended-dimensional correlation matrix, from row 1 to row N and from column 1 to column N. The algorithm execution engine fills the dimension-aligned transitive response column vector into the upper-right sub-vector region of the extended-dimensional correlation matrix, from row 1 to row N and at column N+1. The algorithm execution engine performs a row and column transpose operation on the dimension-aligned transitive response column vector to obtain its transpose matrix. For the underlying operator implementation of the matrix row and column transpose, those skilled in the art can use the transpose matrix function in a general matrix operation library. Transpose matrix calculation is a well-known technique in this field and will not be elaborated further here. The algorithm execution engine fills the transpose matrix of the transitive response column vector into the lower-left sub-vector region of the extended-dimensional correlation matrix, located at row N+1 and from column 1 to column N. The algorithm execution engine sets the value of the bottom right corner cross element in the (N+1)th row and (N+1)th column of the expanded dimension correlation matrix to 0, thereby completing the matrix splicing operation to obtain the expanded dimension correlation matrix.

[0075] ; In the formula, Represents an extended-dimensional correlation matrix; Represents the spatial adjacency matrix; Represents the column vector of the transmitted response after dimension alignment; This represents the transpose of the dimension-aligned column vector of the transmitted response; This represents the bottom right corner intersection element.

[0076] S312, the extended-dimensional correlation matrix merges the spatial adjacency matrix representing static deformation characteristics with the dimensionally aligned transferred response column vector representing dynamic wave propagation characteristics into a unified algebraic space. Through this cross-domain dimensional incorporation operation, the N+1th dimension of the extended-dimensional correlation matrix acts as a virtual node characterizing the seabed foundation bearing capacity. This matrix fully characterizes the static and dynamic coupling state between the topology of the offshore steel platform and the seabed foundation environment, eliminating the technical deficiency that single force data cannot reflect the variation in the bearing capacity of the underlying mudstone foundation. The algorithm execution engine then feeds the generated extended-dimensional correlation matrix into the subsequent singular value decomposition calculation process.

[0077] The algorithm execution engine extracts principal components through singular value decomposition and calculates their spatial deviation relative to the baseline state. The specific steps are as follows: S401, the algorithm execution engine retrieves the extended dimension correlation matrix from the memory block. The extended dimension correlation matrix has an order of N+1 rows and N+1 columns. The algorithm execution engine calls the matrix decomposition operator to perform singular value decomposition on the extended dimension correlation matrix. For the numerical calculation process of singular value decomposition, those skilled in the art can use the two-step QR algorithm or the divide-and-conquer algorithm. Matrix diagonalization and singular value solving are well-known technologies in this field and will not be described in detail here.

[0078] S402, the algorithm execution engine decomposes the extended dimension correlation matrix into the product of a left orthogonal matrix and the transpose of a diagonal singular value matrix and a right orthogonal matrix through the singular value decomposition operation.

[0079] ; In the formula, Represents an extended-dimensional correlation matrix; Represents a left orthogonal matrix; Represents a diagonal singular value matrix; This represents the transpose of a right orthogonal matrix.

[0080] S403, the algorithm execution engine reads the diagonal singular value matrix. The diagonal elements of the diagonal singular value matrix are singular value values ​​arranged in descending order. The algorithm execution engine extracts the maximum singular value located in the first row and first column of the diagonal singular value matrix. The maximum singular value concentrates the principal energy features of the spatial structure in the extended dimension correlation matrix.

[0081] S404, the algorithm execution engine extracts the corresponding first column feature vector from the left orthogonal matrix based on the row index position of the maximum singular value. The algorithm execution engine determines the first column feature vector as the first principal component feature vector. The dimension of the first principal component feature vector is equal to N+1. The first N elements in the first principal component feature vector correspond to the deformation weights of each installation node on the surface of the offshore steel platform, and the N+1th element in the first principal component feature vector corresponds to the power transmission weight of the bottom steel pipe pile. The algorithm execution engine writes the first principal component feature vector into the cache to complete the dimensionality reduction and decoupling extraction of the high-dimensional static-dynamic coupling features.

[0082] In step S405, the algorithm execution engine retrieves a pre-stored baseline feature vector matrix library from internal memory. This library is a set of one-dimensional feature column vectors extracted by performing singular value decomposition on the historical extended-dimensional correlation matrix corresponding to different excitation frequency ranges during dry drilling or shallow standard rock layer operations, under a manually verified safe calibration state where the offshore steel platform structure is free from damage and uneven settlement. The process involves drilling at multiple speeds using a rotary drilling rig across different excitation frequency ranges. Each feature vector in the library has a dimension of N+1. The baseline feature vectors represent the state-space benchmark of the offshore steel platform under conditions of no uneven settlement and no structural damage, encompassing the structural integrity characteristics in terms of static stress and dynamic wave propagation.

[0083] It should be noted that the historical stress data used to generate the baseline feature vector has also undergone tidal and wave environmental stress compensation operations, such as step S108. Therefore, the baseline feature vector is a pure structural intrinsic feature that has been freed from external environmental changes.

[0084] In a preferred embodiment, the edge computing node is equipped with a model recalibration module. After the system confirms that there is no structural damage to the offshore steel platform during regular maintenance, it uses the first principal component feature vector extracted under the current no-load state to cover and update the benchmark feature vector matrix library, so as to eliminate the slow benchmark drift caused by the natural aging of the offshore steel platform.

[0085] S406, the algorithm execution engine retrieves the first principal component eigenvector, calculated in real time, from the cache. Based on the spindle speed parameters in the current unified timestamp data frame, the engine matches and retrieves the target reference eigenvector within the corresponding excitation frequency range from the reference eigenvector matrix library to eliminate reference feature drift caused by the inherent resonant frequency characteristics of the structural dynamic transmission. The engine performs matrix multiplication between the transpose of the first principal component eigenvector and the matched target reference eigenvector to obtain the initial vector dot product. The engine extracts the absolute value of the initial vector dot product as the actual calculated vector dot product to eliminate sign ambiguity generated during eigenvector extraction via singular value decomposition.

[0086] The algorithm execution engine calculates the L2 norm of the first principal component eigenvector and the L2 norm of the reference eigenvector, respectively. For the underlying algebraic algorithms for calculating the eigenvector magnitudes and the vector inner product, those skilled in the art can use standard linear algebra subroutine libraries for implementation. Vector algebra operations are well-known techniques in this field and will not be elaborated upon here.

[0087] S407, the algorithm execution engine calculates the multidimensional spatial angle between the first principal component eigenvector and the reference eigenvector using the inverse cosine function, based on the vector inner product, the L2 norm modulus of the first principal component eigenvector, and the L2 norm modulus of the reference eigenvector. The mathematical formula for calculating the spatial angle is as follows: ; In the formula, This represents the multidimensional spatial angle between the first principal component eigenvector and the baseline eigenvector; Represents the inverse cosine function; This represents the eigenvector of the first principal component. Represents the pre-stored baseline feature vector; This represents the transpose of the eigenvector of the first principal component; This represents finding the absolute value of the dot product of vectors; The second norm of the eigenvectors of the first principal component represents the magnitude of the eigenvectors. The second norm modulus of the baseline eigenvector represents the length of the eigenvector.

[0088] S408, the spatial angle quantitatively characterizes the degree of geometric deflection of the current first principal component eigenvector from the reference eigenvector in multidimensional geometric space. When the bottom steel pipe piles of an offshore steel platform experience hidden settlement due to the degradation of mudstone bearing capacity or fatigue damage to the main structure, the algebraic characteristics of the expanded-dimensional correlation matrix change accordingly, causing the decoupled first principal component eigenvector to deflect in multidimensional space, and the value of the spatial angle increases accordingly. The algorithm execution engine writes the calculated spatial angle into the memory block of the edge computing node for subsequent threshold comparison and judgment operations.

[0089] The algorithm execution engine within the edge computing node compares the spatial angle with a pre-stored judgment threshold. When the spatial angle exceeds the judgment threshold, it determines that the settlement of the underlying steel pipe piles has caused structural instability and triggers an anomaly warning command. The threshold comparison judgment and instability prevention anomaly warning output operation are divided into the following steps: S409, the algorithm execution engine extracts the spatial angle values ​​generated in real time from the memory block. Simultaneously, the algorithm execution engine reads the pre-stored structural instability judgment threshold from the internal memory. The structural instability judgment threshold is a physical angle boundary parameter pre-calibrated and written into the finite element simulation data of the offshore steel platform's design overturning safety factor and structural ultimate bearing capacity.

[0090] S410, the algorithm execution engine constructs a warning indicator judgment model. The engine performs real-time normalized comparison and calculation between the spatial angle value and the structural instability judgment threshold to obtain the structural instability risk index of the offshore steel platform. The mathematical formula for calculating the structural instability risk index is as follows: ; In the formula, Indicator of structural instability risk; This represents the spatial angle between the first principal component eigenvector and the baseline eigenvector; This represents the pre-stored threshold for determining structural instability.

[0091] In step S411, the algorithm execution engine performs a conditional branch decision on the obtained structural instability risk index. If the algorithm determines that the structural instability risk index is greater than 1, it indicates that the deflection of the first principal component eigenvector in multidimensional space exceeds the safety tolerance boundary. Based on this, the algorithm determines that the underlying steel pipe piles have caused uneven settlement due to bearing capacity degradation, and classifies the offshore steel platform as being in an abnormal state of structural instability. The algorithm triggers an anti-instability anomaly warning command in its internal instruction register. The internal data payload of the anti-instability anomaly warning command includes the timestamp of the anomaly, the current structural instability risk index, and the spatial angle deflection value.

[0092] In S412, the algorithm execution engine packages the anti-instability anomaly warning command into a structural status warning data frame. The algorithm execution engine then transmits this data frame to the remote control room server via a wireless network through the communication module integrated into the edge computing node. Simultaneously, the communication module sends a high-level control signal to the audible and visual alarm devices deployed on the offshore steel platform to activate the on-site buzzers and warning lights. For the network protocol transmission of the warning data frame and the level drive control of the audible and visual alarm peripherals, those skilled in the art can use general network socket programming and industrial control bus pin control. The data network transmission and peripheral hardware driver are well-known technologies in the field and will not be elaborated upon here.

[0093] Specific application examples: To further verify the reliability and accuracy of this system and method in actual marine engineering, a specific application example is provided below.

[0094] This embodiment is set at the construction site of a steel platform for an offshore wind power booster station. A data acquisition module (containing 10 strain gauge installation nodes, where N=10, and the number of bottom steel pipe piles K=4) is deployed on site. A 50t crawler crane and a rotary drilling rig are currently operating on the steel platform.

[0095] The process of timestamp alignment and frequency domain separation: At a certain time period, the synchronization scheduling unit within the edge computing node receives multi-source data and completes unified timestamp data frame encapsulation. The algorithm execution engine parses and obtains that the spindle speed parameter of the rotary drilling rig at this time is 24 revolutions per minute (RPM), and the pre-stored rotary drilling bit tooth quantity parameter is 5.

[0096] The algorithm execution engine calculates the cutting fundamental frequency as follows: .

[0097] The algorithm execution engine sets the center frequency of the bandpass filter to 2.0Hz. Combined with the fact that the current wave height threshold is normal and the frequency offset parameter is set to 2Hz, the lower cutoff frequency of the bandpass filter is set to 0Hz (actually taking the minimum tolerance value) and the upper cutoff frequency is set to 4.0Hz.

[0098] Meanwhile, the cutoff frequency of the digital low-pass filter was set to 0.66Hz (the smaller of one-third of 2.0Hz and 1Hz). Using this filtering strategy, the foundation static stress data, unaffected by high-frequency interference from ocean waves and machinery, and the high-frequency excitation response signal containing underlying damping characteristics were successfully separated from the mixed stress time-series signal.

[0099] The process of static correction and dynamic correlation reconstruction: The algorithm execution engine parses the unified timestamp data frame to obtain the two-dimensional coordinates of the 50t crawler crane and rotary drilling rig, retrieves the spatial grid static load offset mapping table, and extracts the comprehensive additional static load stress parameter. The basic static stress data is then substituted into the formula. Eliminating static interference caused by mechanical movement, we obtain the true static load corrected stress vector. .

[0100] Subsequently, the algorithm execution engine extracts the high-frequency excitation response signal around the rotary drilling rig and compares it with the first... The high-frequency excitation response signal of a steel pipe pile. Substituting into the cross-correlation integral function: ; The cross-correlation coefficients of the four steel pipe piles are obtained, and the transitive response column vector is generated.

[0101] Will Transformed into a 10×10 spatial adjacency matrix according to topology. and the transmitted response column vector aligned with the dimensions Combine and substitute into the formula Generate an 11×11 extended dimension correlation matrix. .

[0102] Singular value decomposition and early warning determination process: At this time, the silt layer where the No. 3 bottom steel pipe pile is located undergoes hidden bearing capacity degradation (uneven settlement) due to cutting disturbance.

[0103] The algorithm execution engine performs singular value decomposition on the extended dimension correlation matrix. Extract the first principal component feature vector corresponding to the largest singular value. .

[0104] Retrieve the reference feature vector under the pre-stored safety calibration state Substitute into the formula for calculating spatial angles: ; The current multidimensional spatial angle is calculated. And substitute into the formula. Calculate the structural instability risk index. Preset structural instability judgment threshold. The degree is 12 degrees. The dynamic characteristics changed abruptly due to the settlement of pile number 3, resulting in the calculated... It rose to 15.6 degrees Celsius. The system successfully triggered the abnormal warning command.

[0105] The experimental verification and effect comparison are as follows: To fully verify the technical effects of the present invention, a comparative experiment was conducted between the system of the present invention and a traditional system that relies solely on static stress threshold alarms. The conclusions are as follows: Reference Appendix Figure 6 , attached Figure 6 This demonstrates the data purification capabilities of the algorithm execution engine under the action of an adaptive bandpass filter. Figure 6 It contains the original stress time series signal mixed with mechanical white noise and wave interference, and the separated high-frequency excitation response signal. Figure 6 It is known that the raw signals acquired by traditional methods are chaotic and have an extremely low signal-to-noise ratio; however, this invention uses a filter that is dynamically set by the spindle speed parameter to successfully filter out irrelevant frequency bands and extract high-frequency excitation response signals with regular waveforms and concentrated energy, providing a high-purity data source for subsequent dynamic cross-correlation calculations.

[0106] Reference Appendix Figure 7 , attached Figure 7 This demonstrates the technical effectiveness of the synchronous scheduling unit and the algorithm execution engine working together to solve the problem of physical delay dispersion. Figure 7 The diagram illustrates how the cross-correlation integral value changes with the dynamic offset parameter. Due to the dispersion phenomenon in the steel medium, the energy peak cannot be accurately captured simply based on the theoretical time delay (as shown in the diagram, the value at the origin of the theoretical time delay is not the maximum).

[0107] This invention introduces a time-sliding search window. It successfully captured the true maximum absolute value of the integral at the offset position as the cross-correlation coefficient, effectively absorbing the waveform phase dispersion caused by the physical topology.

[0108] Reference Appendix Figure 8 , attached Figure 8 This demonstrates the advantages of the present invention in the early detection of concealed settlement. Figure 8 The data records a comparison between the structural instability risk index calculated by this invention and the traditional static stress monitoring index during the 60 minutes of slow settlement of the No. 3 steel pipe pile. As shown in Figure 8, due to the redistribution of internal forces in the statically indeterminate structure of the steel platform, the traditional static stress index showed almost no significant fluctuations in the first 40 minutes, only exceeding the threshold when significant settlement occurred. In contrast, this invention integrates static and dynamic coupling characteristics, and the spatial angle of the first principal component eigenvector is extremely sensitive to minor degradation of the structural boundary conditions. The structural instability risk index exceeded the judgment threshold (horizontal dashed line) at the 25th minute, triggering an anomaly warning command nearly 15 minutes in advance, thus improving the safety redundancy of the project.

Claims

1. A real-time stress monitoring data processing system for offshore steel platform structures, characterized in that, include: The data acquisition module is equipped with strain gauges and environmental sensors. The strain gauges measure the stress time series signal on the surface of the offshore steel platform, and the environmental sensors measure sea wave data and tide data. The data acquisition module sends the stress time series signal, the wave data, and the tide data to the edge computing node. The equipment telemetry module is equipped with a real-time positioning antenna and a main control data interface. The real-time positioning antenna reads the two-dimensional coordinates of the 50t crawler crane and rotary drilling rig running on the offshore steel platform. The main control data interface reads the spindle speed parameters of the rotary drilling rig. The equipment telemetry module sends the two-dimensional coordinates and the spindle speed parameters to the edge computing node. The edge computing node is equipped with a memory block, a synchronization scheduling unit, and an algorithm execution engine. The synchronization scheduling unit performs time-stamp alignment on the received stress time series signal, wave data, tide data, two-dimensional coordinates, and spindle speed parameters. The algorithm execution engine performs narrowband filtering, static load offset correction, active excitation transmission response extraction, and spatial deflection early warning calculation based on the time-stamp aligned data. The edge computing node determines the settlement status of the offshore steel platform based on the output of the spatial deflection early warning calculation and outputs an early warning command.

2. The real-time stress monitoring data processing system for offshore steel platform structures according to claim 1, characterized in that, The data acquisition module is arranged at each strain gauge installation node of the offshore steel platform, and the strain gauges are respectively fixed on the surface of the main crossbeam, the surface of the Bailey bridge sheet and the surface of the steel pipe pile of the offshore steel platform. The telemetry module of the equipment is installed inside the 50t crawler crane and the rotary drilling rig; The edge computing node is located inside the terminal box of the offshore steel platform.

3. The real-time stress monitoring data processing system for offshore steel platform structures according to claim 2, characterized in that, The synchronization scheduling unit is internally divided into a high-frequency data buffer queue, a low-frequency data buffer queue, and a medium-frequency data buffer queue. The synchronization scheduling unit stores the stress time series signal into the high-frequency data buffer queue, the wave data and the tide data into the low-frequency data buffer queue, and the two-dimensional coordinates and the spindle speed parameters into the medium-frequency data buffer queue. The synchronization scheduling unit sets a global reference timestamp sequence, reads the target time node in the global reference timestamp sequence, and performs resampling interpolation calculations on the low-frequency data buffer queue and the medium-frequency data buffer queue according to the target time node to generate wave data alignment values, tide level data alignment values, two-dimensional coordinate interpolation results, and spindle speed parameter interpolation results; The synchronization scheduling unit extracts the stress time series signal, wave data alignment value, tide data alignment value, two-dimensional coordinate interpolation result, and spindle speed parameter interpolation result corresponding to the same target time node. It then packages all the extracted parameters corresponding to the same target time node into a unified timestamp data frame and continuously transmits it to the algorithm execution engine to perform the narrowband filtering operation.

4. The real-time stress monitoring data processing system for offshore steel platform structures according to claim 3, characterized in that, The algorithm execution engine continuously reads the unified timestamp data frame and parses and extracts the stress time series signal corresponding to each measurement point position from the unified timestamp data frame; The algorithm execution engine sets the cutoff frequency parameter of the digital low-pass filter, inputs the stress time series signal into the digital low-pass filter to perform calculations, retains the signal waveform data in the stress time series signal with a frequency less than or equal to the cutoff frequency parameter, integrates them into basic static stress data, and stores them in the memory block; The algorithm execution engine extracts the tide level data alignment values ​​under the same target time node, retrieves the preset tide level to strain linear conversion coefficient, and converts the tide level data alignment values ​​into the global seawater buoyancy change reference drift amount. The algorithm execution engine performs a scalar multiplication vector operation on the global seawater buoyancy change benchmark drift amount and the preset node elevation compensation coefficient vector to generate a differentiated drift compensation vector. The algorithm execution engine subtracts the differentiated drift compensation vector from the basic static stress data and overwrites and saves it as tide level compensation static stress data into the memory block.

5. The real-time stress monitoring data processing system for offshore steel platform structures according to claim 4, characterized in that, The algorithm execution engine retrieves the spindle speed parameter interpolation result and the pre-stored rotary drill bit tooth number parameter from the unified timestamp data frame to calculate the true high-frequency cutting fundamental frequency, and sets a bandpass filter in the built-in internal memory, assigning the value of the true high-frequency cutting fundamental frequency to the center frequency variable of the bandpass filter. The algorithm execution engine extracts the wave data alignment values ​​under the same target time node, calculates the current wave height, and compares it with the preset wave height threshold. The algorithm execution engine sets a frequency offset parameter. When the wave height is less than or equal to the wave height threshold, the algorithm execution engine presets the value of the frequency offset parameter as a reference value. When the wave height is greater than the wave height threshold, the algorithm execution engine calculates a dynamic attenuation coefficient and uses the dynamic attenuation coefficient to proportionally reduce the value of the frequency offset parameter. The algorithm execution engine calculates the upper and lower cutoff frequencies of the bandpass filter based on the center frequency variable and the frequency offset parameter, inputs the stress time series signal in the unified timestamp data frame into the bandpass filter for filtering, integrates the waveform components retained by the bandpass filter to generate a high-frequency excitation response signal, and stores the high-frequency excitation response signal in the memory block to complete the narrowband filtering operation.

6. The real-time stress monitoring data processing system for offshore steel platform structures according to claim 5, characterized in that, The algorithm execution engine reads the structural dimension parameters of the offshore steel platform to establish a two-dimensional coordinate system, and divides the surface of the offshore steel platform into an array of two-dimensional coordinate grids according to the set distance step size; The algorithm execution engine applies the virtual static loads corresponding to the 50t crawler crane and the rotary drilling rig to the center node of each two-dimensional coordinate grid. In the finite element simulation model, it extracts the simulated stress values ​​of each strain gauge installation node when each two-dimensional coordinate grid is subjected to the virtual static load. It extracts the difference vector between the simulated stress values ​​and the reference simulated stress values ​​of each strain gauge installation node when there is no virtual static load. It arranges and combines the difference vectors corresponding to a single two-dimensional coordinate grid according to the node order to generate additional static load stress parameters. It binds the boundary coordinate parameters of the two-dimensional coordinate grid with the additional static load stress parameters to generate mapping entries. It summarizes and compiles to generate a spatial grid static load offset mapping table and writes it into the internal memory. The algorithm execution engine parses the unified timestamp data frame to extract the two-dimensional coordinate interpolation result as a search keyword and inputs it into the internal memory. When it determines that the two-dimensional coordinate interpolation result falls within the boundary coordinate parameter range of the target two-dimensional coordinate grid, it extracts the first additional static load stress parameter corresponding to the 50t crawler crane and the second additional static load stress parameter corresponding to the rotary drilling rig, and performs vector addition to merge them to generate a comprehensive additional static load stress parameter. The algorithm execution engine subtracts the comprehensive additional static load stress parameter from the tidal level compensation static stress data in the memory block to obtain the static load correction stress vector, stores it in the data cache to await the call of the correlation matrix reconstruction operation, and completes the static load bias correction operation.

7. The real-time stress monitoring data processing system for offshore steel platform structures according to claim 6, characterized in that, The algorithm execution engine sets the sampling time window length according to the center frequency variable of the bandpass filter, calculates the planar Euclidean distance between the two-dimensional coordinate interpolation result of the rotary drilling rig and the three-dimensional spatial projection coordinates of each strain gauge installation node on the offshore steel platform, determines the strain gauge installation node whose planar Euclidean distance is less than a preset search radius threshold as the target node, retrieves the high-frequency excitation response signal generated by the strain gauge at the target node within the sampling time window length from the memory block, and simultaneously retrieves the corresponding high-frequency excitation response signal generated by the strain gauge at the surface of each steel pipe pile from the memory block. The algorithm execution engine reads the spatial topology model pre-stored in the internal memory to calculate the shortest physical transmission path length, and calculates the theoretical time delay of the vibration mechanical wave to each of the steel pipe piles based on the shortest physical transmission path length and the sound velocity constant of the mechanical wave in the steel. The algorithm execution engine sets a time sliding search window centered on the theoretical time delay, substitutes each dynamic offset parameter in the time sliding search window into the cross-correlation integral function, and performs dynamic advance compensation operation on the time axis for the high-frequency excitation response signal corresponding to the surface of each steel pipe pile. The algorithm execution engine performs cross-correlation integral calculation on the high-frequency excitation response signal at the target node and the high-frequency excitation response signal corresponding to the surface of each steel pipe pile after the dynamic advance compensation operation, and extracts the maximum absolute value of the integral within the time sliding search window as the cross-correlation number. The algorithm execution engine extracts the corresponding cross-correlation coefficients according to the structural number order of each steel pipe pile, arranges them sequentially to generate a transmission response column vector, and completes the active excitation transmission response extraction operation.

8. The real-time stress monitoring data processing system for offshore steel platform structures according to claim 7, characterized in that, The algorithm execution engine retrieves the static load correction stress vector stored in the data cache and initializes a blank two-dimensional matrix in the memory block. When the algorithm execution engine determines that there is a direct structural connection between any two nodes in each of the strain gauge installation nodes based on the spatial topology model, it extracts the two corresponding stress values ​​in the static load correction stress vector, calculates the arithmetic mean of the absolute values ​​of the two stress values, adds the arithmetic mean to the preset structural reference connectivity minimum constant to generate edge connection weights, and fills the corresponding row and column intersection matrix element positions in the two-dimensional square matrix according to the index number of the strain gauge installation node where the corresponding two stress values ​​are located. When the algorithm execution engine determines, based on the spatial topology model, that there is no direct structural connection between any two strain gauge installation nodes, it keeps the value of the corresponding element position in the two-dimensional matrix as zero. The algorithm execution engine sequentially fills the absolute values ​​of the stress values ​​corresponding to each strain gauge mounting node in the static load correction stress vector into the main diagonal position of the two-dimensional square matrix to generate a spatial adjacency matrix.

9. The real-time stress monitoring data processing system for offshore steel platform structures according to claim 8, characterized in that, The algorithm execution engine allocates a matrix storage space in the memory block for storing the extended-dimensional correlation matrix; The algorithm execution engine fills the cross-correlation coefficient in the transmitted response column vector into the newly created aligned column vector corresponding to the position of the strain gauge installation node on the surface of the steel pipe pile, according to the index number of each strain gauge installation node. The engine fills the element values ​​of the other strain gauge installation node positions in the aligned column vector, except for the strain gauge installation node on the surface of the steel pipe pile, with zero, and generates a dimension-aligned transmitted response column vector. The algorithm execution engine extracts the global maximum absolute value from the spatial adjacency matrix and the dimension-aligned transitive response column vector, respectively. It then uses this global maximum absolute value to scale and normalize the corresponding matrix elements and cross-correlation coefficients proportionally, mapping the values ​​to the same dimensionless value range. Next, the algorithm execution engine employs a block matrix combination algorithm to fill the spatial adjacency matrix into the upper-left sub-matrix region of the expanded-dimensional correlation matrix, and to fill the dimension-aligned transitive response column vector into the upper-right sub-vector region of the expanded-dimensional correlation matrix. A row and column transpose operation is performed on the dimension-aligned transitive response column vector to obtain its transpose matrix, which is then filled into the lower-left sub-vector region of the expanded-dimensional correlation matrix. Finally, the lower-right cross-element value of the expanded-dimensional correlation matrix is ​​set to zero to complete the concatenation operation and obtain the expanded-dimensional correlation matrix.

10. The real-time stress monitoring data processing system for offshore steel platform structures according to claim 9, characterized in that, The algorithm execution engine performs singular value decomposition on the expanded dimension correlation matrix, decomposing the expanded dimension correlation matrix into a left orthogonal matrix, a diagonal singular value matrix, and the transpose of a right orthogonal matrix. It extracts the maximum singular value in the diagonal singular value matrix and extracts the corresponding eigenvector from the left orthogonal matrix based on the row index position of the maximum singular value to determine the first principal component eigenvector. The algorithm execution engine retrieves a pre-stored reference feature vector matrix library from its internal memory, matches and retrieves the corresponding target reference feature vector in the library, performs matrix multiplication on the transpose of the first principal component feature vector and the target reference feature vector to obtain an initial vector inner product, extracts the absolute value of the initial vector inner product as the actual calculation vector inner product, calculates the L2 norm of the first principal component feature vector and the L2 norm of the target reference feature vector respectively, and uses the inverse cosine function to calculate the multidimensional spatial angle between the first principal component feature vector and the target reference feature vector based on the actual calculation vector inner product, the L2 norm of the first principal component feature vector, and the L2 norm of the target reference feature vector. The algorithm execution engine performs real-time normalization comparison and calculation with the value of the multi-dimensional spatial angle and the pre-stored structural instability judgment threshold to obtain the structural instability risk index. When the structural instability risk index is greater than 1, it triggers an anti-instability anomaly warning command as the warning command output, thus completing the spatial deflection warning calculation operation.

Images