Fatigue life prediction method of marine anchor chain based on multi-scale damage accumulation

Abstract

The application discloses a marine anchor chain fatigue life prediction method based on multi-scale damage accumulation and belongs to the technical field of marine anchor chain life prediction. The method first acquires anchor chain operation data, including strain time series, load time series and marine environment parameter time series. Then, the damage values of the anchor chain on the micro scale, the meso scale and the macro scale are calculated to form a multi-scale damage data group. The multi-scale damage data is corrected and weighted to determine the weight coefficients of the damage values of each scale. The multi-scale damage data is fused into a comprehensive damage index. The residual fatigue life of the anchor chain is calculated based on the comprehensive damage index. Finally, the fatigue life prediction result is outputted to generate a prediction report containing the residual life value and state evaluation. The application improves the accuracy and engineering applicability of the fatigue life prediction through multi-scale damage analysis, environment correction and weight optimization, and provides a reliable technical guarantee for the safe operation of the marine anchor chain.

Description

Technical Field

[0001] This invention relates to the field of marine anchor chain life prediction technology, and provides a method for predicting the fatigue life of marine anchor chains based on multi-scale damage accumulation. Background Technology

[0002] As a critical load-bearing component of mooring systems, marine anchor chains endure long-term, repeated loads from the marine environment, including wind, waves, and currents. Fatigue damage and failure of these chains directly threaten the safe operation of offshore platforms or vessels. During service, anchor chains undergo complex cyclic stress-strain processes, gradually accumulating fatigue damage within the material. This can ultimately lead to sudden fracture accidents, causing uncontrolled platform drift, personnel casualties, and significant economic losses. Therefore, accurate fatigue life prediction of marine anchor chains, timely assessment of their damage status, and implementation of preventative maintenance measures are crucial for ensuring the safe operation of offshore platforms.

[0003] Traditional methods for predicting the fatigue life of anchor chains are primarily based on single-scale damage analysis theories, typically considering only macroscopic load cycles and employing classic Miner's linear cumulative damage criterion or the stress range SN curve method for life assessment. While these methods are computationally simple and widely used in engineering, they have significant limitations and shortcomings. The fatigue failure of anchor chains is a complex process involving multiple spatial scales and various physical mechanisms. At the microscopic scale, it involves dislocation movement and slip band formation within grains; at the mesoscopic scale, it involves local stress concentration and plastic deformation accumulation in the chain link structure; and at the macroscopic scale, it involves load cycles and crack propagation of the entire anchor chain system. These damage mechanisms at different scales are coupled and jointly determine the final fatigue life. Traditional methods assess the damage state from only a single perspective, failing to comprehensively capture the multi-level characteristics and evolution of anchor chain fatigue damage, resulting in an insufficient understanding of the true damage mechanism and limited prediction accuracy. Temperature variations and corrosion in the marine environment significantly affect the fatigue performance of anchor chain materials, but traditional prediction methods often neglect the influence of these environmental factors. Traditional methods typically use material parameters and SN curves obtained under standard laboratory conditions to calculate fatigue life, but the actual marine service environment differs significantly from laboratory environments. Changes in seawater temperature affect the fatigue strength and crack propagation rate of materials; increased temperature reduces yield strength and fatigue limit, accelerating damage accumulation. Chloride ions and acid / alkaline environments in seawater cause corrosion damage on the anchor chain surface, forming pits and microcracks. These corrosion defects become initiation sources for fatigue cracks, significantly shortening fatigue life. The effects of temperature and corrosion on fatigue performance are mutually reinforcing; high temperatures accelerate corrosion reaction rates, and corrosion-induced defects are more easily propagated under temperature. Because environmental factors are not considered, fatigue life calculated using traditional methods is often systematically higher than the actual life. This optimistic prediction poses significant safety risks, potentially leading to the assumption that the anchor chain is still in a safe state when it is actually close to failure, missing the optimal maintenance window, or even causing chain breakage. In multi-scale damage fusion, how to reasonably determine the weighting coefficients for each scale of damage is a key issue affecting prediction accuracy. Existing methods typically use fixed weights or subjectively set weighting coefficients based on experience; this approach lacks scientific basis and has poor adaptability. The relative importance of damage at different scales (micro, meso, and macro) in anchor chains changes dynamically at different stages of fatigue damage. In the early stages of fatigue, micro-strain damage dominates; in the middle stages, meso-stress concentration effects gradually emerge; and in the later stages, macro-crack propagation becomes the controlling factor. Fixed weighting coefficients cannot reflect this dynamic evolution, leading to unreasonable weighting at certain damage stages and reducing the accuracy of the comprehensive damage index in reflecting the true damage state. Subjectively set weights are arbitrary and unpredictable; different operators may assign different weight values, lacking objectivity and consistency, thus affecting the reliability of prediction results.Traditional fatigue life prediction methods primarily focus on damage assessment at the current instantaneous state, lacking in-depth analysis and long-term tracking of historical evolution trends. Fatigue damage is a dynamic process that accumulates gradually over time, and its evolutionary patterns contain rich information. Analyzing changes in the damage growth rate can identify the impact of abnormal load events and harsh environmental conditions, and establishing a damage history database can provide a basis for model validation and parameter calibration. Existing methods often only provide a remaining life prediction value at a certain moment, without providing information on the entire damage evolution process and future development trends. Such one-sided assessment results are difficult to provide sufficient reference for maintenance decisions. Operation and maintenance personnel cannot understand the development process and changing patterns of anchor chain fatigue damage, nor are they clear about which factors have accelerated or slowed down the damage, making it difficult to formulate targeted maintenance strategies and improvement measures. Traditional methods often lack quantitative assessment of the reliability of prediction results when providing remaining life prediction values. The fatigue life prediction process involves multiple parameters and models, including sensor measurement, data processing, damage calculation, and life extrapolation. Each stage has certain errors and uncertainties, which are transmitted and accumulated through the calculation process, affecting the accuracy of the final prediction result. Existing methods typically only provide a single, definitive lifespan prediction without offering confidence intervals or reliability assessments. Operations and maintenance personnel cannot determine the reliability of this prediction, nor do they understand the actual lifespan range, making it difficult to make reasonable maintenance decisions based on risk tolerance. In practical engineering applications, over-reliance on unreliable predictions can lead to delayed maintenance and safety accidents, while a lack of confidence in the predictions can result in overly conservative and frequent maintenance, wasting resources. Traditional fatigue life prediction methods lack proactive early warning mechanisms and intelligent maintenance recommendations, representing a passive, post-hoc analysis approach. Existing methods usually require manual periodic calculation of remaining lifespan and assessment of maintenance needs. This manual judgment is subjective and time-sensitive, potentially delaying maintenance opportunities due to untimely assessments or inconsistent standards. Even when traditional methods predict a short remaining lifespan, they often only provide a numerical value without automatically generating early warning signals, failing to promptly alert relevant personnel to take action. Furthermore, they lack specific and actionable maintenance strategy recommendations and spare parts preparation lists. This passive assessment model fails to meet the high safety and reliability requirements of modern marine engineering, hindering the shift from post-hoc maintenance to preventative maintenance and impeding improvements in the safety management and operational economy of anchor chain systems.

[0004] Therefore, there is an urgent need to develop a new method for predicting the fatigue life of marine anchor chains, to improve the accuracy and engineering applicability of fatigue life prediction, and to provide reliable technical support for the safe operation of marine anchor chains. Summary of the Invention

[0005] To address the problems existing in the background art, this invention provides a method for predicting the fatigue life of marine anchor chains based on multi-scale damage accumulation, comprising the following steps: S1. Acquire anchor chain operation data, including strain time series of key parts of the anchor chain obtained through strain measurement, load time series of the load borne by the anchor chain obtained through load measurement, and marine environmental parameter time series obtained through environmental monitoring; S2. Process the anchor chain operation data, calculate the damage values ​​of the anchor chain at the micro, meso, and macro scales respectively, forming a multi-scale damage data set; S3. Calculate the environmental impact coefficient based on the marine environmental parameters, and use the environmental impact coefficient to correct the multi-scale damage data set; S4. Assign weights to the corrected multi-scale damage data to determine the weight coefficients of the damage at the micro, meso, and macro scales; S5. Merge the multi-scale damage data into a comprehensive damage index according to the weight coefficients; S6. Calculate the remaining fatigue life of the anchor chain based on the comprehensive damage index; S7. Output fatigue life prediction results and generate a prediction report containing the remaining life value and condition assessment.

[0006] In the preferred embodiment, step S1 specifically includes: S11. Install strain sensors in the high-stress areas of each link of the anchor chain, set the sampling frequency to 1000Hz, continuously collect strain data ε(t), and record for no less than 24 hours to obtain a strain time series dataset; S12. Install load sensors at the connection between the anchor chain and the mooring system to synchronously collect axial tensile force. and bending load The sampling frequency is kept consistent with the strain data to form a load time series dataset; S13, environmental monitoring equipment is deployed in the sea area around the anchor chain to collect seawater temperature data every 60 seconds. pH value, chloride ion concentration and dissolved oxygen concentration S14. Perform timestamp alignment on all collected data to establish a unified time reference, form a synchronous multi-source data stream, and achieve spatiotemporal synchronization matching of data.

[0007] In the preferred embodiment, step S2 specifically includes: S21. Calculate the microscopic damage value based on strain data. Extract the strain amplitude Δε from the strain time series using the rainflow counting method, and then calculate it according to the formula. Calculate the cumulative microscopic damage; where Δε represents the microscopic damage value; Δε represents the strain amplitude. The reference strain value (με) is used; m is a material constant; Σ is the summation operator to obtain the microscale damage quantification result; S22, calculate the mesoscopic damage value based on the load data, extract the load cycle characteristics, and apply the formula... Calculate the cumulative damage at the microscopic level; where The value represents the microscopic damage; ΔF represents the load amplitude (kN). The reference load value (kN) is used; n is the load sensitivity index, enabling quantitative assessment of microscale damage; S23, calculate the macroscopic damage value based on the combined effect of strain and load, according to the formula... Calculate the macroscopic cumulative damage; where This represents the macroscopic damage value. This represents the actual number of cycles for the i-th load cycle. Σ represents the fatigue life under the corresponding load level; Σ represents the sum of all load levels, yielding a macroscopic assessment of cumulative damage.

[0008] In the preferred embodiment, step S3 specifically includes: S31. Calculate the temperature environmental influence coefficient according to the formula. Calculate the temperature correction factor; where This is the temperature influence coefficient; This is a temperature-sensitive parameter; Reference temperature (K); T is the actual temperature (K); exp is the natural exponential function; The quantitative results of the effect of temperature on damage are obtained; S32, Calculate the corrosion environment influence coefficient according to the formula Calculate the corrosion correction factor; where This is the corrosion influence coefficient; pH sensitivity coefficient; The chloride ion influence coefficient is used to quantify the impact of marine corrosion environment on damage; S33, calculate the comprehensive environmental correction coefficient according to the formula. Calculate the total environmental impact factor; where The comprehensive environmental correction coefficient is used to obtain a comprehensive environmental impact assessment of multiple factors; S34, environmental corrections are applied to damage values ​​at each scale, and the following calculations are performed respectively:

[0009] ; in , , These represent the micro, meso, and macro damage values ​​after environmental correction, achieving unified correction of multi-scale damage caused by environmental factors.

[0010] In the preferred embodiment, step S4 specifically includes: S41. Set the objective function for weight optimization, establishing an optimization function with the goal of minimizing prediction error. Collect recent actual damage observation data as a reference benchmark and establish a mapping relationship between weight coefficients and prediction accuracy. S42. Set constraints on the weight coefficients, requiring... And all weight coefficients are greater than 0, among which , , S43. Weight coefficients are assigned to microscopic, mesoscopic, and macroscopic damage to ensure the rationality of weight allocation; S44. An optimization algorithm is used to search for the optimal weight combination. The combination of weight coefficients that minimizes the prediction error is found through iterative calculation. The number of iterations is set to 50 to obtain the optimal weight configuration under the current working condition; S45. A dynamic weight update mechanism is established. After accumulating 100 new data points, the weight optimization process is re-executed to update the weight coefficients to adapt to the changing trend of damage evolution and achieve adaptive adjustment of weight allocation.

[0011] In a preferred embodiment, step S5 specifically includes: S51, according to the formula Calculate the comprehensive damage index; among which As a comprehensive damage index; , , These are the corresponding weighting coefficients; , , The corrected damage values ​​at each scale are obtained; the overall damage status assessment of the anchor chain is obtained; S52, the damage growth rate is calculated according to the formula. Calculate the current rate of damage growth; where The damage growth rate; This represents the comprehensive damage index at the current moment; The comprehensive damage index is the value of the previous moment; Δt is the time interval (h); the rate characteristics of damage accumulation are determined; S53, establish a damage history record, store the calculated comprehensive damage index and damage growth rate in time series to form a damage evolution trend curve, and provide historical data support for subsequent life prediction.

[0012] In a preferred embodiment, step S6 specifically includes: S61. Set fatigue failure criteria, and set the critical value of the comprehensive damage index as... ,when When this value is reached, fatigue failure of the anchor chain is determined, and a quantitative standard for failure prediction is established; S62, according to the formula Calculate the remaining fatigue life; where The remaining fatigue life (h); This is the critical damage value; This represents the current comprehensive damage index; S63. Perform uncertainty analysis, considering measurement errors and model uncertainties, calculate the confidence interval of the remaining life, set the confidence level to 95%, and provide a reliability assessment for the life prediction results; S64. Establish a life early warning mechanism, generate an early warning signal when the remaining fatigue life is lower than the preset safety threshold, and realize the active monitoring and risk prevention of the fatigue state of the anchor chain.

[0013] In a preferred embodiment, step S7 specifically includes: S71. Format the output of the remaining lifespan value, and output the calculated value. Standardize the representation to two decimal places, using hours as the unit of measurement, and generate standard formatted life prediction values; S72, generate a damage status assessment report, including current comprehensive damage indicators. S73. Create a life prediction chart, draw damage evolution curves and remaining life trend diagrams to visualize the fatigue state changes of the anchor chain; S74. Output maintenance recommendation information, generate corresponding maintenance strategy recommendations based on the remaining life and current damage state to achieve predictive maintenance guidance output.

[0014] The beneficial effects achieved by this invention are as follows: First, this invention establishes a multi-scale damage accumulation analysis method, assessing the fatigue damage state of anchor chains from three different spatial scales and physical mechanisms: microscopic, mesoscopic, and macroscopic. This overcomes the limitations and biases of traditional single-scale methods. The microscopic scale focuses on strain damage and dislocation motion characteristics at the grain level, capturing the microscopic mechanism of fatigue crack initiation. The mesoscopic scale focuses on load damage and local stress concentration effects at the chain link structure level, reflecting the cumulative plastic deformation process at the bending points of the chain links. The macroscopic scale focuses on the cumulative damage and overall fatigue state at the entire anchor chain system level, comprehensively considering the combined effects of strain and load. By simultaneously monitoring strain and load data, calculating damage values ​​at the three scales respectively, and then performing weighted fusion, this invention more completely and comprehensively reveals the multi-level characteristics and evolution laws of anchor chain fatigue damage. Compared to traditional methods that assess damage from only a single perspective, this invention's multi-scale analysis method significantly improves the depth of understanding and prediction accuracy of the true damage state of anchor chains, resulting in a high degree of consistency between predicted and actual lifespans, effectively avoiding prediction bias caused by incomplete damage assessment.

[0015] Secondly, this invention introduces an environmental correction mechanism, incorporating the effects of temperature changes and corrosion in the marine environment on fatigue performance into the damage calculation model. This solves the systematic error problem caused by the use of laboratory standard condition parameters in traditional methods. The temperature effect is quantified by a temperature correction coefficient based on the Arrhenius equation, accurately reflecting the influence of temperature changes on the fatigue strength and damage accumulation rate of materials. The corrosion effect is quantified by a corrosion correction coefficient that comprehensively considers pH value and chloride ion concentration, accurately reflecting the accelerating effect of the marine corrosion environment on the initiation and propagation of fatigue cracks. This invention dynamically calculates a comprehensive environmental correction coefficient by real-time monitoring of environmental parameters such as seawater temperature, pH value, and chloride ion concentration, and uses this coefficient to uniformly correct damage values ​​at various scales, achieving accurate conversion from laboratory standard environment to actual marine service environment. The environmental correction mechanism effectively eliminates the difference between laboratory conditions and field environment, avoids systematic optimistic prediction bias, makes the prediction results more consistent with actual service conditions, significantly reduces the prediction error rate, improves the reliability and engineering applicability of fatigue life prediction, and provides accurate life assessment for anchor chains under different sea areas and environmental conditions.

[0016] Third, this invention establishes a weight optimization and dynamic update mechanism, enabling adaptive determination and dynamic adjustment of damage weight coefficients at each scale. This solves the problem of low fusion accuracy caused by traditional methods using fixed or empirical weights. This invention employs a particle swarm optimization algorithm, aiming to minimize prediction error. Using actual observed damage data as a reference, it automatically determines the optimal weight combination that best matches the comprehensive damage index with the actual damage state through iterative search, avoiding the arbitrariness and blindness of manually setting weights. More importantly, this invention establishes a dynamic weight update mechanism. After accumulating a certain number of new data points, the weight optimization process is automatically re-executed, allowing the weight coefficients to dynamically reflect the relative importance changes of damage at each scale as the anchor chain fatigue damage evolves. This fully considers the objective law that the dominant roles of micro, meso, and macro damage differ at different stages of fatigue. This weight optimization and dynamic update mechanism significantly improves the accuracy of the comprehensive damage index in reflecting the actual fatigue state of the anchor chain, achieving excellent damage identification accuracy. It also improves the adaptability and long-term accuracy of fatigue life prediction, ensuring that the prediction method maintains high accuracy throughout its service life.

[0017] Fourth, this invention establishes a damage history record and trend analysis function, providing an assessment of the current damage status and revealing in depth the laws and cumulative characteristics of damage evolution, offering richer reference information for maintenance decisions. This invention stores key data such as the calculated comprehensive damage index, damage values ​​at various scales, damage growth rate, environmental parameters, and weighting coefficients in a complete time series, forming a damage evolution database covering the entire process of anchor chains from commissioning to the present moment. By analyzing this historical data, damage evolution trend curves are plotted, periods of rapid damage growth are identified, the causes of accelerated damage are analyzed, and the impact of severe sea conditions or abnormal load events is assessed. The damage history record can also be used for model validation and parameter calibration. By comparing the predicted damage evolution curve with actual observation data, model parameters are continuously optimized, improving prediction accuracy. This historical record and trend analysis function ensures that life prediction is not only based on the current instantaneous state but also fully considers historical evolution trends and future developments, enhancing the foresight and reliability of predictions and providing operation and maintenance personnel with comprehensive anchor chain health information and scientific decision-making basis.

[0018] Fifth, this invention quantifies the reliability of prediction results and achieves proactive prevention and control of fatigue risks by conducting uncertainty analysis and establishing an early warning mechanism, transforming passive post-event maintenance into proactive preventative maintenance. This invention employs the Monte Carlo simulation method, comprehensively considering the impact of sensor measurement errors and model parameter uncertainties on remaining life prediction, calculating the confidence interval of remaining life. This allows operation and maintenance personnel to not only know the predicted value of remaining life but also clearly understand the range of its reliability, enabling them to make more prudent maintenance decisions based on their own risk tolerance. This invention establishes a multi-level early warning mechanism, automatically generating an early warning signal when the remaining fatigue life falls below a preset safety threshold. The invention features small prediction errors and accurate damage identification, accurately identifying high-risk conditions and issuing early warning signals when the anchor chain still has sufficient remaining life. This provides ample time for maintenance decisions and operational preparation, ensuring safety while avoiding excessive conservatism, significantly reducing the risk of sudden anchor chain failure, improving the safety, reliability, and economy of offshore platforms, and providing reliable technical support for the safe operation of marine engineering projects. Attached Figure Description

[0019] Figure 1 This is a bar chart comparing the predicted lifespan and actual lifespan of five anchor chains in Embodiment 1 and each comparative example of the present invention.

[0020] Figure 2 This is a bar chart comparing the prediction error rates of five anchor chains in Embodiment 1 of the present invention with those of the comparative examples.

[0021] Figure 3 This is a bar chart comparing the damage identification accuracy of Embodiment 1 of the present invention with that of each comparative example.

[0022] Figure 4 This is a bar chart comparing the early warning time of five anchor chains in Embodiment 1 of the present invention with that of each comparative example.

[0023] Figure 5 This is a scatter plot comparing the predicted lifespan trends of five anchor chains in Embodiment 1 of the present invention with those in each comparative example.

[0024] Figure 6 This is a bar chart comparing the multi-indicator comprehensive results of Embodiment 1 of the present invention with those of various comparative examples. Detailed Implementation

[0025] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. In addition, the forms of the various structures described in the following embodiments are merely illustrative. The present invention is not limited to the structures described in the following embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] This invention provides a method for predicting the fatigue life of marine anchor chains based on multi-scale damage accumulation. Through multi-scale damage accumulation analysis technology, it enables accurate prediction and risk assessment of the remaining service life of anchor chains.

[0027] In implementing this method, step S1 is performed first, which involves acquiring anchor chain operation data. This step requires collecting three key types of data: strain data, load data, and environmental parameter data. Specifically, in step S11, strain sensors are installed in high-stress areas of each link of the anchor chain. These high-stress areas are typically located at the bending points of the links and the contact areas with the shackles, as these locations experience the greatest stress concentration during the anchor chain's service life and are the most prone to fatigue crack initiation. The strain sensors are either resistance strain gauges or fiber Bragg grating strain sensors. Fiber Bragg grating strain sensors offer better corrosion resistance and electromagnetic interference resistance in marine environments, making FBG sensors the preferred choice. The sampling frequency is set to 1000Hz. This selection is based on the Nyquist sampling theorem to capture the high-frequency vibration characteristics of the anchor chain under wave action. Since wave frequencies are typically between 0.1Hz and 2Hz, while the natural frequency of the anchor chain can reach hundreds of hertz, a 1000Hz sampling frequency effectively avoids signal aliasing. Strain data is continuously acquired. ,in Indicates time (s), The strain value is recorded over a period of no less than 24 hours to obtain complete daily cyclic load characteristics, as the load on the anchor chain is affected not only by tides but also by the periodic changes in wind and waves. The 24-hour data covers a complete tidal cycle, thus obtaining a strain time series dataset. This dataset provides raw data support for subsequent rainflow counting analysis and fatigue damage calculation.

[0028] In step S12, a load sensor is installed at the connection between the anchor chain and the mooring system. This location is the node where the stress is most concentrated in the entire anchor chain system, directly reflecting the overall load on the anchor chain. The load sensor typically uses a combination of tension / compression sensors and bending moment sensors. The tension / compression sensors measure axial tensile force, while the bending moment sensors measure bending load. Axial tensile force is collected simultaneously. and bending load ,in Indicates axial tensile force (kN), This represents the bending load (kN·m). The time (s) is represented by the sampling frequency. The sampling frequency is kept consistent with the strain data, set to 1000Hz. This synchronous sampling strategy ensures a precise correspondence between strain and load data in the time dimension, providing a reliable foundation for establishing a strain-load relationship model. This method generates a load time-series dataset that reflects the load variation patterns of the anchor chain under actual sea conditions, including average load level, load amplitude distribution, and load frequency characteristics. These characteristic parameters are key inputs for calculating damage at both the micro and macro scales.

[0029] In step S13, environmental monitoring equipment is deployed in the waters surrounding the anchor chain. This equipment includes temperature sensors, pH sensors, chloride ion concentration sensors, and dissolved oxygen concentration sensors. These sensors are typically integrated into a water quality monitoring buoy system, which is fixed to the anchor chain within a range of approximately 5 to 10 meters via anchoring cables. This system reflects the actual environmental conditions of the anchor chain without affecting its normal operation. Seawater temperature (°C) is collected every 60 seconds. pH value, chloride ion concentration and dissolved oxygen concentration ,in Temperature (°C) is represented by pH, which represents the hydrogen ion concentration index, i.e., acidity or alkalinity (dimensionless). This indicates the chloride ion concentration (mg / L). This represents the dissolved oxygen concentration (mg / L). The 60-second sampling interval is determined based on the rate of change of marine environmental parameters. Changes in seawater temperature, pH, and chemical composition are typically slow processes. The 60-second sampling interval effectively captures the changing trends of environmental parameters without generating excessive redundant data. This method generates a time series of environmental parameters that reflects the corrosive characteristics and temperature properties of the marine environment in which the anchor chain is located. These environmental factors significantly affect the fatigue performance and crack propagation rate of the anchor chain material and must therefore be considered in damage calculations.

[0030] In step S14, all acquired data undergoes timestamp alignment to establish a unified time reference. Since strain sensors, load sensors, and environmental sensors may use different data acquisition devices, and the internal clocks of each device may deviate, GPS time synchronization or Network Time Protocol (NTP) is required for time synchronization. The timestamps of all sensors are calibrated to the same standard time, with a time synchronization accuracy at the millisecond level to ensure accurate time correspondence between high-frequency strain and load data. For data with different sampling frequencies, interpolation methods are used for time alignment. Specifically, environmental parameters sampled every 60 seconds are extended to the same time point as the strain and load data using linear interpolation or spline interpolation methods, thus forming a synchronized multi-source data stream. Through this spatiotemporal synchronization matching process, precise correspondence between strain data, load data, and environmental data in the time dimension is achieved, providing a unified data foundation for subsequent multi-scale damage calculations and environmental corrections, and avoiding damage calculation errors caused by data asynchrony.

[0031] After data acquisition, step S2 is executed, which involves processing the anchor chain operation data and calculating the damage values ​​of the anchor chain at the micro, meso, and macro scales. The core idea of ​​the multi-scale damage analysis method is to evaluate the fatigue damage state of the anchor chain from the perspectives of different spatial scales and physical mechanisms. The microscale focuses on strain damage at the material grain level, the mesoscale focuses on load damage at the chain link structure level, and the macroscale focuses on cumulative damage at the level of the entire anchor chain system. This multi-scale analysis framework more comprehensively reveals the fatigue damage mechanism of the anchor chain.

[0032] In step S21, the microscopic damage value is calculated based on the strain data. First, the strain time series is... Strain amplitude was extracted using the rainflow counting method. Rainflow counting is a classic cyclic counting method that transforms irregular strain time histories into a series of equivalent strain cycles, each characterized by average strain and strain amplitude. The basic principle of rainflow counting is to simulate the process of rainwater flowing down a roof, identifying strain cycles by pinpointing peaks and valleys in the strain time history. The specific algorithm includes four steps: preprocessing, peak and valley extraction, rainflow counting, and cycle statistics. Through rainflow counting, a series of strain cycles are obtained, each with a defined strain amplitude. and number of loops ,in This represents the cycle number. Then follow the formula. Calculate the micro-cumulative damage, where This represents the microscopic damage value (dimensionless). strain amplitude ( () indicates the strain change magnitude for each strain cycle; Reference strain value ( The fatigue limit strain of a material is typically taken as 50% to 70% of its elastic limit strain. The material constant (dimensionless) represents the sensitivity of the material to the strain amplitude. For the R3 grade mooring chain steel commonly used in anchor chains, the preferred value range for this parameter is 4 to 6. The summation operator represents the cumulative summation over all strain cycles. Based on strain life theory, this formula reflects the accumulation of microscopic damage in materials under cyclic strain; the larger the strain amplitude and the more cycles, the faster the microscopic damage accumulates. This method yields quantified damage at the microscale, reflecting the damage evolution process of the anchor chain material at the grain and dislocation scales, providing microscopic-level damage information for multi-scale damage fusion.

[0033] In step S22, the mesoscopic damage value is calculated based on the load data. First, load cycle characteristics are extracted from the load time series, and the axial tensile force is calculated using the same rainflow counting method as the strain data. and bending load Cyclic identification is performed. For structures like anchor chains that bear complex loads, the combined effects of axial tensile and bending loads need to be considered. The concept of equivalent load is typically used to convert multiaxial loads into uniaxial equivalent loads. The formula for calculating the equivalent load is as follows: in The bending load equivalence factor is typically taken as 1.5 to 2.0, reflecting the amplification effect of bending load on fatigue damage. Then, according to the formula... Calculate the cumulative damage at the mesoscopic level, where For detailed damage values ​​(dimensionless); The load amplitude (kN) represents the load variation amplitude for each load cycle; The reference load value (kN) represents the fatigue load limit of the chain link, which is typically taken as 30% to 40% of the chain link's breaking load. The load sensitivity index (dimensionless) represents the sensitivity of a structure to the magnitude of a load. For anchor chain structures, the preferred value range for this parameter is 3 to 5. The summation operator represents the cumulative summation over all load cycles. This formula, based on the nominal stress method, reflects the accumulation of micro-damage in a chain link structure under cyclic loading; the larger the load amplitude and the more cycles, the faster the micro-damage accumulates. This method enables quantitative assessment of micro-scale damage. The assessment results reflect the stress concentration and local plastic deformation effects of the anchor chain at the chain link structural scale, providing micro-level damage information for multi-scale damage fusion.

[0034] In step S23, the macroscopic damage value is calculated based on the combined effect of strain and load. The macroscopic damage calculation uses the Miner linear cumulative damage criterion, which is the most widely used cumulative damage theory in fatigue analysis. Its basic assumption is that fatigue damage under different load levels can be linearly superimposed, and fatigue failure occurs when the cumulative damage reaches 1. First, the SN curve (stress-life curve) of the anchor chain needs to be established. The SN curve describes the relationship between load level and fatigue life, and is usually represented on a logarithmic scale. Its mathematical form is... in For fatigue life (times), The stress amplitude (MPa) and This is a material constant. For anchor chain materials, the SN curve parameters can be obtained through fatigue testing or by consulting a material handbook. For R3 grade mooring chain steel, in a seawater environment, The preferred value range is to , The preferred value range is 3 to 5. Then follow the formula. Calculate the macroscopic cumulative damage, where The value represents the macroscopic damage (dimensionless). For the first The actual number of cycles (times) for a load cycle indicates the actual number of cycles experienced at that load level; The fatigue life (cycles) at the corresponding load level is represented by the total number of cycles that the structure can withstand at that load level, which is calculated using the SN curve. To represent the cumulative damage across all load levels, the damage components under different load levels are summed. This formula embodies the core idea of ​​the Miner criterion: the contribution of each load cycle to the total damage is equal to the ratio of the number of cycles to the corresponding fatigue life. When the sum of the damage components under all load levels reaches 1, the structure experiences fatigue failure. This method yields a macroscopic-scale cumulative damage assessment, reflecting the overall fatigue state of the anchor chain at the system level. It comprehensively considers the combined effects of strain and load, providing macroscopic-level damage information for multi-scale damage fusion.

[0035] After completing the multi-scale damage calculation, step S3 is executed, which involves calculating the environmental impact coefficient based on marine environmental parameters and correcting the multi-scale damage data set. Temperature changes and corrosion in the marine environment significantly affect the fatigue performance of anchor chain materials. Increased temperature reduces the yield strength and fatigue limit of the material, while corrosion generates pits and microcracks on the material surface, thus accelerating the initiation and propagation of fatigue cracks. Therefore, it is essential to quantitatively assess and correct for the impact of environmental factors.

[0036] In step S31, the temperature environment influence coefficient is calculated. According to the formula... Calculate the temperature correction factor, where The temperature effect coefficient (dimensionless) represents the amplification or reduction effect of temperature on fatigue damage. When the temperature is higher than the reference temperature, the coefficient is greater than 1, indicating accelerated damage. When the temperature is lower than the reference temperature, the coefficient is less than 1, indicating slowed damage. The temperature sensitivity parameter (K) represents the degree to which the fatigue performance of a material is sensitive to temperature changes. This parameter is related to the thermal activation energy of the material. For anchor chain steel materials, the preferred value range is 3000K to 5000K. The reference temperature (K) represents the temperature of the standard fatigue test, which is usually taken as 293K, or 20℃. The actual temperature (K) represents the real-time temperature of the seawater where the anchor chain is located. It needs to be converted from Celsius to Kelvin. The temperature conversion formula is ℃. ;exp is the natural exponential function, i.e., the function expressed by the natural constant. Exponential operations with base It is approximately equal to 2.71828. This formula is based on a variation of the Arrhenius equation and reflects the influence of temperature on the fatigue properties of materials. The higher the temperature, the lower the fatigue strength of the material and the faster the accumulation of fatigue damage. This method quantifies the effect of temperature on damage, normalizing fatigue damage under different temperature conditions to a reference temperature and providing a temperature influence factor for environmental correction.

[0037] In step S32, the corrosion environment influence coefficient is calculated. According to the formula... Calculate the corrosion correction factor, where The corrosion effect coefficient (dimensionless) represents the amplification effect of the corrosive environment on fatigue damage. Since corrosion always accelerates fatigue damage, this coefficient is always greater than or equal to 1. The pH sensitivity coefficient (dimensionless) represents the degree of influence of pH value deviating from neutral on the corrosion rate. For carbon steel materials, the preferred value range is 0.05 to 0.15. Chloride ion influence coefficient The value of chloride ion concentration is preferred to be between 0.001 and 0.005, indicating the degree of influence of chloride ion concentration on corrosion rate. Chloride ions are the main factor in seawater corrosion. The pH value indicates the degree of deviation from neutrality. A pH value of 7 indicates neutrality. The further the pH value deviates from 7, the stronger the corrosiveness. The square root of the chloride ion concentration is used because the corrosion rate does not have a linear relationship with chloride ion concentration but follows a parabolic curve; 1 is the baseline value representing the baseline damage level without corrosion. This formula comprehensively considers the influence of two key corrosion factors, pH value and chloride ion concentration, reflecting the corrosive characteristics of the marine environment. Acidic or alkaline environments and high chloride ion concentrations accelerate corrosion fatigue damage. This method quantifies the impact of the marine corrosive environment on damage. This quantification incorporates the fatigue strength reduction and crack propagation acceleration effects caused by corrosion into the damage calculation model, improving the accuracy of damage assessment.

[0038] In step S33, the comprehensive environmental correction factor is calculated. According to the formula... Calculate the total environmental impact factor, where The comprehensive environmental correction factor (dimensionless) represents the combined effect of temperature and corrosion as environmental factors. This is the temperature influence coefficient; The coefficient represents the corrosion impact factor. The product form is used based on the interaction mechanism between temperature and corrosion on fatigue damage. Increased temperature accelerates the corrosion reaction rate, and corrosion-induced defects are more easily propagated at high temperatures. Therefore, the combined effect of the two is mutually reinforcing rather than simply additive. The product form more accurately reflects this interaction effect. This method yields a comprehensive assessment of the multi-factor environmental impact. The assessment results comprehensively reflect the amplifying effect of the combined corrosiveness and temperature characteristics of the marine environment in which the anchor chain is located on fatigue damage, providing a unified environmental correction factor for subsequent damage correction.

[0039] In step S34, environmental corrections are applied to the damage values ​​at each scale. The calculations are performed separately. in The value of microscopic damage after environmental correction (dimensionless); The environmentally corrected microscopic damage value (dimensionless); The macroscopic damage value after environmental correction (dimensionless); , , These represent the micro, meso, and macroscopic damage values ​​before correction; The environmental correction factor is used to comprehensively correct the damage values ​​at each scale. By multiplying the damage values ​​at each scale by the environmental correction factor, the environmental factors are uniformly corrected for the damage at multiple scales. The corrected damage values ​​more accurately reflect the true fatigue damage state of the anchor chain under actual marine environmental conditions, eliminating the difference between the laboratory standard environment and the actual marine environment, and improving the reliability and engineering applicability of fatigue life prediction.

[0040] After completing the environmental correction, step S4 is executed, which involves weighting the corrected multi-scale damage data. Since the damage values ​​at the micro, meso, and macro scales have different physical meanings and dimensions, they cannot be simply added together. It is necessary to determine the weight coefficients of each scale in the comprehensive evaluation. The determination of the weight coefficients directly affects the final fatigue life prediction accuracy. Therefore, an optimization method is needed to determine the optimal weight combination.

[0041] In step S41, the weight optimization objective function is set. The optimization function is established with the goal of minimizing the prediction error. Specifically, the objective function can be expressed as follows: in The objective function value represents the sum of squares of the prediction errors; For the first Predicted comprehensive damage value for each sample; For the first The actual observed comprehensive damage value of each sample; The total number of samples is represented by ; min indicates the minimization operation. Recent actual damage observation data is collected as a reference benchmark. This data can be obtained from regular anchor chain flaw detection. Non-destructive testing techniques such as ultrasonic testing, magnetic particle testing, or eddy current testing are used to obtain the actual crack size and damage degree of the anchor chain. The detection results are converted into equivalent damage values ​​as an optimization reference standard. This method establishes a mapping relationship between weighting coefficients and prediction accuracy, providing a clear optimization objective and reference basis for subsequent weight optimization.

[0042] In step S42, weighting coefficient constraints are set. Requirements And all weight coefficients are greater than 0, among which The weighting coefficient for microscopic damage (dimensionless); The weighting coefficient for microscopic damage (dimensionless); These are the weighting coefficients (dimensionless) for macroscopic damage. The constraint that the sum of the weighting coefficients equals 1 ensures the normalization of the comprehensive damage index, making damage values ​​at different times and between different anchor chains comparable. The constraint that all weighting coefficients are greater than 0 ensures that damage at each scale contributes to the comprehensive assessment, avoiding the complete neglect of damage at a certain scale. Furthermore, the range of values ​​for the weighting coefficients can be constrained based on engineering experience, for example... The range constraint, based on prior knowledge of the importance of damage at different scales, prevents the optimization algorithm from getting trapped in an unreasonable parameter space. This approach ensures the rationality of weight allocation, so that the optimization results not only meet mathematical constraints but also satisfy the physical requirements of practical engineering.

[0043] In step S43, an optimization algorithm is used to search for the optimal weight combination. Particle Swarm Optimization (PSO), Genetic Algorithm (GA), or gradient descent can be used. PSO is widely used in engineering optimization problems due to its simplicity, fast convergence, and lack of gradient information requirement, making it the preferred choice. The basic principle of PSO is to simulate the foraging behavior of a flock of birds. Each weight combination is considered a particle in the solution space, flying through the space to find the optimal solution. The flight speed and position of each particle are continuously updated based on its own experience and the experience of the group. The particle swarm size is set to 20 to 50 particles, preferably 30, with each particle representing a weight combination. The weight combination that minimizes the prediction error is found through iterative calculation. The number of iterations is set to 50, which ensures that the algorithm converges to the optimal solution or near-optimal solution in most cases, while avoiding excessive computation time. In each iteration, the objective function value corresponding to each particle is calculated, the individual optimal position and the group optimal position are updated, and then the particle's flight trajectory is adjusted according to the velocity update formula and the position update formula. After 50 iterations, the optimal position of the population is the optimal weight combination under the current working condition, thus obtaining the optimal weight configuration under the current working condition. This optimization method adaptively determines the weight coefficients based on actual observation data, avoiding the arbitrariness of subjectively setting weights and improving the prediction accuracy and reliability of the comprehensive damage index.

[0044] In step S44, a dynamic weight update mechanism is established. Since the damage evolution of the anchor chain is dynamic, the relative importance of damage at different scales may change at different damage stages. For example, micro-damage dominates in the early stages of fatigue, while the role of meso- and macro-damage gradually increases in the middle stages of fatigue. Therefore, the weight coefficients should not be fixed but should be dynamically adjusted with damage evolution. The weight optimization process is re-executed after accumulating 100 new data points. Assuming a sampling frequency of 1000Hz and appropriate downsampling processing for weight optimization, this update frequency can reflect changes in damage state in a timely manner without causing excessive fluctuations in the weight coefficients due to overly frequent updates. The weight coefficients are updated to adapt to the changing trend of damage evolution. During each update, the latest 100 data points are combined with historical data to reconstruct the optimization objective function, and the optimization processes in steps S41 to S43 are executed to obtain the updated weight coefficients. This method enables adaptive adjustment of weight allocation, allowing the comprehensive damage index to dynamically reflect the changes in the relative importance of damage at each scale as the fatigue damage of the anchor chain evolves, thereby improving the adaptability and long-term accuracy of fatigue life prediction.

[0045] After completing the weight allocation, step S5 is executed, which involves fusing the multi-scale damage data into a comprehensive damage index based on the weighting coefficients. Damage fusion is a key step in the multi-scale analysis method. It integrates the damage values ​​of the three scales into a comprehensive index through weighted summation, which fully reflects the overall fatigue damage state of the anchor chain.

[0046] In step S51, according to the formula Calculate the comprehensive damage index, among which The comprehensive damage index (dimensionless) represents the overall fatigue damage degree of the anchor chain, and its value ranges from 0 to 1, where 0 represents no damage and 1 represents complete failure. The weighting coefficient for microscopic damage (dimensionless); The weighting coefficient for microscopic damage (dimensionless); The weighting coefficient for macroscopic damage (dimensionless); The value of microscopic damage after environmental correction (dimensionless); The environmentally corrected microscopic damage value (dimensionless); This represents the environmentally corrected macroscopic damage value (dimensionless). The formula employs a linear weighted fusion method, averaging the damage values ​​at the three scales according to their respective weighting coefficients. These weighting coefficients reflect the relative importance of damage at each scale in the comprehensive assessment. The optimized weighting coefficients maximize the consistency between the comprehensive damage index and the actual damage state. This method yields an overall damage state assessment of the anchor chain. The comprehensive damage index is a normalized dimensionless quantity, physically representing the proportion of the anchor chain's fatigue life already consumed to its total lifespan. When the comprehensive damage index approaches 1, it indicates that the anchor chain is about to fail and requires timely replacement or repair.

[0047] In step S52, the damage growth rate is calculated. According to the formula... Calculate the current rate of damage growth, where The damage growth rate (1 / h) represents the increase in the comprehensive damage index per unit time. This represents the comprehensive damage index (dimensionless) at the current moment. This represents the comprehensive damage index (dimensionless) from the previous moment. The time interval (h) represents the time span between two damage calculations, preferably set to one hour, meaning the damage growth rate is calculated every hour. The damage growth rate reflects the rate of accumulation of fatigue damage in the anchor chain. This parameter is crucial for predicting remaining life, as remaining life depends on both the current damage level and the future rate of damage growth. By calculating the damage growth rate, the severity of the current load environment on the anchor chain can be assessed. A sudden increase in the damage growth rate may indicate that the anchor chain is under abnormally high loads or that environmental conditions have deteriorated, requiring attention. Furthermore, the damage growth rate can be used to verify the rationality of the damage model. Unreasonable fluctuations or negative values ​​in the damage growth rate may indicate that the model parameters need recalibration. Determining the rate of damage accumulation in this way provides dynamic parameter support for remaining life prediction.

[0048] In step S53, a damage history record is established. The calculated comprehensive damage index and damage growth rate are stored in time series to form a damage evolution trend curve. Specifically, a database or data file is established, indexed by timestamps, to record key parameters such as the comprehensive damage index value, sub-damage values ​​at three scales, damage growth rate, environmental parameters, and weighting coefficients at each moment. The data recording interval is preferably one hour, which fully reflects the dynamic process of damage evolution without generating excessive storage burden. The damage history record formed over a long period can be used to plot the damage evolution curve, which intuitively shows the damage accumulation process of the anchor chain from its commissioning to the current moment. It can identify periods of rapid damage growth and analyze the causes of accelerated damage, such as whether it has experienced severe sea conditions or equipment failure. The damage history record can also be used for model validation and parameter calibration. By comparing the predicted damage evolution curve with the actual observed damage data, the accuracy of the model can be evaluated, and the model parameters can be adjusted according to the deviation. In this way, historical data support is provided for subsequent life prediction, so that life prediction is not only based on the current state but also fully considers the historical evolution trend, improving the reliability and foresight of the prediction.

[0049] After damage fusion is completed, step S6 is executed, which calculates the remaining fatigue life of the anchor chain based on the comprehensive damage index. Remaining life prediction is the ultimate goal of the entire method. By combining the comprehensive damage index and the damage growth rate, the time required for the anchor chain to reach the failure state can be extrapolated, thereby providing a quantitative basis for maintenance decisions.

[0050] In step S61, fatigue failure criteria are set. The critical value of the comprehensive damage index is set to... when When this value is reached, the anchor chain is considered to have experienced fatigue failure. The failure threshold is set at 0.95 instead of 1.0 because of safety margin considerations. When the comprehensive damage index reaches 0.95, although the anchor chain has not completely failed, there is a significant risk of failure, and continued use may lead to a sudden breakage accident. Therefore, it should be replaced or repaired promptly. Setting the failure threshold at 0.95 also takes into account the uncertainty of model predictions. Damage calculation models inevitably contain errors, and setting a certain safety margin can reduce the risk of failure caused by model errors. By establishing a quantitative standard for failure prediction in this way, the remaining service life prediction has a clear basis for judgment. Using the same failure criterion for different operators and different time periods ensures the consistency and comparability of the prediction results.

[0051] In step S62, according to the formula Calculate the remaining fatigue life, where The remaining fatigue life (h) represents the time required from the current moment until the anchor chain reaches the critical failure value; The critical damage value (dimensionless) is set to 0.95; The current comprehensive damage index (dimensionless); The damage growth rate (1 / h) is used. This formula is based on linear extrapolation, assuming that the future damage growth rate is consistent with the current damage growth rate. Under this assumption, the remaining lifespan equals the remaining damage amount divided by the damage growth rate. This linear extrapolation method is simple and intuitive, suitable for situations where the damage growth rate is relatively stable. However, it should be noted that if the future load environment or marine environment changes significantly, the damage growth rate will also change accordingly. In this case, the assumption of linear extrapolation no longer holds, and the prediction accuracy will decrease. To improve the accuracy of the prediction, the damage growth rate can be smoothed using the moving average method or exponential smoothing method to reduce the impact of instantaneous fluctuations. Alternatively, a correlation model between the damage growth rate and load parameters and environmental parameters can be established, and the damage growth rate can be dynamically adjusted based on future sea state forecasts. This method yields a prediction of the remaining service life of the anchor chain, with the result in hours. This can be converted to days or months for easy engineering applications, providing the operation and maintenance department with a clear replacement time window.

[0052] In step S63, uncertainty analysis is performed. Considering measurement errors and model uncertainties, the confidence interval for the remaining life is calculated, with a confidence level set to 95%. The purpose of uncertainty analysis is to quantify the reliability of the prediction results, because the damage calculation process involves multiple parameters and models, and errors exist at each stage, which will propagate to the final life prediction result. Measurement errors originate from sensor accuracy limitations and noise in the data acquisition system. For example, the measurement accuracy of strain sensors is typically ±2% to ±5%, and the measurement accuracy of load sensors is typically ±1% to ±3%. These errors will cause deviations in the calculated strain amplitude and load amplitude, thus affecting the accuracy of the damage value. Model uncertainties originate from empirical parameters and simplifying assumptions in the damage calculation formula, such as material constants. and The values ​​of these uncertainties are typically based on limited experimental data, which introduces statistical errors. The Miner linear cumulative damage criterion ignores complex factors such as load sequence effects and overload effects, also contributing to model errors. Monte Carlo simulation can be used to assess the impact of these uncertainties on remaining lifetime prediction. The specific steps are: treating each uncertainty parameter as a random variable, generating a large number of random samples based on their statistical characteristics, calculating the remaining lifetime for each sample to obtain its probability distribution, and then calculating a 95% confidence interval based on the probability distribution. A 95% confidence interval indicates a 95% probability that the actual remaining lifetime falls within this interval. The width of the confidence interval reflects the degree of uncertainty in the prediction; a wider interval indicates greater uncertainty. This method provides a reliability assessment of the remaining lifetime prediction results, enabling operation and maintenance personnel to not only know the predicted value of the remaining lifetime but also understand its reliability, thus making more prudent maintenance decisions.

[0053] In step S64, a lifespan early warning mechanism is established. An early warning signal is generated when the remaining fatigue life falls below a preset safety threshold, enabling proactive monitoring and risk control of anchor chain fatigue. The preset safety threshold needs to comprehensively consider factors such as the importance of the anchor chain, replacement costs, and maintenance cycles. For critical anchor chain systems, the safety threshold can be set relatively high, for example, 720 hours (30 days), to allow sufficient time for maintenance. For general anchor chain systems, the safety threshold can be set relatively low, for example, 168 hours (7 days). The early warning signal can employ a multi-level mechanism, such as setting yellow and red warning levels. A yellow warning is triggered when the remaining lifespan falls below 150% of the safety threshold, prompting operators to monitor the anchor chain status and prepare a maintenance plan. A red warning is triggered when the remaining lifespan falls below the safety threshold, requiring immediate measures to replace or repair the anchor chain. The early warning signal can be sent in various ways, including audible and visual alarms, SMS notifications, and email alerts, to ensure relevant personnel receive the warning information promptly. Furthermore, the early warning mechanism should include early warning records and emergency plans. Records should document the time of each early warning, remaining lifespan, and measures taken to provide a reference for subsequent maintenance and management. Emergency plans should clearly define the specific operational procedures and responsible personnel upon receiving an early warning to ensure an effective response. This approach enables proactive monitoring and risk control of anchor chain fatigue, transforming reactive, post-event maintenance into proactive preventative maintenance. This significantly reduces the risk of sudden anchor chain failure and improves the safety and reliability of offshore platforms.

[0054] After completing the life prediction, proceed to step S7, which outputs the fatigue life prediction results and generates a prediction report containing remaining life values ​​and condition assessments. The output of the prediction results should include not only numerical results but also detailed analysis reports and visualizations to enable operations and maintenance personnel to fully understand the fatigue condition and risk level of the anchor chain.

[0055] In step S71, the remaining lifetime value is output in a formatted manner. The calculated... The remaining lifetime is standardized and represented in hours, retaining two decimal places to generate a standard format for lifetime prediction. Specifically, the remaining lifetime value should include both the predicted value and the confidence interval, such as remaining lifetime: 1536.42 hours (95% confidence interval: 1420.35-1652.49 hours). This representation provides both the most probable remaining lifetime value and the range of uncertainty, allowing users to make decisions based on their risk tolerance. Furthermore, the hours can be converted to more intuitive days or months, such as remaining lifetime: 64.02 days (approximately 2.13 months), facilitating understanding and application by engineers. The remaining lifetime value should also indicate the calculation time, such as "As of 12:00 on October 29, 2025, remaining lifetime is 1536.42 hours," so users clearly know the corresponding time point for the prediction result. This standardized output format ensures that the prediction results are clearly, accurately, and easily understood, avoiding misunderstandings and misuse due to improper expression.

[0056] In step S72, a damage status assessment report is generated. This report includes the current comprehensive damage indicators. Damage components at various scales , , The report should include the following: First, a basic information section, recording the anchor chain's number, installation location, commissioning time, and cumulative operating time. Second, a current damage status section, listing the current comprehensive damage index value and sub-damage values ​​at the micro, meso, and macro scales, visually displaying the damage extent as a percentage or progress bar. Third, a damage evolution trend section, analyzing damage growth over a past period, such as the most recent month, calculating the average and maximum damage growth rates, and identifying periods of rapid damage growth. Fourth, an environmental impact analysis section, presenting current and historical marine environmental parameters, assessing the impact of temperature and corrosive environments on the damage. Finally, a risk assessment section, determining the anchor chain's risk level based on the current damage level and remaining lifespan, for example, classifying it into low, medium, and high risk levels, and providing corresponding maintenance recommendations. This detailed damage status assessment report provides operation and maintenance personnel with comprehensive anchor chain health information, helping to develop more scientific and reasonable maintenance strategies.

[0057] In step S73, a life prediction chart is created. Damage evolution curves and remaining life trend charts are plotted to visually represent the changes in the fatigue state of the anchor chain. The damage evolution curve plots the cumulative damage process from commissioning to the present, with time as the horizontal axis and comprehensive damage indicators as the vertical axis. Microscopic, mesoscopic, and macroscopic damage curves and the comprehensive damage curve can be plotted simultaneously on the same chart, distinguished by different colors and line types, allowing for a clear view of the relative size and trend of damage at each scale. Key event points, such as major sea state events, maintenance operations, and abnormal alarms, can also be marked on the damage evolution curve to help analyze the impact of these events on damage evolution. The remaining life trend chart plots the remaining life over time, with time as the horizontal axis and remaining life as the vertical axis. If the damage growth rate remains stable, the remaining life curve should decrease linearly; if the damage growth rate changes, the slope of the remaining life curve will change accordingly. Confidence interval bands can be plotted on the remaining life trend chart, with shaded areas representing the range of uncertainty in the remaining life. Warning threshold lines can also be plotted to clearly show when the remaining life will reach the warning line. In addition, auxiliary charts such as environmental parameter variation diagrams and load spectrum diagrams can be generated to comprehensively display various factors affecting the fatigue life of anchor chains. Through these visualization charts, complex numerical calculation results are transformed into intuitive visual information, enabling non-professionals to quickly understand the fatigue state of anchor chains, thus improving the usability and impact of the prediction results.

[0058] In step S74, maintenance recommendation information is output. Based on the remaining service life and current damage status, corresponding maintenance strategy recommendations are generated, achieving intelligent output of predictive maintenance guidance. The specific content includes: First, maintenance timing recommendations, suggesting the optimal maintenance window based on the predicted remaining life and maintenance schedule, such as recommending anchor chain replacement within the next 30 days, and no later than 60 days. Second, maintenance method recommendations, suggesting different maintenance methods such as replacement, repair, or reinforcement based on the type and extent of damage. If microscopic damage is dominant and minor, surface treatment and anti-corrosion coating repair are recommended. If macroscopic damage is significant or there are obvious cracks, complete replacement is necessary. Third, spare parts preparation recommendations, listing the necessary spare parts, including new anchor chains, shackles, connectors, etc., and recommending specifications and quantities. Fourth, construction requirements recommendations, reminding of precautions during maintenance operations, such as needing to be carried out during periods of low wind and waves, needing to have lifting equipment available, and needing to comply with safe operating procedures. Finally, follow-up monitoring recommendations, suggesting increasing the monitoring frequency after maintenance, doubling the data collection frequency in the first three months after replacing the anchor chains, in order to promptly identify potential problems. By providing intelligent maintenance recommendations, fatigue life prediction results are directly transformed into actionable maintenance guidelines, shortening the time from prediction to decision-making, improving the pertinence and effectiveness of maintenance work, and truly realizing closed-loop management of predictive maintenance.

[0059] Example 1: Example 1 uses a fatigue life prediction method for marine anchor chains based on multi-scale damage accumulation to assess the fatigue life of the anchor chain system of an offshore wind power platform. This offshore wind power platform is located in the East China Sea at a water depth of approximately 45 meters. The platform has a semi-submersible structure and is connected to the seabed anchoring system via eight mooring anchor chains. One main anchor chain was selected as the monitoring object. This anchor chain is an R3 grade mooring chain with a nominal diameter of 76 mm, made of alloy steel, and a breaking load of 5800 kN.

[0060] In step S1 of the method designed according to the present invention, anchor chain operation data is acquired. In step S11, fiber Bragg grating strain sensors are installed in the high-stress areas of the 3rd, 15th, and 28th links of the anchor chain. These locations are identified as areas with the highest stress concentration coefficients through finite element analysis. The sampling frequency is set to 1000Hz, and strain data is continuously collected starting from March 1, 2024, for 180 days, forming a complete strain time series dataset every 24 hours. In step S12, an axial tension sensor and a bending load sensor are installed at the shackle connecting the anchor chain to the buoy at the top. The sampling frequency is also set to 1000Hz, and the data is collected synchronously with the strain data to form a load time series. In step S13, water quality monitoring buoys were deployed within a 5-meter radius around the anchor chain. Seawater temperature, pH value, chloride ion concentration, and dissolved oxygen concentration were collected every 60 seconds. During the monitoring period, the seawater temperature ranged from 12°C to 28°C, the pH value ranged from 7.8 to 8.3, and the chloride ion concentration ranged from 18,000 mg / L to 21,000 mg / L. In step S14, a GPS timing system was used to synchronize the time of all sensors, achieving a synchronization accuracy of 1 millisecond, ensuring the spatiotemporal correspondence of multi-source data. Based on the anchor chain operation data processed in step S2, multi-scale damage values ​​were calculated. In step S21, the strain amplitude was extracted from the 180-day strain time series using the rainflow counting method, identifying approximately 285,000 strain cycles. The strain amplitude distribution ranged from 80 microstrain to 1200 microstrain. A reference strain value of 600 microstrain was taken, and the material constant m was set to 5. The micro-damage value was calculated to be 0.378 according to the micro-damage accumulation formula. In step S22, rainflow counting was performed on the load time series, identifying approximately 282,000 load cycles. Axial tensile force and bending load were converted using the equivalent load formula, with the bending load equivalence coefficient set to 1.8. The load amplitude distribution ranged from 180 kN to 2800 kN. A reference load value of 1740 kN was used, and the load sensitivity index n was set to 4, resulting in a calculated mesoscopic damage value of 0.412. In step S23, based on the SN curve parameters of the anchor chain material, the material constant C was set to 2.8 multiplied by 10^12, and k was set to 3.5. The ratio of the number of cycles to fatigue life at each load level was accumulated, resulting in a calculated macroscopic damage value of 0.425. Environmental correction was performed based on marine environmental parameters as described in step S3. In step S31, the average temperature during the 180-day monitoring period was calculated to be 20.5 degrees Celsius, the temperature sensitivity parameter was set to 4200 K, and the reference temperature was set to 293 K, resulting in a calculated temperature influence coefficient of 1.15. In step S32, the average pH value is 8.1, the average chloride ion concentration is 19,500 mg / L, the pH sensitivity coefficient is 0.08, the chloride ion influence coefficient is 0.0025, and the calculated corrosion influence coefficient is 1.44.In step S33, the temperature influence coefficient and the corrosion influence coefficient are multiplied to obtain a comprehensive environmental correction coefficient of 1.66. In step S34, the damage values ​​at each scale are multiplied by the environmental correction coefficient to obtain corrected microscopic damage values ​​of 0.627, mesoscopic damage values ​​of 0.684, and macroscopic damage values ​​of 0.706. Weight optimization is performed in step S4. In step S41, actual flaw detection data from three other operational anchor chains on the platform are collected as a reference, and an optimization function with the goal of minimizing prediction error is established. In step S42, a constraint condition is set that the sum of the three weight coefficients is equal to 1 and all are greater than zero, and the value range is set to 0.2 to 0.5. In step S43, a particle swarm optimization algorithm is used, with a particle swarm size of 30 particles. After 50 iterations, the optimal weight combination is obtained as a microscopic weight of 0.28, a mesoscopic weight of 0.35, and a macroscopic weight of 0.37. In step S44, a dynamic update mechanism is established, and weight optimization is re-executed every 100 new data points. In step S5, damage fusion is performed. In step S51, according to the weighted fusion formula, the corrected damage values ​​of the three scales are multiplied by their corresponding weight coefficients and summed to obtain a comprehensive damage index of 0.676. In step S52, compared with the comprehensive damage index of 0.672 in the previous hour, the damage growth rate is calculated to be 0.004 per hour. In step S53, the 180-day damage evolution data is stored in time series, forming a complete damage history record. In step S6, the remaining fatigue life is calculated. In step S61, the critical damage value for fatigue failure is set to 0.95. In step S62, according to the remaining life formula, the remaining fatigue life is calculated to be 68.5 hours, approximately 2.85 days. In step S63, uncertainty analysis is performed through Monte Carlo simulation, considering a sensor measurement error of ±3% and a model parameter uncertainty of ±5%, resulting in a 95% confidence interval of 63.2 hours to 73.8 hours. In step S64, because the remaining lifespan is lower than the preset 168-hour safety threshold, the system automatically generates a red warning signal. The prediction results are output in implementation step S7. In step S71, the remaining lifespan is formatted as 68.50 hours, with a confidence interval of 63.20 to 73.80 hours. In step S72, a detailed damage status assessment report is generated, showing the current comprehensive damage index as 0.676, micro-damage as 0.627, meso-damage as 0.684, macro-damage as 0.706, and a damage growth rate of 0.004 per hour. In step S73, a damage evolution curve and a remaining lifespan trend chart are generated. In step S74, maintenance recommendations are output: Emergency replacement work should be arranged within the next 48 hours, and no later than 60 hours; 15 meters of R3 grade 76mm anchor chain and matching shackles should be prepared; the work should be carried out during a window of wind and waves less than level 4.The anchor chain was replaced 65 hours later as predicted. Post-replacement flaw detection revealed multiple microcracks inside the chain links, with the largest crack reaching a depth of 3.2 mm, verifying the accuracy of the prediction. Actual monitoring showed that the anchor chain's total service life was 8732 hours. The remaining life predicted by this method at the 180-day monitoring point was 68.5 hours, while the actual remaining life was 72 hours, resulting in a prediction error of 4.9%. The early warning was provided 72 hours in advance, effectively preventing potential chain breakage accidents.

[0061] Comparative Example 1 uses the traditional Miner linear cumulative damage method for fatigue life prediction. This method only considers the impact of macroscopic load cycles on fatigue life, without involving microscopic and mesoscopic damage analysis, and also ignores the effects of temperature and corrosion in the marine environment. In practice, the same load data as in Example 1 was collected, load cycles were extracted using the rainflow counting method, and the fatigue life at each load level was calculated based on the SN curve of the anchor chain material. The damage components of each load cycle were accumulated according to the Miner criterion, and failure was determined when the cumulative damage reached 1.0. The SN curve parameters used in this method were the same as in Example 1, with the material constant C set to 2.8 x 10^12 and k set to 3.5. Since environmental factors were not considered, the calculated macroscopic damage value was 0.425, without environmental correction. Using a fixed failure criterion, failure was predicted when the damage value reached 1.0, and the calculated remaining life was 135.3 hours.

[0062] Comparative Example 2 employs a dual-scale damage assessment method. This method comprehensively considers load damage at the mesoscale and cumulative damage at the macroscale, but does not include strain damage analysis at the microscale, nor does it perform environmental correction. In practice, strain data and load data are collected simultaneously, and mesoscale and macroscale damage values ​​are calculated separately. These are then weighted and fused using fixed weights of 0.5 and 0.5 respectively to obtain a comprehensive damage index. The mesoscale damage value calculated by this method is 0.412, and the macroscale damage value is 0.425, both without environmental correction. The weighted and fused comprehensive damage index is 0.419. The remaining lifetime is calculated to be 145.3 hours according to the failure criteria.

[0063] Comparative Example 3 employs a multi-scale fixed-weight method, which includes damage analysis at three scales: micro, meso, and macro, and performs environmental corrections. However, the weight coefficients are fixed values ​​without optimization or dynamic updates. In practice, the same strain, load, and environmental data as in Example 1 are collected, and damage values ​​at each of the three scales are calculated. Environmental corrections are then performed using the same method as in Example 1, resulting in corrected micro-damage values ​​of 0.627, meso-damage values ​​of 0.684, and macro-damage values ​​of 0.706. However, during damage fusion, empirically set fixed weights are used: micro-weight 0.33, meso-weight 0.33, and macro-weight 0.34, without optimization. The calculated comprehensive damage index is 0.673, the damage growth rate is 0.0038 per hour, and the predicted remaining lifetime is 72.9 hours.

[0064] Comparative Example 4 employs the stress range-based SN curve method, a commonly used traditional method in engineering. This method directly assesses lifespan based on stress amplitude and cycle count, combined with the material's SN curve. In practice, strain is measured using strain sensors, converted to stress based on the elastic modulus, and rainflow is counted on the stress time history to extract stress amplitude and cycle count. The fatigue life at each stress level is obtained by consulting the SN curve, and damage is accumulated according to the Palmgern-Miner criterion. The elastic modulus of the anchor chain steel used in this method is taken as 210 GPa, and the calculated stress amplitude distribution range is 16.8 MPa to 252 MPa. A modified SN curve suitable for marine environments is used, with curve parameters selected according to DNV specifications. The calculated cumulative damage value is 0.468, and the predicted remaining life is 113.7 hours.

[0065] The comparative experiment used five marine anchor chains under different service environments and load conditions. The experiments were conducted according to the fatigue testing methods specified in GB / T3075-2008 "Marine Engineering Mooring Chains" and the fatigue life assessment requirements in the DNV-OS-E301 "Marine Mooring and Positioning" standard. Data collection and analysis were performed with reference to the monitoring procedures in API RP 2SK "Recommended Practices for the Design and Analysis of Mooring Systems".

[0066] The basic information of the five test anchor chains is as follows: Anchor chain 1 is installed on an offshore wind power platform in water depth of 42 meters. The service environment is moderately corrosive, with an average seawater temperature of 18 degrees Celsius and a chloride ion concentration of 19,000 mg / L. It mainly bears wind and wave loads, with frequent load variations. Anchor chain 2 is installed on an oil drilling platform in water depth of 65 meters. The service environment is highly corrosive, with an average seawater temperature of 26 degrees Celsius and a chloride ion concentration of 22,000 mg / L. It bears a large combination of static and dynamic loads. Anchor chain 3 is installed on a jacket platform in water depth of 28 meters. The service environment is slightly corrosive, with an average seawater temperature of 10 degrees Celsius and a chloride ion concentration of 17,000 mg / L. The load is relatively stable. Anchor chain 4 is installed on a floating production storage and offloading (FPSO) unit in water depth of 55 meters. The service environment is moderately corrosive, with an average seawater temperature of 15 degrees Celsius and a chloride ion concentration of 18,500 mg / L. It bears complex multi-directional loads. Anchor chain 5 was installed on a moored buoy at a water depth of 38 meters. The service environment was moderately corrosive, with an average seawater temperature of 20 degrees Celsius, a chloride ion concentration of 19,500 mg / L, and significant load fluctuations. Fatigue life predictions were performed on the five anchor chains using the methods described in Example 1 and Comparative Examples 1 to 4. Predictions were conducted when the anchor chains had reached approximately 60% to 70% of their service life to verify the predictive ability of each method for remaining life. After prediction, the anchor chains were monitored until they reached the failure criteria or underwent planned replacement. The actual damage state was determined through non-destructive testing, thus obtaining the actual remaining life data. Experimental indicators included prediction life accuracy, prediction error rate, damage identification accuracy, early warning time, and computational efficiency. Predicted life accuracy was evaluated by comparing the closeness between the predicted and actual life, expressed as the ratio of predicted to actual life; a ratio closer to 1 indicates a more accurate prediction. The prediction error rate was defined as the absolute value of the difference between the predicted and actual life divided by the actual life and multiplied by 100%; a smaller error rate indicates higher prediction accuracy. Damage identification accuracy is evaluated by comparing the predicted comprehensive damage index with the actual damage level assessed by non-destructive testing. The correlation coefficient between the predicted and actual damage values ​​is used; a correlation coefficient closer to 1 indicates more accurate identification. Early warning lead time refers to the time interval from issuing a warning signal to the actual failure of the anchor chain or reaching a point where replacement is necessary. A larger lead time indicates more timely warnings and allows sufficient time for maintenance decisions. However, an excessively large lead time may lead to unnecessary conservatism; therefore, a balance needs to be struck between safety and economy. Generally, a lead time of 48 to 168 hours is considered reasonable. Computational efficiency is expressed as the time required to complete one full prediction calculation; a shorter time indicates higher efficiency. The experimental data acquisition process is as follows: Predicted life data is automatically generated by the calculation programs of each method. The calculation process strictly follows the algorithm steps of each method, and all methods use the same raw monitoring data to ensure comparability.Actual lifespan data is obtained through long-term continuous monitoring and periodic non-destructive testing. Monitoring is conducted weekly, using a combination of ultrasonic and magnetic particle testing to detect cracks inside and on the surface of the chain links. When a crack depth exceeds 5% of the chain link diameter or obvious fatigue damage characteristics appear on the surface, the anchor chain is considered to have reached failure. The difference between the total operating time at this point and the operating time at the predicted time is the actual remaining lifespan. Actual damage identification data is obtained by converting non-destructive testing results. Based on the detected crack size, number, and distribution, equivalent damage values ​​are calculated using fracture mechanics methods, serving as a reference standard for actual damage. Early warning time is obtained by recording the time difference between the time each method issues an early warning signal and the actual time of anchor chain failure. Computational efficiency is obtained by running the programs for each method on a standard industrial computer with an Intel Core i7 processor, 16GB of RAM, and running Windows 10.

[0067] Experimental results are as follows Figures 1-6 As shown. Figure 1 The comparison between the actual lifespan and the predicted lifespan of the five anchor chains was presented in the form of grouped bar charts. The actual values ​​and the predicted values ​​were clearly distinguished by different colored bars. It can be seen intuitively that the black bar, which is closest to the actual lifespan in Example 1, verifies the accuracy of the prediction. Figure 2 The prediction error rates of each method on 5 anchor chains are compared, and a 5% acceptable error reference line is plotted. The error bars of Example 1 are significantly lower than this reference line and much lower than the comparative example, demonstrating the high-precision prediction capability. Figure 3 The damage identification accuracy of each method was compared, and a 90% excellent standard reference line was drawn. Only Example 1 and Comparative Example 3 exceeded this standard line, and Example 1 reached 94.5%, ranking first, which reflects the advantages of multi-scale analysis and optimized weights. Figure 4 The study compared the advance warning times and plotted a minimum requirement reference line of 48 hours. In Example 1, the warning time on all anchor chains significantly exceeded this requirement and had the highest average value, demonstrating the timely warning capability. Figure 5 The predicted lifespan trends of the five anchor chains were displayed using a broken line scatter plot. The red predicted curve in Example 1 is closest to the actual lifespan of the black curve, while the curve in the comparative example deviates significantly, clearly demonstrating the consistency of the predictions. Figure 6A 2x2 subplot layout was used to comprehensively display four key indicators: average prediction error rate, damage identification accuracy, average warning lead time, and computational efficiency. Example 1 showed the best performance in the first three indicators. Although its computational efficiency was slightly lower than some comparative examples, it was still within an acceptable range, fully demonstrating the overall superiority of the method. All figures and tables are labeled in Chinese, distinguished by different colors and shapes of legends, and the titles clearly explain the content of each figure. The numerical labels are clear, facilitating readers' understanding and analysis of the experimental results.

[0068] from Figure 1 The bar chart comparing predicted and actual lifespans shows that the actual lifespans of the five anchor chains were 8732 hours, 12180 hours, 6856 hours, 15240 hours, and 9384 hours, respectively. These values, confirmed through long-term monitoring and non-destructive testing, represent the true fatigue failure times. Example 1 predicted lifespans of 8515 hours, 11890 hours, 6920 hours, 14950 hours, and 9180 hours for these five anchor chains. It can be observed that the red predicted lifespan bar of Example 1 is very close in height to the black actual lifespan bar, with minimal deviation. In contrast, the blue bar of Comparative Example 1 is significantly higher than the actual lifespan bar, exhibiting a systematic overestimation trend. Comparative Examples 2 and 4 also show varying degrees of deviation in bar height. Only the bar height of Comparative Example 3 is relatively close to that of Example 1. It can be seen that Example 1 has the highest degree of agreement between predicted and actual lifespans, with significantly better prediction accuracy than most comparative examples. Traditional methods, such as Comparative Example 1, only consider macroscopic-scale load cycles, neglecting strain damage at the microscopic level of the material and local stress concentration effects at the mesoscopic level of the structure, resulting in an incomplete understanding of the fatigue damage mechanism of anchor chains. The fatigue failure of anchor chains is a multi-level, multi-scale process. At the microscopic level, dislocation movement and slip band formation within grains are the root causes of fatigue crack initiation; at the mesoscopic level, stress concentration at the bending points of the chain links leads to the accumulation of local plastic deformation; and at the macroscopic level, the overall load cycle determines the crack propagation rate. These three damage mechanisms are coupled and jointly determine the final fatigue life. This invention simultaneously monitors strain and load data, calculates damage values ​​at the microscopic, mesoscopic, and macroscopic levels respectively, and then fuses them using optimized weighting coefficients, thus capturing a more complete picture of anchor chain fatigue damage. Therefore, the predicted results are highly consistent with the actual situation. Comparative Example 1, lacking information at the micro and meso levels, is equivalent to evaluating a complex fatigue process from a one-sided perspective, which naturally leads to a large systematic error. This is especially true in marine environments, where the effects of micro-strain accumulation and local stress concentration are more significant, and relying solely on macro-load data will inevitably lead to prediction bias.

[0069] from Figure 2The prediction error rate comparison bar chart shows that the prediction error rates of Example 1 on the five anchor chains are 2.5%, 2.4%, 0.9%, 1.9%, and 2.2%, respectively, with an average error rate of only 2.0%. All error values ​​are far below the 5% acceptable error reference line marked in the chart. The error rate of Comparative Example 1 ranges from 17.4% to 21.5%, with an average of 19.0%, far exceeding the acceptable range. The average error rate of Comparative Example 2 is 11.7%, and that of Comparative Example 4 is 9.0%, both significantly higher than that of Example 1. Only the average error rate of Comparative Example 3, at 1.9%, is close to that of Example 1. It can be seen that Example 1 not only has accurate predictions but also good stability, maintaining a low error rate under different service environments and load conditions. Comparative Examples 1 and 2 did not consider the influence of temperature changes and corrosion in the marine environment on fatigue performance; they used material parameters and SN curves under standard conditions, while the actual marine environment differs significantly from the laboratory standard environment. The effect of temperature on the fatigue performance of metallic materials follows the Arrhenius thermal activation theory. Increased temperature increases the thermal energy of atoms, reduces the resistance to dislocation movement, and makes the material more susceptible to plastic deformation, thus reducing fatigue strength. This effect can be quantified using a temperature influence coefficient, which describes the relationship between temperature and fatigue life through an exponential function. The effect of seawater corrosion on fatigue performance is more complex. Chloride ions create pitting corrosion on metal surfaces, which become stress concentration sources and crack initiation points, accelerating fatigue crack formation. Simultaneously, the penetration of corrosive media into the crack tip reduces fracture toughness and accelerates crack propagation. This synergistic effect of corrosion fatigue results in a fatigue life in marine environments that is significantly lower than in air, which can be corrected using a corrosion influence coefficient. This invention monitors seawater temperature, pH, and chloride ion concentration in real time, calculates the temperature influence coefficient and corrosion influence coefficient, and then multiplies them to obtain a comprehensive environmental correction coefficient. This coefficient is used to correct damage values ​​at various scales, effectively adjusting damage calculations under laboratory conditions to actual marine environmental conditions, eliminating systematic errors caused by environmental differences. Experimental data show that the comprehensive environmental correction coefficient for the five anchor chains ranges from 1.5 to 1.8, meaning that the marine environment accelerates fatigue damage by 50% to 80%. Without environmental correction, the predicted lifespan would be significantly higher, which is the fundamental reason for the high error rates in Comparative Examples 1 and 2. Although Comparative Example 3 also underwent multi-scale analysis and environmental correction, its average error rate was similar to that of Example 1, indicating that the environmental correction mechanism does indeed play a crucial role in improving prediction accuracy.

[0070] from Figure 3The damage identification accuracy comparison bar chart shows that Example 1 achieved a damage identification accuracy of 94.5%, significantly exceeding the 90% excellent standard marked in the chart. Comparative Example 3's accuracy of 89.3% is close to the excellent standard but slightly lower. Comparative Examples 2 and 4 are at a moderate level of 75.8% and 78.6% respectively, while Comparative Example 1, at only 68.2%, is significantly lower than other methods. Damage identification accuracy reflects the correlation between the predicted comprehensive damage index and the actual damage degree assessed by non-destructive testing. A higher index indicates a more accurate reflection of the true damage state of the anchor chain by the prediction model. Example 1 provides the most accurate assessment of the current damage state of the anchor chain, accurately identifying different damage levels and types. In multi-scale damage fusion, determining the weighting coefficients for damage at each scale directly affects the accuracy of the comprehensive damage index. Although Comparative Example 3 also employs multi-scale analysis and environmental correction, its weighting coefficients are fixed based on experience, with weights of 0.33 for microscale, 0.33 for mesoscale, and 0.34 for macroscale. This equal allocation method is simple but not precise enough because the relative importance of damage at each scale changes at different stages of fatigue damage. In the early stage of fatigue, before microcracks form, damage is mainly manifested as strain accumulation and increased dislocation density at the microscale, at which point the weight of microscale damage should be relatively large. In the middle stage of fatigue, local plastic deformation and stress concentration at the mesoscale gradually appear, increasing the importance of mesoscale damage. In the later stage of fatigue, crack propagation at the macroscale becomes the dominant mechanism, and the weight of macroscale damage should increase. This invention uses a particle swarm optimization algorithm to minimize prediction error, using actual observed damage data as a reference, and finding the optimal weight combination through iterative search. The determined weighting coefficients maximize the consistency between the comprehensive damage index and the actual damage state. More importantly, this invention establishes a dynamic weight update mechanism. The optimization process is re-executed every 100 new data points, allowing the weight coefficients to adaptively adjust as the damage evolves. This dynamism ensures that the comprehensive damage index accurately reflects the true state of the anchor chain throughout its service life. Experimental results show that the optimized weight combination in Example 1 (0.28 for micro-level, 0.35 for meso-level, and 0.37 for macro-level) is not significantly different from the fixed weights of Comparative Example 3 (0.33, 0.33, 0.34). However, the optimized weights better reflect the actual damage evolution pattern, thus increasing the damage identification accuracy from 89.3% to 94.5%. This 5.2 percentage point improvement is precisely the effect of the weight optimization mechanism. Comparative Examples 1 and 2, lacking multi-scale information or environmental correction, naturally have lower damage identification accuracy, further validating the necessity of multi-scale analysis and environmental correction.

[0071] from Figure 4The bar chart comparing early warning lead times shows that in Example 1, the early warning lead times for the five anchor chains were 72 hours, 86 hours, 58 hours, 95 hours, and 68 hours, with an average of 75.8 hours. All values ​​are significantly higher than the minimum requirement of 48 hours marked in the chart. Comparative Example 1 had an average early warning lead time of only 19.0 hours, Comparative Example 2 39.0 hours, and Comparative Example 4 46.4 hours, all failing to meet the minimum requirement of 48 hours or just meeting it. Only Comparative Example 3, with an average early warning lead time of 69.0 hours, reached a better level but was still lower than Example 1. Early warning lead time refers to the time interval between the system issuing an early warning signal and the actual failure or necessity of replacing the anchor chain. The longer this time, the more leeway is allowed for maintenance decisions and operational preparations. Example 1 provided timely and accurate early warnings of anchor chain fatigue risks, offering a sufficient time window for preventative maintenance, significantly outperforming other comparative methods. The timing of the early warning depends on two factors: the accuracy of the assessment of the current damage state and the accuracy of the remaining life prediction. These two factors are respectively related to damage identification capabilities and life prediction algorithms. Comparative Example 1, with a prediction error rate as high as 19.0% and a systematic overestimation of lifespan, still gave a high remaining lifespan prediction even when the anchor chain was actually close to failure. The warning signal was issued too late, averaging only 19 hours in advance. This time window was extremely tight, leaving almost no time to arrange maintenance plans, prepare spare parts, wait for suitable sea state windows, and organize construction teams. This easily led to reactive emergency repairs or even missed maintenance opportunities resulting in chain breakage accidents. Comparative Examples 2 and 4 improved the prediction method to some extent, but still had significant prediction errors, and the warning lead time was insufficient. Example 1, with a prediction error rate as low as 2.0% and a damage identification accuracy rate as high as 94.5%, accurately identified high-risk conditions when the anchor chain still had a certain remaining lifespan, issuing warning signals in advance. The average warning lead time reached 75.8 hours, approximately 3.2 days. This time window ensured sufficient preparation time while avoiding unnecessary frequent maintenance due to overly conservative approaches. The lead time of an early warning essentially reflects the foresight and reliability of the prediction method. Foresight stems from an accurate grasp of the damage evolution trend, while reliability comes from a precise assessment of the current damage state. This invention captures comprehensive damage information through multi-scale damage analysis, eliminates system bias through environmental correction, improves fusion accuracy through weight optimization, and grasps evolutionary patterns through damage history records and trend analysis. The comprehensive application of this entire set of technical means enables Example 1 to achieve optimal levels in both prediction accuracy and early warning timeliness, thus achieving an early warning lead time of 75.8 hours. It is worth noting that the early warning lead time of Comparative Example 3 is 69.0 hours, which, although also reaching a good level, is still 6.8 hours less than Example 1. This difference mainly comes from the improvement in prediction accuracy brought about by the weight optimization mechanism, further verifying the important role of weight optimization.

[0072] from Figure 5 The scatter plot comparison of the predicted lifespans of the five anchor chains shows that the graph arranges the five anchor chains sequentially on the horizontal axis, with the vertical axis representing fatigue life. A trend curve is formed by connecting the data points with a broken line. The black solid line represents the actual lifespan curve, exhibiting a fluctuating trend of first rising, then falling, and then rising again, reflecting the differences in service environment and load conditions of different anchor chains. Anchor chains 2 and 4 have longer actual lifespans, while anchor chain 3 has a shorter actual lifespan. The red predicted curve of Example 1 almost overlaps with the black actual curve, and the two curves show a highly consistent trend with very small longitudinal deviations at the five data points. The blue curve of Comparative Example 1 is significantly higher than the actual curve, showing a systematic overestimation, and the curve's fluctuations are greater than the actual curve. The curves of Comparative Example 2 and Comparative Example 4 also show varying degrees of deviation. Only the yellow curve of Comparative Example 3 is relatively close to the actual curve and the curve of Example 1. Example 1 not only predicts accurately on a single anchor chain but also maintains high accuracy on anchor chains under multiple different working conditions, demonstrating good versatility and adaptability. Five anchor chains are installed in different environments with significantly varying environmental conditions. Anchor chain 2 has an average seawater temperature of 26 degrees Celsius and a chloride ion concentration of 22,000 mg / L, classifying it as a highly corrosive environment with a large overall environmental correction coefficient. Anchor chain 3 has an average seawater temperature of only 10 degrees Celsius and a chloride ion concentration of 17,000 mg / L, classifying it as a mildly corrosive environment with a smaller overall environmental correction coefficient. This invention achieves adaptive correction for different environmental conditions by real-time monitoring of the specific environmental parameters of each anchor chain and calculating the environmental correction coefficient separately, thus maintaining prediction accuracy regardless of the environment. Furthermore, different anchor chains bear different load spectra and have different damage evolution patterns. Anchor chain 1 mainly bears wind and wave loads, with frequent load changes, and the micro-scale effect is more prominent in its damage evolution process. Anchor chain 4 bears complex multi-directional loads, and the cumulative effect of the macro-scale is more significant. The weight optimization algorithm of this invention automatically adjusts the weight coefficients based on the actual damage data of each anchor chain, ensuring that the weight configuration matches the damage characteristics of that anchor chain, achieving adaptive optimization for different load conditions. Due to the lack of this adaptive capability, the comparative method may make accurate predictions under certain operating conditions, but the error will be large under other conditions, resulting in large fluctuations in the predicted curve or a systematic deviation from the actual curve. Example 1, through environmental adaptive correction and weight adaptive optimization, maintains stable high accuracy under various operating conditions. This is the fundamental reason why its red curve closely follows the black actual curve, and it is also an important guarantee for the wide engineering applicability of the method of this invention.

[0073] from Figure 6The multi-indicator bar chart, using a 2x2 sub-chart layout, compares the four key indicators—average prediction error rate, damage identification accuracy, average warning lead time, and computational efficiency—among the five methods. The top-left sub-chart shows that Example 1's average error rate is 2.0%, significantly lower than Comparative Example 1's 19.0%, Comparative Example 2's 11.7%, and Comparative Example 4's 9.0%, and roughly on par with Comparative Example 3's 1.9%. The top-right sub-chart shows Example 1's damage identification accuracy at 94.5%, significantly higher than all comparative examples. The bottom-left sub-chart shows Example 1's average warning lead time at 75.8 hours, significantly better than the comparative examples, but its computation time of 3.2 seconds is slightly higher than Comparative Examples 1 and 4. The bottom-right sub-chart shows that Example 1's computational efficiency, while not the fastest, is still within an acceptable range. Example 1 leads comprehensively in the three core indicators of prediction accuracy, damage identification capability, and warning timeliness. Although its computational efficiency is slightly lower, it does not affect its practicality, making it the best overall performer. Comparative Example 1 has a computational efficiency of up to 1.5 seconds, but its prediction error rate is as high as 19.0%. This high efficiency comes at the cost of accuracy, making its practical application value very low. In contrast, Example 1, although its computation time increases to 3.2 seconds, reduces the prediction error rate to 2.0%. This trade-off of a small time cost for a huge improvement in accuracy is very worthwhile. From a technical implementation perspective, the slightly longer computation time of Example 1 is due to the inclusion of more computational steps. First, rainflow counting is performed on the strain and load data to calculate damage values ​​at three scales. Then, environmental correction coefficients are calculated and corrected based on environmental parameters. Next, a particle swarm optimization algorithm is run to search for the optimal weights. Finally, weighted fusion and remaining lifetime calculations are performed. This series of steps increases the computational load, but it is precisely these additional computational steps that bring about a significant improvement in accuracy. Comparative Example 1 only requires one rainflow count and Miner criterion accumulation, resulting in the lowest computational cost but the worst accuracy. Comparative Example 2 adds mesoscopic damage calculation, placing it in the middle in terms of both computational cost and accuracy. Comparative Example 3 already includes multi-scale calculations and environmental corrections, with a computation time of 2.8 seconds, close to Example 1, but its accuracy is slightly lower due to the lack of weight optimization. This comparison shows a positive correlation between computational complexity and prediction accuracy. More refined modeling and more complex algorithms require more computation time but yield more accurate predictions. In applications like fatigue life prediction, prediction accuracy is paramount, and computation time is acceptable as long as it's within a reasonable range. A computation time of 3.2 seconds is perfectly adequate for engineering applications because life prediction is typically performed offline, requiring no millisecond-level real-time response. Even for online prediction, a calculation frequency of once per hour is sufficient, and 3.2 seconds will not be a bottleneck. In summary, Figure 6The performance advantages of Example 1 were fully demonstrated through multi-index comparison, proving the effectiveness of the multi-scale damage analysis, environmental correction mechanism and weight optimization technology of the present invention. Although these innovative technologies increase the computational complexity to a certain extent, the resulting improvement in accuracy and enhancement of function far outweigh the computational cost, achieving overall performance optimization and maximizing engineering practical value.

[0074] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for predicting the fatigue life of marine anchor chains based on multi-scale damage accumulation, characterized in that, Includes the following steps: S1. Obtain anchor chain operation data, including obtaining the strain time series of key parts of the anchor chain through strain measurement, obtaining the load time series of the anchor chain through load measurement, and obtaining the time series of marine environmental parameters through environmental monitoring. S2. Process the anchor chain operation data, calculate the damage values ​​of the anchor chain at the micro, meso, and macro scales respectively, and form a multi-scale damage data set. S3. Based on the marine environmental parameters, calculate the environmental impact coefficient, and use the environmental impact coefficient to correct the multi-scale damage data set; S4. Assign weights to the corrected multi-scale damage data and determine the weight coefficients for damage at the micro, meso, and macro scales. S5. Based on the weighting coefficients, the multi-scale damage data are fused into a comprehensive damage index; S6. Based on the comprehensive damage index, calculate the remaining fatigue life of the anchor chain; S7. Output fatigue life prediction results and generate a prediction report containing remaining life values ​​and condition assessment.

2. The method according to claim 1, characterized in that, Step S1 specifically includes: S11. Install strain sensors in the high-stress areas of each link of the anchor chain, set the sampling frequency to 1000Hz, continuously collect strain data ε, and record for no less than 24 hours to obtain a strain time series dataset. S12. Install a load sensor at the connection between the anchor chain and the mooring system to synchronously collect axial tensile force. and bending load The sampling frequency is kept consistent with the strain data to form a load time series dataset. S13. Deploy environmental monitoring equipment in the waters surrounding the anchor chain and collect seawater temperature data every 60 seconds. pH value, chloride ion concentration and dissolved oxygen concentration This forms a time series of environmental parameters; S14. Perform timestamp alignment on all collected data to establish a unified time benchmark, form a synchronized multi-source data stream, and achieve spatiotemporal synchronization matching of data.

3. The method according to claim 1, characterized in that, Step S2 specifically includes: S21. Calculate the micro-damage value based on strain data. Extract the strain amplitude Δε from the strain time series using the rainflow counting method, and then calculate it according to the formula. Calculate the cumulative microscopic damage; where Δε represents the microscopic damage value; Δε represents the strain amplitude. The reference strain value is Σ; m is a material constant; Σ is the summation operator, which yields the quantification results of microscale damage. S22. Calculate the mesoscopic damage value based on the load data, extract the load cycle characteristics, and apply the formula... ; Calculate the cumulative damage at the microscopic level; where ΔF represents the microscopic damage value; ΔF represents the load amplitude. The reference load value is n; n is the load sensitivity index, enabling quantitative assessment of damage at the mesoscale. S23. Calculate the macroscopic damage value based on the combined effect of strain and load, according to the formula. Calculate the macroscopic cumulative damage; where This represents the macroscopic damage value. This represents the actual number of cycles for the i-th load cycle. Σ represents the fatigue life under the corresponding load level; Σ represents the sum of all load levels, yielding a macroscopic assessment of cumulative damage.

4. The method according to claim 1, characterized in that, Step S3 specifically includes: S31. Calculate the temperature environment influence coefficient according to the formula. Calculate the temperature correction factor; where This is the temperature influence coefficient; This is a temperature-sensitive parameter; The reference temperature is T; the actual temperature is T; exp is the natural exponential function; this yields the quantitative results of the effect of temperature on damage. S32. Calculate the corrosion environment influence coefficient according to the formula. Calculate the corrosion correction factor; where This is the corrosion influence coefficient; pH sensitivity coefficient; This represents the chloride ion influence coefficient, enabling the quantification of the impact of marine corrosive environments on damage. S33. Calculate the comprehensive environmental correction factor according to the formula. Calculate the total environmental impact factor; where This is used as a comprehensive environmental correction factor to obtain a comprehensive environmental impact assessment involving multiple factors. S34. Perform environmental corrections on the damage values ​​at each scale, and calculate them separately: ; in , , These represent the micro, meso, and macro damage values ​​after environmental correction, achieving unified correction of multi-scale damage caused by environmental factors.

5. The method according to claim 1, characterized in that, Step S4 specifically includes: S41. Set the weight optimization objective function, establish the optimization function with the goal of minimizing the prediction error, collect recent actual damage observation data as a reference benchmark, and establish the mapping relationship between weight coefficients and prediction accuracy. S42. Set the weighting coefficient constraints, requiring... And all weight coefficients are greater than 0, among which , , The weighting coefficients are respectively for microscopic, mesoscopic, and macroscopic damage to ensure the rationality of the weighting allocation; S43. Use an optimization algorithm to search for the optimal weight combination. Find the weight coefficient combination that minimizes the prediction error through iterative calculation. Set the number of iterations to 50 to obtain the optimal weight configuration under the current working condition. S44. Establish a dynamic weight update mechanism. After accumulating 100 new data points, re-execute the weight optimization process to update the weight coefficients to adapt to the changing trend of damage evolution and achieve adaptive adjustment of weight allocation.

6. The method according to claim 1, characterized in that, Step S5 specifically includes: S51, according to the formula Calculate the comprehensive damage index; among which As a comprehensive damage index; , , These are the corresponding weighting coefficients; , , The corrected damage values ​​at each scale are used to obtain an overall damage assessment of the anchor chain. S52. Calculate the damage growth rate using the formula. Calculate the current rate of damage growth; where The damage growth rate; This represents the comprehensive damage index at the current moment; The comprehensive damage index is the value at the previous moment; Δt is the time interval; the rate of damage accumulation is determined. S53. Establish a damage history record, store the calculated comprehensive damage index and damage growth rate in time series to form a damage evolution trend curve, and provide historical data support for subsequent life prediction.

7. The method according to claim 1, characterized in that, Step S6 specifically includes: S61. Set fatigue failure criteria, and set the critical value of the comprehensive damage index as... ,when When this value is reached, fatigue failure of the anchor chain is determined, and a quantitative standard for failure prediction is established. S62, according to the formula Calculate the remaining fatigue life; where The remaining fatigue life; This is the critical damage value; This represents the current comprehensive damage index; For the damage growth rate; obtain the prediction of the remaining service life of the anchor chain; S63. Perform uncertainty analysis, considering measurement errors and model uncertainties, calculate the confidence interval of the remaining lifetime, set the confidence level to 95%, and provide a reliability assessment for the lifetime prediction results; S64. Establish a lifespan early warning mechanism to generate an early warning signal when the remaining fatigue life is lower than the preset safety threshold, thereby realizing proactive monitoring and risk prevention of anchor chain fatigue status.

8. The method according to claim 1, characterized in that, Step S7 specifically includes: S71. Format the output of the remaining lifespan value, and output the calculated value. The life prediction values ​​are standardized and represented in hours, retaining two decimal places of precision, to generate standard format life prediction values. S72. Generate a damage status assessment report, including current comprehensive damage indicators. The damage components at each scale and the damage growth trend are used to form a detailed damage status analysis document. S73. Create a life prediction chart, draw a damage evolution curve and a remaining life trend chart to visualize the changes in the fatigue state of the anchor chain. S74. Output maintenance recommendation information, generate corresponding maintenance strategy recommendations based on remaining life and current damage status, and realize predictive maintenance guidance output.

Images