Intelligent prediction method and system for optical cable performance of data center

Abstract

The application provides an optical cable performance intelligent prediction method and system for a data center, relates to the technical field of data center performance analysis, and first constructs an optical cable microstructure disturbance quantization model, divides the optical cable into multiple virtual quantization units along the longitudinal direction, encapsulates the real-time stress and strain state parameters and response sensitivity coefficients of the basic structure units in each unit, inputs the real-time collected Brillouin scattering spectrum frequency shift data stream into the model for disturbance source analysis, generates a physical excitation decomposition sequence, drives the virtual quantization units to perform state evolution iteration operation according to the physical excitation decomposition sequence, obtains a stress and strain state evolution trajectory set, constructs a space-time correlation network of optical cable link disturbance propagation according to the stress and strain state evolution trajectory set, performs network flow characteristic analysis on the space-time correlation network, extracts a warning node set and calculates the future evolution trend of the warning node set, and generates serialized optical cable performance abnormal warning information. The application can predict the optical cable performance abnormality in advance and guarantee the stable operation of the data center optical cable.

Description

Technical Field

[0001] This invention relates to the field of data center performance analysis technology, and more specifically, to a method and system for intelligent prediction of optical cable performance in data centers. Background Technology

[0002] Data center environments are complex, and optical cables are affected by a variety of external factors, such as temperature changes and mechanical stress. These factors may cause minor disturbances in the internal structure of the optical cable, thereby affecting its transmission performance and even causing data transmission failures.

[0003] Currently, monitoring the performance of optical cables mainly relies on traditional testing methods, such as periodic manual inspections and simple online monitoring equipment. Manual inspections are not only inefficient but also fail to detect minute changes within the optical cable in real time. Simple online monitoring equipment typically only acquires certain single parameters of the optical cable, such as optical power, and cannot provide a comprehensive and in-depth understanding of the stress and strain state of the internal structure or the propagation of disturbances. Furthermore, existing monitoring methods lack the ability to predict optical cable performance anomalies, often only issuing alarms after a fault has already occurred or is about to occur, failing to take preventative measures in advance and posing potential risks to the data transmission security of data centers. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide an intelligent prediction method for optical cable performance in data centers, the method comprising:

[0005] A quantum model of perturbation of optical cable microstructure is constructed. The quantum model of perturbation of optical cable microstructure includes multiple virtual quantization units distributed along the longitudinal direction of optical cable. Each virtual quantization unit corresponds to a basic structural unit inside the optical cable. Each virtual quantization unit encapsulates the real-time stress-strain state parameters of the basic structural unit and the response sensitivity coefficient of the basic structural unit to external perturbation.

[0006] The real-time collected Brillouin scattering spectrum frequency shift data stream along the optical cable is input into the optical cable microstructure perturbation quantization model for perturbation source analysis processing, generating a physical excitation decomposition sequence for each virtual quantization unit at the current moment. Each element in the physical excitation decomposition sequence contains the temperature perturbation component and strain perturbation component obtained from the analysis of the position of the corresponding virtual quantization unit.

[0007] The virtual quantization unit inside the optical cable microstructure perturbation quantization model is driven to perform state evolution iteration calculation according to the physical excitation decomposition sequence, so as to obtain the stress-strain state evolution trajectory set of each virtual quantization unit on the continuous time axis. The stress-strain state evolution trajectory set includes the stress tensor change curve and strain tensor change curve of each virtual quantization unit over time.

[0008] Based on the stress-strain state differences between adjacent virtual quantization units in the stress-strain state evolution trajectory set, a spatiotemporal correlation network for optical cable link disturbance propagation is constructed. The spatiotemporal correlation network uses virtual quantization units as nodes and the rate of change of stress-strain state differences between adjacent nodes as the weight value of the node connection. The spatiotemporal correlation network is used to describe the propagation path and propagation rate of disturbance along the longitudinal direction of the optical cable.

[0009] The spatiotemporal correlation network is subjected to network flow characteristic analysis and processing to extract the set of early warning nodes in the network nodes whose stress and strain states have accumulated to exceed the critical threshold. The evolution trend of stress and strain states of each node in the early warning node set within the future time window is calculated. Based on the evolution trend, serialized optical cable performance anomaly early warning information containing the location coordinates of the early warning nodes and the predicted values ​​of stress and strain states is generated.

[0010] Furthermore, embodiments of the present invention also provide an intelligent prediction system for optical cable performance in data centers, comprising:

[0011] A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the above-described intelligent prediction method for optical cable performance in a data center by executing the machine-executable instructions.

[0012] In another aspect, embodiments of the present invention also provide a computer program product, the computer program product including machine-executable instructions, the machine-executable instructions being stored in a computer-readable storage medium, a processor of an intelligent prediction system for optical cable performance in a data center reading the machine-executable instructions from the computer-readable storage medium, the processor executing the machine-executable instructions, causing the intelligent prediction system for optical cable performance in a data center to execute the aforementioned intelligent prediction method for optical cable performance in a data center.

[0013] Based on the above, a quantized model of optical cable microstructure perturbation is constructed, dividing the optical cable longitudinally into multiple virtual quantization units. Each unit encapsulates the real-time stress-strain state parameters and response sensitivity coefficients of the basic structural unit. Real-time acquired Brillouin scattering spectrum frequency shift data stream is input into the model for perturbation source analysis, generating a physical excitation decomposition sequence. This allows the acquisition of the temperature and strain perturbation components experienced by each virtual quantization unit. Based on the physical excitation decomposition sequence, the virtual quantization units are driven to perform state evolution iterative calculations, obtaining a set of stress-strain state evolution trajectories. This set can represent the stress tensor and strain tensor of each unit at any time. By analyzing the changes in the optical cable link disturbance propagation, a spatiotemporal correlation network is constructed. Using virtual quantization units as nodes and the rate of change of stress-strain state differences between adjacent nodes as the edge weight value, the network can describe the propagation path and rate of disturbance along the longitudinal direction of the optical cable. The network flow characteristics of the spatiotemporal correlation network are analyzed and processed to extract the early warning node set and calculate its future stress-strain state evolution trend. Serialized optical cable performance anomaly early warning information is generated, enabling early prediction of optical cable performance anomalies. This allows for timely preventative measures, effectively ensuring the stable operation of data center optical cables and improving the security and reliability of data transmission. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the execution flow of the intelligent prediction method for optical cable performance in data centers provided in an embodiment of the present invention.

[0015] Figure 2 This is a schematic diagram of exemplary hardware and software components of an intelligent prediction system for optical cable performance in data centers provided in an embodiment of the present invention. Detailed Implementation

[0016] Figure 1 This is a flowchart illustrating an intelligent prediction method for optical cable performance in data centers, provided by an embodiment of the present invention. A detailed description follows.

[0017] Step S110: Construct a quantum model of the microstructure disturbance of the optical cable. The quantum model of the microstructure disturbance of the optical cable includes multiple virtual quantization units distributed along the longitudinal direction of the optical cable. Each virtual quantization unit corresponds to a basic structural unit inside the optical cable. Each virtual quantization unit encapsulates the real-time stress-strain state parameters of the basic structural unit and the response sensitivity coefficient of the basic structural unit to external disturbances.

[0018] Step S111: Divide the internal physical structure of the data center optical cable into a continuous sequence of basic structural units along the longitudinal direction. The length of each basic structural unit matches the microstructural feature scale of the optical cable material. Each basic structural unit in the sequence corresponds to a virtual quantization unit. The total number of virtual quantization units is obtained by dividing the total length of the optical cable by the length of the basic structural units.

[0019] For the core backbone optical cable used to connect the core switch cluster and storage area network within a large data center, the technical specifications of its specific model, GYTS type stranded armored optical cable, were obtained. Microscopic mechanical characteristic parameters of the materials constituting each layer of the cable were extracted from the technical specifications. For the optical fiber itself, its microstructural characteristic scale mainly refers to the diameter of the core and cladding; for the high-density polyethylene material of the tight-buffer and sheath layers, its microstructural characteristic scale is expressed as the average length of the molecular chain segments; for the steel wire armor layer, its microstructural characteristic scale is related to its grain size. To ensure that the constructed model can accurately capture the response of the finest internal structure of the optical cable to external disturbances, the minimum value among all the above material microstructural characteristic scales was taken as the length L_unit of the basic structural unit. The total physical length L_total of the optical cable was obtained from the data center's integrated cabling as-built drawings. L_total / L_unit yielded a quotient value; since L_unit is usually much smaller than L_total, this quotient value is not an integer. To form a discretized unit sequence, the quotient is rounded up to obtain the total number of virtual quantization units, N_unit. Thus, a continuous physical optical cable of length L_total is abstracted as a chain sequence of N_unit virtual quantization units of length L_unit, arranged closely in space. Each of these virtual quantization units represents and encapsulates the macroscopic average mechanical properties and behaviors of all microstructures within the corresponding L_unit optical cable physical segment in the model.

[0020] Step S112: Assign a unique spatial index identifier to each virtual quantization unit. The spatial index identifier increases in the direction from the start end to the end end of the optical cable. A mapping relationship is established between each spatial index identifier and the start coordinates and end coordinates of the actual physical section of the optical cable covered by the virtual quantization unit, forming a correspondence table between spatial index and physical location.

[0021] After generating a chain sequence of N_unit virtual quantization units, each virtual quantization unit in the sequence is assigned a globally unique number for identification and reference within the model, namely a spatial index identifier. The starting point of the optical cable (e.g., the port connecting to core switch A) is set as the coordinate zero point, and the direction along the optical cable laying path towards the end (e.g., the port connecting to storage device B) is set as the positive coordinate direction. The spatial index identifier of the first virtual quantization unit at the starting point is set to ID_1, the second unit to ID_2, and so on, until the last unit at the end is set to ID_N_unit, ensuring that the value of the spatial index identifier strictly increases with the longitudinal distance of the optical cable. Subsequently, a precise mapping is established between each spatial index identifier and the physical section of the optical cable it represents. For the virtual quantization unit ID_i with index i, the starting coordinates of the physical section it covers are Pos_start_i = (i-1) × L_unit, and the ending coordinates are Pos_end_i = i × L_unit. Specifically, for the last unit ID_N_unit, its endpoint coordinate Pos_end_N_unit is forcibly set to the total length L_total of the optical cable to correct for minor length errors that may be introduced by rounding up, ensuring strict alignment between the model's end and the actual physical end. All the above mapping relationships—that is, each spatial index identifier ID_i and its corresponding start coordinate Pos_start_i and end coordinate Pos_end_i—are organized into a structured spatial index-physical location mapping table and stored in the model's memory data structure. The establishment of this mapping table allows any subsequent analysis results based on the virtual quantization unit ID, such as stress anomalies or strain accumulation, to be quickly and accurately deduced, through table lookup operations, the specific geographical location of the anomaly on the optical cable, such as its distance from the starting end.

[0022] Step S113: Initialize a mechanical state memory inside each virtual quantization unit. The mechanical state memory contains three storage areas: the first storage area is used to store the axial stress component value of the virtual quantization unit at the current moment; the second storage area is used to store the radial stress component value of the virtual quantization unit at the current moment; and the third storage area is used to store the shear strain component value of the virtual quantization unit at the current moment.

[0023] After allocating spatial index identifiers, a mechanical state memory is instantiated for each virtual quantization unit in its corresponding memory object. This memory is not a simple variable, but a container with a specific data structure used to dynamically record and update the mechanical state of the optical cable micro-segment represented by the unit at each moment. The memory is logically divided into three independent storage areas with clear physical meaning. The first storage area, labeled σ_a(i, t), stores the axial stress component value of the virtual quantization unit with index i at the current moment t. The direction of this component is parallel to the optical cable axis, reflecting the tensile or compressive forces acting on the optical cable along its length. During model initialization, this area is assigned an initial value σ_a_init, which is calculated using material mechanics formulas based on the pre-stressed tension at the cable's factory and the initial traction force during laying. The second storage area, labeled σ_r(i, t), stores the radial stress component value of the virtual quantization unit with index i at the current moment t. The direction of this component is perpendicular to the optical cable axis, pointing towards the cable core, reflecting the radial compression force caused by external lateral pressure or internal material expansion acting on the optical cable. During initialization, this region is assigned an initial value σ_r_init, which is calculated based on the hydrostatic pressure of the optical cable's environment (such as burial depth or pipeline pressure). The third storage region is labeled γ(i,t) and is used to store the shear strain component value of the virtual quantization unit at index i at the current time t. This component is a dimensionless quantity representing the angular deformation of the optical cable's microstructure caused by torsion or displacement. During initialization, assuming the optical cable is in an ideal, torsion-free state, the value in this region is set to zero.

[0024] Step S114: Initialize a response sensitivity coefficient matrix within each virtual quantization unit. The response sensitivity coefficient matrix includes a temperature perturbation response coefficient and a strain perturbation response coefficient. The temperature perturbation response coefficient represents the equivalent stress change caused by a unit temperature change at the virtual quantization unit, and the strain perturbation response coefficient represents the equivalent stress change caused by a unit mechanical strain at the virtual quantization unit.

[0025] Next, a response sensitivity coefficient matrix is ​​initialized in the memory object for each virtual quantization unit. This response sensitivity coefficient matrix is ​​used to quantify the conversion relationship between external environmental disturbances and internal mechanical responses, and it is in the form of a two-dimensional vector containing two elements. The first element is labeled C_T(i), representing the temperature disturbance response coefficient of the virtual quantization unit at index i. The physical meaning of this temperature disturbance response coefficient is how much equivalent stress Δσ_T will be generated inside the unit when the temperature at the location of the unit changes by a unit ΔT (e.g., an increase of 1 Kelvin), due to the constraint of the thermal expansion and contraction effect of the optical cable materials (such as optical fiber, coating layer, sheath). The relationship is Δσ_T = C_T(i) × ΔT, with the unit being Pascals per Kelvin. The specific value of C_T(i) is not an empirical guess, but is predetermined through finite element simulation or analytical calculation based on the material composition of the optical cable segment constituting the unit, the geometric dimensions of each layer structure, and the corresponding thermal expansion coefficients of the materials, and is fixed as a parameter inside the unit during model initialization. The second element, labeled C_S(i), represents the strain perturbation response coefficient of the virtual quantization unit at index i. The physical meaning of this coefficient is the equivalent stress Δσ_S generated within the unit when the optical cable segment covered by the unit undergoes a mechanical deformation Δε per unit length (i.e., strain is 1). The relationship is Δσ_S = C_S(i) × Δε, with units in Pascals. The value of C_S(i) is primarily determined by the Young's modulus of the optical cable material, which is also pre-calculated and solidified using material mechanics formulas. These two coefficients together constitute the "sensing" sensitivity of the virtual quantization unit to external physical fields (temperature and strain fields), and are key parameters for subsequent decomposition of temperature and strain components from the Brillouin shift.

[0026] Step S115: Establish a state evolution rule function library within each virtual quantization unit. The state evolution rule function library includes a stress relaxation evolution function and a strain accumulation evolution function. The stress relaxation evolution function calculates the stress attenuation at the next moment based on the current stress state and the material viscoelastic parameters. The strain accumulation evolution function calculates the strain increase at the next moment based on the current strain state and the external excitation intensity.

[0027] Furthermore, a state evolution rule function library is established for each virtual quantized unit within its memory object. This library encapsulates the dynamic rules describing how the state of the micro-unit changes over time under external excitation and internal dissipation. The library contains two core functions. The first is the stress relaxation evolution function, denoted as F_relax(σ_a(i,t), σ_r(i,t), P_visco(i)). The input to this stress relaxation evolution function is the axial stress component σ_a(i,t) and radial stress component σ_r(i,t) of the unit at the current time t, and a material viscoelastic parameter vector P_visco(i). P_visco(i) contains constants describing the stress relaxation characteristics of the unit material, such as the relaxation time spectrum. The first function calculates the amount of stress gradually decreasing due to viscous flow within the material over a time step Δt, based on viscoelastic mechanics theory (such as the generalized Maxwell model). The output is the estimated stress decrease for the next time step: Δσ_relax = F_relax(σ_a(i,t), σ_r(i,t), P_visco(i)). The second function is the strain accumulation evolution function, denoted as F_acc(γ(i,t), τ_external(i,t), τ_boundary(i,t)). The inputs to this function are the shear strain component γ(i,t) of the element at the current time t, the equivalent stress τ_external(i,t) acting on the element analytically obtained from the external excitation, and the boundary interaction stress τ_boundary(i,t) transmitted from adjacent elements. The function internally evaluates the increase in irreversible shear strain that the element will produce within a time step Δt under the current stress level, based on the plastic flow law or creep theory. The output is the estimated strain increment for the next moment: Δγ_acc=F_acc(γ(i,t), τ_external(i,t), τ_boundary(i,t)).

[0028] Step S116: Based on the virtual quantization unit with completed core attribute configuration, configure the boundary interaction interface and timing control module, assemble the chain model and complete parameter initialization and consistency verification.

[0029] Finally, after configuring the internal attributes (including the mechanical state memory, response sensitivity coefficient matrix, and state evolution rule function library) of all N_unit virtual quantization units, the overall model is assembled and verified. First, a boundary interaction interface is configured for each virtual quantization unit. This interface includes a data input and a data output. The input receives the axial stress component value σ_a(i-1, t) from its left-adjacent virtual quantization unit (index i-1); the output sends its updated axial stress component value σ_a(i, t+Δt) to its right-adjacent virtual quantization unit (index i+1). For the unit ID_1 at the head of the chain, which has no units to its left, its input is set to receive a fixed boundary condition, such as the free end stress σ_a(0, t) = 0. For the unit ID_N_unit at the tail of the chain, its output is either empty or sends data to the virtual boundary. Next, a time-stepping control module is configured within each virtual quantization unit. This module contains a timer synchronized with the Brillouin optical time-domain reflectometer data acquisition interval Δt. At each Δt time interval, the module triggers the stress relaxation evolution function F_relax and strain accumulation evolution function F_acc of this unit to perform one calculation. Then, all virtual quantization units are chained together in the order of spatial index identifiers ID_1 to ID_N_unit, so that the output of the ID_i unit is connected to the input of the ID_(i+1) unit, forming a complete chained virtual quantization unit sequence model. Parameter initialization and consistency verification are performed on the constructed chained model. All units are traversed, and σ_a(i,0), σ_r(i,0), and γ(i,0) in the mechanical state memory are set to the preset initial values ​​σ_a_init, σ_r_init, and 0, respectively. At the same time, it is verified whether the response sensitivity coefficients C_T(i) and C_S(i) of each unit are within the reasonable threshold range given in the material handbook, for example, whether C_T(i) is between the preset minimum value C_T_min and maximum value C_T_max. The boundary interface connection relationship between adjacent units is verified to ensure that the data flow is correct. After successful verification, the initial state and parameters of all current units are stored in memory, the model is built, and it enters the state of waiting to receive external input data.

[0030] Step S120: Input the real-time acquired Brillouin scattering spectrum frequency shift data stream along the optical cable into the optical cable microstructure perturbation quantization model for perturbation source analysis processing, and generate the physical excitation decomposition sequence of each virtual quantization unit at the current moment. Each element in the physical excitation decomposition sequence contains the temperature perturbation component and strain perturbation component obtained by analyzing the position of the corresponding virtual quantization unit.

[0031] After completing the construction of the quantization model of the fiber optic cable microstructure perturbation, the receiving and processing of external sensing data began. A Brillouin optical time-domain reflectometer deployed at the beginning of the fiber optic cable continuously emitted probe light pulses into the cable at fixed time intervals Δt, and received the backscattered Brillouin light along the cable. The raw data stream continuously output by the instrument, i.e., the Brillouin spectrum curve at each spatial sampling point, was input into the completed model in real time. The perturbation source resolution module inside the model processed the input data stream, aiming to separate the physical excitation components caused by temperature and strain from the coupled Brillouin frequency shift signal, and accurately allocate these components to each corresponding virtual quantization unit.

[0032] Step S121: Receive the raw data stream of frequency shift of Brillouin scattering spectrum along the optical cable continuously collected by the Brillouin optical time domain reflectometer. The raw data stream contains the Brillouin spectrum curve at each spatial sampling point. Each Brillouin spectrum curve consists of a frequency value and the corresponding scattered light intensity value.

[0033] The Brillouin optical time-domain reflectometer (BDR) collects M spatial sampling points at equal intervals along the longitudinal direction of the optical cable in each measurement cycle. For each sampling point j (j from 1 to M), the instrument records a Brillouin spectrum curve. This Brillouin spectrum curve is a two-dimensional data sequence consisting of a series of discrete frequency values ​​f_{j, k} (k from 1 to K, representing the k-th frequency sampling point) and the backscattered light intensity value I_{j, k} corresponding to each frequency value. Therefore, the raw data stream output in one measurement cycle can be structured as follows: for j=1 to M, there is a set {(f_{j, 1}, I_{j, 1}), (f_{j, 2}, I_{j, 2}), ..., (f_{j, K}, I_{j, K})}, which is continuously and in real time pushed to the data input interface of the optical cable microstructure perturbation quantization model.

[0034] Step S122: Perform peak detection processing on the Brillouin spectrum curve at each spatial sampling point location, find the frequency value corresponding to the maximum value of scattered light intensity as the Brillouin frequency shift center frequency at that spatial sampling point location, and obtain the Brillouin frequency shift center frequency curve distributed along the longitudinal direction of the optical cable.

[0035] For each received spectral curve data at sampling point j, the model calls the peak detection algorithm for processing. The algorithm iterates through the K data pairs of that sampling point, compares all scattered light intensity values ​​I_{j, 1} to I_{j, K}, and finds the maximum value I_{j, max}. The frequency value f_{j, max} corresponding to this maximum value I_{j, max} is the Brillouin shift center frequency ν_B(j) at the sampling point j. After performing the above peak detection operation on all M sampling points in sequence, a set of center frequency values ​​corresponding to the spatial positions is obtained: ν_B(1), ν_B(2), ..., ν_B(M). Arranging this set of values ​​according to the spatial order of the sampling points constitutes the Brillouin shift center frequency curve ν_B(x, t) distributed along the longitudinal direction of the optical cable at time t of the current measurement period, where x represents the discrete spatial position.

[0036] Step S123: Read the reference Brillouin frequency shift center frequency curve of the optical cable under the condition of no external disturbance from the pre-stored optical cable reference database, and subtract the reference Brillouin frequency shift center frequency curve from the real-time acquired Brillouin frequency shift center frequency curve point by point to obtain the Brillouin frequency shift change curve distributed along the longitudinal direction of the optical cable.

[0037] From the persistent storage of the model, the reference measurement data of the optical cable when the initial installation and commissioning were completed and no external stress or temperature abnormalities were confirmed are retrieved. This reference data also contains a set of Brillouin frequency shift center frequency values ​​distributed along the longitudinal direction of the optical cable, denoted as ν_B0(1), ν_B0(2), ..., ν_B0(M). The frequency shift center frequency value ν_B(j,t) at the current time t obtained in step S122 is subtracted point by point from the corresponding reference value ν_B0(j). For each sampling point j, the Brillouin frequency shift change Δν_B(j,t) = ν_B(j,t) - ν_B0(j) is calculated. After completing the above calculation for all j = 1 to M, a new curve is obtained, namely the Brillouin frequency shift change curve Δν_B(x,t) distributed along the longitudinal direction of the optical cable at the current time t. This Brillouin frequency shift change curve quantifies the overall offset of the Brillouin frequency shift at each position of the optical cable relative to the reference state at the current time. This offset is the result of the combined effect of temperature and strain.

[0038] Step S124: The Brillouin frequency shift change curve is segmented and mapped according to the spatial index identifier of the virtual quantization unit. Each virtual quantization unit corresponds to a segment of optical cable physical interval. The average value of the Brillouin frequency shift change of all spatial sampling points in the optical cable physical interval is taken as the representative value of the overall Brillouin frequency shift change of the virtual quantization unit.

[0039] Based on the spatial index and physical location correspondence table established in step S112, each virtual quantization unit ID_i covers the physical interval from the starting coordinate Pos_start_i to the ending coordinate Pos_end_i. The sampling point j of the Brillouin optical time-domain reflectometer also corresponds to a specific physical location. The model determines which sampling points fall within the interval covered by ID_i through coordinate comparison. Assuming there are Q sampling points whose position coordinates are within the range [Pos_start_i, Pos_end_i], the index set of these sampling points is denoted as J_i = {j_{i, 1}, j_{i, 2}, ..., j_{i, Q}}. From the Δν_B(x, t) curve obtained in step S123, the Brillouin frequency shift values ​​Δν_B(j_{i, 1}, t), Δν_B(j_{i, 2}, t), ..., Δν_B(j_{i, Q}, t) of all sampling points belonging to the set J_i are extracted. Calculate the arithmetic mean of these values ​​to obtain the representative value of the overall Brillouin frequency shift change of the virtual quantization unit ID_i at the current time t: Δν_B(i,t) = (Δν_B(j_{i,1},t) + Δν_B(j_{i,2},t) + ... + Δν_B(j_{i,Q},t)) / Q. Perform the above mapping and averaging operation on all virtual quantization units from i=1 to N_unit to obtain a sequence of frequency shift change values ​​corresponding to the unit index: {Δν_B(1,t), Δν_B(2,t), ..., Δν_B(N_unit,t)}.

[0040] Step S125: Based on the linear coupling relationship between the Brillouin frequency shift change and temperature and strain, establish the frequency shift change decomposition equation for each virtual quantization unit. The frequency shift change decomposition equation includes a first frequency shift component caused by temperature disturbance and a second frequency shift component caused by strain disturbance. The sum of the two components is equal to the overall Brillouin frequency shift change representative value of the virtual quantization unit.

[0041] Based on the fundamental principles of Brillouin scattering, within a certain temperature and strain range, the change in Brillouin frequency shift has a linear relationship with both temperature and strain changes. For any virtual quantization unit ID_i, its overall Brillouin frequency shift change Δν_B(i,t) can be decomposed into a portion Δν_T(i,t) caused by temperature change and a portion Δν_S(i,t) caused by strain change, satisfying the equation Δν_B(i,t) = Δν_T(i,t) + Δν_S(i,t). This is the foundational equation for the next step of quantitative decomposition.

[0042] Step S126: Call the response sensitivity coefficient matrix encapsulated inside each virtual quantization unit, and extract the temperature perturbation response coefficient and strain perturbation response coefficient of the virtual quantization unit from the response sensitivity coefficient matrix. The temperature perturbation response coefficient represents the amount of Brillouin frequency shift caused by a unit temperature change, and the strain perturbation response coefficient represents the amount of Brillouin frequency shift caused by a unit strain change.

[0043] For each virtual quantization unit ID_i, two key coefficients are extracted from its internally encapsulated response sensitivity coefficient matrix. The first is the temperature-frequency shift coefficient C_νT(i), which physically represents the change in Brillouin frequency shift of the unit caused by a unit temperature change (e.g., 1 Kelvin), in gigahertz per Kelvin. The second is the strain-frequency shift coefficient C_νS(i), which physically represents the change in Brillouin frequency shift of the unit caused by a unit strain change (strain = 1), in gigahertz. These two coefficients are also pre-determined through calibration experiments or theoretical calculations of the optical cable material and are fixed as static parameters within the unit.

[0044] Step S127: Combine the frequency shift change decomposition equation of each virtual quantization unit with the response sensitivity coefficient matrix to construct a system of two linear equations with two unknowns, namely the temperature change and strain change experienced by the virtual quantization unit at the current moment.

[0045] According to the decomposition equation Δν_B(i,t)=Δν_T(i,t)+Δν_S(i,t) in step S125, and the coefficient definition in step S126, the frequency shift component Δν_T(i,t) caused by temperature is equal to the temperature change ΔT(i,t) multiplied by the temperature-frequency shift coefficient C_νT(i), that is, Δν_T(i,t)=C_νT(i)×ΔT(i,t). The frequency shift component Δν_S(i,t) caused by strain is equal to the strain change Δε(i,t) multiplied by the strain-frequency shift coefficient C_νS(i), that is, Δν_S(i,t)=C_νS(i)×Δε(i,t). Substituting these two relationships into the decomposition equation, we obtain a linear equation in two variables concerning the unknowns ΔT(i,t) and Δε(i,t): C_νT(i)×ΔT(i,t)+C_νS(i)×Δε(i,t)=Δν_B(i,t). For each virtual quantization unit ID_i, the coefficients C_νT(i), C_νS(i), and the constant term Δν_B(i,t) in the above equation are all known quantities.

[0046] Step S128: Solve the system of two linear equations for each virtual quantization unit to obtain the current temperature change and strain change values ​​of the virtual quantization unit, thus forming the physical excitation decomposition result of the virtual quantization unit.

[0047] For each virtual quantization unit ID_i, the model solves the two-variable linear equation established in step S127. Since this equation has two unknowns but only one equation, direct solution is difficult. Here, auxiliary conditions are introduced, utilizing the independence of temperature and strain responses in Brillouin shift and the physical properties of the optical cable material. A typical auxiliary condition assumes that within a very short sampling interval Δt, the temperature field changes relatively smoothly in a local region, while the strain field may change more drastically. A second independent equation is constructed through spatial smoothing or temporal difference. For example, it can be assumed that the temperature changes of three adjacent units are approximately equal, thus solving the individual temperature and strain equations of the three units simultaneously. Alternatively, using known optical cable material properties, temperature and strain can be separated under certain specific conditions (such as the gain spectrum shape at a specific frequency). Here, the model employs a pre-calibration method: independent temperature and strain control experiments are conducted on the same batch of optical cable samples in the laboratory to accurately measure their C_νT(i) and C_νS(i), and it is verified that the equation has a unique solution within the normal operating range. Therefore, for each element ID_i, directly solving the simultaneous equations ΔT(i,t)=(Δν_B(i,t)-C_νS(i)×Δε(i,t)) / C_νT(i) and Δε(i,t)=(Δν_B(i,t)-C_νT(i)×ΔT(i,t)) / C_νS(i) effectively achieves decoupling through iteration or based on prior knowledge (such as the low-frequency characteristics of the temperature field). In the simplified model, matrix inversion can be directly performed based on the calibrated coefficients to obtain the unique solutions for ΔT(i,t) and Δε(i,t). After solving, for each element ID_i, a pair of values ​​is obtained: the temperature change ΔT(i,t) (unit: Kelvin) and the strain change Δε(i,t) (dimensionless). This pair of values ​​constitutes the decomposition result of the physical excitation experienced by the element at the current time t.

[0048] Step S129: Arrange the physical excitation decomposition results of all virtual quantization units at the current moment into a physical excitation decomposition sequence in ascending order of spatial index identifiers. Each element in the physical excitation decomposition sequence corresponds to a virtual quantization unit and includes its temperature change and strain change.

[0049] The physical excitation decomposition results of all N_unit virtual quantization units calculated in step S128 are arranged in the order of spatial index identifiers ID_1, ID_2, ..., ID_N_unit. This forms the physical excitation decomposition sequence at the current time t. The i-th element in the physical excitation decomposition sequence is a data structure Element(i,t), which contains two fields: temperature_delta=ΔT(i,t) and strain_delta=Δε(i,t). This physical excitation decomposition sequence completely describes the quantized values ​​of the external physical field (temperature field and strain field) perturbations experienced by each micro-segment (virtual quantization unit) along the entire optical cable at the current measurement time t. This physical excitation decomposition sequence will be used as the core driving input and passed to the model for the next state evolution iteration calculation.

[0050] Step S130: Drive the virtual quantization unit inside the optical cable microstructure perturbation quantization model to perform state evolution iteration calculation according to the physical excitation decomposition sequence, and obtain the stress-strain state evolution trajectory set of each virtual quantization unit on the continuous time axis. The stress-strain state evolution trajectory set includes the stress tensor change curve and strain tensor change curve of each virtual quantization unit over time.

[0051] After obtaining the physical excitation decomposition sequence at the current time t, it is used as an external driving force and input into the fiber optic cable microstructure perturbation quantization model. The model utilizes the state evolution rule function library within each virtual quantization unit, combined with the stress-strain state of the unit at the current time and the boundary interaction information received from neighboring units, to perform a one-step forward time iteration calculation, updating the state of all units. By repeatedly executing the above iterative process on the physical excitation decomposition sequences of multiple consecutive measurement times, the model gradually accumulates and records the stress and strain evolution history of each virtual quantization unit over the entire observation time window, ultimately forming a complete set of stress-strain state evolution trajectories.

[0052] Step S131: Read the values ​​in the stress-strain state parameter memory of all virtual quantization units at the current moment from the perturbation quantization model of the optical cable microstructure, including the current values ​​of the axial stress component, radial stress component and shear strain component of each virtual quantization unit, to form the stress-strain state matrix at the current moment.

[0053] Before performing the first iteration, the model first reads the stress-strain states of all virtual quantized units ID_1 to ID_N_unit at the current time t from their mechanical state memory. Specifically, it reads three state variables for each unit: axial stress component σ_a(i,t), radial stress component σ_r(i,t), and shear strain component γ(i,t). This data is then organized into an N_unit matrix S(t) with 3 columns, ordered by unit index i. The i-th row of the matrix represents the state of unit ID_i, and the three columns correspond to σ_a(i,t), σ_r(i,t), and γ(i,t), respectively. This matrix S(t) serves as the initial baseline for the state evolution at the next time step.

[0054] Step S132: Input the physical excitation decomposition sequence at the current moment as an external driving force into the state evolution rule function library of each virtual quantization unit. For each virtual quantization unit, calculate the temperature change and strain change in its physical excitation decomposition result with the corresponding coefficients in the response sensitivity coefficient matrix of the virtual quantization unit to obtain the estimated value of the stress increment generated by the external excitation in the next time step.

[0055] The physical excitation decomposition sequence at the current time t generated in step S129, i.e., ΔT(i,t) and Δε(i,t) for each element, are input into the state evolution rule function library of the corresponding element ID_i. Within the function library, the stress increment directly caused by external excitation is first calculated. For temperature perturbations, the equivalent stress increment caused by temperature, Δσ_T(i,t) = C_T(i) × ΔT(i,t), is calculated using the element's temperature perturbation response coefficient C_T(i) (from step S114). For strain perturbations, the equivalent stress increment caused by strain, Δσ_S(i,t) = C_S(i) × Δε(i,t), is calculated using the element's strain perturbation response coefficient C_S(i). These two parts are combined to obtain the estimated total stress increment caused by external excitation within time step Δt for this element, Δσ_ext(i,t) = Δσ_T(i,t) + Δσ_S(i,t). The addition here is based on the principle of dimensional consistency. The dimensions of C_T(i) and C_S(i) are both Pascal per unit. The stress increment obtained by multiplying by their respective changes is unified to Pascal.

[0056] Step S133: Call the stress relaxation evolution function in the state evolution rule function library of each virtual quantization unit, input the current value of the axial stress component and the current value of the radial stress component of the virtual quantization unit at the current moment, as well as the material viscoelastic parameters, and calculate the estimated value of the stress attenuation at the next moment due to the internal relaxation effect of the material.

[0057] Simultaneously, the stress relaxation evolution function F_relax from the function library of each unit's state evolution rules is invoked. The inputs to this stress relaxation evolution function are: the axial stress component σ_a(i,t), the radial stress component σ_r(i,t) at the current time t, and the material viscoelastic parameter vector P_visco(i) pre-fixed within the unit. Based on the viscoelastic constitutive model, the function F_relax calculates the natural stress decay over time Δt due to internal stress relaxation phenomena (molecular chain rearrangement, internal friction dissipation, etc.). The output of this stress relaxation evolution function is the estimated stress decay value for the next time step: Δσ_relax(i,t) = F_relax(σ_a(i,t), σ_r(i,t), P_visco(i)). This natural stress decay value also has Pascal dimensions.

[0058] Step S134: Call the strain cumulative evolution function in the state evolution rule function library of each virtual quantization unit, input the current value of the shear strain component of the virtual quantization unit at the current moment and the stress state information received from the boundary condition interface of the adjacent virtual quantization unit, and calculate the estimated value of the increase in strain at the next moment due to the mechanical interaction of adjacent units.

[0059] Then, the strain accumulation evolution function F_acc from the function library of each element's state evolution rule is called. The inputs of this strain accumulation evolution function include: the shear strain component γ(i,t) at the current time t, the equivalent stress τ_external(i,t) acting on the element obtained from the external excitation (this equivalent stress can be calculated by combining the stress increment Δσ_ext(i,t) from step S132 with the element's geometric characteristics), and, crucially, the stress state information received from the boundary condition interface of the adjacent virtual quantized element. Here, the axial stress σ_a(i-1,t) received from the left adjacent element (ID_(i-1)) is passed in through the interface, and its difference from the axial stress of this element constitutes the boundary shear stress τ_boundary(i,t) that drives the accumulation of shear strain. Based on plasticity or creep theory, the function F_acc comprehensively evaluates the amount of irreversible shear strain increase that will occur in time Δt under the current shear strain and driving stress. The output of this function is the estimated increase in the dependent variable at the next time step: Δγ_acc(i,t) = F_acc(γ(i,t), τ_external(i,t), τ_boundary(i,t)).

[0060] Step S135: Subtract the estimated stress decay from the estimated stress increment of each virtual quantization unit to obtain the updated values ​​of the axial stress component and radial stress component at the next moment for that virtual quantization unit. Add the estimated strain increase to the current value of the shear strain component to obtain the updated value of the shear strain component at the next moment for that virtual quantization unit.

[0061] Based on the above calculations, the state of each element ID_i is initially updated. The estimated values ​​of the new axial stress component σ_a(i, t+Δt) and radial stress component σ_r(i, t+Δt) are obtained by adding the stress increment caused by external excitation to the current stress value, and then subtracting the stress decay caused by internal relaxation. Since the relaxation and excitation response mechanisms of axial and radial stresses are similar, a unified stress update formula is used here. In practical applications, tensor decomposition of Δσ_ext can be performed according to the directionality, but the principle is the same: σ_a(i, t+Δt)_pre=σ_a(i, t)+Δσ_ext(i, t)-Δσ_relax(i, t); σ_r(i, t+Δt)_pre=σ_r(i, t)+Δσ_ext(i, t)-Δσ_relax(i, t). The estimated value of the new shear strain component γ(i,t+Δt) is obtained by adding the strain increment generated by the driving stress to the shear strain at the current moment: γ(i,t+Δt)_pre=γ(i,t)+Δγ_acc(i,t). The above estimated value is calculated based on the current state of the element itself and the local excitation, and the closed-loop feedback of stress interaction between elements has not yet been considered.

[0062] Step S136: Based on the stress-strain update estimate of a single time step, perform stress interaction and boundary correction between elements, complete multi-time step iterative calculation, and generate a set of stress-strain state evolution trajectories.

[0063] After obtaining the initial update estimates for all elements, an inter-element interaction and correction step needs to be executed to ensure that the mechanical state is spatially continuous and conforms to physical laws. This is because the update in step S135 is relatively isolated, while the stresses of adjacent elements in an actual optical cable interact. Therefore, immediately after obtaining the initial update values, a sub-process for inter-element stress interaction and boundary correction is executed to correct the initial update values, obtain the final state values, and write them back to memory. Subsequently, steps S131 to S136 are repeated for the physical excitation decomposition sequence at the next time point t+Δt, and the final determined state values ​​at each time point are stored in the historical record, ultimately forming the complete evolution trajectory of each element on a continuous time axis.

[0064] Step S136-1: Through the boundary condition interface of each virtual quantization unit, the updated axial stress component value of this unit is transmitted to the adjacent virtual quantization unit on the right, and the updated axial stress component value is received from the boundary condition interface of the adjacent virtual quantization unit on the left, so as to realize the interactive transmission of stress state between units.

[0065] After calculating the initial axial stress component update value σ_a(i, t+Δt)_pre, each virtual quantization unit ID_i sends this initial value to its right-side adjacent unit ID_(i+1) through the output of its boundary condition interface. Simultaneously, it receives its initial update value σ_a(i-1, t+Δt)_pre from its left-side adjacent unit ID_(i-1) through its input. For the chain-head unit ID_1, its left-side boundary conditions are determined by a preset external environment (e.g., free end, zero stress); for the chain-tail unit ID_N_unit, there are no units to its right, and its output data is discarded or used for boundary monitoring. Through this step, each unit obtains the initial axial stress update values ​​for itself and its left-side adjacent units.

[0066] Step S136-2: Based on the received axial stress component update values ​​transmitted from the left adjacent virtual quantization unit, recalculate the boundary stress gradient of each virtual quantization unit. The boundary stress gradient is the difference between the axial stress component update values ​​of the left adjacent unit and the axial stress component update values ​​of this unit divided by the length of the basic structural unit.

[0067] Each virtual quantization unit ID_i recalculates the boundary stress gradient acting on it using the received preliminary axial stress update value σ_a(i-1, t+Δt)_pre from its left-adjacent unit and its own calculated preliminary axial stress update value σ_a(i, t+Δt)_pre. This boundary stress gradient reflects the rate of change of stress along the optical cable axis and is a key factor driving shear deformation and stress wave propagation in the unit. The gradient value G(i, t+Δt) is calculated as G(i, t+Δt) = (σ_a(i, t+Δt)_pre - σ_a(i-1, t+Δt)_pre) / L_unit. This boundary stress gradient has the dimension of Pascals per meter, representing the stress change per unit length.

[0068] Step S136-3: Feed back the recalculated boundary stress gradient to the strain cumulative evolution function of each virtual quantization unit, correct the estimated increase in strain for the next time step, and obtain the corrected updated value of the shear strain component.

[0069] After calculating the more accurate boundary stress gradient G(i, t + Δt), it is used as the new boundary driving condition and fed back to the element's strain accumulation evolution function F_acc. In fact, the τ_boundary used in step S134 is calculated based on the stress gradient of the previous time step, but now the stress gradient of the current time step is obtained, so the strain increment needs to be corrected. The element calls the correction function F_acc_correct(γ(i, t), G(i, t + Δt), ...), which recalculates the actual increase in shear strain Δγ_acc_correct(i, t) within the time interval Δt based on the new stress gradient. This results in the corrected updated shear strain component value γ(i, t + Δt) = γ(i, t) + Δγ_acc_correct(i, t). Simultaneously, since changes in shear strain react upon axial and radial stresses, the stress values ​​also need to be fine-tuned. For example, σ_a(i, t+Δt) = σ_a(i, t+Δt)_pre - K × Δγ_acc_correct(i, t), where K is a coefficient reflecting stress-strain coupling. After this corrective closed-loop process, the consistent and self-consistent stress-strain state of all elements at time t+Δt is finally determined.

[0070] Step S136-4: Write the updated values ​​of the axial stress component, radial stress component, and corrected shear strain component of all virtual quantization units back to their respective stress-strain state parameter memory, overwriting the original values ​​at the current time, and complete the state evolution iteration of one time step.

[0071] The finalized state values ​​σ_a(i, t+Δt), σ_r(i, t+Δt), and γ(i, t+Δt), after inter-unit interaction correction, are written back to the mechanical state memory of each virtual quantization unit ID_i, overwriting the original σ_a(i, t), σ_r(i, t), and γ(i, t), respectively. At this point, the model has completed a full time step of state evolution iteration from time t to time t+Δt.

[0072] Step S136-5: Drive the physical excitation decomposition sequence corresponding to multiple consecutive sampling time points in sequence. After each iteration of a time step, store the stress and strain state parameters of all virtual quantization units at the corresponding time into the historical record, and finally form a set of stress tensor curves and strain tensor curves of each virtual quantization unit on the continuous time axis.

[0073] The model continuously receives the physical excitation decomposition sequence at subsequent times t+2Δt, t+3Δt, ..., and repeats the iterative process of steps S131 to S136-4. After completing each iteration of a time step and obtaining a new set of state values, the system appends and stores the final state values ​​(σ_a, σ_r, γ) of all N_units at that time as a time slice to a dedicated historical database. Over time, for each virtual quantized unit ID_i, its axial stress component values ​​at different times constitute a time sequence {σ_a(i, t0), σ_a(i, t0+Δt), σ_a(i, t0+2Δt), ...}, which is the axial stress variation curve of that unit over time. Similarly, the radial stress variation curve and the shear strain variation curve over time can be obtained. All the curves of all units together constitute the set of stress-strain state evolution trajectories. This set of stress-strain state evolution trajectories exists in the form of a data structure, for example, it can be a three-dimensional array with dimensions of element index, time index, and state type index, which completely records the micromechanical evolution history of the optical cable from the initial moment to the current moment.

[0074] Step S140: Construct a spatiotemporal correlation network for the propagation of optical cable link disturbances based on the stress-strain state differences between adjacent virtual quantization units in the set of stress-strain state evolution trajectories. The spatiotemporal correlation network uses virtual quantization units as nodes and the rate of change of stress-strain state differences between adjacent nodes as the weight value of the node connection. The spatiotemporal correlation network is used to describe the propagation path and propagation rate of disturbances along the longitudinal direction of the optical cable.

[0075] After obtaining the stress-strain evolution trajectory of each virtual quantization unit on a continuous time axis, the spatial correlation between these trajectories is further analyzed to reveal how disturbance energy and information propagate longitudinally along the optical cable. A spatiotemporal correlation network is constructed, abstracting each virtual quantization unit as a node in the network, while the connecting edges between nodes represent the propagation characteristics of disturbances between adjacent units. The dynamic weights of this spatiotemporal correlation network can quantify the instantaneous efficiency and path of disturbance propagation.

[0076] Step S141: Extract the axial stress component variation curve of each virtual quantization unit on the complete time axis from the set of stress-strain state evolution trajectories to obtain a set of spatially ordered axial stress time series curves. The order of the curves in the sequence is completely consistent with the order of the spatial index identifiers of the virtual quantization units.

[0077] From the set of stress-strain state evolution trajectories generated in step S136-5, the data on the axial stress component changes over time for all virtual quantized units are specifically extracted. For the unit with index i, its axial stress time series curve is denoted as σ_a(i, τ), where τ takes all historical time points t0, t0+Δt, t0+2Δt, ... . The above curves are arranged in the order of spatial index identifiers i=1, 2, ..., N_unit to form a sequence, where each curve describes the stress fluctuation history at the corresponding spatial location.

[0078] Step S142: For each pair of adjacent virtual quantization units, the axial stress time series curve of the left unit and the axial stress time series curve of the right unit are subtracted point by point over time to obtain the time series variation curve of the axial stress difference between the two units. This time series variation curve reflects the stress state transfer effect during the process of the disturbance propagating from the left unit to the right unit.

[0079] For each pair of adjacent virtual quantized units, i.e., units with indices i and i+1, the following operation is performed: At each identical time point τ, the difference between the axial stress value σ_a(i, τ) of the left unit ID_i and the axial stress value σ_a(i+1, τ) of the right unit ID_(i+1) is calculated. This difference Δσ_{i, i+1}(τ) = σ_a(i+1, τ) - σ_a(i, τ) constitutes a new time series, namely the axial stress difference time series curve. This axial stress difference time series curve reflects the change in stress state caused by the perturbation propagating from unit i to unit i+1 at each time point. If the perturbation propagation is smooth, Δσ may exhibit regular fluctuations; if the propagation is hindered, Δσ may show abrupt changes or abnormal accumulation.

[0080] Step S143: Calculate the first derivative of the time-series curve of the axial stress difference for each adjacent unit in the time dimension to obtain the time-series curve of the axial stress difference change rate. The axial stress difference change rate characterizes the rate of evolution of the stress state difference between the two units over time, i.e., the instantaneous rate of disturbance transmission.

[0081] The time-series curve Δσ_{i, i+1}(τ) of the axial stress difference between each pair of adjacent elements obtained in step S142 is differentiated in the time dimension. Since the data is a sequence of discrete time points, the derivative is approximated using the finite difference method. At each time point τ_t, the rate of change R_{i, i+1}(τ_t) = (Δσ_{i, i+1}(τ_t) - Δσ_{i, i+1}(τ_{t-1})) / Δt, where Δt is the time interval between adjacent time points. This yields a new time-series curve R_{i, i+1}(τ), which is the time-series curve of the axial stress difference rate of change. The magnitude of this time-series curve represents the rate of change of the stress state difference between two adjacent elements per unit time. Physically, this can be interpreted as the instantaneous rate at which the disturbance is transmitted from element i to element i+1. A higher rate indicates that the disturbance energy is transmitted faster in that section of the optical cable; if the rate is zero or negative, it may indicate that the disturbance is blocked or reflected at this point.

[0082] Step S144: Take the value of each time point on the time series curve of the axial stress difference change rate of each adjacent unit as the instantaneous weight value of the edge of the node at the corresponding time. Take the virtual quantization unit as the network node, the directed connection between adjacent units as the network edge, and the instantaneous weight value to mark the attribute of the edge, and construct a spatiotemporal correlation network snapshot sequence that evolves dynamically over time.

[0083] Based on the above calculation results, a spatiotemporal correlation network is constructed. Each virtual quantization unit ID_i is defined as a network node. For each pair of adjacent units (i, i+1), a directed edge from node i to node i+1 is defined, representing the direction of perturbation propagation. For each historical time point τ_t, the value R_{i, i+1}(τ_t) at that moment is extracted from the time series curve R_{i, i+1}(τ) obtained in step S143, and it is used as the instantaneous weight w_{i→i+1}(τ_t) = R_{i, i+1}(τ_t) of the directed edge from node i to node i+1 at time τ_t. This operation is repeated for all adjacent unit pairs and time points to obtain a series of network snapshots that change over time. Each snapshot corresponds to a time point τ_t and contains N_unit-1 directed edges, each edge having a weight value w_{i→i+1}(τ_t).

[0084] Step S145: For each time point spatiotemporal correlation network snapshot, the spatial index identifier of the virtual quantization unit is used as the coordinate of the network node. The network nodes are arranged in a one-dimensional straight line. The connection edges between nodes only exist between nodes with adjacent indices, forming a weighted network with a chain-like topology.

[0085] In each network snapshot at any given time point, the node layout is determined by its physical location. Since the fiber optic cable itself is a one-dimensional linear structure, the network topology is fixed as a chain structure. Nodes are logically arranged in a straight line from left to right according to their spatial index identifiers ID_1, ID_2, ..., ID_N_unit. Edges in the network exist only between directly adjacent nodes, i.e., the edge set is {(ID_1, ID_2), (ID_2, ID_3), ..., (ID_{N_unit-1}, ID_N_unit)}. There are no cross-node connections. Thus, the network at each time point is a simple chain-weighted directed graph with N_unit nodes and N_unit-1 edges. The edge weight w_{i→i+1}(τ_t) quantifies the instantaneous efficiency or rate at which a disturbance propagates from unit i to unit i+1 at time τ_t.

[0086] Step S146: Based on the spatiotemporal correlation network snapshot sequence, perform network weight normalization and smoothing optimization to generate the final spatiotemporal correlation network for optical cable link disturbance propagation.

[0087] After obtaining the original dynamic chain-weighted network snapshot sequence, it is further normalized and smoothed to eliminate the influence of measurement noise and transient anomalies, resulting in a more stable and representative spatiotemporal correlation network for subsequent analysis.

[0088] For example, in step S146-1, in the chain-weighted network at each time point, the instantaneous weight values ​​of the nodes are normalized so that all weight values ​​are mapped to the range of zero to one. The normalized weight values ​​represent the relative efficiency of the disturbance propagating from the left node to the right node at the corresponding time. The larger the weight value, the smoother the disturbance propagation.

[0089] For each network snapshot at time point τ_t, collect the weight values ​​{w_{i→i+1}(τ_t)|i=1toN_unit-1} of all its N_unit-1 edges. Find the minimum value w_min(τ_t) and the maximum value w_max(τ_t) among these weights. Then, perform min-max normalization on the weight of each edge to obtain the normalized weight w'_{i→i+1}(τ_t)=(w_{i→i+1}(τ_t)-w_min(τ_t)) / (w_max(τ_t)-w_min(τ_t)). The normalized weight value w'_ is restricted to the interval [0, 1]. A value of 1 represents that at the current time, the disturbance transmission efficiency of this segment of the optical cable is the highest, that is, the stress difference changes the fastest; a value of 0 represents the lowest transmission efficiency, which may mean that this is a node or boundary of disturbance propagation.

[0090] Step S146-2: Stack the chain-weighted network snapshots of multiple consecutive time points in chronological order to construct a three-dimensional spatiotemporal correlation network data structure. The first dimension of this spatiotemporal correlation network data structure is the virtual quantization unit node index, the second dimension is the adjacent node edge index, the third dimension is the time point index, and the data element is the normalized perturbation propagation efficiency value.

[0091] The network snapshots from all time points after normalization are integrated to construct a three-dimensional data structure W. The first dimension (x-axis) of this three-dimensional data structure corresponds to the node index i (from 1 to N_unit), the second dimension (y-axis) corresponds to the edge index (since each edge is uniquely determined by its adjacent nodes (i, i+1), the edge index can also be represented by i, ranging from 1 to N_unit-1), and the third dimension (z-axis) corresponds to the time point index k (corresponding to time τ_k). The data element W(i, k) represents the normalized perturbation propagation efficiency value w'_{i→i+1}(τ_k) on the edge from node i to node i+1 at time τ_k. This three-dimensional array completely describes the evolution of perturbation propagation efficiency in both spatial (node ​​position) and temporal dimensions.

[0092] Step S146-3: Perform spatial smoothing on the three-dimensional spatiotemporal correlation network data structure. For each time point, perform a moving average on the weight values ​​of adjacent edges along the node index direction to eliminate abnormal weight fluctuations caused by local noise and obtain a spatiotemporal distribution map of disturbance transmission efficiency after spatial smoothing.

[0093] To eliminate random noise in the spatial dimension, the three-dimensional data structure W is spatially smoothed. For each fixed time point k (i.e., each layer of network snapshot), a moving average is applied to the weights of adjacent edges along the direction of node index i (i.e., the spatial direction). For example, a three-point moving average can be used: for edge index i (2≤i≤N_unit-2), its smoothed weight W_s(i,k)=(W(i-1,k)+W(i,k)+W(i+1,k)) / 3. For boundary edge indices 1 and N_unit-1, a two-point average is used. After this operation, a new three-dimensional data structure W_s is obtained, in which the data is smoother in the spatial dimension, highlighting large-scale perturbation propagation patterns and suppressing isolated measurement noise points.

[0094] Step S146-4: Perform time smoothing on the spatially smoothed perturbation transfer efficiency spatiotemporal distribution map. For each node connection, perform a moving average of the weight values ​​of consecutive time points along the time axis to eliminate weight abrupt changes caused by instantaneous disturbances, and obtain the time-smoothed stable perturbation transfer efficiency spatiotemporal evolution map.

[0095] Furthermore, the spatially smoothed data W_s is smoothed in the time dimension. For each fixed edge index i (i.e., spatial location), a moving average is applied to the continuous weight values ​​along the time point index k. For example, a three-point moving average can also be used: for time point k (2≤k≤K-1, K is the total number of time points), the smoothed weight W_st(i,k)=(W_s(i,k-1)+W_s(i,k)+W_s(i,k+1)) / 3. For the start and end time points, a two-point average is used. After both temporal and spatial smoothing, the final three-dimensional data structure W_st is obtained, which is the spatiotemporal correlation network of optical cable link disturbance propagation.

[0096] Step S146-5: The spatiotemporal evolution map of the stable disturbance propagation efficiency after time smoothing is used as the final spatiotemporal correlation network of optical cable link disturbance propagation. This spatiotemporal correlation network of optical cable link disturbance propagation graphically shows the spatiotemporal distribution characteristics of the disturbance energy along the longitudinal direction of the optical cable, the change in propagation speed, and the propagation efficiency.

[0097] The final generated W_st, the spatiotemporal correlation network of optical cable link disturbance propagation, is a highly structured data model. It can be visualized as a two-dimensional image or dynamic video, where the horizontal axis represents the longitudinal position of the optical cable (node ​​or edge index), the vertical axis represents time, and the color or brightness represents the normalized disturbance propagation efficiency W_st(i, k). This figure clearly reveals: from which spatial point the disturbance energy begins to appear (the color brightens), at what speed (the slope of the longitudinal stripes) and in which direction it propagates along the optical cable, at which locations the propagation efficiency changes abruptly (the color changes drastically), and how these propagation patterns evolve over time.

[0098] Step S150: Perform network flow characteristic analysis on the spatiotemporal correlation network, extract the set of early warning nodes in the network nodes whose stress and strain states have accumulated to exceed the critical threshold, calculate the evolution trend of stress and strain states of each node in the early warning node set within the future time window, and generate serialized optical cable performance anomaly early warning information containing the location coordinates of the early warning nodes and the predicted values ​​of stress and strain states based on the evolution trend.

[0099] Finally, based on the constructed spatiotemporal correlation network of optical cable link disturbance propagation and the original set of stress-strain state evolution trajectories, in-depth network flow characteristic analysis and trend prediction are conducted. By analyzing whether the "flow" (i.e., stress accumulation) of nodes in the network exceeds their carrying capacity, potential weak links and risk points are identified. Furthermore, the future state of these risk points is extrapolated, ultimately generating forward-looking, quantitative early warning information to guide data center operations and maintenance personnel to take early intervention measures.

[0100] Step S151: Extract the axial stress component variation curve of each virtual quantization unit node at all time points from the spatiotemporal correlation network to obtain the stress history trajectory of each node. The stress history trajectory records all stress state changes experienced by the node from the initial moment to the current moment.

[0101] From the set of stress-strain state evolution trajectories generated in step S136-5, the complete axial stress time series data {σ_a(i, τ)} for each virtual quantization unit ID_i is extracted again. This curve is the stress history trajectory of this node, which completely depicts the entire historical fluctuation of the axial tensile and compressive stress borne by this micro-telescopic cable segment from the beginning of monitoring to the current moment, and is the basis for fatigue damage assessment.

[0102] Step S152: Perform cumulative damage calculation on the stress history trajectory of each node, use the rainflow counting method to identify and extract stress cycles in the trajectory, count the number of stress cycles of different amplitudes experienced by each node within the historical time window, and form a stress cycle amplitude distribution histogram.

[0103] The rainflow counting method is applied to the stress history trajectory σ_a(i, τ) of each node ID_i. This method simplifies a complex, random stress-time history into a series of full and semi-cycles with different amplitudes and means. The algorithm scans the stress curve segment by segment, identifies closed stress-strain hysteresis loops, and records the stress amplitude Δσ_cycle and mean value corresponding to each loop. For node ID_i, after rainflow counting, a set of cycle information is obtained: for the m-th cycle, its stress amplitude is Δσ_cycle(i, m). The amplitudes of all cycles are counted, and the amplitude range is divided into several intervals (e.g., 5 MPa per interval). The number of cycles falling into each interval is counted. Finally, a histogram H_i(Δσ_bin) of the stress cycle amplitude distribution of the node is formed, where Δσ_bin represents a certain amplitude interval, and the value of H_i represents the number of cycles occurring within that amplitude interval.

[0104] Step S153: Based on the fatigue characteristic curve of the optical cable material, map each stress cycle amplitude in the stress cycle amplitude distribution histogram of each node to the amount of micro-damage caused to the material by that cycle, and sum up the micro-damage of all stress cycles to obtain the cumulative fatigue damage value of the node.

[0105] Consulting the SN curve (stress-life curve) of this type of optical cable material, this curve describes the number of cycles N_f(Δσ) that the material can withstand under a specific stress amplitude Δσ. According to the Mainner linear cumulative damage criterion, the damage caused by a cycle with a stress amplitude of Δσ is D_cycle = 1 / N_f(Δσ). For node ID_i, traversing each amplitude interval Δσ_bin in its stress cycle amplitude distribution histogram H_i, the damage caused by each cycle within this interval is 1 / N_f(Δσ_bin). The total damage caused by this interval is D_bin = H_i(Δσ_bin) × (1 / N_f(Δσ_bin)). Summing up the damage D_bin from all amplitude intervals, we obtain the total cumulative fatigue damage value D_total(i) = Σ_{all Δσ_bin}D_bin for this node from the initial time to the current time. D_total(i) is a dimensionless cumulative quantity; theoretically, fatigue failure occurs when it reaches 1.

[0106] Step S154: Preset a critical damage threshold for each virtual quantization unit node. The critical damage threshold is determined by the ultimate tensile strength of the optical cable material and the design safety factor. Compare the cumulative fatigue damage value of each node with the critical damage threshold of that node.

[0107] Based on the design standards and material properties of the optical cable, a safe critical damage threshold D_critical is set. This value is typically much less than 1, for example, 0.1 or 0.2. It is calculated by dividing the ultimate tensile strength by a safety factor (e.g., 10) and then mapping it to the corresponding lifetime N_f_safe on the SN curve. Therefore, D_critical = 1 / N_f_safe. The cumulative fatigue damage value D_total(i) of each node calculated in step S153 is compared with the preset critical damage threshold D_critical.

[0108] Step S155: When the cumulative fatigue damage value of a certain virtual quantization unit node exceeds its critical damage threshold, the node is marked as a potential risk node, and the spatial index identifier, the current cumulative fatigue damage value, and the specific time point when the threshold is exceeded are recorded.

[0109] For any node ID_i, if D_total(i) > D_critical, then the optical cable micro-segment represented by this node is determined to have entered the fatigue accumulation warning zone, posing a high risk of failure. This node is marked as a potential risk node, and a risk record R_i is generated, containing the fields: node index i, current cumulative damage value D_total(i), and the time point τ_exceed when the threshold is first exceeded.

[0110] Step S156: Perform spatial clustering analysis on all virtual quantization units marked as potential risk nodes to detect whether multiple potential risk nodes that are consecutively adjacent in spatial location constitute a risk node cluster. If the number of consecutively adjacent nodes exceeds the preset cluster size threshold, the entire cluster is marked as a cumulative damage exceeding the standard segment.

[0111] Iterate through the set of all nodes marked as potential risk nodes, checking if their spatial index identifiers are consecutive. For example, nodes ID_10, ID_11, ID_12, and ID_13 are found to be potential risk nodes, forming a consecutive cluster of length 4. A cluster size threshold is preset, for example, 3. Since the number of nodes in this cluster, 4, is greater than the threshold 3, the consecutive fiber optic cable segment covered by ID_10 to ID_13 is marked as a "cumulative damage exceeding the limit segment." The starting index of this segment is 10, and the ending index is 13. This step helps identify potential, large-scale weakened areas, rather than just isolated points.

[0112] Step S157: Extract the curves of shear strain components of all nodes in each cumulative damage exceeding the standard section from the spatiotemporal correlation network as a function of time, and calculate the slope of the rate of change of each curve in the most recent time window. The slope of the rate of change reflects the deterioration rate of the strain state of the node. The larger the value of the slope of the rate of change, the more severe the deterioration of the strain state.

[0113] For each cumulative damage exceeding the limit segment identified in step S156, such as segment [10, 13], extract the time-series curves γ(i, τ) of the shear strain components of all nodes ID_10, ID_11, ID_12, and ID_13 within that segment. Take the most recent time window, such as the period of the most recent 24 hours [τ_now-T_win, τ_now]. Perform linear regression fitting on the data of each curve within this time window to obtain a straight line y=slope_i×τ+intercept_i. The slope of this line, slope_i, is the slope of the strain change rate of that node within the most recent time window, in units of per second (or per sampling period). A positive slope_i indicates that the strain is continuously increasing; the larger the value, the faster the rate of strain deterioration.

[0114] Step S158: Sort the slope of the rate of change of all nodes in the section where the cumulative damage exceeds the standard, and select the top few nodes with the largest slope of the rate of change as key early warning nodes. The key early warning nodes represent the local locations where the strain state deteriorates most drastically in the section where the cumulative damage exceeds the standard.

[0115] Sort the slopes of the strain rate of change of all nodes in the section with excessive cumulative damage from largest to smallest. For example, the slope sorting result in the section [10, 13] is slope_12>slope_10>slope_13>slope_11. Preset a number K (e.g., K=2), select the K nodes with the largest slopes, namely ID_12 and ID_10, as the "critical early warning nodes" in this section. These nodes are the local weak points on the optical cable with the highest stress concentration, the fastest damage accumulation, and the highest probability of being the first to break.

[0116] Step S159: Map the spatial index identifier of each key early warning node back to the actual physical location coordinates of the optical cable to obtain the actual geographical location information of the key early warning node, including the distance of the key early warning node from the starting end of the optical cable and the specific laying environment section in which it is located.

[0117] Using the spatial index and physical location mapping table established in step S112, the index ID_i of the critical early warning node is converted into actual physical coordinates. For the critical early warning node ID_12, the physical range it covers is found to be [Pos_start_12, Pos_end_12]. Its center position or starting point position Pos_start_12 is the distance of the node from the beginning of the optical cable. Simultaneously, by combining the data center's GIS (Geographic Information System) or structured cabling drawings, the specific environment of this location can be determined, such as "located at the bend of the overhead cable tray between computer room A and computer room B" or "at the entrance through the underground cable tray." All of the above geographical environment information is recorded.

[0118] Step S1510: Combine the spatial index identifier, actual geographical location information, current cumulative fatigue damage value, and slope of strain state change rate of the key early warning node to form an early warning node information entry. All early warning node information entries together constitute the early warning node set.

[0119] A structured warning information entry is generated for each critical warning node. For example, for node ID_12, the entry includes: node index = 12, geographical location = "12 meters from the starting point Pos_start, located at the bend of the cable tray at Exit A of the equipment room", current cumulative damage value = D_total(12), and strain deterioration rate = slope_12. The entries for all critical warning nodes are aggregated to form a set, namely the warning node set. This warning node set directly points to the most dangerous and urgent microscopic locations on the optical cable.

[0120] Step S1511: Calculate the evolution trend of stress and strain state of each node in the early warning node set within the future time window, and generate serialized optical cable performance anomaly early warning information containing the location coordinates of the early warning node and the predicted values ​​of stress and strain state based on the evolution trend.

[0121] For example, in step S15111, extract the axial stress component time series data and shear strain component time series data of each early warning node within the historical time window from the set of stress-strain state evolution trajectories. The length of the historical time window includes at least a number of recent complete stress cycle cycles.

[0122] For each node ID_i in the early warning node set, extract its most recent axial stress time series data σ_a(i, τ) and shear strain time series data γ(i, τ) from the historical database. The selected historical time window length T_history should be long enough to ensure that at least several complete stress fluctuation cycles are captured, thus enabling meaningful spectral analysis and trend extrapolation. For example, if the main stress fluctuation cycle of the optical cable is 24 hours (varying with data center temperature control or business load), then T_history should be no less than 72 hours.

[0123] Step S15112: Perform spectral analysis on the time series data of the axial stress components of each early warning node, extract the main frequency components and corresponding amplitudes of the stress fluctuations at that node through Fourier transform, and identify the dominant period and energy distribution characteristics of the stress fluctuations.

[0124] A Fast Fourier Transform (FFT) is performed on the historical axial stress time-series data σ_a(i, τ) of node ID_i. This converts the time-domain signal to a frequency-domain signal, resulting in a spectrum. The horizontal axis of the spectrum represents frequency f (period T = 1 / f), and the vertical axis represents the amplitude A(f) of the corresponding frequency component. The frequency components f1, f2, ... with the largest amplitudes are identified from the spectrum; their corresponding periods T1, T2, ... are the dominant periods of stress fluctuation. For example, significant amplitudes of the 24-hour and 12-hour periodic components may be observed, indicating that stress fluctuations are mainly driven by daytime temperature control and changes in workload. This information will be used to construct a predictive model.

[0125] Step S15113: Based on the identified dominant cycle and fluctuation energy distribution characteristics, construct a stress fluctuation extrapolation model for each early warning node. The stress fluctuation extrapolation model adopts the autoregressive moving average method and uses historical stress data to predict the stress values ​​at several future time points.

[0126] Based on the spectral analysis results of step S15112, a time series prediction model can be constructed for each node. For example, an autoregressive moving average model can be used. The model identification process uses historical data σ_a(i, τ) to estimate the model parameters (autoregressive order p and moving average order q, and the corresponding coefficients). After the model is established, by inputting the stress values ​​of the past p times, the stress value σ_a_pred(i, τ_now + Δt) at the next time moment can be recursively predicted. Through iteration, a series of predicted values ​​{σ_a_pred(i, τ_now + Δt), σ_a_pred(i, τ_now + 2Δt), ..., σ_a_pred(i, τ_now + T_future)} can be obtained within a future time window T_future.

[0127] Step S15114: Perform trend decomposition on the time series data of shear strain components of each early warning node, decomposing the strain time series data into a long-term trend term and a periodic fluctuation term. The long-term trend term reflects the slow accumulation process of strain, and the periodic fluctuation term reflects the periodic fluctuation of strain.

[0128] The historical shear strain time series data γ(i, τ) of node ID_i is decomposed. For example, the STL (Seasonal Trend Decomposition) method can be used, which decomposes the original time series γ(i, τ) into three parts: a long-term trend term Trend(i, τ), a periodic fluctuation term Seasonal(i, τ), and a residual term Residue(i, τ). That is, γ(i, τ) = Trend(i, τ) + Seasonal(i, τ) + Residue(i, τ). Here, the Trend term reflects the slow but continuous increase in strain due to irreversible processes such as creep; the Seasonal term reflects the reversible strain fluctuations caused by periodic changes in temperature or load.

[0129] Step S15115: Perform linear regression fitting on the long-term trend term of each early warning node to obtain the slope and intercept parameters of the trend line. Extrapolate the values ​​of the long-term trend term within the future time window based on the slope and intercept parameters to obtain the future strain benchmark value sequence.

[0130] Linear regression is performed on the decomposed long-term trend term Trend(i, τ) within the historical time window to fit a straight line Trend(i, τ) = k_trend_i × τ + b_trend_i. The slope k_trend_i represents the average rate of long-term strain accumulation. Using this linear equation, the long-term trend value Trend_pred(i, τ') = k_trend_i × τ' + b_trend_i is predicted for each time point τ' within the future time window T_future. This yields a set of future strain baseline values.

[0131] Step S15116: Perform phase and amplitude analysis on the periodic fluctuation term of each early warning node, extract the current phase and average amplitude of the periodic fluctuation term, assume that the phase and amplitude of the periodic fluctuation term remain stable within the future time window, and generate a prediction sequence of the future periodic fluctuation term.

[0132] The decomposed periodic fluctuation term Seasonal(i, τ) is analyzed. Its waveform template within a complete cycle (e.g., 24 hours) is extracted. The current phase φ_now is obtained. It is assumed that this periodic fluctuation will continue to maintain the same waveform, amplitude, and period within the future time window T_future. Therefore, the predicted value of the periodic fluctuation term Seasonal_pred(i, τ') at the future time τ' is equal to the Seasonal value of the historical phase at the same time. For example, if the period is 24 hours, predicting the periodic fluctuation within the next 24 hours is simply repeating the waveform of the most recent cycle.

[0133] Step S15117: Add the future strain benchmark value sequence and the future periodic fluctuation term prediction sequence point by point to obtain the shear strain component prediction time series data of each early warning node within the future time window. This shear strain component prediction time series data includes the combined effects of long-term accumulation and periodic fluctuation.

[0134] The two sets of prediction sequences obtained in steps S15115 and S166 are added together point by point, while the remainder can be ignored (considered as noise). The shear strain prediction sequence γ_pred(i, τ') = Trend_pred(i, τ') + Seasonal_pred(i, τ') for node ID_i in the future time window T_future is obtained, where each τ' belongs to the future time window.

[0135] Step S15118: Input the axial stress component prediction time series data and shear strain component prediction time series data of each early warning node into the cumulative damage calculation formula. The cumulative damage calculation formula is based on the rainflow counting method and the material fatigue characteristic curve to predict the amount of additional damage that may occur in the future time window.

[0136] The axial stress prediction sequence σ_a_pred(i, τ') at node ID_i is considered as the future stress load spectrum. Rainflow counting is applied to perform cyclic counting on this predicted load spectrum, obtaining a series of predicted stress cycle amplitudes. Combined with the material's SN curve, the damage caused by each predicted cycle is calculated. The damage from these predicted cycles is accumulated to obtain the expected cumulative fatigue damage ΔD_pred(i) within the future time window T_future.

[0137] Step S15119: Add the current accumulated fatigue damage value of each early warning node to the estimated additional damage amount within the future time window to obtain the predicted accumulated fatigue damage value of the node at the end of the future time window. The predicted accumulated fatigue damage value represents the degree of damage that may be reached in the future.

[0138] The current cumulative fatigue damage value of node ID_i is D_total(i) (from step S153). Then, the predicted cumulative fatigue damage value at the end of the future time window T_future is D_total_pred(i) = D_total(i) + ΔD_pred(i).

[0139] Step S151110: Based on the ratio of the cumulative fatigue damage prediction value of each early warning node at the end of the future time window to the critical damage threshold of the early warning node, generate the damage exceeding risk coefficient of the node in the future time window. The damage exceeding risk coefficient is used to quantify the probability that the future damage will exceed the critical threshold.

[0140] The risk coefficient Risk(i) is calculated as Risk(i) = D_total_pred(i) / D_critical. If Risk(i) > 1, it indicates that the cumulative damage to this node will exceed the safety threshold within a future time window, resulting in an extremely high risk of failure. The value of Risk(i), such as 1.5 or 2.0, represents the severity of the exceedance. Finally, a serialized optical cable performance anomaly warning message is generated. This message contains multiple entries, each corresponding to a key warning node, including: the node's geographical coordinates (obtained by index mapping), the current cumulative damage value D_total(i), the predicted future damage value D_total_pred(i), the risk coefficient Risk(i), and the predicted stress-strain evolution trend curve data points. This information is sorted from high to low risk coefficients to form a priority-based warning report, which is provided to the data center operation and maintenance management system to facilitate timely maintenance or reinforcement measures, preventing service losses due to optical cable interruptions.

[0141] For example, after driving the virtual quantization units inside the optical cable microstructure perturbation quantization model to perform state evolution iteration calculations according to the physical excitation decomposition sequence to obtain the set of stress-strain state evolution trajectories of each virtual quantization unit on a continuous time axis, the method further includes:

[0142] Step S210: Extract the stress-strain state parameters of each virtual quantization unit from the set of stress-strain state evolution trajectories and perform optical cable microstructure disturbance energy spectrum construction processing to generate an optical cable microstructure disturbance energy spatiotemporal spectrum containing the disturbance energy accumulation density and disturbance energy propagation flux at each spatial location point in the longitudinal direction of the optical cable.

[0143] Based on the generated set of stress-strain state evolution trajectories, a spatiotemporal spectrum of disturbance energy of optical cable microstructure can be further constructed to observe the accumulation and flow of disturbances more intuitively from an energy perspective.

[0144] Step S211: Extract the axial stress component variation curve and shear strain component variation curve of each virtual quantization unit on the complete time axis from the stress-strain state evolution trajectory set. Multiply the axial stress component value of each virtual quantization unit at each time point with the shear strain component value at that time point to obtain the instantaneous value of the unit volume strain energy density of the virtual quantization unit at that time point. Arrange the instantaneous values ​​of the unit volume strain energy density of all virtual quantization units at all time points according to the spatial index identifier order and the time point order to form the initial spatiotemporal distribution matrix of strain energy density.

[0145] For each virtual quantization unit ID_i and each time point τ_t, its unit volume strain energy density u(i,t) is approximately equal to the product of axial stress σ_a(i,t) and axial strain. However, the model already includes shear strain γ; a more refined model could consider shear strain energy. For simplification, we assume the main energy is related to axial stress and shear strain, and let u(i,t) = σ_a(i,t) × γ(i,t). This is an approximation used to characterize the relative change in energy density. For all i = 1...N_unit and all t = 1...T, calculate u(i,t). Then construct an N_unit matrix with T columns, where the element U(i,t) = u(i,t). This matrix is ​​the initial spatiotemporal distribution matrix of the strain energy density.

[0146] Step S212: Perform spatial gradient calculation on the strain energy density spatiotemporal distribution matrix. Calculate the difference in instantaneous strain energy density per unit volume between adjacent virtual quantization units at each time point along the spatial dimension. Divide the difference by the physical distance between adjacent virtual quantization units to obtain the spatial gradient value of the strain energy density of the corresponding optical cable section at that time point. Arrange the spatial gradient values ​​of the strain energy density of all adjacent sections at all time points according to the spatial section order and the time point order to form the spatiotemporal distribution matrix of the strain energy density spatial gradient.

[0147] For each fixed time point t, along the spatial dimensions i=1...N_unit-1, calculate the strain energy density difference ΔU(i,t)=U(i+1,t)-U(i,t) between adjacent elements. Divide this difference by the element spacing L_unit to obtain the spatial gradient of strain energy density Grad_x(i,t)=ΔU(i,t) / L_unit. Grad_x(i,t) represents the steepness of the energy density change between positions i and i+1 at time t. After calculation for all i and t, a matrix Grad_x with (N_unit-1) rows and T columns is obtained, which is the spatiotemporal distribution matrix of the strain energy density spatial gradient.

[0148] Step S213: Perform time gradient calculation on the strain energy density spatiotemporal distribution matrix, calculate the difference in instantaneous values ​​of strain energy density per unit volume between adjacent time points in each spatial location row along the time dimension, divide the difference by the time interval between adjacent time points to obtain the strain energy density time change rate value in the corresponding time interval at that spatial location point, and arrange the strain energy density time change rate values ​​of all time intervals of all spatial location points in order of spatial location and time interval to form a strain energy density time change rate spatiotemporal distribution matrix.

[0149] For each fixed spatial location i, along the time dimension t=1...T-1, calculate the strain energy density difference ΔU(i,t)=U(i,t+1)-U(i,t) between adjacent time points. Divide this difference by the time interval Δt to obtain the strain energy density time-varying rate of change Grad_t(i,t)=ΔU(i,t) / Δt. Grad_t(i,t) represents the rate at which the energy density increases or decreases with time at location i. After calculation for all i and t, an N_unit matrix with (T-1) columns is obtained, which is the spatiotemporal distribution matrix of the strain energy density time-varying rate of change.

[0150] Step S214: Construct a continuity equation for the propagation of disturbance energy based on the spatiotemporal distribution matrix of the spatial gradient of strain energy density and the spatiotemporal distribution matrix of the time rate of change of strain energy density. Solve the continuity equation to obtain the flux vector of the propagation of disturbance energy at each time point in each spatial interval.

[0151] In continuum mechanics, energy conservation is described by the continuity equation: the rate of decrease in energy density equals the divergence of energy flux. In the one-dimensional case, we have... Where q is the energy flux (energy passing through a unit cross-section per unit time). Based on this, the flux q can be deduced using discrete Grad_t and Grad_x. , Therefore, q(x,t) can be obtained by... The flux is obtained through spatial integration. In the discrete form, q(i,t) is recursively calculated starting from the boundary. Boundary conditions are set, for example, at the starting point i=1 of the optical cable, the flux q(1,t)=0 or is determined by external input. Then, for i from 2 to N_unit, we have q(i,t)=q(i-1,t)-Grad_t(i-1,t)×L_unit. Thus, the energy flux at the boundary of each spatial interval can be calculated.

[0152] Step S215: Perform cumulative integration on the time axis on the instantaneous value of unit volume strain energy density of each virtual quantization unit at all time points to obtain the total cumulative strain energy of each virtual quantization unit from the initial time to the current time. Arrange the total cumulative strain energy of all virtual quantization units in order according to the spatial index identifier to form a disturbance energy cumulative density distribution curve distributed along the longitudinal direction of the optical cable.

[0153] For each virtual quantization unit ID_i, its strain energy density u(i,t) at all time points is accumulated over time, i.e., summed and multiplied by the time interval Δt to obtain the total accumulated strain energy of that unit. Plotting E_acc(i) in the order of i=1...N_unit yields the cumulative density distribution curve of the disturbance energy. The vertical axis of this curve represents the energy accumulated per unit volume, and the horizontal axis represents the longitudinal position of the optical cable. The peak points on the curve represent the locations where the disturbance energy has historically accumulated the most.

[0154] Step S216: Perform cumulative integration on the spatial axis on the disturbance energy propagation flux vector of each spatial interval at all time points to obtain the total amount of disturbance energy that has passed through each spatial interval from the starting end of the optical cable to the location of the spatial interval. Arrange the total amount of energy passed through all spatial intervals in the order of the spatial intervals to form a disturbance energy cumulative flux distribution curve distributed along the longitudinal direction of the optical cable.

[0155] For each edge (i.e., the interface between adjacent units), its energy flux q(i,t) (representing the energy flowing from unit i to i+1) has been calculated in step S214. For each spatial location i (representing the interface between unit i and i+1), the flux q(i,t) at all time points is accumulated over time to obtain the total energy E_flux(i) = Σ_{t=1toT}q(i,t) × Δt from the beginning to the current time (note that q itself is the power density, multiplied by time to obtain the energy density). Then, starting from the beginning of the optical cable, E_flux(i) is spatially accumulated. The accumulated flux from the beginning to location i is defined as C_flux(i) = Σ_{k=1toi}E_flux(k). Plotting C_flux(i) as a curve in the order of i=1...N_unit-1 yields the distribution curve of the accumulated flux of the disturbance energy. The cumulative flux distribution curve of the disturbance energy reflects the total amount of energy that accumulates as the disturbance propagates through each cross section. Steep rises in the curve indicate a large flow of energy, while flatter curves indicate a smaller flow of energy.

[0156] Step S217: The cumulative density distribution curve of the disturbance energy and the cumulative flux distribution curve of the disturbance energy are fused together to draw a dual vertical axis joint distribution map. The disturbance energy accumulation area is identified according to the local peak position of the cumulative density curve in the joint distribution map, and the disturbance energy propagation obstruction area is identified according to the local steep rise position of the cumulative flux curve. The spatial coordinate range of the identified accumulation area and obstruction area is marked to form a list of abnormal disturbance energy sections of the optical cable microstructure.

[0157] Plot the cumulative density distribution curve obtained in step S215 and the cumulative flux distribution curve obtained in step S216 on the same graph, sharing the horizontal axis (optical cable location), but using two different vertical axes, left and right. Analyze this joint distribution graph. In a certain location interval, if the cumulative density curve shows a local peak (indicating energy accumulation at this location), while the cumulative flux curve slows down or flattens at this location and downstream (indicating that energy is difficult to continue propagating forward), then this interval is very likely to be a point of obstruction and accumulation of disturbance energy propagation, i.e., an abnormal segment. Identify all continuous intervals that meet the criteria of "local peak in cumulative density" and "decreasing growth rate of cumulative flux," and record their start and end coordinates (e.g., from Pos_start_a to Pos_end_b) to form a list of abnormal segments of optical cable microstructure disturbance energy, which complements the set of early warning nodes.

[0158] In one exemplary embodiment, an intelligent prediction system for optical cable performance in data centers is provided. This system can be a terminal, server, etc., and its internal structure diagram can be as follows: Figure 2As shown, this intelligent prediction system for fiber optic cable performance in data centers includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides the environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements an intelligent prediction method for fiber optic cable performance in data centers. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, or a button, trackball, or touchpad set on the shell of the intelligent prediction system for optical cable performance in data centers, or an external keyboard, touchpad, or mouse, etc.

[0159] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for intelligent prediction of optical cable performance in data centers, characterized in that, The method includes: A quantum model of perturbation of optical cable microstructure is constructed. The quantum model of perturbation of optical cable microstructure includes multiple virtual quantization units distributed along the longitudinal direction of optical cable. Each virtual quantization unit corresponds to a basic structural unit inside the optical cable. Each virtual quantization unit encapsulates the real-time stress-strain state parameters of the basic structural unit and the response sensitivity coefficient of the basic structural unit to external perturbation. The real-time collected Brillouin scattering spectrum frequency shift data stream along the optical cable is input into the optical cable microstructure perturbation quantization model for perturbation source analysis processing, generating a physical excitation decomposition sequence for each virtual quantization unit at the current moment. Each element in the physical excitation decomposition sequence contains the temperature perturbation component and strain perturbation component obtained from the analysis of the position of the corresponding virtual quantization unit. The virtual quantization unit inside the optical cable microstructure perturbation quantization model is driven to perform state evolution iteration calculation according to the physical excitation decomposition sequence, so as to obtain the stress-strain state evolution trajectory set of each virtual quantization unit on the continuous time axis. The stress-strain state evolution trajectory set includes the stress tensor change curve and strain tensor change curve of each virtual quantization unit over time. Based on the stress-strain state differences between adjacent virtual quantization units in the stress-strain state evolution trajectory set, a spatiotemporal correlation network for optical cable link disturbance propagation is constructed. The spatiotemporal correlation network uses virtual quantization units as nodes and the rate of change of stress-strain state differences between adjacent nodes as the weight value of the node connection. The spatiotemporal correlation network is used to describe the propagation path and propagation rate of disturbance along the longitudinal direction of the optical cable. The spatiotemporal correlation network is subjected to network flow characteristic analysis and processing to extract the set of early warning nodes in the network nodes whose stress and strain states have accumulated to exceed the critical threshold. The evolution trend of stress and strain states of each node in the early warning node set within the future time window is calculated. Based on the evolution trend, serialized optical cable performance anomaly early warning information containing the location coordinates of the early warning nodes and the predicted values ​​of stress and strain states is generated.

2. The intelligent prediction method for optical cable performance in data centers according to claim 1, characterized in that, The construction of the fiber optic cable microstructure perturbation quantization model includes: The internal physical structure of the data center optical cable is divided longitudinally into a continuous sequence of basic structural units. The length of each basic structural unit matches the microstructural feature scale of the optical cable material. Each basic structural unit in the sequence corresponds to a virtual quantization unit. The total number of virtual quantization units is obtained by dividing the total length of the optical cable by the length of the basic structural unit. Each virtual quantization unit is assigned a unique spatial index identifier, which increments in the direction of the optical cable from the start end to the end. Each spatial index identifier is mapped to the start coordinates and end coordinates of the actual physical section of the optical cable covered by the virtual quantization unit, forming a correspondence table between spatial index and physical location. A mechanical state memory is initialized within each virtual quantization unit. The mechanical state memory contains three storage areas: the first storage area is used to store the axial stress component value of the virtual quantization unit at the current moment; the second storage area is used to store the radial stress component value of the virtual quantization unit at the current moment; and the third storage area is used to store the shear strain component value of the virtual quantization unit at the current moment. Within each virtual quantization unit, a response sensitivity coefficient matrix is ​​initialized. The response sensitivity coefficient matrix includes a temperature perturbation response coefficient and a strain perturbation response coefficient. The temperature perturbation response coefficient represents the equivalent stress change caused by a unit temperature change at the virtual quantization unit, and the strain perturbation response coefficient represents the equivalent stress change caused by a unit mechanical strain at the virtual quantization unit. Within each virtual quantization unit, a state evolution rule function library is established. The state evolution rule function library contains stress relaxation evolution function and strain accumulation evolution function. The stress relaxation evolution function calculates the stress decay at the next moment based on the current stress state and material viscoelastic parameters. The strain accumulation evolution function calculates the strain increase at the next moment based on the current strain state and external excitation intensity. Based on the virtual quantization unit that has completed the core attribute configuration, the boundary interaction interface and timing control module are configured, the chain model is assembled, and parameter initialization and consistency verification are completed.

3. The intelligent prediction method for optical cable performance in data centers according to claim 2, characterized in that, The configuration boundary interaction interface and timing control module assemble the chain model and complete parameter initialization and consistency verification, including: A boundary condition interface is set for each virtual quantization unit. The boundary condition interface is used to receive stress and strain state information transmitted from adjacent virtual quantization units. The boundary condition interface includes an input end and an output end. The input end receives the stress state information of the virtual quantization unit on the left and the output end sends the stress state information of the current unit to the virtual quantization unit on the right. A time step control module is configured inside each virtual quantization unit. The time step control module controls the calling frequency of the state evolution rule function library according to the preset simulation time step parameter. The simulation time step parameter is consistent with the acquisition time interval of the Brillouin scattering spectrum frequency shift data stream. The initial stress and strain state parameter memory inside each virtual quantization unit is initialized. The initial value of the axial stress component of all virtual quantization units is set to the pre-stress value when the optical cable leaves the factory, the initial value of the radial stress component is set to the equivalent stress value corresponding to the ambient atmospheric pressure, and the initial value of the shear strain component is set to zero. All virtual quantization units are chained together in the order of their spatial index identifiers. The boundary condition interfaces between adjacent virtual quantization units are interconnected to form a chain-like virtual quantization unit sequence model. The completed chain-like virtual quantization unit sequence model is then subjected to consistency verification. This involves checking whether the response sensitivity coefficient matrix of each virtual quantization unit is within a preset reasonable range and whether the connection of the boundary condition interfaces between adjacent virtual quantization units is correct. After verification, the model parameters are fixed and the model is ready to receive external input data.

4. The intelligent prediction method for optical cable performance in data centers according to claim 1, characterized in that, The process involves inputting the real-time acquired Brillouin scattering spectrum frequency shift data stream along the optical cable into the optical cable microstructure perturbation quantization model for perturbation source analysis, generating a physical excitation decomposition sequence for each virtual quantization unit at the current moment, including: Receive raw data stream of frequency shift of Brillouin scattering spectrum along the optical cable continuously collected by Brillouin optical time domain reflectometer. The raw data stream contains Brillouin spectrum curves at each spatial sampling point. Each Brillouin spectrum curve consists of a frequency value and a corresponding scattered light intensity value. Peak detection processing is performed on the Brillouin spectrum curve at each spatial sampling point location. The frequency value corresponding to the maximum value of scattered light intensity is found as the Brillouin frequency shift center frequency at that spatial sampling point location, thus obtaining the Brillouin frequency shift center frequency curve distributed along the longitudinal direction of the optical cable. The reference Brillouin frequency shift center frequency curve of the optical cable under the condition of no external disturbance is read from the pre-stored optical cable reference database. The reference Brillouin frequency shift center frequency curve is subtracted from the real-time acquired Brillouin frequency shift center frequency curve point by point to obtain the Brillouin frequency shift change curve distributed along the longitudinal direction of the optical cable. The Brillouin frequency shift change curve is segmented and mapped according to the spatial index identifier of the virtual quantization unit. Each virtual quantization unit corresponds to a segment of optical cable physical interval. The average value of the Brillouin frequency shift change of all spatial sampling points in the optical cable physical interval is taken as the representative value of the overall Brillouin frequency shift change of the virtual quantization unit. Based on the linear coupling relationship between Brillouin frequency shift change and temperature and strain, a frequency shift change decomposition equation is established for each virtual quantization unit. The frequency shift change decomposition equation includes a first frequency shift component caused by temperature perturbation and a second frequency shift component caused by strain perturbation. The sum of the two components is equal to the overall Brillouin frequency shift change representative value of the virtual quantization unit. The response sensitivity coefficient matrix encapsulated within each virtual quantization unit is invoked, and the temperature perturbation response coefficient and strain perturbation response coefficient of the virtual quantization unit are extracted from the response sensitivity coefficient matrix. The temperature perturbation response coefficient represents the amount of Brillouin frequency shift caused by a unit temperature change, and the strain perturbation response coefficient represents the amount of Brillouin frequency shift caused by a unit strain change. The frequency shift change decomposition equation of each virtual quantization unit is combined with the response sensitivity coefficient matrix to construct a system of two linear equations with two unknowns, namely the temperature change and strain change experienced by the virtual quantization unit at the current moment. Solve the system of two linear equations for each virtual quantization unit to obtain the current temperature change and strain change values ​​for that virtual quantization unit, thus forming the physical excitation decomposition result for that virtual quantization unit. The physical excitation decomposition results of all virtual quantization units at the current moment are arranged in ascending order of spatial index identifiers to form the physical excitation decomposition sequence at the current moment. Each element in the physical excitation decomposition sequence corresponds to a virtual quantization unit and includes its temperature change and strain change.

5. The intelligent prediction method for optical cable performance in data centers according to claim 1, characterized in that, The virtual quantization units within the optical cable microstructure perturbation quantization model are driven to perform state evolution iteration calculations based on the physical excitation decomposition sequence, resulting in a set of stress-strain state evolution trajectories for each virtual quantization unit on a continuous time axis, including: The values ​​of the stress-strain state parameters of all virtual quantized units at the current moment are read from the perturbation quantization model of the optical cable microstructure, including the current values ​​of the axial stress component, radial stress component and shear strain component of each virtual quantized unit, to form the stress-strain state matrix at the current moment. The physical excitation decomposition sequence at the current moment is used as an external driving force input into the state evolution rule function library of each virtual quantization unit. For each virtual quantization unit, the temperature change and strain change in its physical excitation decomposition result are calculated with the corresponding coefficients in the response sensitivity coefficient matrix of the virtual quantization unit to obtain the estimated value of the stress increment generated by the external excitation in the next time step of the virtual quantization unit. Call the stress relaxation evolution function in the state evolution rule function library of each virtual quantization unit, input the current value of the axial stress component and the current value of the radial stress component of the virtual quantization unit at the current moment, as well as the material viscoelastic parameters, and calculate the estimated value of the stress decay at the next moment due to the internal relaxation effect of the material. Call the strain cumulative evolution function in the state evolution rule function library of each virtual quantization unit, input the current value of the shear strain component of the virtual quantization unit at the current moment and the stress state information received from the boundary condition interface of the adjacent virtual quantization units, and calculate the estimated amount of strain increase at the next moment due to the mechanical interaction of adjacent units. Subtract the estimated stress decay from the estimated stress increment of each virtual quantization unit to obtain the updated values ​​of the axial stress component and radial stress component at the next moment for that virtual quantization unit. Add the estimated strain increase to the current value of the shear strain component to obtain the updated value of the shear strain component at the next moment for that virtual quantization unit. Based on the single-time-step stress-strain update estimate, the stress interaction and boundary correction between elements are performed to complete the multi-time-step iterative calculation and generate a set of stress-strain state evolution trajectories.

6. The intelligent prediction method for optical cable performance in data centers according to claim 5, characterized in that, The stress interaction and boundary correction between the execution units complete multi-time-step iterative calculations and generate a set of stress-strain state evolution trajectories, including: Through the boundary condition interface of each virtual quantization unit, the updated axial stress component value of this unit is transmitted to the adjacent virtual quantization unit on the right, and the updated axial stress component value of the adjacent virtual quantization unit on the left is received from its boundary condition interface, so as to realize the interactive transmission of stress state between units. Based on the received axial stress component update values ​​transmitted from the left adjacent virtual quantization unit, the boundary stress gradient of each virtual quantization unit is recalculated. The boundary stress gradient is the difference between the axial stress component update value of the left adjacent unit and the axial stress component update value of this unit divided by the length of the basic structural unit. The recalculated boundary stress gradient is fed back to the strain cumulative evolution function of each virtual quantization unit to correct the estimated increase in strain in the next time step, thus obtaining the corrected updated values ​​of the shear strain components. Write the updated values ​​of axial stress components, radial stress components, and corrected shear strain components of all virtual quantization units back to their respective stress-strain state parameter memory, overwriting the original values ​​at the current time, and complete the state evolution iteration of one time step. The physical excitation decomposition sequence corresponding to multiple consecutive sampling time points is driven sequentially. After each iteration of a time step, the stress and strain state parameters of all virtual quantization units at the corresponding time are stored in the historical record, ultimately forming a set of stress tensor curves and strain tensor curves of each virtual quantization unit on the continuous time axis.

7. The intelligent prediction method for optical cable performance in data centers according to claim 1, characterized in that, The construction of a spatiotemporal correlation network for optical cable link disturbance propagation based on the stress-strain state differences between adjacent virtual quantization units in the set of stress-strain state evolution trajectories includes: Extract the axial stress component variation curve of each virtual quantization unit on the complete time axis from the set of stress-strain state evolution trajectories to obtain a set of spatially ordered axial stress time series curves. The order of the curves in the sequence is completely consistent with the order of the spatial index identifiers of the virtual quantization units. For each pair of adjacent virtual quantization cells, the axial stress time series curve of the left cell is subtracted from the axial stress time series curve of the right cell at time points to obtain the time series variation curve of the axial stress difference between the two cells. This time series variation curve reflects the stress state transfer effect during the process of disturbance propagating from the left cell to the right cell. Calculate the first derivative of the time-series curve of the axial stress difference for each adjacent unit in the time dimension to obtain the time-series curve of the axial stress difference rate of change. The axial stress difference rate of change characterizes the rate of evolution of the stress state difference between the two units over time, i.e. the instantaneous rate of disturbance transmission. Each time point value on the time series curve of the rate of change of the axial stress difference of each adjacent unit is used as the instantaneous weight value of the edge of the node at the corresponding time. The virtual quantization unit is used as the network node, the directed connection between adjacent units is used as the network edge, and the attribute of the edge is marked by the instantaneous weight value. A spatiotemporal correlation network snapshot sequence that evolves dynamically with time is constructed. For each time point of the spatiotemporal correlation network snapshot, the spatial index identifier of the virtual quantization unit is used as the coordinate of the network node. The network nodes are arranged in a one-dimensional straight line, and the connection edges between nodes only exist between nodes with adjacent indices, forming a weighted network with a chain-like topology. Based on the spatiotemporal correlation network snapshot sequence, network weight normalization and smoothing optimization are performed to generate the final spatiotemporal correlation network for optical cable link disturbance propagation.

8. The intelligent prediction method for optical cable performance in data centers according to claim 1, characterized in that, The process of performing network flow characteristic analysis on the spatiotemporal correlation network to extract a set of early warning nodes in the network whose accumulated stress-strain state exceeds a critical threshold includes: The axial stress component variation curve of each virtual quantization unit node at all time points is extracted from the spatiotemporal correlation network to obtain the stress history trajectory of each node. The stress history trajectory records all stress state changes experienced by the node from the initial moment to the current moment. Cumulative damage calculation is performed on the stress history trajectory of each node. The rainflow counting method is used to identify and extract stress cycles in the trajectory. The number of stress cycles of different amplitudes experienced by each node within the historical time window is counted to form a stress cycle amplitude distribution histogram. Based on the fatigue characteristic curve of the optical cable material, each stress cycle amplitude in the stress cycle amplitude distribution histogram of each node is mapped to the amount of micro-damage caused to the material by that cycle, and the cumulative fatigue damage value of the node is obtained by summing up the micro-damage of all stress cycles. A critical damage threshold is preset for each virtual quantization unit node. The critical damage threshold is determined by the ultimate tensile strength of the optical cable material and the design safety factor. The cumulative fatigue damage value of each node is compared with the critical damage threshold of that node. When the cumulative fatigue damage value of a virtual quantization unit node exceeds its critical damage threshold, the node is marked as a potential risk node, and the spatial index identifier, current cumulative fatigue damage value, and the specific time point when the threshold is exceeded are recorded. Spatial clustering analysis is performed on all virtual quantization units marked as potential risk nodes to detect whether multiple potential risk nodes that are consecutively adjacent in spatial location constitute a risk node cluster. If the number of consecutively adjacent nodes exceeds the preset cluster size threshold, the entire cluster is marked as a segment with excessive cumulative damage. The curves of shear strain components of all nodes in each cumulative damage exceeding the standard section are extracted from the spatiotemporal correlation network and the changes over time. The slope of the rate of change of each curve in the most recent time window is calculated. The slope of the rate of change reflects the deterioration rate of the strain state of the node. The larger the slope of the rate of change, the more severe the deterioration of the strain state. The slope of the rate of change of all nodes in the section where the cumulative damage exceeds the standard is sorted, and the top few nodes with the largest slope of the rate of change are selected as key early warning nodes. The key early warning nodes represent the local locations where the strain state deteriorates most drastically in the section where the cumulative damage exceeds the standard. Map the spatial index identifier of each key early warning node back to the actual physical location coordinates of the optical cable to obtain the actual geographical location information of the key early warning node, including the distance of the key early warning node from the starting end of the optical cable and the specific laying environment section in which it is located. The spatial index identifier, actual geographical location information, current cumulative fatigue damage value, and slope of strain state change rate of key early warning nodes are combined to form an early warning node information entry. All early warning node information entries together constitute the early warning node set.

9. A smart prediction system for optical cable performance in data centers, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the intelligent prediction method for optical cable performance in a data center as described in any one of claims 1 to 8 by executing the machine-executable instructions.

10. A computer program product, characterized in that, The computer program product includes machine-executable instructions stored in a computer-readable storage medium. The processor of the intelligent prediction system for optical cable performance in a data center reads the machine-executable instructions from the computer-readable storage medium and executes the machine-executable instructions, causing the intelligent prediction system for optical cable performance in a data center to perform the intelligent prediction method for optical cable performance in a data center as described in any one of claims 1 to 8.

Images