A movable collateral financing risk assessment system and method

Abstract

The present application relates to the technical field of financial technology, in particular to a kind of movable collateral financing risk assessment system and method, comprising: through the multiple types of sensors deployed in the storage environment of collateral, real-time physical state data set is obtained, based on material characteristic degradation submodel and environmental stress accumulation submodel, fusion history loss data mapping table, construct degradation coupling model, calculate the real-time value coefficient of collateral, combined with grey correlation degree model, generate dynamic collateral rate deviation index, when deviation exceeds dynamic threshold interval, based on the geometric variation coefficient of surface deformation feature matrix and internal structure integrity index calculation disposal emergency level, call corresponding gradient disposal strategy template, generate and execute gradient disposal instruction set containing automatic replenishment collateral, interest rate adjustment and forced liquidation.The present application can realize the dynamic accurate evaluation and credible disposal of the risk state of collateral, improve the safety and response efficiency of movable financing business.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of financial technology, and in particular to a movable collateral financing risk assessment system and method. BACKGROUND

[0002] With the rapid development of supply chain finance, small and medium-sized enterprise financing and other scenarios, movable collateral financing as a flexible credit enhancement method is widely used in industrial raw materials, agricultural product storage, manufacturing inventory and other fields. In order to protect the rights and interests of the pledgee, accurately assessing the value change and risk exposure of the collateral becomes the key. In recent years, with the maturity of emerging technologies such as Internet of Things sensors, smart contracts and blockchains, a movable financing risk control system based on real object monitoring and intelligent judgment has gradually been built, providing a technical foundation for realizing online monitoring and dynamic management of the state of the collateral.

[0003] However, the existing technology still has several key deficiencies: first, most methods only focus on a single dimension such as inventory quantity or market valuation, lacking substantive evaluation of physical state changes of the collateral (such as structural fatigue, corrosion degradation); second, under the influence of multiple factors, the collateral value assessment lacks a dynamic modeling and degradation coupling mechanism, making it difficult to accurately reflect the real depreciation process; third, risk response mostly adopts static rules or manual intervention, and has not yet realized an automatic gradient disposal strategy triggered based on deviation index; fourth, the data processing process is isolated and scattered, lacking a closed-loop execution system driven by consensus mechanism and smart contract, making it difficult to guarantee the transparency and synergy of risk disposal. SUMMARY

[0004] The present application provides a movable collateral financing risk assessment system and method, providing a movable collateral financing risk assessment method with high-dimensional data fusion and intelligent response mechanism to realize precise monitoring and controllable management throughout the process.

[0005] A movable collateral financing risk assessment method, comprising the following steps:

[0006] S1: Real-time physical state data is obtained by deploying monitoring equipment in the storage environment of the collateral, including surface deformation characteristics, internal density distribution and environmental corrosion parameters;

[0007] S2: The physical state data is input into a dynamic assessment model, and an real-time value coefficient of the collateral is output, wherein the dynamic assessment model is obtained by training based on historical loss data, including a material property degradation sub-model and an environmental stress accumulation sub-model;

[0008] S3: The real-time value coefficient is multi-dimensionally fused with enterprise credit score and industry risk index to generate a dynamic collateral rate deviation index;

[0009] S4: if the deviation index exceeds a preset threshold, triggering the smart contract to execute a gradient handling strategy, the gradient handling strategy including a priority combination of automatic collateral replenishment instructions, interest rate adjustment instructions and forced liquidation instructions.

[0010] Optionally, the S1 includes:

[0011] S11: real-time collection of surface point cloud data by a three-dimensional laser scanner array deployed in the collateral storage environment, differential calculation using a preset reference geometric model to generate surface deformation features with millimeter-level precision;

[0012] S12: inputting the surface deformation features into a Gaussian filter for motion noise elimination, simultaneously collecting transverse and longitudinal wave velocity ratio data through an ultrasonic density probe array, separating the internal density distribution by wavelet decoupling algorithm, and continuously monitoring the sulfide concentration, humidity and temperature parameters in the environment using an electrochemical corrosion sensor group, generating environmental corrosion parameters after Kalman filtering processing;

[0013] S13: spatiotemporal alignment of the preprocessed surface deformation features, internal density distribution and environmental corrosion parameters in time series, integration through a data fusion gateway to generate physical state data, wherein the physical state data includes a surface deformation feature matrix with a time stamp, an internal density distribution tensor and an environmental corrosion parameter vector.

[0014] Optionally, the S2 includes:

[0015] S21: inputting the surface deformation feature matrix in the physical state data into a material property degradation sub-model, extracting a material fatigue degree sequence in the time dimension through a pre-trained LSTM neural network, and calculating a density gradient change rate based on the internal density distribution tensor to generate a first degradation factor;

[0016] S22: inputting the environmental corrosion parameter vector into an environmental stress accumulation sub-model, aligning the time sequence relationship of sulfide concentration, humidity and temperature parameters using dynamic time warping algorithm, fitting the corrosion rate curve through Weibull distribution, and combining the spatial distribution weight of the surface deformation feature matrix to generate a second degradation factor;

[0017] S23: inputting the first degradation factor and the second degradation factor into a degradation coupling module, using the historical loss data mapping table saved in the transfer learning framework to adaptively weight and fuse them, wherein the weight coefficient is dynamically adjusted according to the timeliness of the physical state data, and outputting a collateral real-time value coefficient.

[0018] Optionally, the construction method of the historical loss data mapping table saved in the transfer learning framework includes:

[0019] S24: Extracting a historical surface deformation feature matrix, a historical internal density distribution tensor and a historical environmental corrosion parameter vector from the physical state data, performing space-time alignment with artificial accelerated degradation data in an accelerated aging experiment database, and generating a degradation feature comparison table;

[0020] S25: Inputting the historical environmental corrosion parameter vector into an environmental stress accumulation sub-model, calculating a theoretical corrosion rate curve, and performing residual analysis with an actually monitored environmental corrosion parameter vector to generate an environmental stress correction coefficient;

[0021] S26: Based on the degradation feature comparison table and the environmental stress correction coefficient, constructing a cross-scene mapping relationship between field service data and accelerated aging experiment data through a generative adversarial network to generate a historical loss data mapping table, wherein the historical loss data mapping table includes a fatigue conversion factor of the surface deformation feature matrix, a structural attenuation coefficient of the internal density distribution tensor, and a time-varying acceleration factor of the environmental corrosion parameter vector.

[0022] Optionally, the specific method of constructing a cross-scene mapping relationship through a generative adversarial network in S26 includes:

[0023] S261: Inputting the historical surface deformation feature matrix in the degradation feature comparison table and the simulated surface deformation matrix in the accelerated aging experiment data into a generator network, extracting a cross-scene fatigue conversion factor through a convolution attention mechanism, and simultaneously inputting the historical internal density distribution tensor and the experimental density data into a discriminator network to calculate a distribution difference value of the structural attenuation coefficient;

[0024] S262: Dynamically weighting the time-varying acceleration factor output by the generator network using the environmental stress correction coefficient to construct a loss function:

[0025] S263: Performing time-domain convolution on the historical environmental corrosion parameter vector and the environmental parameter sequence of the accelerated aging experiment to generate an environmental parameter mapping channel, and jointly optimizing the adversarial network with the loss function until the distribution difference value of the structural attenuation coefficient is lower than a preset tolerance threshold, and outputting a historical loss data mapping table including a calibrated surface deformation feature matrix fatigue conversion factor, an internal density distribution tensor structural attenuation coefficient, and an environmental corrosion parameter vector time-varying acceleration factor.

[0026] Optionally, S3 includes:

[0027] S31: Inputting the real-time value coefficient of the pledge, the enterprise credit score and the industry risk index into a data standardization module respectively, and performing dimensionless processing on the three through a range method to generate a standardized pledge real-time value coefficient, a standardized enterprise credit score and a standardized industry risk index;

[0028] S32: input the standardized pledge real-time value coefficient, the standardized enterprise credit score and the standardized industry risk index into a dynamic weight distributor, dynamically adjust the weight proportion based on the sulfide concentration value in the environmental corrosion parameter, when the sulfide concentration exceeds the preset critical value, increase the weight of the standardized pledge real-time value coefficient by 15%-20%, and construct a hybrid weighted matrix by using the analytic hierarchy process and the entropy method, output the weighted pledge value parameter, the weighted enterprise credit parameter and the weighted industry risk parameter;

[0029] S33: input the weighted pledge value parameter, the weighted enterprise credit parameter and the weighted industry risk parameter into a grey correlation degree model, calculate the dynamic deviation degree of the three from the preset reference index, correct the deviation threshold interval according to the geometric variation coefficient of the surface deformation characteristic matrix, and finally generate a dynamic pledge rate deviation index.

[0030] Optionally, the S33 of correcting the deviation threshold interval according to the geometric variation coefficient of the surface deformation characteristic matrix comprises:

[0031] S331: extract the maximum deformation gradient value at each time point from the surface deformation characteristic matrix, calculate the geometric variation coefficient by the sliding window method, and generate a deformation stability index;

[0032] S332: input the deformation stability index into a threshold dynamic adjustment module, correct the deviation threshold interval according to the product factor of the number of days of exceeding the sulfide concentration in the environmental corrosion parameter and the structural integrity index of the internal density distribution tensor, according to the following formula

[0033] S333: input the corrected deviation threshold interval into the grey correlation degree model, when the dynamic pledge rate deviation index exceeds the new upper limit, trigger a threshold out-of-range alarm signal, and perform a logical AND operation between the threshold out-of-range alarm signal and the pledge real-time value coefficient as the output condition of the final dynamic pledge rate deviation index.

[0034] Optionally, the S4 comprises:

[0035] S41: if the dynamic pledge rate deviation index exceeds the preset threshold, verify the trigger condition validity by a blockchain consensus node, package the dynamic pledge rate deviation index, the pledge real-time value coefficient and the environmental corrosion parameter to generate a risk event data package, and calculate a disposal emergency level based on the geometric variation coefficient of the surface deformation characteristic matrix;

[0036] S42: call a gradient disposal strategy template from a strategy library according to the disposal emergency level, dynamically compare the pledge real-time value coefficient with a preset pledge rate reference curve, trigger a threshold by the structural integrity index of the internal density distribution tensor correction instruction, and generate a gradient disposal instruction set including a priority identifier;

[0037] S43: input the gradient disposal instruction set into the smart contract execution engine, perform instruction validity verification according to the digital signature state of the pledge participation node, sequence the automatic supplementary pledge instruction, the interest rate adjustment instruction and the forced liquidation instruction according to the disposal emergency level, generate a blockchain transaction hash with a timestamp, and synchronize the execution of the gradient disposal strategy through a distributed ledger.

[0038] Optionally, the gradient disposal instruction set includes a first priority instruction, a second priority instruction and a third priority instruction, wherein:

[0039] The first priority instruction is to activate the automatic supplementary pledge instruction if the structural integrity index of the internal density distribution tensor decreases by more than 10%;

[0040] The second priority instruction is to start the interest rate adjustment instruction if the sulfide concentration in the environmental corrosion parameter exceeds the standard for 24 consecutive hours;

[0041] The third priority instruction is to execute the forced liquidation instruction if the maximum deformation variable of the surface deformation feature matrix breaks through the safety boundary.

[0042] A movable property pledge financing risk assessment system for implementing the movable property pledge financing risk assessment method described above, comprising the following modules:

[0043] A physical state acquisition module for acquiring the surface deformation features, internal density distribution and environmental corrosion parameters of the pledged property in real time through a three-dimensional laser scanner array, an ultrasonic density probe array and an electrochemical corrosion sensor group, and generating physical state data;

[0044] A degradation coupling module for extracting material fatigue degree sequences, structural integrity indexes and corrosion rate characteristics based on material property degradation sub-models and environmental stress accumulation sub-models, and performing adaptive weighted fusion of degradation factors combined with historical loss data mapping tables to output real-time value coefficients of the pledged property;

[0045] A data standardization module for receiving real-time value coefficients of the pledged property, enterprise credit scores and industry risk indexes, and performing normalization processing on them using the range method to generate standardized results;

[0046] A dynamic weight allocation module for dynamically adjusting the weight proportions of the three types of standardized indicators based on the sulfide concentration value in the environmental corrosion parameter, and constructing a hybrid weighting matrix through the analytic hierarchy process and the entropy method to output weighted pledged property value parameters, weighted enterprise credit parameters and weighted industry risk parameters;

[0047] The threshold dynamic adjustment module is used for dynamically correcting the deviation degree judgment threshold by combining the geometric variation coefficient of the surface deformation feature matrix, the structural integrity index of the internal density distribution tensor and the abnormal duration of the environmental corrosion parameter.

[0048] The deviation degree calculation module is used for calculating the dynamic staking rate deviation degree index based on the grey correlation degree model, and judging whether it exceeds the corrected deviation degree threshold interval.

[0049] The risk response module is used for generating a risk event data packet through a blockchain consensus node when the dynamic staking rate deviation degree index exceeds the threshold value, calculating a disposal emergency level, and calling a matched gradient disposal instruction set.

[0050] The smart contract execution module is used for inputting the gradient disposal instruction set into a smart contract execution engine, verifying the validity of the instruction based on the digital signature state of the staking participation node, and performing trusted execution on the automatic supplementary staking object instruction, the interest rate adjustment instruction and the forced liquidation instruction through a distributed ledger.

[0051] The beneficial effects of the present application are as follows:

[0052] In the present application, a three-dimensional laser scanner array, an ultrasonic density detector array and an electrochemical corrosion sensor group are deployed in the staking object storage environment to respectively collect surface deformation features, internal density distribution and environmental corrosion parameters, and multiple algorithms such as Gaussian filter, wavelet decoupling algorithm and Kalman filter are used for data preprocessing, and then a unified physical state data set is constructed through sliding window and time alignment to ensure the high consistency and stability of the extracted data in the space-time dimension, so that the health status of the staking object is perceived and dynamically tracked with millimeter-level precision.

[0053] In the present application, the material property degradation sub-model and the environmental stress accumulation sub-model are fused, the LSTM network is pre-trained to extract the material fatigue degree sequence, the Weibull distribution is introduced to model the corrosion rate, the fatigue degree conversion factor, the structure attenuation coefficient and the time-varying acceleration factor extracted from the historical loss data mapping table are combined, and the degradation coupling module is used to realize multi-factor dynamic fusion. Meanwhile, the data standardization module, the dynamic weight distributor and the grey correlation degree model work together to effectively combine the real-time value coefficient of the staking object, the enterprise credit score and the industry risk index to generate a quantifiable dynamic staking rate deviation degree index, and the geometric variation coefficient of the surface deformation feature matrix and the structural integrity index of the internal density tensor are used to dynamically adjust the judgment threshold, which comprehensively improves the discrimination sensitivity and scene adaptability of the staking risk.

[0054] The application, when the dynamic staking rate deviation index exceeds the dynamic threshold, the system verifies the validity of the triggering event through the blockchain consensus node, combines the geometric coefficient of variation extracted in the risk event data packet to evaluate the emergency treatment level, and calls the corresponding gradient treatment strategy template. Based on the internal density tensor, corrosion duration and deformation overrun, three types of gradient disposal instruction sets of automatic supplementary pledge, interest rate adjustment and forced liquidation are generated, and distributed, time-stamped and trusted transactions are completed through the smart contract execution engine, realizing the full-link closed-loop control of "data sensing-risk assessment-strategy generation-contract execution", and significantly improving the response speed and automation level of risk control in the process of staking financing. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the present application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are only a part of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0056] Figure 1 The method flowchart of the embodiment of the present application is shown in the figure.

[0057] Figure 2 The system flowchart of the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0058] The present application will be described in detail below in combination with the drawings and specific embodiments. It should be noted that in order to make the embodiments more detailed, the following embodiments are the best, preferred embodiments, and other alternative ways can also be used by those skilled in the art to implement them; and the drawings are only used to describe the embodiments more specifically, and are not intended to specifically limit the present application.

[0059] It should be noted that in the specification, "one embodiment", "embodiment", "exemplary embodiment", "some embodiments" and the like indicate that the described embodiments can include specific features, structures or characteristics, but not necessarily every embodiment includes the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in combination with an embodiment, it should be within the knowledge of those skilled in the related art to realize this feature, structure or characteristic in combination with other embodiments (whether or not explicitly described).

[0060] In general, terms can be understood, at least partly, from usage in the context. For example, depending at least in part on the context, the term "one or more" as used herein can describe any feature, structure, or characteristic in the singular or can describe combinations of features, structures, or characteristics, in the plural, both singular and plural. Further, the term "based on" can be understood as not necessarily of a set of exclusive factors, but, instead, as allowing for existence of additional factors not necessarily explicitly described.

[0061] As shown in Figure 1 A movable collateral pledge financing risk assessment method, comprising the following steps:

[0062] S1: Real-time acquisition of physical state data through a monitoring device deployed in the collateral storage environment, the physical state data including surface deformation characteristics, internal density distribution, and environmental corrosion parameters;

[0063] S2: Inputting the physical state data into a dynamic assessment model to output a real-time value coefficient of the collateral, wherein the dynamic assessment model is obtained by training based on historical loss data, including a material property degradation sub-model and an environmental stress accumulation sub-model;

[0064] S3: Multi-dimensional fusion of the real-time value coefficient with enterprise credit score and industry risk index to generate a dynamic collateral rate deviation index;

[0065] S4: If the deviation index exceeds a preset threshold, triggering an intelligent contract to execute a gradient disposal strategy, the gradient disposal strategy including a priority combination of automatic collateral replenishment instructions, interest rate adjustment instructions, and forced liquidation instructions.

[0066] S1 includes:

[0067] S11, surface deformation feature acquisition: real-time acquisition of collateral surface point cloud data through a three-dimensional laser scanner array deployed in the collateral storage environment .

[0068] Performing point-by-point difference calculation on the collected point cloud data and a preset reference geometric model to generate a surface deformation feature matrix , the calculation formula of which is:

[0069] ;

[0070] wherein, is the point cloud data at the current time point, is the reference model point cloud data, is the surface deformation feature matrix with millimeter-level precision, is the number of sampling points in the point cloud.

[0071] S12, internal density distribution and environment corrosion parameter extraction: the surface deformation feature matrix Input Gaussian filter to eliminate high-frequency motion noise generated during scanning. After filtering, the smoothed deformation feature matrix is obtained : ;

[0072] where, is a Gaussian kernel with a standard deviation of , and * represents convolution operation;

[0073] Synchronous acquisition of internal acoustic response signals of the pledge, and use of an ultrasonic density detector array to obtain P / S wave velocity ratio data , where is the longitudinal wave velocity, is the transverse wave velocity;

[0074] The above data is processed using a wavelet decoupling algorithm to strip out the internal density distribution reflecting the material structure homogeneity In addition, the deployed electrochemical corrosion sensor group continuously monitors the sulfide concentration , humidity and temperature in the environment. These data are input into the Kalman filter for state estimation to generate a multivariate environmental corrosion parameter vector , denoted as:

[0075] ;

[0076] where, is the filtered sulfide concentration, is the filtered humidity value, is the filtered temperature value.

[0077] S13, physical state data set generation: the smoothed surface deformation feature matrix , internal density distribution and environmental corrosion parameter vector are synchronized and aligned in time series, processed uniformly through a data fusion gateway, and a physical state data set is constructed , denoted as: , where, is a set of time series of data acquisition;

[0078] S2 includes:

[0079] S21, first degradation factor generation: the surface deformation feature matrix in the physical state data is input into the material property degradation sub-model, and a pre-trained LSTM (Long Short Term Memory) neural network is called to extract the material fatigue degree sequence in the time dimension , which is generated in the following way: , wherein, is a material fatigue degree sequence, represents a pre-trained LSTM neural network with parameters , and represents a surface deformation feature sub-sequence with a time window of ;

[0080] At the same time, the internal density distribution is used to calculate the density gradient change rate in three-dimensional space, and the calculation formula is:

[0081] ;

[0082] Finally, the material fatigue degree sequence and the density gradient change rate are fused to generate a first degradation factor , which is represented as: ;

[0083] wherein, : the first degradation factor, is a fusion weight coefficient, and Norm is a normalization function;

[0084] S22, second degradation factor generation: input the environmental corrosion parameter vector into the environmental stress accumulation sub-model, and use the dynamic time warping algorithm to align the time sequence relationship to obtain the aligned time sequence ;

[0085] Then, the corrosion rate is fitted using the Weibull distribution function, and the corrosion rate function is defined as:

[0086] wherein, is the corrosion rate, is the scale parameter and shape parameter of the Weibull distribution, which is determined by least squares fitting.

[0087] Combined with the normalized weight function of the surface deformation feature space distribution, the second degradation factor is generated, which is represented as: ;

[0088] S23, collateral real-time value coefficient output: input the first degradation factor and the second degradation factor into the degradation coupling module, which is based on the historical loss data mapping table stored in the transfer learning framework., to realize the nonlinear mapping and weight fusion of the two degradation factors.

[0089] The fusion function is represented as: ;

[0090] wherein, is the real-time value coefficient of the collateral, is the historical loss data mapping table, including training data between different combinations of degradation factors and estimated values in the past, is the weight coefficient vector dynamically adjusted according to the timeliness of the physical state data, is an adaptive weighted fusion function generated by a transfer learning model.

[0091] The construction method of the historical loss data mapping table saved in the transfer learning framework includes:

[0092] S24, Degradation feature comparison table generation: Extract the historical surface deformation feature matrix , the historical internal density distribution tensor and the historical environmental corrosion parameter vector from the physical state data of the historical stage, perform spatio-temporal alignment processing on the above historical data and the artificial accelerated degradation data in the accelerated aging experiment database, establish a corresponding relationship using an index matching function , represented as:

[0093] ;

[0094] wherein, is the degradation feature comparison table, recording the mapping relationship between the historical actual state and the accelerated degradation response.

[0095] S25, Environmental stress correction coefficient calculation: Input the above historical environmental corrosion parameter vector into the environmental stress accumulation sub-model to calculate the theoretical corrosion rate curve , which is in the form of:

[0096] ;

[0097] At the same time, extract the actual monitored environmental corrosion parameter vector , and calculate the corresponding actual corrosion rate . Based on the two, residual analysis is performed, and the residual function is defined as:

[0098] ;

[0099] The residual sequence is time-integrated or least squares fitted to generate the environmental stress correction coefficient , represented as: ;

[0100] wherein, is the theoretical corrosion rate curve, is the actual corrosion rate, is the environmental stress correction factor, is the corrosion rate residual.

[0101] S26, historical loss data mapping table generation: based on the degradation feature comparison table in S24 and the environmental stress correction factor calculated in S25 , learn the field service data using the generative adversarial network and the cross-scene mapping function between the accelerated aging test data , , is expressed as:

[0102] wherein, represents the cross-scene mapping relationship, is the experimental equivalent data after style transfer; the discriminator of the GAN is used to minimize the statistical distance between the experimental data and the mapped data.

[0103] On this basis, the historical loss data mapping table is constructed , the structure of which includes:

[0104] fatigue conversion factor of surface deformation feature matrix : used to convert to material fatigue;

[0105] structure attenuation coefficient of internal density distribution tensor : used to quantitatively describe the density homogeneity degradation;

[0106] time-varying acceleration factor of environmental corrosion parameter vector : used to map the equivalent influence of the actual environment on the accelerated aging process, and finally generate the mapping table as follows:

[0107] ;

[0108] The specific method of constructing the cross-scene mapping relationship by the generative adversarial network in S26 includes:

[0109] S261, cross-scene fatigue degree and structure attenuation difference extraction: input the historical surface deformation feature matrix in the degradation feature comparison table and the simulated surface deformation matrix in the accelerated aging test data into the generator network , use convolution attention mechanism for encoding enhancement to extract the fatigue conversion factor , ; wherein: is a generator network, and the parameters are , CA( ) represents a convolutional attention mechanism that enhances cross-domain deformation feature interaction, is a fatigue conversion factor for the surface deformation feature matrix.

[0110] At the same time, the historical internal density distribution tensor and the experimental density tensor are input into the discriminator network , and the output is the distribution difference value of the structure attenuation coefficient , which is represented as:

[0111] ;

[0112] wherein, is a discriminator network, and the parameters are , is the distribution difference value of the structure attenuation coefficient, which measures the distinguishability of density degradation between two scenarios.

[0113] S262: Time-varying acceleration factor weighting and loss function construction: Call the environmental stress correction coefficient obtained in S25 , weight the time-varying acceleration factor output by the generator network to obtain the weighted factor , which is represented as: ;

[0114] Construct a composite loss function for training the generative adversarial network , which includes the following three parts:

[0115] ;

[0116] wherein, is the standard adversarial loss, which measures the ability of the discriminator to distinguish between true and false data, represents the distribution difference loss of the structure attenuation coefficient, represents the time-varying acceleration factor deviation, is the weighting coefficient, which is used to balance the importance of each loss.

[0117] S263, joint optimization and mapping table output: Perform time-domain convolution operation on the historical environmental corrosion parameter vector and the environmental parameter sequence of the accelerated aging experiment to obtain the environmental parameter mapping channel , which is represented as: wherein, is the environmental convolution kernel constructed based on the experimental environmental response mode, and * represents the time convolution operation.

[0118] mapping channel of environmental parameters with the loss function are used together to optimize the training process of the adversarial generative network, and the parameters of the generator and the discriminator are iteratively updated until the following convergence conditions are met:

[0119] ;

[0120] wherein, is a preset tolerance threshold representing the upper limit of the acceptable difference in structural decay distribution. When the above condition is met, the historical loss data mapping table is output ;

[0121] S3 includes:

[0122] S31, standardization processing: the real-time value coefficient of the pledge and the enterprise credit score , industry risk index input data standardization module, using the range method for normalization, the calculation formula is as follows:

[0123] ;

[0124] ;

[0125] ;

[0126] wherein, is the standardized real-time value coefficient of the pledge, is the standardized enterprise credit score, is the standardized industry risk index, is the minimum and maximum value of the pledge value coefficient within the historical range, is the extreme value range of the enterprise credit score, is the extreme value range of the industry risk index.

[0127] S32, dynamic weight adjustment and hybrid weighting: the above three standardization results , , input dynamic weight distributor. According to the sulfide concentration value in the environmental corrosion parameter, the weighting strategy is adjusted in real time.

[0128] If the current sulfide concentration satisfies: ;

[0129] then increase the weight of the standardized real-time value coefficient of the pledge by 15%-20%, set the original initial weight vector as , and the updated weight is ;

[0130] The judgment matrix is generated by analytic hierarchy process and entropy method respectively and information vector , a hybrid weighted matrix is constructed , through linear fusion:

[0131] ;

[0132] wherein, is the fusion regulatory factor, , weighted vectors obtained by analytic hierarchy process and entropy method respectively.

[0133] The final output: , represents the weighted collateral value parameter;

[0134] , represents the weighted enterprise credit parameter;

[0135] , represents the weighted industry risk parameter.

[0136] S33, dynamic collateral rate deviation index calculation: input the above three weighted parameters into the grey correlation degree model, and carry out correlation analysis with the preset reference vector , calculate the grey correlation coefficient , and calculate the deviation of the three parameters :

[0137] ;

[0138] wherein, is the dynamic collateral rate deviation index, is the current dynamic adjusted weight coefficient, is the weighted collateral value parameter, is the weighted enterprise credit parameter, is the weighted industry risk parameter;

[0139] In addition, in order to adapt to the volatility of the state of the collateral, the geometric coefficient of variation CV is calculated by using the surface deformation characteristic matrix , and the deviation threshold interval is corrected ;

[0140] ;

[0141] wherein, is the standard deviation and mean of the surface deformation characteristic matrix, is the fluctuation sensitivity adjustment factor, is the corrected deviation threshold upper limit.

[0142] In S33, the deviation threshold interval is corrected according to the geometric variation coefficient of the surface deformation feature matrix, including:

[0143] S331, Geometric Variation Coefficient and Deformation Stability Index Generation: From Surface Deformation Feature Matrix Extract the maximum deformation gradient value at each time point , its expression is:

[0144] ;

[0145] Sliding window method is used to In continuous time windows Calculate the geometric coefficient of variation CV , the calculation formula is:

[0146] ;

[0147] in, is the standard deviation of the maximum deformation gradient value in the time window, is the mean of the maximum deformation gradient value in the time window, C is the geometric coefficient of variation.

[0148] CV Defined as deformation stability index : ;

[0149] S332, deviation threshold interval correction: the above deformation stability index The input threshold dynamic adjustment module combines the following two factors to perform threshold correction:

[0150] 1. Number of days when the sulfide concentration in the environmental corrosion parameters exceeds the standard ;

[0151] 2. Internal density distribution tensor Calculated Structural Integrity Index , which is defined as:

[0152] ;

[0153] in, is the reference density field (e.g., the initial uniform state), is the spatial region covered by the density tensor, It is the structural integrity index, reflecting the degree of density degradation.

[0154] Use the product of the two factors above , combined with Correct the original deviation threshold interval The modified deviation threshold interval :

[0155] ;

[0156] ;

[0157] wherein, is a dynamic threshold correction coefficient (empirical setting), is a modified deviation threshold interval, is a deformation stability index.

[0158] S333, deviation threshold linkage output mechanism: input the modified deviation threshold interval into the grey correlation degree model, when the calculated dynamic staking rate deviation index satisfies:

[0159] ;

[0160] that is, when it exceeds the new upper limit, a threshold out-of-range alarm signal is triggered, otherwise ;

[0161] the alarm signal is then logically ANDed with the real-time value coefficient of the staking asset (that is, the alarm is valid only when the value coefficient has not fallen below the minimum protection value), and the minimum protection value threshold is set to , and the final output judgment is:

[0162] ;

[0163] wherein, is the final dynamic staking rate deviation index, is the threshold out-of-range alarm signal, is the real-time value coefficient of the staking asset, is the set minimum acceptable value protection lower limit.

[0164] S4 includes:

[0165] S41, risk event data packet generation and emergency level calculation: when the dynamic staking rate deviation index satisfies ;

[0166] a trigger condition validity verification is performed by a blockchain consensus node , and after reaching an agreement through multi-node signature, a risk event data packet including the following contents is generated , wherein, is the dynamic staking rate deviation index, is the real-time value coefficient of the collateral, is the environmental corrosion parameter vector, is the trigger event timestamp.

[0167] based on the surface deformation feature matrix of the geometric coefficient of variation CV , the treatment emergency level is defined , which is defined as: ;

[0168] wherein, is the preset level division threshold, is the treatment emergency level (3 is the highest).

[0169] S42, gradient treatment instruction set generation: according to the treatment emergency level , the corresponding gradient treatment strategy template is called from the preset strategy library. Then the current real-time value coefficient of the collateral is compared with the historical training derived collateral rate benchmark curve to obtain the deviation curve :

[0170] ;

[0171] Then the difference value is combined with the structural integrity index calculated by the internal density distribution tensor to modify the trigger threshold of each treatment instruction: wherein, is the original instruction trigger threshold, is the sensitivity adjustment coefficient, is the modified instruction trigger threshold.

[0172] Finally, the gradient treatment instruction set with priority identification is generated:

[0173] ;

[0174] wherein, CMD is the th treatment instruction (such as automatic replenishment of collateral, interest rate adjustment, forced liquidation) and its priority identification .

[0175] S43, blockchain execution of gradient treatment strategy: input the above gradient treatment instruction set into the smart contract execution engine ;

[0176] The execution engine first verifies the digital signature status of all initiators and collateral-related parties, ensuring that the participating nodes' signatures are valid. The verification function is:

[0177] ;

[0178] Then, according to the emergency level of disposal The following three types of instructions are sequenced:

[0179] (1) Automatic collateral replenishment instruction;

[0180] (2) Interest rate adjustment instruction;

[0181] (3) Forced liquidation instruction;

[0182] After sequencing, an execution chain with time stamp is generated:

[0183] ;

[0184] The execution chain is packaged into a blockchain transaction hash , and written into a distributed ledger :

[0185] ;

[0186] Finally, the multi-party synchronization and trusted execution of the gradient disposal strategy are completed.

[0187] The gradient disposal instruction set includes three priority control mechanisms, as follows:

[0188] First priority instruction: automatic collateral replenishment instruction;

[0189] If the structural integrity index of the internal density distribution tensor decreases, and the decrease compared to the previous cycle value exceeds 10%, the automatic collateral replenishment instruction is immediately activated. The decision logic is expressed as:

[0190] ;

[0191] Where, is the structural integrity index calculated at the current time, is the structural integrity index of the previous cycle, is the first priority automatic collateral replenishment instruction, and the priority identifier is .

[0192] Second priority instruction: interest rate adjustment instruction;

[0193] If the sulfide concentration in the environmental corrosion parameter continuously exceeds the set threshold and reach 24 consecutive hours, i.e.:

[0194] ;

[0195] wherein, is the time at which the sulfide concentration is monitored, is the preset upper limit of the sulfide concentration safety, is the starting time point of continuous monitoring, is the second priority rate adjustment instruction, and the priority is identified as .

[0196] The third priority instruction is a forced liquidation instruction.

[0197] If the maximum deformation variable in the surface deformation feature matrix exceeds the set safety boundary value , the forced liquidation instruction is triggered. The judgment condition is:

[0198] ;

[0199] wherein, is the maximum value in the surface deformation feature matrix at the current time point, is the preset maximum safety deformation variable threshold, is the third priority forced liquidation instruction, and the priority is identified as .

[0200] As shown in Figure 2 , a movable property pledge financing risk assessment system for realizing the movable property pledge financing risk assessment method described above comprises the following modules:

[0201] A physical state acquisition module is used to acquire the surface deformation features, internal density distribution and environmental corrosion parameters of the pledged property in real time through a three-dimensional laser scanner array, an ultrasonic density detector array and an electrochemical corrosion sensor group, and generate a physical state data set.

[0202] A degradation coupling module is used to extract material fatigue degree sequences, structural integrity indexes and corrosion rate features based on material property degradation sub-models and environmental stress accumulation sub-models, and to perform adaptive weighted fusion of degradation factors combined with a historical loss data mapping table, and output a real-time value coefficient of the pledged property.

[0203] A data standardization module is used to receive the real-time value coefficient of the pledged property, the enterprise credit score and the industry risk index, and to perform normalization processing on them using the range method to generate a standardization result.

[0204] ​The dynamic weight distribution module is used for dynamically adjusting the weight proportion of the three types of standardized indexes based on the sulfide concentration value in the environmental corrosion parameter, and constructing a hybrid weighting matrix through the analytic hierarchy process and the entropy method, and outputting the weighted pledge value parameter, the weighted enterprise credit parameter and the weighted industry risk parameter.

[0205] The threshold dynamic adjustment module is used for dynamically correcting the deviation degree judgment threshold by combining the geometric variation coefficient of the surface deformation feature matrix, the structural integrity index of the internal density distribution tensor and the abnormal duration of the environmental corrosion parameter.

[0206] The deviation degree calculation module is used for calculating the dynamic pledge rate deviation degree index based on the grey correlation degree model, and judging whether it exceeds the corrected deviation degree threshold interval.

[0207] The risk response module is used for generating a risk event data packet through a blockchain consensus node when the dynamic pledge rate deviation degree index exceeds the threshold value, calculating a disposal emergency level, and calling a matched gradient disposal instruction set.

[0208] The smart contract execution module is used for inputting the gradient disposal instruction set into a smart contract execution engine, verifying the validity of the instruction based on the digital signature state of the pledge participation node, and performing trusted execution on the automatic supplementary pledge material instruction, the interest rate adjustment instruction and the forced liquidation instruction through a distributed ledger.

[0209] The present application encompasses any alternative, modification, equivalent method and scheme made on the essence and scope of the present application. In order to make the public have a thorough understanding of the present application, specific details are described in the following preferred embodiments of the present application, and the present application can also be fully understood without the description of these details to those skilled in the art. In addition, in order to avoid unnecessary confusion to the essence of the present application, well-known methods, processes, procedures, elements and circuits are not described in detail.

[0210] The above is only the preferred embodiment of the present application, and it should be pointed out that for ordinary skilled in the art, without departing from the principle of the present application, a number of improvements and refinements can be made, and these improvements and refinements should be regarded as the protection scope of the present application.

Claims

1. A movable property pledge financing risk assessment method, characterized in that: The following steps are involved: S1: Real-time physical status data is acquired through monitoring equipment deployed in the collateral storage environment. The physical status data includes surface deformation characteristics, internal density distribution, and environmental corrosion parameters. Specifically, it includes: S11: A 3D laser scanner array deployed in the collateral storage environment collects surface point cloud data in real time, performs differential calculations using a preset reference geometric model, and generates surface deformation features with millimeter-level accuracy. S12: Inputting the surface deformation characteristics into a Gaussian filter to eliminate motion noise, and simultaneously collecting transverse and longitudinal wave velocity ratio data through an ultrasonic density detector array, combining with a wavelet decoupling algorithm to separate the internal density distribution, and using an electrochemical corrosion sensor group to continuously monitor the sulfide concentration, humidity and temperature parameters in the environment, and generating environmental corrosion parameters after Kalman filtering; S13: performing spatiotemporal alignment on the pre-processed surface deformation features, internal density distribution, and environmental corrosion parameters according to a time series, and integrating them through a data fusion gateway to generate physical state data, wherein the physical state data includes a surface deformation feature matrix with a timestamp, an internal density distribution tensor, and an environmental corrosion parameter vector; S2: Input the physical state data into a dynamic assessment model to output a real-time value coefficient of the pledged assets, wherein the dynamic assessment model is trained based on historical loss data and includes a material property degradation sub-model and an environmental stress accumulation sub-model; specifically, the following steps are involved: S21: Inputting the surface deformation feature matrix in the physical state data into the material property degradation sub-model, extracting the material fatigue sequence in the time dimension through a pre-trained LSTM neural network, and calculating the density gradient change rate based on the internal density distribution tensor to generate a first degradation factor; S22: Inputting the environmental corrosion parameter vector into the environmental stress accumulation sub-model, using a dynamic time warping algorithm to align the temporal relationship of sulfide concentration, humidity, and temperature parameters, fitting the corrosion rate curve using a Weibull distribution, and combining the spatial distribution weight of the surface deformation feature matrix to generate a second degradation factor; S23: Input the first degradation factor and the second degradation factor into the degradation coupling module. Using the historical loss data mapping table stored in the transfer learning framework, the two are adaptively weighted and fused. The weight coefficient is dynamically adjusted according to the timeliness of the physical state data, and the real-time value coefficient of the pledged asset is output. The method for constructing the historical loss data mapping table stored in the transfer learning framework includes: S24: extracting a historical surface deformation feature matrix, a historical internal density distribution tensor, and a historical environmental corrosion parameter vector from the physical state data, performing spatiotemporal alignment with the artificial accelerated degradation data in the accelerated aging experiment database, and generating a degradation feature comparison table; S25: Inputting the historical environmental corrosion parameter vector into the environmental stress accumulation sub-model, calculating the theoretical corrosion rate curve, and performing residual analysis with the actual monitored environmental corrosion parameter vector to generate an environmental stress correction coefficient; S26: Based on the degradation feature comparison table and the environmental stress correction coefficient, a cross-scenario mapping relationship between the field service data and the accelerated aging test data is constructed through a generative adversarial network to generate the historical loss data mapping table, wherein the historical loss data mapping table includes a fatigue conversion factor of the surface deformation feature matrix, a structural attenuation coefficient of the internal density distribution tensor, and a time-varying acceleration factor of the environmental corrosion parameter vector; S3: The real-time value coefficient is integrated with the enterprise credit score and industry risk index in multiple dimensions to generate a dynamic pledge rate deviation indicator; S4: If the deviation index exceeds the preset threshold, the smart contract is triggered to execute a gradient disposal strategy, which includes a priority combination of automatic collateral replenishment instructions, interest rate adjustment instructions, and forced liquidation instructions.

2. A movable property pledge financing risk assessment method according to claim 1, characterized in that: The specific method of constructing a cross-scene mapping relationship through the adversarial generative network in S26 includes: S261: Inputting the historical surface deformation feature matrix in the degradation feature comparison table and the simulated surface deformation matrix in the accelerated aging experimental data into the generator network, extracting the fatigue conversion factor across scenes through the convolutional attention mechanism, and simultaneously inputting the historical internal density distribution tensor and the experimental density data into the discriminator network to calculate the distribution difference value of the structural attenuation coefficient; S262: Dynamically weight the time-varying acceleration factor output by the generator network using the environmental stress correction coefficient to construct a loss function: S263: Perform time-domain convolution on the historical environmental corrosion parameter vector and the environmental parameter sequence of the accelerated aging experiment to generate an environmental parameter mapping channel, and jointly optimize the adversarial generative network with the loss function until the distribution difference value of the structural attenuation coefficient is lower than the preset tolerance threshold, and output a historical loss data mapping table including the calibrated surface deformation feature matrix fatigue conversion factor, the internal density distribution tensor structural attenuation coefficient and the time-varying acceleration factor of the environmental corrosion parameter vector.

3. A movable property pledge financing risk assessment method according to claim 2, characterized in that: The S3 includes: S31: Input the real-time value coefficient of the pledge, the enterprise credit score, and the industry risk index into the data standardization module respectively, and normalize the dimensions of the three using the range method to generate a standardized real-time value coefficient of the pledge, a standardized enterprise credit score, and a standardized industry risk index; S32: Input the standardized collateral real-time value coefficient, standardized enterprise credit score, and standardized industry risk index into a dynamic weight allocator, dynamically adjust the weight ratio based on the sulfide concentration value in the environmental corrosion parameter, and increase the weight of the standardized collateral real-time value coefficient by 15%-20% when the sulfide concentration exceeds a preset critical value. Simultaneously, a hybrid weighted matrix is ​​constructed using the analytic hierarchy process and entropy method to output the weighted collateral value parameter, weighted enterprise credit parameter, and weighted industry risk parameter. S33: Input the weighted pledge value parameter, weighted enterprise credit parameter and weighted industry risk parameter into the grey correlation model, calculate the dynamic deviation of the three from the preset benchmark indicator, correct the deviation threshold interval according to the geometric variation coefficient of the surface deformation feature matrix, and finally generate a dynamic pledge rate deviation indicator.

4. A movable property pledge financing risk assessment method according to claim 3, characterized in that: Correcting the deviation threshold interval according to the geometric variation coefficient of the surface deformation feature matrix in S33 includes: S331: extracting the maximum deformation gradient value at each time point from the surface deformation feature matrix, calculating the geometric variation coefficient using a sliding window method, and generating a deformation stability index; S332: Input the deformation stability index into the threshold dynamic adjustment module, and according to the product factor of the number of days when the sulfide concentration exceeds the standard in the environmental corrosion parameter and the structural integrity index of the internal density distribution tensor, correct the deviation threshold interval according to the following formula S333: The revised deviation threshold interval is input into the grey correlation model. When the dynamic pledge rate deviation index exceeds the new upper limit, a threshold crossing alarm signal is triggered, and a logical AND operation is performed on the threshold crossing alarm signal and the real-time value coefficient of the pledged assets as the output condition of the final dynamic pledge rate deviation index.

5. A movable property pledge financing risk assessment method according to claim 4, characterized in that: The S4 includes: S41: If the dynamic pledge rate deviation index exceeds a preset threshold, the validity of the trigger condition is verified through the blockchain consensus node, the dynamic pledge rate deviation index, the real-time value coefficient of the pledged assets, and the environmental corrosion parameters are packaged to generate a risk event data package, and the emergency level of the disposal is calculated based on the geometric variation coefficient of the surface deformation feature matrix; S42: Based on the emergency level of the disposal, a gradient disposal strategy template is called from the strategy library, the real-time value coefficient of the pledged assets is dynamically compared with a preset pledge rate benchmark curve, the instruction trigger threshold is modified by the structural integrity index of the internal density distribution tensor, and a gradient disposal instruction set including a priority identifier is generated; S43: Input the gradient disposal instruction set into the smart contract execution engine, verify the validity of the instructions based on the digital signature status of the pledge participating nodes, arrange the automatic collateral replenishment instructions, interest rate adjustment instructions and forced liquidation instructions in a time sequence according to the disposal urgency level, generate a blockchain transaction hash with a timestamp, and synchronously execute the gradient disposal strategy through the distributed ledger.

6. A movable property pledge financing risk assessment method according to claim 5, characterized in that: The gradient handling instruction set includes first priority instructions, second priority instructions, and third priority instructions, wherein: First priority command: If the structural integrity index of the internal density distribution tensor drops by more than 10%, the automatic replenishment of collateral command is activated; Second priority instruction: If the sulfide concentration in the environmental corrosion parameter exceeds the standard for 24 consecutive hours, the interest rate adjustment instruction is activated; Third priority instruction: If the maximum deformation of the surface deformation feature matrix exceeds the safety limit, a forced liquidation instruction is executed.

7. A movable property pledge financing risk assessment system, used to implement a movable property pledge financing risk assessment method according to any one of claims 1 to 6, characterized in that: Includes the following modules: Physical state acquisition module: This module uses a 3D laser scanner array, an ultrasonic density detector array, and an electrochemical corrosion sensor group to acquire the surface deformation characteristics, internal density distribution, and environmental corrosion parameters of the pledged assets in real time, and generates a physical state data set. Degradation Coupling Module: This module extracts material fatigue series, structural integrity index, and corrosion rate characteristics based on the material property degradation sub-model and the environmental stress accumulation sub-model. It then performs adaptive weighted fusion of degradation factors using a historical loss data mapping table to output the real-time value coefficient of the collateral. Data standardization module: used to receive the real-time value coefficient of the pledge, corporate credit score and industry risk index, and normalize them using the range method to generate standardized results; Dynamic Weight Allocation Module: This module dynamically adjusts the weight ratios of the three standardized indicators based on the sulfide concentration value in the environmental corrosion parameters. It also constructs a hybrid weighted matrix using the analytic hierarchy process and entropy method, and outputs weighted collateral value parameters, weighted corporate credit parameters, and weighted industry risk parameters. Threshold dynamic adjustment module: used to dynamically modify the deviation judgment threshold by combining the geometric variation coefficient of the surface deformation feature matrix, the structural integrity index of the internal density distribution tensor, and the abnormal duration of the environmental corrosion parameters; Deviation calculation module: used to calculate the dynamic pledge rate deviation index based on the grey correlation model and determine whether it exceeds the revised deviation threshold range; Risk Response Module: When the dynamic pledge rate deviation indicator exceeds the threshold, it generates a risk event data packet through the blockchain consensus node, calculates the emergency level of the disposal, and calls the matching gradient disposal instruction set; Smart contract execution module: used to input the gradient disposal instruction set into the smart contract execution engine, verify the validity of the instructions based on the digital signature status of the pledge participating nodes, and perform reliable execution of automatic collateral replenishment instructions, interest rate adjustment instructions and forced liquidation instructions through the distributed ledger.

Images