Urban rail vehicle preventive maintenance cycle adaptive adjustment system considering life cycle cost

Abstract

This invention discloses an adaptive adjustment system for preventive maintenance cycles of urban rail vehicles that considers the entire life cycle cost. The system includes a data acquisition and preprocessing module, a real-time degradation assessment and prediction module, a dynamic life cycle cost calculation module, and an adaptive optimization and decision-making module. Through real-time degradation assessment and prediction, the system adjusts maintenance cycles based on the actual health evolution trend of components. Unlike existing technologies that only focus on reliability or single-use costs, this invention incorporates residual value loss and downtime losses into dynamic calculations, aiming for optimal life cycle cost. This maximizes economic benefits while ensuring a safety baseline. The system not only extrapolates the future from the current state but also continuously corrects model parameters through a strategy closed-loop verification mechanism, enabling the system to have self-learning capabilities that evolve with the vehicle's entire life cycle, ensuring the accuracy of long-term decisions.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for urban rail transit vehicles, specifically to an adaptive adjustment system for preventive maintenance cycles of urban rail vehicles that considers the entire life cycle cost. Background Technology

[0002] Urban rail vehicles are the core hub equipment of urban public transportation systems, and their safety, stability, and efficiency directly determine the operational quality and service level of urban rail transit. With the continuous expansion of the domestic urban rail network, the increasing number of vehicles, and the growing service life of vehicles, the aging and deterioration of core components are becoming increasingly prominent, leading to a continuous increase in maintenance and management pressure. Preventive maintenance, as a core means of preventing urban rail vehicle failures in advance, extending equipment service life, and ensuring operational continuity, is a core component of the current urban rail operation and maintenance system.

[0003] Currently, the vast majority of urban rail transit operators in China adopt a fixed-cycle planned preventive maintenance model. This model, based on vehicle manufacturer's technical manuals and industry standards, uses vehicle mileage and cumulative operating time as the sole criterion to set uniform, fixed maintenance cycles for regular inspections, maintenance, and replacements of key components such as bogies, wheelset bearings, traction inverters, and braking systems. While this model is simple and convenient to implement, and suitable for early network conditions and uniform vehicle service status, its technical shortcomings are becoming increasingly prominent in today's complex and ever-changing operating environment. Specifically, it suffers from the following core problems: 1. The dual contradiction of over-maintenance and under-maintenance is prominent. The actual deterioration rate of urban rail vehicle components is not a constant value, but is affected by a combination of dynamic factors such as track gradient, curve radius, track smoothness, passenger load, high and low temperature climate, humid and dusty environment, and start-stop frequency. A uniform and fixed maintenance cycle cannot adapt to the differentiated deterioration state of individual vehicles and individual components: For vehicles with good operating conditions and healthy components, fixed-cycle maintenance will lead to frequent disassembly and inspection, over-maintenance, and premature replacement, significantly increasing manpower, material, and downtime maintenance costs, resulting in serious waste of maintenance resources; For vehicles that are in heavy load, harsh road conditions, and extreme climate environments for a long time, the component deterioration rate is faster, and the fixed cycle cannot match their aging progress, which can easily lead to maintenance delays and under-maintenance, causing hidden faults such as excessive component wear and functional failure, and thus resulting in safety and operational risks such as vehicle operation failures, line delays, and temporary shutdowns.

[0004] 2. Maintenance decisions lack a holistic consideration of lifecycle cost (LCC). Existing traditional maintenance models and some existing condition-based maintenance technologies have significant limitations in their decision-making objectives: most solutions focus solely on "maximizing vehicle operational reliability," indiscriminately shortening maintenance cycles and increasing maintenance frequency, leading to persistently high operation and maintenance costs; a few solutions only pursue "lowest cost per maintenance," blindly extending maintenance cycles and ignoring hidden costs such as subsequent fault repair, equipment downtime, and lifespan depreciation. Existing technologies generally sever the correlation between preventative maintenance costs, post-fault repair costs, operational downtime losses, and component residual value depreciation, failing to construct a multi-dimensional cost equilibrium accounting system from the perspective of the entire service life of the vehicle and components, thus failing to achieve the optimal synergy between safety, reliability, and economy.

[0005] 3. The maintenance cycle adjustment mechanism is outdated and lacks intelligence. Existing maintenance technologies with cycle adjustment capabilities are mostly reactive, adjusting only after a component fails, health indicators severely exceed limits, or maintenance costs surge abnormally. Furthermore, the adjustment logic often relies on single state thresholds or the experience of maintenance personnel, lacking the ability to predict the sequential evolution of component degradation. Simultaneously, existing systems lack a dynamic adaptive iteration mechanism, failing to continuously optimize maintenance strategies based on long-term vehicle service conditions and state changes. This results in low cycle adjustment accuracy, significant lag, and difficulty in adapting to the dynamic degradation patterns throughout the vehicle's lifecycle.

[0006] 4. The operation and maintenance system lacks closed-loop evolution capabilities, resulting in insufficient long-term decision-making accuracy. Traditional operation and maintenance systems operate on an open-loop decision-making model. After maintenance strategies are executed, there is no data feedback or parameter correction process. This prevents the accumulation of component degradation and cost data under different operating conditions and maintenance strategies. Model parameters remain fixed over time, and as vehicle service life increases, prediction and decision-making biases accumulate, making it difficult to adapt to the degradation characteristics of older vehicles. Therefore, those skilled in the art have provided an adaptive adjustment system for the preventive maintenance cycle of urban rail vehicles that considers the entire lifecycle cost to address the problems mentioned in the background. Summary of the Invention

[0007] The purpose of this invention is to provide an adaptive adjustment system for the preventive maintenance cycle of urban rail vehicles that takes into account the total life cycle cost, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution: An adaptive adjustment system for preventive maintenance cycles of urban rail vehicles, considering full life-cycle costs, includes: The data acquisition and preprocessing module is used to collect real-time operating status data, environmental data, and historical maintenance data of key components of urban rail vehicles, and to perform data cleaning and feature extraction. The real-time degradation assessment and prediction module is used to build a component degradation model based on preprocessed data, calculate the current health index, and predict the degradation trajectory within a set future time period. The dynamic lifecycle cost accounting module is used to dynamically calculate the comprehensive lifecycle cost (LCC) based on the predicted degradation trajectory and the preset maintenance cycle. The adaptive optimization and decision-making module is used to solve for the optimal preventive maintenance cycle with the goal of minimizing the total life cycle cost (LCC) and the operational reliability threshold of key vehicle components as constraints, and then issue maintenance instructions.

[0009] As a further aspect of the present invention: in the real-time degradation assessment and prediction module, a TCN-LSTM model based on the fusion of temporal convolutional network and long short-term memory network is used to predict the degradation trajectory, and environmental and load covariates are used as inputs to the model attention mechanism to output the expected health index H(t) and its confidence interval at any future time t.

[0010] As a further aspect of the present invention: in the dynamic lifecycle cost accounting module, the comprehensive LCC accounting model is as follows: ; Where T represents the preventive maintenance cycle to be optimized; For preventative maintenance costs; The fault repair cost is calculated based on the probability of exceeding the health index limit output by the degradation trajectory prediction module. Downtime loss costs include operational losses caused by both planned and unplanned downtime; Residual value depreciation cost represents the waste of a component's remaining lifespan due to improper maintenance timing.

[0011] As a further aspect of the present invention: the residual value depreciation cost The calculation formula is: where,

[0012] in, H(T) represents the cost of purchasing new components, and H(T) represents the predicted health index at the end of period T. This is the critical health index threshold for component failure.

[0013] As a further aspect of the present invention: in the adaptive optimization and decision-making module, the objective function and constraint condition are: minLCC(T)

[0014] Where Hth is the critical health index threshold for component failure, and α is the system's preset reliability confidence level.

[0015] As a further aspect of the present invention: the adaptive optimization and decision-making module uses an improved whale optimization algorithm that introduces a nonlinear convergence factor to solve for the optimal preventive maintenance cycle.

[0016] As a further aspect of the present invention, it also includes a strategy closed-loop verification module, which compares the optimal cycle output by the adaptive optimization and decision-making module with the current execution cycle. If the difference exceeds the set lag threshold, the cycle adjustment is triggered, and the adjusted actual operating data and cost data are fed back to the data acquisition and preprocessing module and the full life cycle cost dynamic accounting module to realize the closed-loop self-updating of model parameters.

[0017] As a further aspect of the present invention: for mechanical wear-related components and electrical fatigue-related components, the system can adaptively adjust the cost accounting weight and deterioration prediction parameters; for electrical fatigue-related components, the rainflow counting method is introduced to statistically measure thermal fatigue damage parameters, enhance the accuracy of fatigue deterioration fitting, and adapt to the failure mechanisms of different types of core components.

[0018] Compared with the prior art, the beneficial effects of the present invention are: By conducting real-time degradation assessment and prediction, the maintenance cycle adjustment is based on the actual health status evolution trend of components, avoiding both "over-maintenance" and "under-maintenance." Unlike existing technologies that only focus on reliability or single-use cost, this invention incorporates residual value loss and downtime losses into dynamic accounting, aiming for optimal life-cycle cost. Under the premise of ensuring safety, it maximizes economic benefits. The system not only extrapolates the future based on the current state but also continuously corrects model parameters through a strategy closed-loop verification mechanism, enabling the system to have self-learning capabilities that evolve with the entire vehicle life cycle, ensuring the accuracy of long-term decisions. Attached Figure Description

[0019] Figure 1 A flowchart illustrating the process framework of an adaptive adjustment system for preventive maintenance cycles of urban rail vehicles, taking into account the total lifecycle cost. Detailed Implementation

[0020] Please see Figure 1 In this embodiment of the invention, the adaptive adjustment system for preventive maintenance cycles of urban rail vehicles, which considers the entire life cycle cost, includes the following steps:

[0021] 1. Data Acquisition and Preprocessing Module: This module forms the data foundation of the system. It is used to collect comprehensive and multi-dimensional operational data of key core components of urban rail vehicles, covering three major categories: status perception data, environmental condition data, and historical operation and maintenance data. The data collection objects include, but are not limited to, easily worn and high-risk core components such as wheelset bearings, bogies, traction inverter IGBT modules, braking units, and traction motors.

[0022] The operational status data includes real-time status parameters such as component vibration signals, temperature signals, voltage and current signals, wear, fatigue damage parameters, operating mileage, number of start-stop cycles, and operating speed; environmental condition data includes external influencing parameters such as track gradient, curve radius, track smoothness, ambient temperature and humidity, dust concentration, rain and snow weather conditions, and passenger flow load; historical maintenance data includes historical maintenance records such as component maintenance time, maintenance type, maintenance cost, fault records, repair plans, replacement records, and service life.

[0023] Meanwhile, this module incorporates standardized data preprocessing algorithms to address issues such as noise interference, missing data, anomalous mutations, and inconsistent dimensions in the original acquired data. It employs wavelet thresholding to reduce signal noise, interpolation to repair missing data, the 3σ criterion to eliminate anomalous mutations, and normalization to unify the dimensions of data across all dimensions. Finally, based on time-domain and frequency-domain signal analysis methods, it extracts core degradation features such as kurtosis, root mean square (RMS), peak factor, junction temperature fluctuation amplitude, and fatigue damage frequency, constructing a standardized high-dimensional feature dataset to provide accurate data support for subsequent degradation prediction and cost accounting.

[0024] 2. Real-time Deterioration Assessment and Prediction Module: This module is the core of the system's state awareness, used to accurately assess the current health status of components and predict their long-term degradation trajectory based on a preprocessed standardized feature dataset. This module innovatively employs a TCN-LSTM fusion deep learning model, combined with an attention mechanism to achieve multi-covariate weighted optimization, overcoming the prediction deficiencies of single models.

[0025] Among them, the Temporal Convolutional Network (TCN) has a powerful ability to extract local features, which can accurately capture short-term local degradation features and abrupt fault information in vibration and temperature signals; the Long Short-Term Memory Network (LSTM) can effectively mine the long-term dependencies of time-series data and accurately fit the temporal patterns of slow aging and continuous degradation of components. The fusion of the two models can take into account both short-term anomaly perception and long-term trend prediction, and significantly improve the accuracy of degradation prediction.

[0026] Meanwhile, this module introduces an attention mechanism, using dynamic covariates such as ambient temperature and humidity, line operating conditions, passenger load, and operating frequency as attention weight inputs. Weights are dynamically allocated based on the degree of impact of different operating conditions on component degradation, weakening the interference of irrelevant variables and strengthening the role of core influencing factors. The model ultimately outputs the expected component health index H(t) at any future time t, along with a 95% confidence interval. The uncertainty of the quantitative prediction results provides data support for subsequent reliability constraint determination. The health index ranges from 0 to 1, with an initial health index of 1 for brand-new components and a value of 0 indicating component failure when the health index falls below a threshold.

[0027] 3. The Life Cycle Cost (LCC) dynamic accounting module is the core basis for system decision-making. It breaks through the limitations of traditional single cost accounting, integrates four types of core loss costs from the perspective of the entire life cycle of components, and constructs a dynamic and iterative LCC accounting model. It can dynamically calculate the corresponding comprehensive life cycle cost in real time according to different maintenance cycles T to be evaluated, and accurately quantify the relationship between maintenance cycle and operation and maintenance economy.

[0028] The core formula for LCC accounting in this invention is as follows:

[0029] In the formula, T represents the preventive maintenance cycle to be optimized, in units of operating days or 10,000 kilometers; the specific definitions and accounting logic of each cost item are as follows: ①Preventive maintenance costs This refers to the planned costs of routine inspections, maintenance, disassembly and testing, and replacement of regular consumables, which are directly related to the frequency of maintenance throughout the entire life cycle. The longer the maintenance cycle T, the fewer planned maintenance sessions per unit of service life, and the lower the total cost of preventive maintenance; the two are negatively correlated.

[0030] ② Fault repair costs This refers to the post-failure maintenance costs incurred after a component experiences an unplanned failure, including troubleshooting, damage repair, parts replacement, and fault verification. This cost is dynamically calculated based on the probability of the health index exceeding its limit, as output by the degradation prediction module. The longer the maintenance cycle T, the deeper the component's long-term service degradation, the higher the probability of the health index falling below the failure threshold, the greater the risk of failure, and the more exponentially the repair costs.

[0031] ③ Downtime loss costs This includes two categories: planned downtime losses and unplanned downtime losses. Planned downtime losses are minor operational losses caused by vehicle downtime and capacity reallocation during scheduled maintenance. Unplanned downtime losses are significant operational and brand losses caused by sudden component failures leading to vehicle downtime, route delays, passenger evacuation and rescue operations, and schedule adjustments. Unplanned downtime losses are far higher than planned downtime losses and increase substantially with unreasonably extended maintenance cycles and increased failure risks.

[0032] ④ Residual value depreciation cost This innovative cost dimension, distinct from traditional technologies, quantifies the wasted cost of components' remaining lifespan due to improper maintenance timing. If the maintenance cycle T is too short, components in good health and still possessing remaining service life are replaced prematurely, resulting in a waste of equipment residual value. The precise calculation formula is as follows:

[0033] In the formula, V0 is the original equipment manufacturer (OEM) cost of the new component, and H(T) is the component's predicted health index at the end of period T. This refers to the critical health index threshold for component failure. A higher health index and a shorter maintenance cycle result in higher residual value depreciation costs.

[0034] Extensive field testing has verified that the LCC(T) curve of this invention exhibits a standard "U-shaped" characteristic: if the maintenance cycle is too short, the residual value loss and preventive maintenance costs are too high, resulting in a high overall cost; if the maintenance cycle is too long, the costs of fault repair and downtime losses surge, leading to a continuous increase in overall costs; and there is a unique globally optimal cost minimum point in the middle interval of the curve, providing a core basis for optimizing the maintenance cycle.

[0035] 4. Adaptive Optimization and Decision Module: This module is the core of the system's decision output. With minimizing the total life cycle cost as the optimization objective and vehicle operational reliability as the safety constraint, it constructs a constrained nonlinear optimization model. Through improved intelligent algorithms, it iteratively solves the problem and outputs the optimal preventive maintenance cycle that adapts to the current vehicle operating conditions and component status.

[0036] The optimization model constructed in this module is as follows: Objective function: minLCC(T) Constraints:

[0037] In the formula, The critical health index threshold for component failure is preset according to the failure mechanism of different components; α is the system's preset reliability confidence level, which is set to 95% under normal operating conditions and can be increased to 98% for core safety components, ensuring that optimization decisions always meet the bottom line of safe operation.

[0038] To address the issues of slow convergence, susceptibility to local optima, and unbalanced search capabilities in traditional optimization algorithms, this invention employs an improved whale optimization algorithm incorporating a nonlinear convergence factor to solve the optimization model. By balancing the global search capability and local exploitation capability of the nonlinear convergence factor algorithm, the search scope is expanded in the early stages to traverse the global feasible region, avoiding local optimum traps; in the later stages, the search scope is narrowed for precise iterative convergence, significantly improving the accuracy and efficiency of solving for the optimal cycle, ultimately outputting the globally optimal maintenance cycle Topt.

[0039] 5. Strategy Closed-Loop Verification Module: This module is the core of the system's self-evolution, enabling closed-loop iterative optimization throughout the entire process. It addresses the issues of traditional system model rigidity and the decay of decision-making accuracy over service time. The module includes a built-in customizable lag adjustment threshold to avoid operational strategy confusion caused by frequent small fluctuations in maintenance cycles.

[0040] The specific working logic is as follows: the optimal cycle Topt obtained by solving is compared with the maintenance cycle Tcurrent currently being executed by the vehicle to calculate the relative deviation of the cycle; if the relative deviation exceeds the preset lag threshold (default 10%, which can be customized according to the component type), the current maintenance strategy is determined to be lagging and ineffective, and the system automatically triggers the maintenance cycle adjustment command to update to the optimal cycle Topt; if the deviation is within the threshold range, the existing cycle is maintained to avoid frequent adjustments.

[0041] Meanwhile, the module feeds back all-dimensional data, including actual component degradation data, actual maintenance costs, fault records, and downtime losses, during the execution of the new maintenance cycle to the data acquisition and preprocessing module, degradation prediction module, and LCC calculation module. It then fine-tunes the TCN-LSTM model network parameters and cost accounting weight coefficients in real time, enabling continuous iterative updates of model parameters. This allows the system to adapt to the state evolution patterns of the vehicle throughout its entire life cycle and possesses long-term self-learning and self-optimization capabilities.

[0042] Example 1: This embodiment applies the system to the optimization of preventive maintenance cycles for wheelset bearings, a core running gear component of urban rail vehicles. Wheelset bearings are the core load-bearing and running gear components of the vehicle; deterioration can easily lead to jamming and derailment risks, thus requiring extremely high reliability. The specific implementation process is as follows: High-frequency vibration envelope spectrum signals, real-time temperature data, vehicle mileage, operating speed, start-stop frequency, and operating condition data such as track curve radius, track smoothness, ambient temperature and humidity, and vehicle passenger load were collected for the wheelset bearings. Simultaneously, historical maintenance data such as the bearing's factory parameters, maintenance records, fault history, and replacement cycle were retrieved. A wavelet soft thresholding denoising algorithm was used to remove environmental noise from the vibration signals, and core degradation features such as vibration kurtosis, root mean square value, and temperature range were extracted. After normalization, a standardized feature dataset was constructed.

[0043] The preprocessed feature dataset is input into the trained TCN-LSTM fusion model. The TCN network extracts local wear and micro-impact degradation features from the bearing vibration signal, while the LSTM network captures the temporal evolution of long-term bearing wear. An attention mechanism is used to assign high weights to harsh operating conditions such as high loads, small-radius curves, and high temperature and humidity, enhancing the fitting of the influence of these conditions on degradation. The model outputs the bearing health index H(t) and its 95% confidence interval for each future day. The initial health index is 1, and a critical failure health index is set. =0.2.

[0044] Using 10,000 kilometers as the cycle unit T, four types of costs were calculated item by item: When the cycle is shortened, bearings are repaired and replaced prematurely without obvious wear, resulting in a significant increase in residual value loss costs and preventive maintenance costs; when the cycle is lengthened, bearing wear accumulates more rapidly, the probability of health index exceeding limits increases significantly, and the costs of fault repair and train downtime losses rise exponentially. By fitting the calculation formula, a U-shaped curve of wheelset bearing LCC(T) was obtained to determine the optimal cost range.

[0045] A constrained optimization model was constructed, with the constraint that the reliability probability of the bearing health index ≥ 0.2 at the end of the maintenance cycle is no less than 95%. Using the manufacturer's default 600,000 km as the initial cycle, an improved whale optimization algorithm was used for iterative solution, balancing global search and local precise iteration. The optimal maintenance cycle Topt = 520,000 km was finally determined. Compared to a fixed cycle, this avoids the resource waste of over-maintenance and eliminates the failure risk of under-maintenance, reducing the overall lifecycle cost by more than 12%.

[0046] With a lag adjustment threshold set at 10% and a current maintenance cycle of 600,000 kilometers, if the relative deviation of the cycle exceeds the threshold, the system automatically updates the next maintenance cycle to 520,000 kilometers. After the new cycle is completed, the system collects actual bearing wear data, maintenance costs, and operating status data for this cycle, and iteratively updates the TCN-LSTM model feature weights and LCC cost calculation coefficients to complete one closed-loop optimization.

[0047] Example 2

[0048] This embodiment is applied to the maintenance cycle optimization of the IGBT module of the traction inverter, a core electrical component of urban rail vehicles. The failure mechanism of the IGBT module is thermal fatigue damage. Sudden failures can directly lead to vehicle power interruption, resulting in extremely high downtime losses. This embodiment differs from the mechanical component maintenance logic in Embodiment 1. The specific differences and implementation process are as follows: The system focuses on collecting electrical status data such as junction temperature fluctuation amplitude, gate voltage fluctuation, coolant flow rate, switching frequency, and load current of IGBT modules, as well as operating condition data such as vehicle load, ambient temperature, and continuous operating time, while discarding mechanical vibration characteristic parameters.

[0049] Based on the thermal fatigue failure mechanism of IGBTs, the rainflow counting method is introduced on the basis of the TCN-LSTM model to statistically analyze the number and amplitude of junction temperature cycle fluctuations, calculate the cumulative thermal fatigue damage parameters, accurately fit the fatigue deterioration law of electrical components, and improve the accuracy of thermal fatigue failure prediction.

[0050] Given that a sudden failure of an IGBT module can cause significant losses such as vehicle power interruption, line shutdown, and passenger evacuation and rescue, it greatly increases the cost of fault repair. Downtime loss costs The penalty coefficient amplifies the cost weight corresponding to the failure risk.

[0051] Under the same 95% reliability constraint, the system adaptively outputs a more conservative optimal maintenance cycle. Compared with mechanical components, the maintenance cycle is appropriately shortened, prioritizing the avoidance of high downtime losses and operational risks, and achieving the economic optimization of "high reliability and low failure risk" for electrical components.

[0052] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any equivalent modifications, substitutions, or optimizations that can be made by those skilled in the art without creative effort based on the technical solutions disclosed in the present invention fall within the scope of protection of the present invention.

[0053] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An adaptive adjustment system for preventive maintenance cycles of urban rail vehicles considering full life-cycle costs, characterized in that: include: The data acquisition and preprocessing module is used to collect real-time operating status data, environmental data, and historical maintenance data of key components of urban rail vehicles, and to perform data cleaning and feature extraction. The real-time degradation assessment and prediction module is used to build a component degradation model based on preprocessed data, calculate the current health index, and predict the degradation trajectory within a set future time period. The dynamic lifecycle cost accounting module is used to dynamically calculate the comprehensive lifecycle cost (LCC) based on the predicted degradation trajectory and the preset maintenance cycle. The adaptive optimization and decision-making module is used to solve for the optimal preventive maintenance cycle with the goal of minimizing the total life cycle cost (LCC) and the operational reliability threshold of key vehicle components as constraints, and then issue maintenance instructions.

2. The adaptive adjustment system for preventive maintenance cycle of urban rail vehicles considering full life-cycle costs as described in claim 1, characterized in that, In the real-time degradation assessment and prediction module, a TCN-LSTM model based on the fusion of temporal convolutional network and long short-term memory network is used to predict the degradation trajectory. The environmental and load covariates are used as inputs to the model attention mechanism, and the expected health index H(t) and its confidence interval are output at any future time t.

3. The adaptive adjustment system for preventive maintenance cycle of urban rail vehicles considering full life-cycle costs as described in claim 1, characterized in that, In the dynamic lifecycle cost accounting module, the comprehensive LCC accounting model is as follows: ; Where T represents the preventive maintenance cycle to be optimized; For preventative maintenance costs; The fault repair cost is calculated based on the probability of exceeding the health index limit output by the degradation trajectory prediction module. Downtime loss costs include operational losses caused by both planned and unplanned downtime; Residual value depreciation cost represents the waste of a component's remaining lifespan due to improper maintenance timing.

4. The adaptive adjustment system for preventive maintenance cycle of urban rail vehicles considering full life-cycle costs as described in claim 3, characterized in that, The residual value depreciation cost The calculation formula is: where, in, H(T) represents the cost of purchasing brand new parts, and H(T) represents the predicted health index of the parts at the end of maintenance cycle T. This is the critical health index threshold for component failure.

5. The adaptive adjustment system for preventive maintenance cycle of urban rail vehicles considering full life-cycle costs as described in claim 1, characterized in that, In the adaptive optimization and decision-making module, the objective function and constraint condition are: minLCC(T) in, The critical health index threshold for component failure is α, which is the system's preset reliability confidence level, with a value of not less than 95%, used to ensure the operational safety baseline for maintenance decisions.

6. The adaptive adjustment system for preventive maintenance cycle of urban rail vehicles considering full life-cycle costs as described in claim 5, characterized in that, The adaptive optimization and decision-making module uses an improved whale optimization algorithm that introduces a nonlinear convergence factor to solve for the optimal preventive maintenance cycle.

7. The adaptive adjustment system for preventive maintenance cycle of urban rail vehicles considering full life-cycle costs as described in claim 1, characterized in that, For mechanical wear-related components and electrical fatigue-related components, the system can adaptively adjust the cost accounting weights and deterioration prediction parameters. For electrical fatigue-related components, the rainflow counting method is introduced to statistically measure thermal fatigue damage parameters, enhance the accuracy of fatigue deterioration fitting, and adapt to the failure mechanisms of different types of core components.

8. The adaptive adjustment system for preventive maintenance cycle of urban rail vehicles considering full life-cycle cost according to any one of claims 1-7, characterized in that, It also includes a strategy closed-loop verification module, which compares the optimal cycle output by the adaptive optimization and decision-making module with the current execution cycle. If the difference exceeds the set lag threshold, the cycle adjustment is triggered, and the adjusted actual running data and cost data are fed back to the data acquisition and preprocessing module and the full life cycle cost dynamic accounting module to realize the closed-loop self-updating of model parameters.

Images