A method for calculating wave energy distribution and structural response of breakwater based on energy concentration theory

Abstract

The present application relates to electric digital data processing technology, disclose a kind of based on the calculation method of wave energy distribution and structure response of breakwater wave energy based on energy aggregation theory, comprising: obtaining the multi-dimensional time series state data stream and discrete topological atlas of the physical field evolution of the wave load of breakwater, data stream is divided into sub-tensor block and the first-order time series difference integral value is calculated, to extract the energy space aggregation and time series sparsity of wave physical field Energy aggregation characteristic parameter;Establish numerical comparison rule to generate bit mask sequence, and trigger logic shielding path at the state register level of processor, so that arithmetic logic unit skips the non-active sub-tensor block marked as resting state in the calculation process, the present application uses bit mask freezing mechanism to block invalid data flow to high-power floating-point operation unit, so that computing power resources and data fluctuation state reach nonlinear adaptive alignment;Massive interactive computing can effectively relieve the occupancy pressure of memory bus bandwidth.

Description

Technical Field

[0001] This invention belongs to the field of electrical digital data processing technology, and in particular relates to a method for calculating the wave energy distribution and structural response of breakwaters based on energy concentration theory. Background Technology

[0002] Currently, in numerical calculations for marine engineering, high-dimensional matrices are typically constructed and iterative calculations are used to achieve digital reconstruction, thereby forming a data pool that characterizes the spatiotemporal evolution. As the computational scale increases, marine environmental data exhibits multidimensional and nonlinear characteristics. Combined with the breakwater topology, this leads to the generation of large-scale sparse matrices in the underlying processing flow. Existing technologies employ an iterative mode driven by a globally synchronized clock. This mode assumes that all data blocks of the spatiotemporal state matrix are in an active update state within the surface time step. Due to the spatial clustering and temporal sparsity of the physical field energy distribution, the stress state of various parts of the breakwater is in a physically resting state within a specific period. Existing computational logic lacks the ability to distinguish low-gradient data, resulting in processor resources being consumed by sub-matrix blocks that have not undergone state transitions, causing ineffective occupation of memory bus bandwidth.

[0003] Simplifying the physical model by adding hardware configurations cannot eliminate redundant logic in the underlying matrix operations. Simplifying the model reduces analytical accuracy. In addition to the limitations of hardware structure physical modeling, the mismatch between the underlying computing power scheduling logic and data characteristics is the key factor restricting the solution efficiency. For example, Chinese invention patent application CN121411827A discloses a branch processing system and method based on a structured node SIMT architecture. By inserting explicit node markers into the branch code through the compiler and controlling the thread activity mask, irrelevant branch instructions are skipped in the scenario of full branch consistency. Such logical branch marking schemes only handle the program's preset logical jumps and cannot cope with the data stream of physical field with continuous fluctuation characteristics in breakwater monitoring. Faced with massive sensor data in non-ideal engineering environments, this architecture lacks the ability to perceive the dynamic evolution characteristics of the physical field and is difficult to autonomously identify and separate physical resting regions from the original time series signals, resulting in an imbalance in the allocation of computing power resources when processing high-gradient physical mutations and low-gradient environmental noise.

[0004] Therefore, how to reconstruct the underlying matrix operation path to achieve adaptive alignment between computing resources and physical state abrupt changes has become the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention aims to solve the problems of redundant underlying matrix operation paths and mismatch between computing resources and physical state changes in high-dimensional nonlinear data processing.

[0006] In this technical solution, a method for calculating the wave energy distribution and structural response of breakwaters based on energy-concentrating theory includes the following steps: Step S1: Obtain the multidimensional time-series state data stream characterizing the evolution of the physical field of wave load on the breakwater and the discrete topological map of the breakwater structure; complete the discretization encoding of the multidimensional time-series state data stream; and initialize the data analysis pool in the memory space. Step S2: Divide the multidimensional temporal state data stream into multiple sub-tensor blocks, calculate the first-order temporal difference integral value for each sub-tensor block within the time window, and extract the energy-gathering characteristic parameters that characterize the spatial clustering and temporal sparsity of wave physical field energy fluctuation gradient. Step S3: Establish a numerical comparison rule between the energy-gathering feature parameter and the calculated activation threshold, and generate a bitmask sequence for marking data activity; when the energy-gathering feature parameter is lower than the calculated activation threshold, allocate a low-level freeze bitmask to the corresponding sub-tensor block in the status register; when the energy-gathering feature parameter is not lower than the calculated activation threshold, allocate a high-level activation bitmask to the corresponding sub-tensor block. In step S4, the arithmetic logic unit completes directional addressing based on the bitmask sequence, skips the floating-point operation instructions of the sub-tensor block carrying the low-level frozen bitmask through the logic shielding path, maintains the cache reference value of the sub-tensor block in the previous time window, and only performs dimensionality reduction mapping calculation on the sub-tensor block carrying the high-level active bitmask, and reassembles to generate the energy-gathering feature matrix characterizing the wave energy distribution of the breakwater. Step S5: Perform topological constraint convolution operation on the energy-concentrating feature matrix based on the connectivity properties of the discrete topological map, and linearly map the operation result into a discrete array of structural responses characterizing the physical state of the breakwater, so as to synchronously output the calculation results of the breakwater wave energy distribution and structural response.

[0007] Preferably, step S1 includes: step S11, obtaining the geometric boundary parameters and interaction moment characteristic values ​​of the breakwater structure, and abstracting the geometric boundary parameters and interaction moment characteristic values ​​into a discrete topological mesh composed of topological nodes and topological edges connecting adjacent topological nodes, so as to construct a digital solution framework.

[0008] Preferably, step S2 includes: step S21, within the sliding time window, performing a first-order temporal difference operation on the temporal feature vector contained in the sub-tensor block, and performing cumulative integration processing on the obtained absolute difference value in the time dimension to obtain a local feature index characterizing the intensity of numerical evolution.

[0009] Preferably, step S3 includes: step S31, comparing the energy-gathering feature parameter with a preset change-sensitive threshold; if the energy-gathering feature parameter is greater than or equal to the change-sensitive threshold, then setting the corresponding bit in the bit mask sequence to a first logic state; if the energy-gathering feature parameter is less than the change-sensitive threshold, then setting the corresponding bit in the bit mask sequence to a second logic state.

[0010] Preferably, step S4 includes: step S41, using the logic masking instruction generated by the bitmask sequence to filter node data in the instruction scheduling stage of the central processing unit, thereby blocking the transmission path of redundant data to the floating-point arithmetic unit.

[0011] Preferably, after step S5, step S6 is also included, which involves extracting the numerical features of the stress distribution nodes in the breakwater structure, calling the topology constraint-driven feature alignment logic, performing coordinate mapping and fusion of the dynamic feature flow of the energy-concentrating feature matrix and the discrete topology map, and outputting the structural loss probability data of the breakwater.

[0012] Preferably, after step S2, there is also step S22, which involves constructing an adjacency matrix based on the adjacency attributes of the discrete topological graph, and using a clustering algorithm to divide the adjacency matrix into multiple sub-matrix blocks, each sub-matrix block corresponding to a sub-tensor block, in order to complete the spatial decoupling of the multidimensional temporal state data stream.

[0013] Preferably, the process of completing the dimensionality reduction mapping calculation in step S4 includes: step S42, calculating the covariance matrix of the sub-tensor block with high-level activation bit mask, and realizing feature compression by removing feature dimensions whose internal eigenvalues ​​are lower than the preset dimensionality reduction threshold.

[0014] Preferably, after step S4, step S43 is also included, whereby after the sub-tensor block is skipped in the logical masking path, the addressing instruction is updated to directly jump to the starting storage address of the next sub-tensor block that is assigned a high-level activation bitmask.

[0015] Compared with existing technologies, the wave energy distribution and structural response calculation method for breakwaters based on energy focusing theory of the present invention has the following advantages: 1. In the distribution of wave energy in breakwaters, by introducing a tensor block bitmask freezing mechanism at the register level, the spatial clustering of data fluctuations is identified by using first-order time-series differential integration. This allows the arithmetic logic unit to automatically skip inactive tensor blocks in a frozen state during the iteration process. This mechanism changes the underlying computing power consumption from global full calculation to local effective iteration, and blocks the transmission path of invalid data to high-power floating-point arithmetic units from the hardware instruction level, thereby reducing the computing load and power consumption of the central processing unit.

[0016] 2. The digital analysis stage and the tensor compression stage work together to divide the high-dimensional matrix into multiple sub-tensor blocks with a preset step size based on the physical spatial adjacency attribute of the topological nodes. Combined with the feature mapping logic within the sliding time window, it achieves precise isolation of data in the resting region. This method significantly reduces the amount of high-dimensional nonlinear data transported on the memory bus, alleviates the memory bandwidth throughput pressure in the process of massive spatiotemporal interactive computing, and ensures that the system maintains stable response latency when processing large-scale sparse matrices.

[0017] 3. The solution uses topology-constrained feature alignment and position mapping logic to deeply integrate dynamic feature flow and static boundary map in digital space. By using logical comparison instructions generated by bitmasks, the system automatically filters redundant node data in a disconnected state without losing the model's analytical accuracy, and reassembles it to generate a high-density effective energy distribution matrix. This process reduces the computational complexity from exponentially expanding with the node size to linearly increasing with the size of the local high-activity region, thus overcoming the computational power ceiling of traditional full-node traversal algorithms. Attached Figure Description

[0018] Figure 1 This is a flowchart of the calculation of wave energy distribution and structural response of breakwater based on the energy-concentrating theory of this invention; Figure 2 This is the underlying hardware architecture and data flow logic diagram of the digital computation framework of this invention. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0020] It should be noted that all directional and positional terms used in this invention, such as: up, down, left, right, front, back, vertical, horizontal, inner, outer, top, low, lateral, longitudinal, center, etc., are only used to explain the relative positional relationship and connection between components in a specific state (as shown in the accompanying drawings). They are only for the convenience of describing this invention and do not require that this invention be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention. In addition, the descriptions of "first," "second," etc., in this invention are for descriptive purposes only and should not be construed as indicating or implying their relative importance or implicitly indicating the number of technical features indicated.

[0021] In the description of this invention, unless otherwise explicitly specified and limited, the terms installation, connection, and linking should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections; they can refer to direct connections or indirect connections through an intermediate medium; they can refer to the internal connection of two components. For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances.

[0022] In the description of this specification, references to the terms "an embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example, and the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0023] A method for calculating wave energy distribution and structural response of breakwaters based on energy-concentrating theory includes the following steps: Step S1: Obtain the multidimensional time-series state data stream characterizing the evolution of the physical field of wave load on the breakwater and the discrete topological map of the breakwater structure; complete the discretization encoding of the multidimensional time-series state data stream; and initialize the data analysis pool in the memory space. Step S2: Divide the multidimensional temporal state data stream into multiple sub-tensor blocks, calculate the first-order temporal difference integral value for each sub-tensor block within the time window, and extract the energy-gathering characteristic parameters that characterize the spatial clustering and temporal sparsity of wave physical field energy fluctuation gradient. Step S3: Establish a numerical comparison rule between the energy-gathering feature parameter and the calculated activation threshold, and generate a bitmask sequence for marking data activity; when the energy-gathering feature parameter is lower than the calculated activation threshold, allocate a low-level freeze bitmask to the corresponding sub-tensor block in the status register; when the energy-gathering feature parameter is not lower than the calculated activation threshold, allocate a high-level activation bitmask to the corresponding sub-tensor block. In step S4, the arithmetic logic unit completes directional addressing based on the bitmask sequence, skips the floating-point operation instructions of the sub-tensor block carrying the low-level frozen bitmask through the logic shielding path, maintains the cache reference value of the sub-tensor block in the previous time window, and only performs dimensionality reduction mapping calculation on the sub-tensor block carrying the high-level active bitmask, and reassembles to generate the energy-gathering feature matrix characterizing the wave energy distribution of the breakwater. Step S5: Perform topological constraint convolution operation on the energy-concentrating feature matrix based on the connectivity properties of the discrete topological map, and linearly map the operation result into a discrete array of structural responses characterizing the physical state of the breakwater, so as to synchronously output the calculation results of the breakwater wave energy distribution and structural response.

[0024] Preferably, step S1 includes: step S11, obtaining the geometric boundary parameters and interaction moment characteristic values ​​of the breakwater structure, and abstracting the geometric boundary parameters and interaction moment characteristic values ​​into a discrete topological mesh composed of topological nodes and topological edges connecting adjacent topological nodes, so as to construct a digital solution framework.

[0025] Preferably, step S2 includes: step S21, within the sliding time window, performing a first-order temporal difference operation on the temporal feature vector contained in the sub-tensor block, and performing cumulative integration processing on the obtained absolute difference value in the time dimension to obtain a local feature index characterizing the intensity of numerical evolution.

[0026] Preferably, step S3 includes: step S31, comparing the energy-gathering feature parameter with a preset change-sensitive threshold; if the energy-gathering feature parameter is greater than or equal to the change-sensitive threshold, then setting the corresponding bit in the bit mask sequence to a first logic state; if the energy-gathering feature parameter is less than the change-sensitive threshold, then setting the corresponding bit in the bit mask sequence to a second logic state.

[0027] Preferably, step S4 includes: step S41, using the logic masking instruction generated by the bitmask sequence to filter node data in the instruction scheduling stage of the central processing unit, thereby blocking the transmission path of redundant data to the floating-point arithmetic unit.

[0028] Preferably, after step S5, step S6 is also included, which involves extracting the numerical features of the stress distribution nodes in the breakwater structure, calling the topology constraint-driven feature alignment logic, performing coordinate mapping and fusion of the dynamic feature flow of the energy-concentrating feature matrix and the discrete topology map, and outputting the structural loss probability data of the breakwater.

[0029] Preferably, after step S2, there is also step S22, which involves constructing an adjacency matrix based on the adjacency attributes of the discrete topological graph, and using a clustering algorithm to divide the adjacency matrix into multiple sub-matrix blocks, each sub-matrix block corresponding to a sub-tensor block, in order to complete the spatial decoupling of the multidimensional temporal state data stream.

[0030] Preferably, the process of completing the dimensionality reduction mapping calculation in step S4 includes step S42, which calculates the covariance matrix of the sub-tensor block with high-level activation bit mask, and achieves feature compression by removing feature dimensions whose internal eigenvalues ​​are lower than the preset dimensionality reduction threshold.

[0031] Preferably, after step S4, step S43 is also included, whereby after the sub-tensor block is skipped in the logical masking path, the addressing instruction is updated to directly jump to the starting storage address of the next sub-tensor block that is assigned a high-level activation bitmask.

[0032] Example 1: In a monitoring scenario of an ultra-deepwater port protection project containing over 10,000 discrete topological nodes, when the computing system faces a nonlinear strong storm surge environment, the multi-dimensional sensor array pushes a massive multi-dimensional time-series state data stream to the back-end digital computation framework. If a matrix full-iteration mode driven by a globally synchronized clock is adopted, the floating-point unit of the central processing unit will cause invalid memory bus bandwidth usage and computational latency due to processing a large number of low-gradient data blocks in a physically resting state. To address the aforementioned computational power consumption dilemma, the memory controller loads the acquired multi-dimensional time-series state data stream into a preset data analysis pool in the memory space. The central processing unit divides the high-dimensional matrix into multiple sub-tensor blocks with a preset step size based on the adjacency attribute of the discrete topological graph, and retrieves historical frame data within the sliding time window. The first-order temporal difference integration operation is performed on each sub-tensor block according to the following formula: ,in, For sub-tensor blocks in a sliding time window The cumulative fluctuation gradient within the range is the energy-focusing characteristic parameter. For the corresponding time node The sub-tensor block matrix data, The system uses a sliding time window; it executes logical comparison instructions to determine the energy-focusing characteristic parameters of each sub-tensor block. Whether the preset calculation activation threshold is exceeded, if the determination result is no, the mask generation unit allocates a low-level freeze bit mask for the sub-tensor block in the processor's status register, and if the determination result is yes, it allocates a high-level activation bit mask.

[0033] When performing tensor feature dimensionality reduction operations, the arithmetic logic unit performs directional addressing based on the bitmask sequence of the status register. For sub-tensor blocks carrying low-level frozen bitmasks, the hardware instruction scheduler masks the internal high-order nonlinear transformations and floating-point multiply-accumulate instructions, and maintains the cache base value of the sub-tensor block in the previous time window unchanged. The processor only performs dimensionality reduction mapping calculations on high-gradient sub-tensor blocks carrying high-level active bitmasks to generate energy-concentrated feature matrices. This process utilizes the local energy accumulation effect to establish the logic masking path of the underlying arithmetic logic unit, transforming the computational complexity from an exponential increase with node size to a linear increase with the size of the local active region. This avoids the consumption of memory bandwidth by global floating-point operations while ensuring the quantization accuracy of the structural response.

[0034] Example 2: A computing server equipped with a 2.4GHz processor and a 128-bit floating-point vector operation bandwidth was used to extract the multidimensional time-series state data matrix of the breakwater under nonlinear wave loads through a dynamic simulation model based on the finite element analysis method. This simulation model, based on the three-dimensional Navier-Stokes equations, was used to simulate the response process of a 500m long breakwater structure containing 12,000 discrete elements. To verify the stability of the method in non-ideal engineering environments, Gaussian white noise with a signal-to-noise ratio of 20dB and power frequency interference harmonics with a frequency of 50Hz were actively superimposed on the original signal stream with a sampling frequency of 100Hz during the data acquisition stage. An experimental input set simulating the distribution of real sensor monitoring data was constructed. The experimental group included the sample group of this invention, control group 1, and control group 2. The sample group of this invention adopted a complete bit mask freezing mechanism, control group 1 adopted a global synchronization matrix full iteration mode, and control group 2 did not allocate bit masks during the calculation process but only performed conventional tensor dimensionality reduction. The calculation activation threshold was set according to the energy distribution law, and its value balanced computational efficiency and detection sensitivity. When the calculation activation threshold approached the lower limit of 1.0, low-energy noise falsely triggered the mask activation bits, resulting in a decrease in the computational power saving effect. When it approached the upper limit of 20.0, the response signal of the key structure was filtered, resulting in an increase in quantization error. For specific strong storm surge conditions, the calculation activation threshold was set to 5.0.

[0035] In the operation of the prototype of this invention, the central processing unit divides the multidimensional temporal state data stream into sub-tensor blocks with a step size of 32 and sets a sliding time window. The time interval is 10ms. The focusing characteristic parameters of each sub-tensor block are calculated according to the formula. : ,in, For sub-tensor blocks in a sliding time window Internally extracted focusing characteristic parameters, For the corresponding time node The sub-tensor block matrix data, Using a sliding time window, the energy-concentrating characteristic parameters are extracted in the local region of the wave-facing surface where wave energy accumulates. The measured value is 14.28, while in the resting region of the back wavefront, the energy-concentrating characteristic parameter... The measured value is 0.65; according to the numerical comparison rules, the system's status register contains the energy-concentrating characteristic parameters. A low-level freeze bit mask is assigned to the 0.65 sub-tensor block. The arithmetic logic unit skips the floating-point operation instructions in this region through the hardware instruction scheduler and maintains the cache base value of this region in a continuous 100 time step iterations.

[0036] When processing noisy input, the control group consistently maintained a memory bus bandwidth utilization rate above 85% due to its inability to isolate redundant calculations in the resting region, resulting in a single-cycle structural response calculation delay of 42.6ms. In contrast, the present invention, under the same signal-to-noise ratio conditions, identified and masked 72.4% of invalid data blocks using a bitmask sequence, reducing the measured memory bus bandwidth utilization rate to 19.5% and the single-cycle calculation delay to 11.2ms. Furthermore, the root mean square error of the output structural response discrete array remained stable within 2.8% compared to the ideal value without noise. This improved the performance regarding wave load intensity. The gradient verification results show that when the wave load increases from 20% of the normal operating condition to 100% of the impact condition, the computational efficiency improvement ratio of the sample group of this invention is positively correlated with the area ratio of the topological resting region. In contrast, the second control group, due to the lack of a bitmask-driven logic shielding path, experiences an exponential increase in computation time with increasing load. When the wave load exceeds 120% of the rated load, the second control group experiences memory overflow. The sample group of this invention, by freezing the dynamic reallocation of the bitmask, maintains the continuous operation of the system through directional addressing of arithmetic logic units while maintaining the solution accuracy. (Energy focusing characteristic parameters...) By logically coupling with the bitmask sequence, the energy locality characteristics at the physical level are mapped to the instruction masking rules of the underlying hardware. In high-dimensional spatiotemporal data processing tasks, the system achieves nonlinear adaptive alignment between computing resources and physical evolution state through directional iteration of effective nodes and logical isolation of resting nodes.

[0037] Example 3: This example combines Figures 1 to 2 This paper describes a method for calculating wave energy distribution and structural response of breakwaters based on energy focusing theory. Figure 1 As shown, label S1 is connected to the processing box that completes discretization encoding and initializes the data analysis pool via an irregular wavy curve; label S2 is connected to the processing box that calculates the first-order time-series difference integral value to extract the energy-focusing feature parameters via an irregular wavy curve; label S3 is connected to the processing box that establishes numerical comparison rules and generates bitmask sequences via an irregular wavy curve; label S4 is connected to the processing box that completes dimensionality reduction mapping calculation and reassembles the energy-focusing feature matrix of wave energy distribution via an irregular wavy curve; and label S5 is connected to the processing box that completes topological constraint convolution operation and maps to a discrete array via an irregular wavy curve. The above processing boxes are connected sequentially by downward-pointing unidirectional arrows.

[0038] like Figure 2As shown, the backend digital computation framework points to the memory space and the distributed sensing array containing the multi-dimensional sensor array, indicated by the dashed box. The distributed sensing array points to the data analysis pool inside the memory space via a unidirectional arrow. This data analysis pool contains a discretized encoding matrix for the multi-dimensional time-series state data stream. The memory space also contains a discrete topology map containing a connected adjacency matrix. The memory space directly points to the arithmetic logic unit inside the central processing unit via a connection marked with a memory bus. Inside the central processing unit, the status register containing the bit mask sequence points to the branching mechanism via a connection marked with a bit reset mechanism. The predictor's instruction scheduling stage points to a logically shielded path containing jumps to inactive addressing spaces via a connection marked with skip resting tensors. Simultaneously, the instruction scheduling stage points to a directed addressing path containing high-gradient computation requests via a connection marked with active node scheduling. Both the logically shielded path and the directed addressing path point downwards to the arithmetic logic unit containing floating-point arithmetic units. The arithmetic logic unit points downwards to feature alignment and linear mapping. Feature alignment and linear mapping ultimately point to the breakwater physical state output, which contains a discrete array of structural responses and structural loss probability data.

[0039] Example 4: In the scenario of real-time solution of breakwater structure response deployed with distributed high-performance computing nodes, the computing system faces memory bus conflict caused by the influx of massive nonlinear wave load data. The clock frequency of the system's central processing unit is set to 2.8GHz, and a 16GB data analysis pool is allocated in the memory space to store the multidimensional time-series state data matrix after discretization and encoding. Due to the background thermal noise interference in the sensor link, the system needs to determine the calculation activation threshold before real-time solution.

[0040] The system acquires raw signal streams under physical resting conditions for 100 sampling periods, loads them into a data analysis pool, and discretizes them into multiple sub-tensor blocks. The central processing unit calculates the average energy variance of these sub-tensor blocks, and uses this variance value as the background noise benchmark. Based on the preset false alarm rate control factor Determine the activation threshold for calculation The specific calculation formula is as follows: ,in, To calculate the activation threshold, its dimensions are consistent with the original signal energy. This is the sensitivity adjustment coefficient. The background noise baseline variance is used as the reference value; the activation threshold will be calculated through this process. Anchored to the physical noise level of the current hardware environment, the generated bitmask sequence can be stripped of non-essential energy fluctuations. During the real-time computation phase, the processor receives the digitally analyzed wave state matrix, divides it into sub-tensor blocks with a step size of 16, and sets a sliding time window. Maintaining a step size mapping relationship of 5 times with the sensor sampling period, the central processing unit extracts the focused energy feature parameters of each sub-tensor block. Then, the internal logic comparator is used to... With calculation of activation threshold Perform bit-by-bit comparison; when determining the energy focusing characteristic parameters Below the calculated activation threshold When the mask generation unit writes a 0 flag bit into the specified address space of the status register, it writes a 1 flag bit to generate a bit mask sequence.

[0041] When the arithmetic logic unit processes the instruction pipeline, the hardware instruction scheduler periodically reads the bitmask sequence in the status register. For address paths where a 0 flag is detected, the hardware instruction scheduler triggers the internal branch predictor to modify the address pointed to by the program counter, skipping the high-dimensional tensor shrinking instruction and floating-point multiply-accumulate operation instruction corresponding to the current sub-tensor block. This process utilizes the hardware-level masking path of the status register to block inactive data from entering the high-power arithmetic unit, enabling the processor to perform dimensionality reduction mapping calculations only for active sub-tensor blocks marked as 1. This achieves targeted allocation of computing resources at the underlying physical register level. To ensure that the overall state matrix can penetrate the system hierarchy and map to the underlying computing unit, the computing framework embeds a direct memory access controller in the operating system kernel mode. This controller maps the acquired matrix data blocks to contiguous physical page spaces and strips the standard thread scheduling logic by calling the underlying parallel computing driver interface, forcibly aligning the flag bits in the status register to the predicate mask pins of the single instruction multiple data pipeline. The arithmetic logic unit (ALU) deploys a single instruction multiple data stream (SMIDD) vector processing module. The bitmask sequence output is stored in the status register and used as the predicate operand of the vector processing module. When loading the wave state matrix data frame, the predicate operand is used to synchronously control multiple parallel hardware arithmetic channels. Floating-point multiply-accumulate instructions are executed for physical channels with a bit value of 1, while physical channels with a bit value of 0 are controlled to skip the multiply-accumulate logic gate array and force the output of the register latched data from the previous clock cycle. The predicate mask procedure enables the ALU to block the level flipping of the resting data area at the underlying logic gate level while maintaining the objective condition of global instruction pipeline clock synchronization. When the system processes breakwater topology data with spatial sparsity, it solves the problem of computational mis-triggering caused by hardware noise. The proportion of invalid instruction cycles in a single iteration loop is reduced by 68.5%, the instantaneous throughput pressure of the memory bus is relieved, and the calculation accuracy remains stable under non-ideal conditions containing random noise interference. This verifies the implementation stability of the bitmask freezing mechanism based on energy concentration theory in engineering applications.

[0042] Example 5: In a protective engineering monitoring environment deployment scenario equipped with a distributed sensor array, the system initiates a process targeting a background noise baseline. The calibration procedure is used to determine the calculation activation threshold. The starting benchmark is to monitor the average power spectral density of the sensor array output signal and confirm that the power fluctuation rate is below the silent index over 50 consecutive sampling periods. After achieving a quantization level of 5.0%, the multidimensional time-series state data stream collected at this point is defined as the resting baseline set and loaded into the data analysis pool. The central processing unit iteratively adjusts the false alarm rate control factor. This freezes the bitmask sequence in the reference set to 99.5%, thus determining the activation threshold. To enable the computing system to operate in a zero-instruction-skip state before processing wave loads, in the above iterative calibration procedure, the processor sets the initial search scalar of the control factor to 1.0, and linearly accumulates it upwards in a fixed step size of 0.1 in each clock check cycle, synchronously comparing the full mask sequence state under the step size. Once the arithmetic logic unit determines that the proportion of frozen masks in the set crosses the 99.5% control line, the system immediately triggers a hardware interrupt and stops optimization, thereby fixing the current factor value.

[0043] When the system faces the scenario of migrating computing nodes with a hardware instruction set architecture, the physical step size of the sub-tensor block is determined through a bus matching procedure, and the central processing unit retrieves the bit width parameters of the hardware's single instruction multiple data register. ,in Given the bit width of the processor's vector register, the processor issues a bit width alignment instruction in the status register. Based on the payload length of a single burst transfer on the bus and the instruction prefetch depth, the tensor step size under the current hardware environment is calculated. The hardware instruction scheduler synchronizes the output cycle of the mask generation unit with the clock phase of the processor's instruction pipeline, so that the program counter jump span generated by the arithmetic logic unit when reading the bit mask sequence covers the corresponding resting data address space. The memory bus bandwidth is maintained below the preset power safety threshold when processing spatially sparse matrices.

[0044] Example 6: In the digital solution scenario of a port breakwater with heterogeneous node density, the system completes the pre-deployment topology clustering calibration procedure to determine the step size parameters of the sub-tensor blocks. The central processing unit retrieves the adjacency matrix of each node in the discrete topology graph and calculates the node association density in the address space. The processor determines the step size parameter that matches the single instruction cycle processing capability of the hardware vector arithmetic unit. ,in, The step size parameter for the sub-tensor block. For node association distribution density, step size parameter The value of allows the memory access span of the corresponding sub-tensor block to be physically aligned with the processor cache line. The computing system determines the discretization encoding specification of the sub-tensor block in memory space through this calibration process, thus enabling the energy-focusing feature parameter to be used. The extraction process meets the high-concurrency bandwidth requirements of the underlying bus transmission.

[0045] When the system faces a momentary interruption of the communication link, the mask generation unit operates within a sliding time window. The instruction counter self-test process is initiated at the start of the process, and the mask generation unit monitors the feedback signal of the status register. When continuous... If no update instruction for the corresponding sub-tensor block is detected within a clock cycle, then The preset clock cycle threshold, As a sliding time window, the system issues a bit reset instruction in the specified bit space of the status register and marks the bit mask sequence corresponding to the sub-tensor block as 0. The hardware instruction scheduler modifies the jump offset of the program counter according to the marked bit. The arithmetic logic unit skips the operation instructions under the current addressing path and maintains the cache base value of the sub-tensor block. This process cuts off the path of abnormal data flow to the floating-point operation unit through the bit reset mechanism of the status register. The computing system maintains a stable output state in the timing input environment containing abnormal data packets.

[0046] In the digital processing flow of structural dynamic characteristic mapping, the system's central processing unit (CPU) retrieves the connected adjacency matrix of the discrete topological graph stored in memory and converts it into a convolution operator with sparse topological constraints. The processor reads the energy-gathering feature matrix generated by the tensor dimensionality reduction step and uses the arithmetic logic unit (ALU) to perform topological constraint convolution operations in the feature space. The spatial receptive field size of this convolution operation is aligned with the physical influence radius of the breakwater node. The weight distribution of the convolution kernel is dynamically decayed according to the connectivity step size between nodes in the discrete topological graph, so that the ALU only gathers feature data with physical and mechanical transmission paths when performing tensor product processing. Through this topological constraint convolution operation, the energy-gathering feature matrix is ​​converted into an intermediate state tensor that characterizes the structural response trend. When performing tensor product processing, the ALU uses a zero-value boundary filling algorithm for non-closed nodes to maintain the consistency of the operation dimension. At the same time, its dynamic decay model is set to inverse square logic, that is, the feature transmission weight coefficient is inversely reduced according to the square of the minimum number of connected edges between two nodes in the topological graph, thus clarifying the specific quantization decay law of the convolution kernel.

[0047] To convert the intermediate state tensor into an observable physical quantitative index, the system executes a linear mapping procedure based on physical dimensions to generate a discrete array of structural responses. The central processing unit retrieves the strain scaling factor matrix preset in memory. And the intermediate state tensor and strain proportionality coefficient matrix are combined. Perform fixed-point multiplication and addition operations, where To characterize the quantized constant matrix of the material's mechanical stiffness properties, before performing fixed-point multiplication and addition operations, the strain proportionality coefficient matrix is ​​adjusted based on the energy characteristic matrix. Online dynamic calibration is performed based on the nonlinear stiffness degradation principle of continuum mechanics. When the breakwater structure is subjected to reciprocating wave impacts, the material stiffness in the local area decreases exponentially with the increase of accumulated strain energy. The current time-series differential integral value of each node corresponding to the energy-concentrating feature matrix is ​​extracted, and the local strain ratio coefficient of the node is calculated and updated according to the exponential decay model. ,in, This indicates that the local strain proportionality coefficient of the dynamic calibration node has been completed. This indicates that the static initial parameters of the nodes are retrieved from the discrete topology graph. This indicates that the empirical attenuation constant of the material is obtained through a pre-set standard hydrodynamic impact fatigue test bench. The term "focusing energy characteristic parameter" represents the physical quantity of transient deformation displacement on the surface of the breakwater, as the original multidimensional time-series state data stream collected by the underlying sensor array actually characterizes the physical quantity of the breakwater surface. Its first-order time-series difference essentially reflects the instantaneous strain rate inside the material, while the cumulative integral within the sliding time window is equivalent to the cumulative mechanical strain energy dissipated by the material under alternating load. Therefore, the focusing energy characteristic parameter has the physical dimension of being reasonable as a measure of local fatigue damage that directly triggers the exponential decay of material stiffness. The terms obtained from the traversal calculation are extracted... The transient stiffness scaling matrix is ​​reconstructed by combining the transient stiffness scaling matrix and then used to replace the static strain scaling coefficient matrix. It participates in fixed-point multiplication and addition operations, which transform the abstract tensor features of the digital space into a discrete array of structural responses that characterizes the physical quantities of breakwater deformation. This allows the numerical sequence output by the computing system to directly correspond to the measured displacement parameters of each node of the breakwater. The single linear mapping process achieves near real-time synchronization with the physical state changes through the parallel addressing mechanism of the hardware instruction pipeline.

[0048] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. A method for calculating wave energy distribution and structural response of breakwaters based on energy-concentrating theory, characterized in that, Includes the following steps: Step S1: Obtain the multidimensional time-series state data stream characterizing the evolution of the physical field of wave load on the breakwater and the discrete topological map of the breakwater structure; complete the discretization encoding of the multidimensional time-series state data stream; and initialize the data analysis pool in the memory space. Step S2: Divide the multidimensional temporal state data stream into multiple sub-tensor blocks, calculate the first-order temporal difference integral value for each sub-tensor block within the time window, and extract the energy-gathering characteristic parameters that characterize the spatial clustering and temporal sparsity of wave physical field energy fluctuation gradient. Step S3: Establish a numerical comparison rule between the energy-gathering feature parameter and the calculated activation threshold, and generate a bitmask sequence for marking data activity; when the energy-gathering feature parameter is lower than the calculated activation threshold, allocate a low-level freeze bitmask to the corresponding sub-tensor block in the status register; when the energy-gathering feature parameter is not lower than the calculated activation threshold, allocate a high-level activation bitmask to the corresponding sub-tensor block. In step S4, the arithmetic logic unit completes directional addressing based on the bitmask sequence, skips the floating-point operation instructions of the sub-tensor block carrying the low-level frozen bitmask through the logic shielding path, maintains the cache reference value of the sub-tensor block in the previous time window, and only performs dimensionality reduction mapping calculation on the sub-tensor block carrying the high-level active bitmask, and reassembles to generate the energy-gathering feature matrix characterizing the wave energy distribution of the breakwater. Step S5: Perform topological constraint convolution operation on the energy-concentrating feature matrix based on the connectivity properties of the discrete topological map, and linearly map the operation result into a discrete array of structural responses characterizing the physical state of the breakwater, so as to synchronously output the calculation results of the breakwater wave energy distribution and structural response.

2. The method for calculating wave energy distribution and structural response of breakwaters based on energy focusing theory according to claim 1, characterized in that, Step S1 includes: Step S11, obtaining the geometric boundary parameters and interaction moment characteristic values ​​of the breakwater structure, and abstracting the geometric boundary parameters and interaction moment characteristic values ​​into a discrete topological mesh composed of topological nodes and topological edges connecting adjacent topological nodes, so as to construct a digital solution framework.

3. The method for calculating wave energy distribution and structural response of breakwaters based on energy focusing theory according to claim 1, characterized in that, Step S2 includes: Step S21, within the sliding time window, performing a first-order temporal difference operation on the temporal feature vector contained in the sub-tensor block, and performing cumulative integration processing on the obtained absolute difference value in the time dimension to obtain a local feature index characterizing the intensity of numerical evolution.

4. The method for calculating wave energy distribution and structural response of breakwaters based on energy focusing theory according to claim 1, characterized in that, Step S3 includes: Step S31, comparing the energy-gathering feature parameter with a preset change-sensitive threshold; if the energy-gathering feature parameter is greater than or equal to the change-sensitive threshold, setting the corresponding bit in the bit mask sequence to the first logic state; if the energy-gathering feature parameter is less than the change-sensitive threshold, setting the corresponding bit in the bit mask sequence to the second logic state.

5. The method for calculating wave energy distribution and structural response of breakwaters based on energy-concentrating theory according to claim 1, characterized in that, Step S4 includes: Step S41, using the logic masking instructions generated by the bitmask sequence to filter the node data in the resting state in the instruction scheduling stage of the central processing unit, blocking the transmission path of redundant data to the floating point arithmetic unit.

6. The method for calculating wave energy distribution and structural response of breakwaters based on energy focusing theory according to claim 1, characterized in that, After step S5, step S6 is also included, which extracts the numerical features of stress distribution nodes in the breakwater structure, calls the feature alignment logic driven by topology constraints, completes coordinate mapping and fusion of the dynamic feature flow of the energy-concentrating feature matrix and the discrete topology map, and outputs the structural loss probability data of the breakwater.

7. The method for calculating wave energy distribution and structural response of breakwaters based on energy focusing theory according to claim 1, characterized in that, After step S2, there is also step S22, which constructs an adjacency matrix based on the adjacency attributes of the discrete topology graph, and uses a clustering algorithm to divide the adjacency matrix into multiple sub-matrix blocks, each sub-matrix block corresponding to a sub-tensor block, so as to complete the spatial decoupling of the multidimensional temporal state data stream.

8. The method for calculating wave energy distribution and structural response of breakwaters based on energy focusing theory according to claim 1, characterized in that, The process of completing the dimensionality reduction mapping calculation in step S4 includes step S42, which calculates the covariance matrix of the sub-tensor block with high-level activation bit mask, and achieves feature compression by removing feature dimensions whose eigenvalues ​​are lower than the preset dimensionality reduction threshold inside the sub-tensor block.

9. The method for calculating wave energy distribution and structural response of breakwaters based on energy focusing theory according to claim 1, characterized in that, After step S4, step S43 is also included, where after the sub-tensor block is skipped in the logic masking path, the addressing instruction is updated to directly jump to the starting storage address of the next sub-tensor block that is assigned a high-level activation bitmask.

Images