A vehicle extended warranty service risk management and control method and system

Abstract

The application discloses a kind of vehicle extended warranty service risk management and control method and system, it is related to vehicle extended warranty service technical field, including, the multi-source heterogeneous data of vehicle key position is collected, and separation processing is carried out, generates dynamic update vehicle health state dataset;Vehicle health state dataset is analyzed in non-standard, in superreal number space each component failure probability distribution is fitted, through lexicographic comparison distribution tail high order moment, generate risk priority list;Mixed strategy Nash equilibrium solution is converted into executable control instruction, automatically executes extended warranty claim, generates closed loop pay-off record;Based on closed loop pay-off record, construct deep reinforcement learning model, simulate new risk mode through generative adversarial network, optimize dynamic game matrix, generate new equilibrium strategy set.The application is accurately aligned with error compensation mechanism through multi-source data, significantly improves the measurement accuracy of torque output, battery capacity retention rate and thermal management efficiency parameters.

Description

Technical Field

[0001] This invention relates to the field of vehicle extended warranty service technology, and in particular to a method and system for risk management of vehicle extended warranty services. Background Technology

[0002] In recent years, risk management technologies for vehicle extended warranty services have gradually shifted from static models based on mileage or time to dynamic analysis relying on onboard sensor data. Existing technologies mainly construct failure probability models through a single dimension and employ traditional statistical methods for risk assessment. With the development of vehicle-to-everything (V2X) technology, some solutions have attempted to integrate multi-source data, but the data fusion accuracy is insufficient, and the problem of quantifying tail risks caused by multi-system coupling effects remains unresolved. Furthermore, existing dynamic strategy adjustment mechanisms largely rely on fixed threshold triggers, lacking the ability to adaptively optimize for higher-order moment characteristics and asymmetric risk distributions.

[0003] The shortcomings of existing technologies are mainly reflected in two aspects: First, the time synchronization and error compensation mechanism of multi-source heterogeneous data is not perfect, which leads to temperature drift and noise interference in key parameters (such as the true value of torque), affecting the accuracy of the failure probability density function; Second, the risk quantification method is limited to standard probability space analysis, which makes it difficult to effectively capture the tail risk characteristics of extreme failure modes. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a risk management method for vehicle extended warranty services to solve the problems of low accuracy in failure risk prediction and lag in dynamic strategy adjustment caused by insufficient fusion of multi-source heterogeneous data in vehicle extended warranty services.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for risk management of vehicle extended warranty services, comprising,

[0008] Collect multi-source heterogeneous data from key parts of the vehicle, separate and process them, and generate a dynamically updated vehicle health status dataset.

[0009] The vehicle health status dataset is subjected to non-standard analysis. The failure probability distribution of each component is fitted in the hyperreal space. The higher-order moments of the distribution tail are compared by lexicographical order to generate a risk priority list.

[0010] A dynamic game matrix is ​​constructed based on a risk priority list, and the policy space is constrained by a confidence interval contraction algorithm to calculate the mixed policy Nash equilibrium solution.

[0011] The Nash equilibrium solution of the hybrid strategy is transformed into executable control instructions. A KL divergence monitoring loop is established through a risk entropy feedback mechanism to automatically execute extended warranty claims and generate closed-loop claims records.

[0012] A deep reinforcement learning model is constructed based on closed-loop payout records. A novel risk pattern is simulated through adversarial generative networks, the dynamic game matrix is ​​optimized, and a new set of equilibrium strategies is generated.

[0013] As a preferred embodiment of the vehicle extended warranty service risk management method of the present invention, the step of generating a dynamically updated vehicle health status dataset includes the following steps:

[0014] The torque output curve, battery degradation rate, and thermal management efficiency are collected by the intelligent monitoring terminals of the powertrain, battery pack, and electronic control unit, respectively. The data are transmitted to the vehicle data storage node via different bus protocols, where timestamp alignment and signal filtering are performed. Kalman filters are used to eliminate temperature drift errors, and wavelet packet transform is used to reduce noise in the thermal management efficiency signal, generating a dynamically updated vehicle health status dataset.

[0015] As a preferred embodiment of the vehicle extended warranty service risk management method described in this invention, the generation of the risk priority list includes the following steps:

[0016] The mean and variance of the cycle life of the power battery and the slope and curvature of the wear rate of the gearbox are extracted from the time-domain feature matrix of the health status dataset. Combined with the total energy of the frequency domain energy spectrum and the amplitude of the harmonic components, a set of core risk indicators is constructed.

[0017] The Weibull and Gamma distributions were fitted by maximum likelihood estimation and moment matching method, respectively, and the first N-order moment sequences of each failure probability distribution were calculated.

[0018] The moment sequence is mapped to the hyperreal number space, the tail dominant index set is filtered by a free ultrafilter, and the moment values ​​are compared in lexicographical order to generate a component failure risk ranking list.

[0019] As a preferred embodiment of the vehicle extended warranty service risk management method of the present invention, the construction of the dynamic game matrix includes the following steps.

[0020] The moment sequence of the power battery failure probability distribution is mapped to the defender strategy set, and the moment sequence of the gearbox wear rate distribution is mapped to the attacker strategy set.

[0021] High-risk moment indices of power battery and transmission failure probability distribution are screened based on dynamic confidence interval contraction mechanism to generate compact game submatrices.

[0022] As a preferred embodiment of the vehicle extended warranty service risk management method described in this invention, the execution of extended warranty claims refers to mapping the strategies in the compact game submatrix to crankshaft clearance adjustment, clutch plate replacement cycle and battery replacement budget parameters, publishing them to the service node through the vehicle network, measuring piston ring wear data, generating an actual risk distribution histogram, comparing it with the theoretical distribution using KL divergence, triggering weight updates and blockchain smart contract verification, and completing the digital currency compensation closed loop.

[0023] As a preferred embodiment of the vehicle extended warranty service risk management method described in this invention, the optimized dynamic game matrix includes the following steps.

[0024] Extract the cost volatility parameters of historical maintenance work orders, input them into a deep reinforcement learning model, and update the policy network weights.

[0025] By synthesizing virtual failure probability distribution moment sequences through adversarial generative networks, and combining them with real maintenance data, an extended risk priority list is generated.

[0026] Based on KL divergence screening, risk patterns that deviate from the actual distribution are selected, and the defender strategy weights and attacker trigger probabilities are adjusted.

[0027] As a preferred embodiment of the vehicle extended warranty service risk management method described in this invention, the generation of a new set of equilibrium strategies includes the following steps.

[0028] The reconstructed defender's policy payoff function and attacker's policy loss function are input into the minimization-maximization principle solver to calculate the hybrid policy Nash equilibrium solution.

[0029] An instruction set for the allocation strategy is generated based on the equilibrium solution, and then fed back to the KL divergence monitoring loop to form a new set of equilibrium strategies.

[0030] Secondly, this invention provides a vehicle extended warranty service risk management system, comprising,

[0031] The data acquisition module collects multi-source heterogeneous data from key parts of the vehicle, separates and processes it, and generates a dynamically updated vehicle health status dataset.

[0032] The risk modeling module performs non-standard analysis on the vehicle health status dataset, fits the failure probability distribution of each component in the hyperreal space, and generates a risk priority list by comparing the higher-order moments of the distribution tails in lexicographical order.

[0033] The strategy optimization module constructs a dynamic game matrix based on a risk priority list, uses a confidence interval contraction algorithm to constrain the strategy space, and calculates the mixed strategy Nash equilibrium solution.

[0034] The claims execution module transforms the hybrid strategy Nash equilibrium solution into executable control instructions, establishes a KL divergence monitoring loop through a risk entropy feedback mechanism, automatically executes extended warranty claims, and generates closed-loop payment records.

[0035] The risk evolution module constructs a deep reinforcement learning model based on closed-loop payout records, simulates new risk patterns through adversarial generative networks, optimizes the dynamic game matrix, and generates a new set of equilibrium strategies.

[0036] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the vehicle extended warranty service risk management method as described in the first aspect of the present invention.

[0037] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the vehicle extended warranty service risk management method as described in the first aspect of the present invention.

[0038] The beneficial effects of this invention are as follows: Through precise alignment and error compensation mechanisms of multi-source data, the measurement accuracy of torque output, battery capacity retention rate, and thermal management efficiency parameters is significantly improved, providing a high-fidelity data foundation for constructing the failure probability density function; the hyperreal space sorting mechanism based on non-standard analysis methods overcomes the limitations of traditional moment matching methods, achieving precise prioritization of tail risks in power battery cycle life and transmission wear rate through the filtering of high-order moment-dominated indices and lexicographical comparison; combined with dynamic confidence interval contraction and compact game submatrix generation, it can adaptively focus on extreme failure modes, optimizing the cost-effectiveness of proactive maintenance and preventative repair strategies; through KL divergence-driven closed-loop compensation verification and blockchain smart contract execution, it effectively reduces the moral hazard and misjudgment probability of extended warranty services, achieving dynamic closed-loop optimization of risk management. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A flowchart of risk management methods for vehicle extended warranty services.

[0041] Figure 2 A schematic diagram for risk modeling.

[0042] Figure 3This is a diagram illustrating strategy optimization.

[0043] Figure 4 This is a diagram illustrating the evolution of risk. Detailed Implementation

[0044] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0045] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0046] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0047] Reference Figures 1-4 This embodiment provides a method for risk management of vehicle extended warranty services, including the following steps:

[0048] S1. The powertrain's intelligent monitoring terminal collects the raw voltage signal output from the crankshaft speed sensor, converts the analog signal into a digital torque value via a built-in AD converter, forming a torque output curve data stream, and transmits it in real time to the vehicle's data storage node via the CAN bus protocol; the battery pack's intelligent monitoring terminal measures the charge / discharge curve of individual cells through a voltage / current sensor array, calculates the battery capacity retention rate within the current cycle, expressed as:

[0049]

[0050] Where R is the battery capacity retention rate, and C cor C represents the actual discharge capacity of the battery during the current cycle. rea This refers to the battery's rated capacity.

[0051] Based on the battery capacity retention rate within the current cycle, a sequence of battery degradation rate parameters, including temperature compensation, is generated and transmitted to the vehicle data storage node via the LIN bus. The intelligent monitoring terminal of the electronic control unit reads the coolant flow meter pulse signal and temperature sensor data for thermal management, and calculates the thermal management efficiency in real time based on a preset heat exchange efficiency formula. The expression is:

[0052]

[0053] Where η is the thermal management efficiency, Q act Q represents the actual heat exchange rate. max This represents the theoretical maximum heat exchange capacity.

[0054] The data is transmitted to the vehicle data storage node via the FlexRay bus. The vehicle data storage node receives the torque output curve data stream from the powertrain's intelligent monitoring terminal, the battery degradation rate parameter sequence from the battery pack's intelligent monitoring terminal, and the thermal management system efficiency parameters from the electronic control unit's intelligent monitoring terminal. After timestamp alignment via the NTP protocol, the data is merged into a raw set of vehicle operating parameters containing torque, speed, battery capacity, and thermal efficiency values.

[0055] The powertrain torque output curve data stream in the original vehicle operating parameter set is parsed into a timestamp-torque value binary sequence. Simultaneously, the engine compartment temperature values ​​recorded by the powertrain temperature sensor are extracted to form a temperature time series. A triplet dataset, including timestamps, original torque values, and temperature values, is generated by linear interpolation to precisely align the sampling rate of the temperature time series with the timestamps of the torque output curve data stream. A Kalman filter state vector, including the true torque value and temperature drift, is defined, and the state transition matrix, control input matrix, and observation matrix are initialized. The process noise covariance matrix, observation noise covariance, initial state estimate, and initial error covariance are set. Kalman filtering iterations are performed by traversing the triplet dataset in chronological order, calculating the prior state estimate, updating the prior error covariance, and correcting the prior state estimate using Kalman gain to update the error covariance and eliminate baseline offset caused by temperature drift. The true torque value component within the updated state vector during iteration is extracted and replaced with the corresponding value in the original torque output curve, generating an error-compensated intermediate dataset containing timestamps, compensated torque values, and temperature values.

[0056] The time-domain signal sequence of thermal management efficiency of the electronic control unit is extracted from the intermediate dataset of error compensation. The time-domain signal is input into the wavelet packet transform processing layer, and decomposed using the Daubechies wavelet basis function to obtain a set of detail coefficients. The noise standard deviation is calculated based on the set of detail coefficients, and a dynamic soft threshold is determined by combining the signal length. The absolute value reduction processing is performed on the detail coefficients exceeding the dynamic soft threshold to obtain the corrected detail coefficients. The corrected detail coefficients and the original approximation coefficients are input into the inverse wavelet packet transform algorithm to reconstruct the time-domain signal, generating a filtered signal sequence that retains the effective frequency band. The mean and standard deviation are calculated on the filtered signal sequence, and the Z-score standardization formula is applied to generate a standardized sequence with zero mean and unit variance. The standardized sequence is divided into a time-domain feature matrix, and the frequency domain energy spectrum is stored synchronously. Finally, they are merged into a dynamically updated vehicle health status dataset containing the time-domain feature matrix and the frequency domain energy spectrum.

[0057] To further explain, the dynamic soft threshold is calculated by taking the median absolute value of the highest frequency detail coefficients after Daubechies wavelet packet decomposition, and then dividing it by a correction factor to obtain an estimate of the initial noise standard deviation. Based on the signal length, the base threshold is calculated using Stein unbiased risk estimation, and an attenuation factor is introduced to compensate for the logarithmic term of the long signal. The base threshold is obtained by weighting and modulating it according to the energy proportion of the noise frequency band, combined with the frequency domain energy spectrum.

[0058] S2. Input the time-domain feature matrix and frequency-domain energy spectrum from the dynamically updated vehicle health status dataset into the risk quantification engine. Extract the mean and variance sequences within the window from the power battery cycle life time-domain feature matrix, and simultaneously calculate the total energy of the frequency-domain energy spectrum for the corresponding window to form a set of core risk indicator parameters for the power battery. Calculate the slope and curvature sequences for each window from the transmission wear rate time-domain feature matrix, and extract the harmonic component amplitudes from the frequency-domain energy spectrum to generate a set of core risk indicator parameters for the transmission wear rate. Input the set of core risk indicator parameters for the power battery cycle life into the maximum likelihood estimation algorithm to construct the Weibull distribution log-likelihood function, expressed as:

[0059]

[0060] Where L(λ,k) is the likelihood function, representing the probability of observed data given the scale parameter λ and shape parameter k. λ is the scale parameter of the Weibull distribution, controlling the scale of the Weibull distribution. k is the shape parameter of the Weibull distribution, controlling the shape of the Weibull distribution. When k<1, it indicates that the failure rate decreases over time (early failure); when k=1, it follows an exponential distribution (constant failure rate); when k>1, it indicates that the failure rate increases over time (wear-out failure). t is the time window index, T is the time window length, and μ... bat (t) represents the average cycle lifetime within the time window t.

[0061] The shape and scale parameters in the log-likelihood function of the Weibull distribution are solved iteratively using the Newton-Raphson method. Based on the solved shape and scale parameters of the Weibull distribution, the cycle life failure probability density function of the power battery is defined as follows:

[0062]

[0063] Among them, f bat (x) represents the failure probability density of the power battery at cycle life x, where x is the cycle life of the power battery.

[0064] Applying the moment matching method to the set of core risk indicator parameters of gearbox wear rate, and analytically solving the simultaneous gamma distribution moment equations to obtain the α shape parameter and β rate parameter, the gearbox wear rate failure probability density function is constructed, expressed as:

[0065]

[0066] Among them, f gear (y) represents the failure probability density of the gearbox at wear rate y, where y is the wear rate, α is the shape parameter of the gamma distribution, when α < 1, it indicates that the peak is near the origin, when α > 1, it indicates that the peak shifts to the right and the distribution is more symmetrical, β is the rate parameter of the gamma distribution, and Γ(α) is the gamma function.

[0067] The support sets of the power battery cycle life failure probability density function and the transmission wear rate failure probability density function are truncated into finite intervals. The first N moments of each distribution (N is a preset positive integer) are calculated to generate the power battery cycle life failure probability distribution moment sequence and the transmission wear rate failure probability distribution moment sequence. Using a non-standard analysis method, the power battery cycle life failure probability distribution moment sequence and the transmission wear rate failure probability distribution moment sequence are mapped to equivalence classes in the hyperreal space. The index set that satisfies the tail dominance condition is screened by a free ultrafilter in the hyperreal space. In the hyperreal space, according to the lexicographical total order relation, the Nth moment of the power battery cycle life failure probability distribution moment sequence and the transmission wear rate failure probability distribution moment sequence in the index set are compared first. If the Nth moment is equivalent, the (N-1)th moment to the 1st moment are compared in reverse order until the strict partial order relation between the distributions is determined. The output result is a sequence of power battery cycle life failure probability distribution and transmission wear rate failure probability distribution sorted by risk level, where the tail quality (dominated by higher-order moments) determines the sorting priority, forming a component failure risk priority list.

[0068] S3 extracts the first moment (mean) and second moment (variance) from the cycle life failure probability distribution moment sequence of the power battery, quantifies them into a battery health score, and triggers an active maintenance strategy when the battery health score is lower than the critical value. The active maintenance strategy includes measures such as equalization charging and temperature control. Based on the third moment (skewness) feature of the cycle life failure probability distribution of the power battery, abnormal degradation patterns are identified. When a right-skewed distribution is detected, a preventive maintenance strategy is initiated, focusing on checking electrode expansion and electrolyte loss. The risk exposure is calculated based on the fourth moment of the cycle life failure probability distribution of the power battery, and a risk reserve allocation strategy is generated. The active maintenance strategy, preventive maintenance strategy, and risk reserve allocation strategy are mapped to a set of defender strategies.

[0069] The system analyzes the moment sequence of the transmission wear rate failure probability distribution. When the first moment exceeds the mechanical-electrical coupling safety boundary of the first moment (mean) of the transmission wear rate failure probability distribution, a battery capacity sag strategy is activated. This strategy intentionally limits the available SOC window through the BMS to offset torque fluctuations caused by mechanical wear. When the second moment of the transmission wear rate shows abrupt changes, a transmission wear acceleration strategy is implemented. This introduces controlled slippage conditions into the shift logic, causing wear to concentrate on replaceable synchronizer components. For the abnormal third moment of the transmission wear rate distribution, a thermal management failure strategy is implemented. This dynamically adjusts the coolant pump speed to maintain the transmission oil temperature within a controllable degradation range. The battery capacity sag strategy, transmission wear acceleration strategy, and thermal management failure strategy are mapped to an attacker strategy set.

[0070] Furthermore, the critical value for battery health rating is obtained by using the average health value of similar batteries before failure.

[0071] The risk exposure is calculated by multiplying the difference between the sample kurtosis (unbiased estimate) of the power battery cycle life failure probability distribution and the baseline kurtosis of the standard normal distribution, by the ratio of the difference between the historical maximum allowable kurtosis in the failure case library and the sample kurtosis of the power battery cycle life failure probability distribution, and a risk amplification factor.

[0072] The moment values ​​of each order are extracted from the moment sequence of the failure probability distribution of the power battery cycle life, and a defender strategy benefit function is constructed to reflect the cost-effectiveness of different maintenance strategies. The expression is as follows:

[0073] D ij =F bat (m bat (N),p i );

[0074] Among them, D ij The payoff values ​​for the defender choosing strategy i and the attacker choosing strategy j, where i is the index of the defender's chosen strategy and j is the index of the attacker's chosen strategy. bat The revenue mapping function for the failure risk of power batteries is a function that transforms the statistical characteristics (moments) of the failure probability distribution into actionable strategy revenues, m. bat (N) represents the Nth moment of the power battery failure probability distribution, characterizing the distribution characteristics of failure risk (such as mean, variance, skewness, etc.), p i is the weighting coefficient of defender strategy i, used to adjust the priority of the strategy in the payoff calculation.

[0075] The moment values ​​(such as mean, variance, skewness, etc.) of each order are extracted from the moment sequence of the failure probability distribution of the gearbox wear rate. An attacker strategy loss function is constructed, which triggers the probability and associates the threat level of different failure modes. The expression is as follows:

[0076] M ij =F gear (m gear (N),q j );

[0077] Among them, M ij F represents the loss value when the defender chooses strategy i and the attacker chooses strategy j. gear The loss mapping function for gearbox wear risk transforms the statistical characteristics of the wear rate distribution into a strategy loss, m. gear (N) is the Nth moment of the failure probability distribution of the gearbox wear rate, which is used to describe the distribution characteristics of the wear rate (such as tail risk), and is the trigger probability of the attacker choosing strategy j (such as the probability of a sudden drop in battery capacity).

[0078] Real-time monitoring of the power battery health index is performed. Based on the real-time health index and the tail risk of the failure probability distribution, a dynamic risk response threshold is defined. When the battery health index drops to the dynamic risk response threshold, a dynamic confidence interval contraction mechanism is activated. By analyzing the higher-order moment characteristics of the failure probability distribution of the power battery and transmission, high-risk moment indices that deviate significantly from the norm are screened out and merged to form a set of dominant higher-order moment indices reflecting extreme failure modes. The dominant indices are prioritized for risk using hyperreal number order relations. The top-ranked subset of indices is selected to generate a compacted evaluation dimension. The sets of defensive and offensive strategies are filtered, and proactive maintenance strategies, preventive maintenance strategies, and corresponding strategies for sudden drop in battery capacity and accelerated wear of the transmission, which are strongly correlated with the high-risk moment characteristics, are retained to generate a compacted game submatrix.

[0079] Furthermore, the following steps were taken to analyze the higher-order moment characteristics of the failure probability distribution of the power battery and transmission, and to screen out the high-risk moment index that significantly deviates from the norm:

[0080] After the dynamic confidence interval contraction mechanism is activated, fourth-order and higher moment values ​​(such as kurtosis and hyperkurtosis) are extracted from the cycle life failure probability distribution of the power battery and the wear rate failure probability distribution of the transmission as high-order moment features characterizing the extreme risk at the tail of the distribution. The high-order moment sequence of the power battery is compared step by step with the theoretical moment values ​​of the preset benchmark distribution (such as the theoretical moment values ​​of the Weibull distribution) to identify the index positions where the kurtosis of the power battery exceeds three times the standard deviation of the theoretical value. At the same time, abnormal interval indexes in the fifth-order moment of the transmission wear rate distribution that show asymmetric peak characteristics are detected. After merging the two types of indexes, the Pearson correlation coefficient between the two types of indexes is calculated to screen out the joint index subset where the moment values ​​of the power battery and transmission failure modes both deviate within the same time window. Based on the temporal continuity and spatial distribution density of the joint index subset, the kernel density estimation method is used to generate a set of high-order moment dominant indexes reflecting extreme failure modes.

[0081] S4. Map the active maintenance strategy and preventive maintenance strategy selected from the compact game submatrix into a specific maintenance operation instruction set, where the active maintenance strategy corresponds to the crankshaft bearing clearance adjustment parameter, the preventive maintenance strategy corresponds to the clutch plate replacement cycle parameter, and the battery pack risk reserve fund allocation strategy corresponds to the single battery replacement budget ratio parameter; publish the operation instruction set to the authorized service network node through the communication protocol of the vehicle network control center to generate a real-time control instruction set.

[0082] After receiving the real-time control command set at the authorized service network node, the laser interferometry unit of the multispectral detector is activated to perform a micron-level three-dimensional topographic scan of the piston ring surface, acquire radial wear depth distribution data, and generate piston ring ellipticity error parameters through polynomial fitting. Combined with the current mileage, the actual wear rate distribution is calculated, generating an actual risk distribution histogram. The actual risk distribution histogram is then compared with the power battery cycle life failure probability density function using KL divergence calculation. Local maxima of distribution differences are extracted using the sliding window method. When the KL divergence value of the crankcase wear rate exceeds the dynamic boundary value of the KL divergence, a weight update mechanism for the compact game submatrix is ​​triggered. The specific operation steps of the weight update mechanism are as follows:

[0083] Based on the Euclidean distance between the piston ring ellipticity error parameter and historical engineering parameters within the current testing cycle, the iterative initial value of the Weibull distribution shape parameter is recalculated to generate an updated power battery cycle life failure probability distribution moment sequence. The hyperreal number sorting process of the non-standard analysis method is re-executed, and an updated component failure risk priority list is output. The sampling sequence of the drive motor insulation impedance value is monitored synchronously. When the impedance value recorded in multiple consecutive testing cycles is lower than the insulation impedance threshold that triggers extended warranty claims, the claim condition verification logic of the blockchain smart contract is activated, as follows:

[0084] The system calls the hash value of the extended warranty terms stored in the distributed ledger, matches it with the claims rules corresponding to the current vehicle identification number, verifies the causal relationship chain between thermal management efficiency and insulation degradation through zero-knowledge proof, and generates a digitally signed claims instruction when the pre-set confidence level is met. After the digitally signed claims instruction is verified by the consensus node, it triggers the automatic clearing process of the digital wallet. Based on the replacement part cost matrix in the current repair work order, it calculates the encrypted value of the amount payable, completes the transfer of digital currency to the repair shop's account through a cross-chain atomic swap protocol, and generates a closed-loop claims record containing the transaction hash value and timestamp.

[0085] S5. Using the transaction hash and timestamp in the closed-loop compensation record as index keys, extract the cost matrix of historical maintenance work orders and the component replacement cost matrix of the current period from the distributed ledger, and generate a cost volatility parameter sequence by comparison; input the cost volatility parameter sequence into a deep reinforcement learning model with a memory enhancement mechanism, retrieve the game strategy trajectory under similar maintenance scenarios in the historical compensation record, extract the policy gradient direction using the experience replay mechanism, and update the policy network weight parameters; at the same time, synchronize the maintenance data of the service outlets, standardize the component replacement type, labor cost and material loss parameters in the maintenance work order into the payoff unit of the game matrix, and map and match them with the encrypted amount value in the closed-loop compensation record to generate a training sample set for the adversarial generative network.

[0086] The generator of the adversarial generative network receives the weight coefficients of the current compact game submatrix and a list of component failure risk priorities. It extracts the spatial correlation features of the risk distribution through convolutional layers and synthesizes a virtual failure probability distribution moment sequence simulating new risk patterns by combining it with a generative adversarial loss function. The discriminator receives the power battery cycle life failure probability distribution moment sequence and the gearbox wear rate failure probability distribution moment sequence from real maintenance data. It calculates the distribution similarity score through a multilayer perceptron and feeds it back to the generator to optimize the generation strategy of the virtual distribution. The updated virtual failure probability distribution moment sequence is re-executed with a non-standard analysis method of hyperreal number sorting through a continuous distribution comparison algorithm to generate an extended set of the component failure risk priority list containing simulated risk events.

[0087] The expanded component failure risk priority list and the actual risk distribution histogram in the closed-loop compensation record are subjected to KL divergence calculation. Risk patterns with deviations exceeding the dynamic boundary values ​​of KL divergence are screened out. The weight coefficients of the defender's strategy payoff function in the compact game submatrix are adjusted, and the trigger probability of the attacker's strategy loss function is simultaneously corrected. The parameter mapping relationship between the defender's strategy payoff function and the attacker's strategy loss function is reconstructed. The reconstructed defender's strategy payoff function and attacker's strategy loss function are input into the minimax principle solver to calculate a new hybrid strategy equilibrium solution, generating an equilibrium strategy set that includes proactive maintenance strategy, preventive maintenance strategy, and risk reserve fund allocation strategy.

[0088] This embodiment also provides a vehicle extended warranty service risk management system, including:

[0089] The data acquisition module collects multi-source heterogeneous data from key parts of the vehicle, separates and processes it, and generates a dynamically updated vehicle health status dataset.

[0090] The risk modeling module performs non-standard analysis on the vehicle health status dataset, fits the failure probability distribution of each component in the hyperreal space, and generates a risk priority list by comparing the higher-order moments of the distribution tails in lexicographical order.

[0091] The strategy optimization module constructs a dynamic game matrix based on a risk priority list, uses a confidence interval contraction algorithm to constrain the strategy space, and calculates the mixed strategy Nash equilibrium solution.

[0092] The claims execution module transforms the hybrid strategy Nash equilibrium solution into executable control instructions, establishes a KL divergence monitoring loop through a risk entropy feedback mechanism, automatically executes extended warranty claims, and generates closed-loop payment records.

[0093] The risk evolution module constructs a deep reinforcement learning model based on closed-loop payout records, simulates new risk patterns through adversarial generative networks, optimizes the dynamic game matrix, and generates a new set of equilibrium strategies.

[0094] This embodiment also provides a computer device applicable to the risk management method for vehicle extended warranty services, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the risk management method for vehicle extended warranty services as proposed in the above embodiment.

[0095] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0096] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the risk management method for vehicle extended warranty services as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0097] In summary, this invention significantly improves the measurement accuracy of torque output, battery capacity retention rate, and thermal management efficiency parameters through multi-source data precise alignment and error compensation mechanisms, providing a high-fidelity data foundation for constructing the failure probability density function. The hyperreal space sorting mechanism based on non-standard analysis methods overcomes the limitations of traditional moment matching methods, achieving precise prioritization of tail risks in power battery cycle life and transmission wear rate through the filtering of high-order moment-dominated indices and lexicographical comparison. Combined with dynamic confidence interval shrinkage and compact game submatrix generation, it can adaptively focus on extreme failure modes, optimizing the cost-effectiveness of proactive maintenance and preventative repair strategies. Through KL divergence-driven closed-loop compensation verification and blockchain smart contract execution, it effectively reduces the moral hazard and misjudgment probability of extended warranty services, achieving dynamic closed-loop optimization of risk management.

[0098] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A vehicle extended warranty service risk management method, characterized in that: The application relates to a vehicle health state dynamic updating method based on deep reinforcement learning. Collecting multi-source heterogeneous data of key parts of a vehicle, performing separation processing, and generating a dynamic updating vehicle health state data set; Performing non-standard analysis on the vehicle health state data set, fitting failure probability distribution of each component in a hyperreal number space, comparing high-order moments of distribution tails through lexicographic order, and generating a risk priority list; Building a dynamic game matrix based on the risk priority list, constraining a strategy space through a confidence interval contraction algorithm, calculating a mixed strategy Nash equilibrium solution, and converting the mixed strategy Nash equilibrium solution into executable control instructions; Through a risk entropy value feedback mechanism, establishing a KL divergence monitoring loop, automatically executing an extended warranty claim, and generating a closed-loop compensation record; Based on the closed-loop compensation record, building a deep reinforcement learning model, simulating new risk patterns through an adversarial generative network, optimizing the dynamic game matrix, and generating a new equilibrium strategy set.

2. The method of claim 1, wherein: The method for generating the dynamic updating vehicle health state data set comprises the following steps that, Torque output curves, battery attenuation rates and thermal management efficiencies of a power assembly, a battery pack and an electronic control unit are collected through intelligent monitoring terminals, transmitted to a vehicle-mounted data storage node through different bus protocols, subjected to time stamp alignment and signal filtering processing, temperature drift errors are eliminated through a Kalman filter, thermal management efficiency signals are denoised through wavelet packet transformation, and a dynamic updating vehicle health state data set is generated.

3. The method of claim 1, wherein: The method for generating the risk priority list comprises the following steps that, Mean value, variance of power battery cycle life and slope and curvature of transmission wear rate are extracted from a time domain feature matrix in the health state data set, combined with total energy and harmonic component amplitude of a frequency domain energy spectrum, and a core risk index set is constructed; Weibull distribution and gamma distribution are fitted through maximum likelihood estimation and moment matching method respectively, and a front N-order moment sequence of each failure probability distribution is calculated; The moment sequence is mapped to a hyperreal number space, a tail dominant index set is screened through a free hyperfilter, moment values are compared in lexicographic order according to priority, and a component failure risk ranking list is generated.

4. The method of claim 1, wherein: The method for constructing the dynamic game matrix comprises the following steps that, Moment sequences of power battery failure probability distribution are mapped to a defender strategy set, and moment sequences of transmission wear rate distribution are mapped to an attacker strategy set; Based on a dynamic confidence interval contraction mechanism, high-risk moment indexes of power battery and transmission failure probability distribution are screened, and a tight game sub-matrix is generated.

5. The method of claim 1, wherein: The method for executing the extended warranty claim refers to that strategies in the tight game sub-matrix are mapped to crankshaft gap adjustment, clutch plate replacement cycle and battery replacement budget parameters, the strategies are published to a service node through vehicle networking, piston ring wear data are measured, an actual risk distribution histogram is generated, and a KL divergence comparison is made with a theoretical distribution, weight updating and blockchain smart contract verification are triggered, and a digital currency compensation closed loop is completed.

6. The method of claim 1, wherein: The method for optimizing the dynamic game matrix comprises the following steps that, Cost fluctuation rate parameters of historical maintenance work orders are extracted, input into a deep reinforcement learning model, and strategy network weights are updated; Virtual failure probability distribution moment sequences are synthesized through an adversarial generative network, combined with real maintenance data to generate an extended risk priority list; Based on KL divergence, risk patterns deviating from actual distribution are screened, and defender strategy weights and attacker trigger probability are adjusted.

7. The method of claim 1, wherein: The generating of the new equilibrium strategy set comprises the following steps, The reconstructed defender strategy payoff function and the attacker strategy loss function are input into a minimax principle solver to calculate a mixed strategy Nash equilibrium solution; Based on the equilibrium solution, an instruction set for the allocation strategy is generated and fed back to a KL divergence monitoring loop to form a new equilibrium strategy set.

8. A vehicle extended warranty service risk management system based on any one of the vehicle extended warranty service risk management methods according to claims 1-7, characterized in that: It comprises, A data acquisition module acquires multi-source heterogeneous data of key parts of a vehicle, separates and processes the data, and generates a dynamically updated vehicle health status dataset; A risk modeling module performs non-standard analysis on the vehicle health status dataset, fits the failure probability distribution of each component in a hyperreal number space, compares the high-order moments of the distribution tails by lexicographic order, and generates a risk priority list; A strategy optimization module constructs a dynamic game matrix based on the risk priority list, constrains the strategy space using a confidence interval contraction algorithm, and calculates a mixed strategy Nash equilibrium solution; A claim execution module converts the mixed strategy Nash equilibrium solution into executable control instructions, establishes a KL divergence monitoring loop through a risk entropy feedback mechanism, automatically executes the warranty claim, and generates a closed-loop claim record; A risk evolution module constructs a deep reinforcement learning model based on the closed-loop claim record, simulates new risk patterns through a generative adversarial network, optimizes the dynamic game matrix, and generates a new equilibrium strategy set. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the vehicle warranty service risk management and control method of any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the vehicle warranty service risk management and control method of any one of claims 1-7.

Images