A blockchain-based lease platform risk early warning method

Abstract

The application discloses a kind of based on blockchain's leasing platform risk early warning method, it is related to blockchain and smart contract technical field, including the following steps: S1, the spatial compliance constraint data of lease object is chained with the initial control authority benchmark of edge control end and stored evidence;S2, output space constraint deviation characteristics;S3, based on the improved Flowformer model introduces fluctuation evolution mechanism, in turn executes critical deviation measure, time evolution diffusion and nonlinear phase transition activation, deduces risk implicit fluctuation;S4 generates degraded control authority benchmark candidate value;S5, generates off-chain final state settlement data;S6, generates degraded execution voucher;S7, output equipment physical operating parameter.The application overcomes the limitation of traditional method risk control decoupling, linear degradation distortion and lack of on-chain authentication, provides a safe adaptive solution for leasing platform risk early warning.

Description

Technical Field

[0001] This invention relates to the field of blockchain and smart contract technology, and in particular to a risk warning method for a blockchain-based leasing platform. Background Technology

[0002] With the continuous growth of the scale of spatiotemporal trajectory data of leased assets and the frequency of edge control interactions, traditional leasing platform risk warning systems face severe trust and timeliness challenges in real-time monitoring and dynamic intervention. Existing risk warning methods typically decouple trajectory monitoring from device control. Off-chain monitoring platforms use rule engines to determine spatiotemporal trajectories and trigger alarms, while on-chain blockchains are only used for post-event evidence storage. There is a significant execution delay and state disconnect between the two. This disconnect mechanism ignores the deep causal relationship between real-time trajectory deviation characteristics and device physical control permissions, making it difficult for warning signals to directly drive devices to execute adaptive degradation control, thus limiting the immediacy of risk intervention. In addition, traditional degradation control strategies mostly use static threshold judgments and linear decay mappings, without introducing a fluctuation evolution mechanism to decouple the implicit fluctuations of risk from temporal dynamic divergence. This fails to capture the nonlinear phase transition characteristics in the risk accumulation process, resulting in delayed degradation control response and a tendency for over-intervention or under-intervention. Meanwhile, the degradation execution at the edge control terminal lacks pre-verification and endorsement at the blockchain level. Existing methods are unable to verify whether the baseline candidate value of degradation control authority exceeds the degradation boundary threshold. Furthermore, the integrity verification of off-chain final settlement data and the over-boundary verification of degradation authority are not verified in a two-dimensional manner. This makes the edge control terminal highly vulnerable to attacks that forge instructions or tamper with parameters, resulting in the risk of unauthorized execution. It is impossible to achieve a closed-loop anti-tampering protection throughout the entire chain, from the perception of spatiotemporal deviation to on-chain rights confirmation and then to secure execution at the edge.

[0003] Therefore, how to provide a risk warning method for a blockchain-based leasing platform is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] This invention proposes a blockchain-based risk warning method for leasing platforms. Through a closed-loop process of off-chain risk deduction and on-chain dual-dimensional verification based on blockchain and edge control, spatial constraint deviation features are extracted by performing spatial alignment and nonlinear mapping on real-time spatiotemporal trajectory data and spatial compliance constraint data of leased assets. These spatial constraint deviation features and real-time spatiotemporal trajectory data are then input into an improved Flowformer model to extract nonlinear risk features. A fluctuation evolution mechanism is introduced, sequentially performing critical deviation measurement, temporal evolution diffusion, and nonlinear phase transition activation on the nonlinear risk features to deduce the implicit fluctuations of risk and output a dynamic risk warning integral through integral calculation. The dynamic risk warning integral is used as a modulation parameter, and a nonlinear degradation mapping is performed with reference to an anchored initial control authority benchmark to generate candidate values ​​for degradation control authority benchmarks. The temporal mutation characteristics based on spatial constraint deviation features drive the generation of off-chain final state settlement data. Finally, the hash digests of the candidate values ​​for degradation control authority benchmarks and the off-chain final state settlement data are encapsulated and submitted to the blockchain. The blockchain verifies the integrity of the off-chain final settlement data and checks whether the candidate value for the degradation control authority benchmark exceeds the degradation boundary threshold. After dual-dimensional verification, a degradation execution credential is generated by signing the hash digest. Based on the successful verification of the degradation execution credential, the edge control terminal establishes the candidate value for the degradation control authority benchmark as the target degradation control authority benchmark and maps it to the physical operating parameters of the device. This invention overcomes the limitations of traditional methods, such as risk control decoupling, linear degradation distortion, and lack of on-chain rights confirmation, providing a secure and adaptive solution for risk warning in leasing platforms.

[0005] A risk warning method for a blockchain-based rental platform according to an embodiment of the present invention specifically includes: S1. Store the spatial compliance constraint data of the leased property and the initial control authority benchmark of the edge control terminal on the blockchain, and simultaneously anchor the initial control authority benchmark to the edge control terminal. Generate a downgrade boundary threshold based on the initial control authority benchmark and store it on the blockchain. S2. Obtain real-time spatiotemporal trajectory data of the leased object and input it into the offline monitoring chain. Perform feature mapping on the real-time spatiotemporal trajectory data based on the spatial compliance constraint data, and output the spatial constraint deviation features. S3. Based on spatial constraint deviation characteristics and real-time spatiotemporal trajectory data, nonlinear risk characteristics are extracted using an improved Flowformer model; a fluctuation evolution mechanism is introduced, and critical deviation measurement, temporal evolution diffusion, and nonlinear phase transition activation are executed sequentially to deduce the implicit fluctuations of risk; finally, the dynamic risk warning integral is output through integral calculation. S4. The downgrade strategy is activated by the dynamic risk warning score, and the dynamic risk warning score is used as a modulation parameter to perform downgrade mapping with reference to the anchored initial control authority benchmark, generating candidate values ​​for the downgrade control authority benchmark. S5. Driven by real-time spatiotemporal trajectory data, the on-chain intervention of critical phase transition is generated, and the off-chain final state settlement data is generated based on dynamic risk warning points. The candidate value of the downgrade control authority benchmark and the hash digest of the off-chain final state settlement data are submitted to the blockchain in parallel. S6. The blockchain verifies the integrity of the off-chain final settlement data based on the hash digest and verifies whether the candidate value of the downgrade control authority benchmark exceeds the downgrade boundary threshold. After the two-dimensional verification is confirmed, the hash digest is signed to generate a downgrade execution certificate. S7. The edge control terminal responds to the trust verification of the off-chain final settlement data and the corresponding degradation execution certificate. Based on the endorsement of the degradation execution certificate, the candidate value of the degradation control authority benchmark is established as the target degradation control authority benchmark and mapped to the device physical operating parameters.

[0006] Optionally, S1 specifically includes: S11. Collect the spatial geofence coordinates of the leased object, the historical spatiotemporal trajectory of the leased object, and the upper limit of the initial device operating parameters of the edge control terminal. Package the spatial geofence coordinates, the historical spatiotemporal trajectory of the leased object, and the upper limit of the initial device operating parameters into an initial block and upload it to the blockchain for evidence storage. Write the upper limit of the initial device operating parameters into the local security register of the edge control terminal to complete the anchoring. S12. Extract the upper limit of the initial device operating parameters anchored in the local security register, extract the historical spatiotemporal trajectory of the leased item stored in the blockchain, statistically analyze the boundary crossing distribution characteristics of the historical spatiotemporal trajectory of the leased item relative to the spatial geofence coordinates, adaptively calculate the benchmark tolerance deviation ratio based on the boundary crossing distribution characteristics, and output the degradation boundary threshold by combining the product operation of the benchmark tolerance deviation ratio and the upper limit of the initial device operating parameters. S13. Encapsulate the degradation boundary threshold into a boundary constraint block and upload it to the blockchain for evidence storage, and establish a hash mapping index on the blockchain between the initial upper limit of device operating parameters and the degradation boundary threshold.

[0007] Optionally, S2 specifically includes: S21. Obtain the real-time spatiotemporal trajectory data of the leased object and input it into the off-chain monitoring. Extract the spatial geofence coordinates stored in the blockchain and align the real-time spatiotemporal trajectory data with the spatial geofence coordinates to output the aligned trajectory sequence. S22. Calculate the shortest distance from each trajectory point in the alignment trajectory sequence to the coordinate boundary of the spatial geographic fence, statistically analyze the evolutionary distribution characteristics of the shortest distance, and adaptively extract the dynamic deviation metric benchmark based on the evolutionary distribution characteristics. S23. Perform nonlinear mapping on the shortest distance of each trajectory point based on the dynamic deviation metric benchmark, and output the spatial constraint deviation characteristics.

[0008] Optionally, the improved Flowformer model includes a spatiotemporal flow feature embedding layer, a nonlinear manifold mapping layer, and a risk integral calculation layer: The spatiotemporal flow feature embedding layer is used to perform spatiotemporal alignment and feature fusion of spatial constraint deviation features and real-time spatiotemporal trajectory data, and inject position awareness and local temporal evolution information to output a spatiotemporal flow embedding tensor. The nonlinear manifold mapping layer is used to map the spatiotemporal flow embedding tensor to a high-dimensional latent space to generate a basic latent space feature tensor, and simultaneously generate a residual tensor of the same dimension; apply temporal causal constraints to the basic latent space feature tensor to perform global feature aggregation, and generate an enhanced latent space feature tensor; fuse the enhanced latent space feature tensor with the residual tensor of the same dimension to output a nonlinear risk feature tensor. The implied volatility extrapolation layer is used to introduce a volatility evolution mechanism, decouple and extract the risk evolution state and risk critical boundary from the nonlinear risk feature tensor; calculate the feature dynamic divergence based on the deviation of the risk evolution state from the risk critical boundary; drive the time-series diffusion evolution with the feature dynamic divergence to generate time-series evolution features; activate and fuse the time-series evolution features through a dual-segment heterogeneous phase transition to output the risk implied volatility tensor; The risk integral solution layer is used to perform adaptive aggregation on the risk implicit volatility tensor along the time dimension to output a risk evolution feature vector, and to perform dimensionality reduction mapping and interval normalization on the risk evolution feature vector to output a dynamic risk warning integral.

[0009] Optionally, the fluctuation evolution mechanism specifically includes: The nonlinear risk feature tensor is extracted by exponential decay memory field sliding convolution along the time dimension, and state-dependent gated selective aggregation is applied to the convolution extraction result. Then, hyperbolic tangent activation and scaling affine mapping are performed on the aggregation result to generate a dynamic state sequence. The nonlinear risk feature tensor is sequentially subjected to feature dimension extreme value comparison extraction, hard-limited truncation amplification and additive offset and scaling mapping to generate the risk critical boundary; Calculate the element-by-element comparison and division of the dynamic state sequence relative to the risk critical boundary on the corresponding feature channel to obtain the feature space approximation ratio; perform a logarithmic transformation on the approximation ratio to generate the logarithmic elasticity deviation. Based on the logarithmic elasticity deviation, backward difference is performed to extract the momentum deviation rate of change in the time dimension. The statistical mean of the momentum deviation rate of change in the time dimension is extracted and used to perform centering detrending. Then, quadratic moment accumulation with decay memory is performed on the detrending result to calculate the characteristic dynamic divergence. The adaptive early warning time window span is calculated based on the dynamic divergence of the features. The dynamic divergence of the features is used as the evolutionary variance parameter. The inverse step size operator is constructed in combination with the adaptive early warning time window span to generate the time deviation step size. Using the evolution variance parameter as the fluctuation scale and the time aberration step size as the walk interval, the long-term equilibrium state and recovery decay coefficient are calculated based on the statistical mean and detrending results. A stepwise random walk with mean recovery constraint and nonlinear scaling mapping are performed to generate time-series evolution features. The temporal evolution features are activated by a dual-segment heterogeneous phase transition of critical threshold gating and overcritical exponential surge, respectively generating subcritical gated response tensors and overcritical exponential surge response tensors, and the routing mask generated by threshold decision is extracted. Based on the routing mask, bitwise selective fusion is performed on the subcritical gated response tensor and the overcritical exponential surge response tensor to calculate the future risk uncertainty spillover value, and the risk implied volatility tensor is aggregated and output according to multi-scale time windows.

[0010] Optionally, the exponentially decaying memory field sliding convolution extraction specifically includes: Extract the feature channel dimension span of the nonlinear risk feature tensor and calculate the local variance of the nonlinear risk feature tensor in the time dimension. Based on the feature channel dimension span and the local variance in the time dimension, dynamically calculate the receptive field window length to construct the initial convolution kernel. Based on the time distance of each element in the initial convolution kernel relative to the current calculation time, an exponential decay function is applied to generate exponential decay constraint weights. The exponential decay constraint weights are then multiplied element-wise with the internal weights of the initial convolution kernel to generate a time decay convolution kernel. The time decaying convolution kernel is trunculated along the time dimension of the nonlinear risk feature tensor with a step size of 1. The truncated local time window features are multiplied element-wise with the time decaying convolution kernel and summed to obtain the local time decaying response value. The center time feature aligned with the current calculation time in the local time window features is extracted as the residual memory component. The local time decay response value and the residual memory component are added bitwise to generate the memory field convolution compensation result. The convolution extraction result is generated by traversing and sliding along the time dimension and splicing the memory field convolution compensation results corresponding to each time.

[0011] Optionally, S4 specifically includes: S41. Extract the dynamic risk warning integral and the upper limit of the initial equipment operating parameters anchored to the local security register of the edge control terminal, and extract the historical time series distribution characteristics of the nonlinear risk feature tensor to adaptively extract the dynamic security benchmark integral. S42. Calculate the relative deviation of the dynamic risk warning integral from the dynamic safety benchmark integral, calculate the dynamic scaling factor based on the relative deviation, multiply the dynamic risk warning integral and the dynamic scaling factor, and output the degradation attenuation coefficient. S43. Using the degradation attenuation coefficient as a modulation parameter, perform attenuation product operation on the upper limit of the initial equipment operating parameters, and output the intermediate degradation parameter value. S44. Calculate the attenuation difference between the initial upper limit of equipment operating parameters and the intermediate degradation parameter values, and calculate the joint penalty weight of the attenuation difference and the dynamic risk warning score; S45. Perform a weighted operation on the attenuation difference based on the joint penalty weight, output the penalty suppression amount, perform a difference operation on the intermediate degradation parameter value and the penalty suppression amount, and output the candidate value of the degradation control authority baseline.

[0012] Optionally, S5 specifically includes: S51. Extract spatial constraint deviation features, statistically analyze the temporal abrupt change distribution features of spatial constraint deviation features, calculate the phase transition triggering coefficient based on the temporal abrupt change distribution features, perform adaptive boundary judgment based on the temporal abrupt change distribution features, and output the phase transition signal. S52. Based on phase transition signal triggering, the phase transition triggering coefficient is jointly mapped with the dynamic risk warning integral to output off-chain final state settlement data, and a hash operation is performed on the off-chain final state settlement data to output a hash digest. S53. Extract the baseline candidate value for downgraded control permissions, encapsulate the baseline candidate value for downgraded control permissions and hash digest into a binary structure, and submit the binary structure to the blockchain for evidence storage.

[0013] Optionally, S6 specifically includes: S61. Extract and parse the binary structure on the blockchain to obtain the off-chain final settlement data, hash digest and downgrade control permission baseline candidate value. Perform hash operation on the off-chain final settlement data to output the reconstructed hash digest. Compare the reconstructed hash digest with the hash digest and output the integrity verification result. S62. Extract the degradation boundary threshold of the blockchain evidence storage, calculate the degradation attenuation degree of the baseline candidate value of degradation control authority relative to the upper limit of the initial device operating parameters, perform an out-of-bounds comparison between the degradation attenuation degree and the degradation boundary threshold, and output the out-of-bounds verification result. S63. Perform a logical AND operation based on the integrity check result and the out-of-bounds check result. If the logical AND operation passes, sign the hash digest and output the downgrade execution credential.

[0014] Optionally, S7 specifically includes: S71. Extract the downgrade execution credential and hash digest. Perform a decryption operation on the downgrade execution credential based on the blockchain public key and output the decrypted hash digest. Perform a consistency comparison between the decrypted hash digest and the hash digest and output the credential verification result. S72. Based on the successful verification of credentials, extract the candidate values ​​of the downgraded control permission benchmark cache in the local cache of the edge control terminal, and establish the candidate values ​​of the downgraded control permission benchmark as the target downgraded control permission benchmark. S73. Extract the initial device operating parameter upper limit anchored to the local security register of the edge control terminal, calculate the normalized degradation ratio of the target degradation control permission baseline candidate value relative to the initial device operating parameter upper limit, perform a product operation on the normalized degradation ratio and the initial device operating parameter upper limit, and output the device physical operating parameters.

[0015] The beneficial effects of this invention are: (1) This invention achieves deep decoupling and nonlinear evolution representation of the spatiotemporal deviation characteristics of leased assets and the implicit fluctuations of risk by constructing a risk inference process jointly driven by spatial compliance constraint feature mapping and an improved Flowformer model. The spatiotemporal alignment and feature fusion of spatial constraint deviation characteristics and real-time spatiotemporal trajectory data are used to extract the spatiotemporal flow embedding tensor. Global feature aggregation is performed based on high-dimensional latent space mapping and temporal causal constraints to generate an enhanced latent space feature tensor, which is then fused with a residual tensor of the same dimension to output a nonlinear risk feature tensor. A fluctuation evolution mechanism is introduced, sequentially performing critical deviation measurement, temporal evolution diffusion, and nonlinear phase transition activation on the nonlinear risk feature tensor. The dynamic divergence of the features drives a random walk with mean recovery constraints, generating a risk implicit fluctuation tensor and integrating it to output a dynamic risk warning integral. This process transforms discrete trajectory offsets into a continuous latent space dynamic evolution field, accurately capturing the nonlinear phase transition and dynamic divergence characteristics in the risk accumulation process, and providing a risk quantification indicator with high sensitivity and advanced warning capabilities.

[0016] (2) This invention establishes an adaptive degradation mapping and cross-chain trust constraint system for risk points to physical operating parameters of equipment by adopting dynamic risk warning integral modulation and on-chain two-dimensional verification and confirmation process. Based on the dynamic risk warning integral, the degradation attenuation coefficient and joint penalty weight are calculated, and the attenuation product and weighted difference operation are performed on the upper limit of the initial equipment operating parameters to generate the baseline candidate value of degradation control authority; based on the temporal mutation distribution characteristics of spatial constraint deviation features, the on-chain intervention critical phase transition is triggered to generate off-chain final state settlement data. The blockchain parses the binary structure to verify the integrity of the off-chain final state settlement data, calculates the degradation attenuation degree of the baseline candidate value of degradation control authority and performs an out-of-bounds comparison with the degradation boundary threshold, and generates a degradation execution certificate by signing the hash digest after the two-dimensional verification is passed; after the edge control end decrypts and compares the certificate, it calculates the product of the normalized degradation ratio and the upper limit of the initial equipment operating parameters to output the physical operating parameters of the equipment. This system achieves adaptive and precise generation of degradation control strategies and tamper-proof secure execution through dynamic modulation of risk scores, on-chain integrity verification, and hard constraints on out-of-bounds thresholds. It ensures that the output physical operating parameters of the devices are strictly controlled by the blockchain and accurately match the real-time risk evolution status. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of a blockchain-based risk warning method for rental platforms proposed in this invention; Figure 2 This is a flowchart illustrating the working principle of the improved Flowformer model for a blockchain-based risk warning method for rental platforms proposed in this invention. Detailed Implementation

[0018] The invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figure 1 and Figure 2 A risk warning method for a blockchain-based rental platform, specifically including: S1. Store the spatial compliance constraint data of the leased property and the initial control authority benchmark of the edge control terminal on the blockchain, and simultaneously anchor the initial control authority benchmark to the edge control terminal. Generate a downgrade boundary threshold based on the initial control authority benchmark and store it on the blockchain. S2. Obtain real-time spatiotemporal trajectory data of the leased object and input it into the offline monitoring chain. Perform feature mapping on the real-time spatiotemporal trajectory data based on the spatial compliance constraint data, and output the spatial constraint deviation features. S3. Based on spatial constraint deviation characteristics and real-time spatiotemporal trajectory data, nonlinear risk characteristics are extracted using an improved Flowformer model; a fluctuation evolution mechanism is introduced, and critical deviation measurement, temporal evolution diffusion, and nonlinear phase transition activation are executed sequentially to deduce the implicit fluctuations of risk; finally, the dynamic risk warning integral is output through integral calculation. S4. The downgrade strategy is activated by the dynamic risk warning score, and the dynamic risk warning score is used as a modulation parameter to perform downgrade mapping with reference to the anchored initial control authority benchmark, generating candidate values ​​for the downgrade control authority benchmark. S5. Driven by real-time spatiotemporal trajectory data, the on-chain intervention of critical phase transition is generated, and the off-chain final state settlement data is generated based on dynamic risk warning points. The candidate value of the downgrade control authority benchmark and the hash digest of the off-chain final state settlement data are submitted to the blockchain in parallel. S6. The blockchain verifies the integrity of the off-chain final settlement data based on the hash digest and verifies whether the candidate value of the downgrade control authority benchmark exceeds the downgrade boundary threshold. After the two-dimensional verification is confirmed, the hash digest is signed to generate a downgrade execution certificate. S7. The edge control terminal responds to the trust verification of the off-chain final settlement data and the corresponding degradation execution certificate. Based on the endorsement of the degradation execution certificate, the candidate value of the degradation control authority benchmark is established as the target degradation control authority benchmark and mapped to the device physical operating parameters.

[0020] In this embodiment, S1 specifically includes: S11. Collect the spatial geofence coordinates of the leased object, the historical spatiotemporal trajectory of the leased object, and the upper limit of the initial device operating parameters of the edge control terminal. Package the spatial geofence coordinates, the historical spatiotemporal trajectory of the leased object, and the upper limit of the initial device operating parameters into an initial block and upload it to the blockchain for evidence storage. Write the upper limit of the initial device operating parameters into the local security register of the edge control terminal to complete the anchoring. The upper limit of the initial device operating parameters is set to 3000 revolutions per minute.

[0021] S12. Extract the upper limit of the initial device operating parameters anchored by the local security register, extract the historical spatiotemporal trajectory of the leased item stored in the blockchain, and construct the boundary crossing distribution characteristics by statistically analyzing the boundary crossing frequency and distance of the historical spatiotemporal trajectory of the leased item relative to the spatial geofence coordinates. Based on the boundary crossing distribution characteristics, calculate the product of the boundary crossing frequency ratio and the average boundary crossing distance as the benchmark tolerance deviation ratio. Combine the product of the benchmark tolerance deviation ratio and the upper limit of the initial device operating parameters to output the degradation boundary threshold. The benchmark tolerance deviation ratio is set to 0.15. Multiply 0.15 by 3000 rpm to output the degradation boundary threshold of 450 rpm.

[0022] S13. Encapsulate the degradation boundary threshold into a boundary constraint block and upload it to the blockchain for evidence storage. Establish a hash mapping index on the blockchain between the initial upper limit of the device operating parameters and the degradation boundary threshold, so that subsequent verification steps can directly extract the degradation boundary threshold by calling the hash mapping index to perform out-of-bounds comparison.

[0023] In this embodiment, S2 specifically includes: S21. Obtain the real-time spatiotemporal trajectory data of the leased object and input it into the off-chain monitoring. Extract the spatial geofence coordinates stored in the blockchain, align the real-time spatiotemporal trajectory data with the spatial geofence coordinates, and output the aligned trajectory sequence.

[0024] S22. Calculate the shortest distance from each trajectory point in the alignment trajectory sequence to the spatial geographic fence coordinate boundary, statistically analyze the evolutionary distribution characteristics of the shortest distance, calculate the mean and standard deviation of the shortest distance, and subtract twice the standard deviation from the mean to obtain the dynamic deviation measurement benchmark, where the standard deviation of the shortest distance is set to 5.0 meters.

[0025] S23. Perform nonlinear mapping on the shortest distance of each trajectory point based on the dynamic deviation metric benchmark, calculate the difference between the shortest distance of each trajectory point and the dynamic deviation metric benchmark, input the difference into the hyperbolic tangent activation function tanh to calculate the mapping value, and multiply the mapping value by the modulus of the shortest distance to output the spatial constraint deviation feature.

[0026] In this embodiment, the improved Flowformer model includes a spatiotemporal flow feature embedding layer, a nonlinear manifold mapping layer, and a risk integral calculation layer: The spatiotemporal flow feature embedding layer is used to align spatial constraint deviation features with real-time spatiotemporal trajectory data with timestamps and concatenate channels. The concatenation result is multiplied by the embedding weight matrix and a position encoding vector is added. The position encoding vector is set to have a dimension of 128. After processing by the modified linear unit ReLU activation function, the spatiotemporal flow embedding tensor is output. A nonlinear manifold mapping layer is used to multiply the spatiotemporal flow embedding tensor by an up-dimensional transformation matrix to map it to a high-dimensional latent space to generate a basic latent space feature tensor. The high-dimensional latent space is set to 256 dimensions, and the spatiotemporal flow embedding tensor is simultaneously copied and multiplied by the residual mapping matrix to generate a residual tensor of the same dimension. A temporal causal mask is applied to the basic latent space feature tensor to mask future time information, and global self-attention aggregation is performed to generate an enhanced latent space feature tensor. The enhanced latent space feature tensor is then fused with the residual tensor of the same dimension by element-wise addition to output a nonlinear risk feature tensor. The implied volatility extrapolation layer is used to calculate the mean of the nonlinear risk feature tensor along its feature dimension as the risk evolution state. The maximum value of the nonlinear risk feature tensor along its feature dimension is calculated as the risk critical boundary. The risk evolution state is calculated and divided by the risk critical boundary to obtain the approximation ratio. The natural logarithm of the approximation ratio is taken, and its backward difference mean in the time dimension is calculated. A 0.01-second offset is added to the backward difference mean to output the feature dynamic divergence. The feature dynamic divergence is used as the evolution variance parameter, multiplied by a standard normal distribution random number, and then a nonlinear... The time dimension mean of the risk feature tensor is used to generate temporal evolution features through iterative diffusion. The temporal evolution features are then input into the hyperbolic tangent activation function tanh to generate a subcritical gated response tensor. The temporal evolution features are then input into the exponential activation function to generate a critical exponential surge response tensor. The critical threshold is set to 1.0. The portion of the temporal evolution features greater than 1.0 is multiplied by the critical exponential surge response tensor, and the portion less than or equal to 1.0 is multiplied by the subcritical gated response tensor. These components are then concatenated and fused to output the risk-implied volatility tensor. The risk integral solution layer is used to calculate the weighted average of the risk implied volatility tensor along the time dimension. The time decay weight is set to 0.9, and the risk evolution feature vector is output. The risk evolution feature vector is multiplied by the dimensionality reduction matrix and mapped to a 1-dimensional scalar. The 1-dimensional scalar is input into the Sigmoid activation function and compressed to the interval between 0 and 1. It is then multiplied by 1000 to output the dynamic risk warning integral.

[0027] In its implementation, this invention deeply inherits and modifies the traditional Flowformer model, inheriting its efficient parallel computing capabilities in long-sequence temporal dependency modeling through its streaming attention mechanism. At the same time, it structurally reconstructs the model to address the pain points of nonlinear mutations and implicit fluctuations in the evolution of leased asset risk.

[0028] Traditional methods, relying solely on static rules or shallow temporal networks, fail to capture the dynamic divergence and phase transition characteristics of risk accumulation, easily leading to delayed warnings and missed detections. This invention, however, firstly integrates spatial constraint deviation features with real-time spatiotemporal trajectory depth through a spatiotemporal flow feature embedding layer and a nonlinear manifold mapping layer. It innovatively preserves residual tensors simultaneously in high-dimensional latent space mapping, combining temporal causal constraints and residual fusion to output a nonlinear risk feature tensor, effectively solving the problem of losing high-dimensional topology and fine-grained spatiotemporal semantics in traditional feature extraction. The most crucial innovation lies in the pioneering implicit volatility inference layer, breaking through the traditional assumption of stationary temporal series and introducing a volatility evolution mechanism. This decouples the risk evolution state and critical boundary from nonlinear risk features and calculates the feature dynamic divergence. The dynamic divergence drives temporal diffusion evolution to generate temporal evolution features, which are then precisely captured through dual-segment heterogeneous phase transition activation and fusion, outputting a risk implicit volatility tensor. Finally, the risk integral calculation layer performs adaptive aggregation and normalization to output a dynamic risk warning integral. This process completely solves the technical problem that traditional early warning methods cannot quantify the implicit fluctuations and nonlinear phase transitions of risks. It achieves a leap from the perception of surface trajectory deviations to the inference of deep risk fluctuations, ensuring that the output dynamic risk early warning integral has the ability to perceive and accurately quantify gradual violations and sudden boundary crossings in complex operating environments. It provides a highly robust and highly sensitive trust-driven anchor point for subsequent adaptive degradation control.

[0029] In this embodiment, the fluctuation evolution mechanism specifically includes: The nonlinear risk feature tensor is subjected to sliding truncation with a step size of 1 step and a window length of 5 steps along the time dimension. Each element in the window is multiplied by an exponentially decaying weight with a base of 0.8 and summed to obtain the convolution extraction result. The convolution extraction result is input into the Sigmoid activation function to generate gating coefficients. The gating coefficients are multiplied by the convolution extraction result and summed along the time dimension to complete selective aggregation. The aggregation result is input into the hyperbolic tangent activation function tanh and multiplied by a scaling factor of 1.5 plus an offset of 0.1 to generate a dynamic state sequence. Extract the maximum value along the feature dimension of the nonlinear risk feature tensor, retain the part with the maximum value greater than 1.0 and set the part less than or equal to 1.0 to 1.0 to perform hard-limited truncation and amplification, add the offset constant 0.05 to the truncation and amplification result and multiply by the scaling factor 2.0 to generate the risk critical boundary; The dynamic state sequence is divided by the risk critical boundary to obtain the feature space approximation ratio. The natural logarithm of the approximation ratio is taken to generate the log elasticity deviation. The backward difference of the log elasticity deviation at adjacent time steps is calculated to extract the momentum deviation rate of change. The statistical mean of the momentum deviation rate of change in the time dimension is calculated. The log elasticity deviation is subtracted from the product of the statistical mean and the time step to perform centering and detrending. The detrending result is multiplied by the decay coefficient 0.9, squared step by step, and accumulated to perform quadratic moment accumulation with decay memory to calculate the feature dynamic divergence. The adaptive warning time window span is calculated by dividing the value 1.0 by the feature dynamic divergence and adding 0.01 to the sum and rounding down without conditions. The feature dynamic divergence is used as the evolution variance parameter. The adaptive warning time window span is divided by 100 to calculate the generation time deviation of the inverse step size operator. The evolutionary variance parameter multiplied by the standard normal distribution random number is used as the fluctuation scale, the time aberration step size is used as the walk interval, the statistical mean is used as the long-term equilibrium state, and the detrending result is multiplied by 0.05 as the recovery decay coefficient. The walk value at the current time is calculated as the walk value at the previous time plus the recovery decay coefficient multiplied by the difference between the long-term equilibrium state and the walk value at the previous time plus the fluctuation scale multiplied by the time aberration step size. A stepwise random walk with mean recovery constraint is performed, and the walk result is input into the hyperbolic tangent activation function tanh to perform nonlinear scaling mapping to generate time series evolution features. The portion of the temporal evolution features less than or equal to the critical threshold of 1.0 is input into the hyperbolic tangent activation function tanh to generate a subcritical gated response tensor. The portion of the temporal evolution features greater than the critical threshold of 1.0 is subtracted by 1.0 and then input into the exponential activation function to generate a critical exponential surge response tensor. The positions of the temporal evolution features greater than 1.0 are set to 1 and the positions less than or equal to 1.0 are set to 0 to extract the routing mask. The routing mask is multiplied by the critical exponential surge response tensor, plus 1, minus the routing mask multiplied by the subcritical gated response tensor, and bitwise selective fusion is performed to calculate the future risk uncertainty spillover value. The future risk uncertainty spillover value is averaged according to the window length in 3 steps and 5 steps respectively, and then concatenated to output the risk implied volatility tensor.

[0030] In this embodiment, exponential decay memory field sliding convolution extraction specifically includes: Extract the feature channel dimension span of the nonlinear risk feature tensor and calculate the local variance of the time dimension. Multiply the feature channel dimension span by 0.2, add the local variance by 0.8, round up and add 3. Dynamically calculate the receptive field window length to construct an initial convolutional kernel with internal weights initialized to 0.5. Calculate the time distance between each element in the initial convolution kernel and the current calculation time. Multiply the time distance by the decay coefficient 0.3 and input it into the exponential decay function to generate exponential decay constraint weights. Multiply the exponential decay constraint weights with the internal weights of the initial convolution kernel element by element to generate the time decay convolution kernel. The time decaying convolution kernel is trunculated along the time dimension of the nonlinear risk feature tensor with a step size of 1. The truncated local time window features are multiplied element by element by the time decaying convolution kernel and summed to obtain the local time decaying response value. The center time feature aligned with the current calculation time in the local time window features is extracted as the residual memory component. The local time decay response value and the residual memory component are added bit by bit to generate the memory field convolution compensation result. The memory field convolution compensation results at each time point are spliced ​​together to generate the convolution extraction result by traversing and sliding along the time dimension.

[0031] In this embodiment, S4 specifically includes: S41. Extract the dynamic risk warning integral and the initial device operating parameter upper limit anchored to the local security register of the edge control terminal. The initial device operating parameter upper limit is set to 3000 rpm. Extract the historical time series distribution characteristics of the nonlinear risk feature tensor. Calculate the sum of the mean and standard deviation of the historical time series distribution characteristics. Use the adaptive extraction result of the mean of the historical time series distribution characteristics plus twice the standard deviation of the historical time series distribution characteristics as the dynamic safety benchmark integral. Set the initial benchmark value of the dynamic safety benchmark integral to 500 points. S42. Calculate the difference between the dynamic risk warning integral and the dynamic safety benchmark integral. Divide the difference by the dynamic safety benchmark integral to output the relative deviation. Input the relative deviation into the hyperbolic tangent activation function tanh and add the bias constant 1.0 to calculate the dynamic scaling factor. Perform a product operation between the dynamic risk warning integral and the dynamic scaling factor to output the degradation attenuation coefficient. S43. Using the degradation attenuation coefficient as a modulation parameter, perform attenuation product operation on the upper limit of the initial equipment operating parameters, and multiply the upper limit of the initial equipment operating parameters by the degradation attenuation coefficient to output the intermediate degradation parameter value. S44. Calculate the attenuation difference between the upper limit of the initial equipment operating parameters and the intermediate degradation parameter values. Calculate the product of the attenuation difference and the dynamic risk warning integral. Divide the product by the sum of the dynamic safety benchmark integral and the offset factor 0.01, and output the joint penalty weight. S45. Perform a weighted operation on the attenuation difference based on the joint penalty weight, multiply the joint penalty weight by the attenuation difference to output the penalty suppression amount, subtract the penalty suppression amount from the intermediate downgrade parameter value to perform an interpolation operation, and output the candidate value of the downgrade control authority baseline.

[0032] In the specific implementation process, this invention has deeply inherited and modified the traditional equipment degradation control methods. It has inherited the basic framework of executing safety constraints based on parameter upper limits, and at the same time, it has thoroughly reconstructed the pain points of traditional degradation strategies that rely on static threshold hard cut-off, resulting in equipment state step jitter, frequent erroneous intervention, and inability to adapt to the nonlinear evolution of risks.

[0033] Traditional methods typically employ a fixed ratio or linear mapping for coarse-grained permission reduction, neglecting the dynamic coupling between risk fluctuations and equipment physical load. This invention, however, firstly extracts the historical temporal distribution characteristics of a nonlinear risk feature tensor to adaptively extract a dynamic safety benchmark integral, establishing a benchmark anchor point for degradation mapping that evolves dynamically with the operating environment. This solves the problem that traditional static benchmarks cannot adapt to complex working conditions. The more fundamental innovation lies in the construction of a two-order nonlinear degradation mapping mechanism of "relative deviation modulation and joint penalty suppression." On one hand, it generates a dynamic scaling factor by calculating the relative deviation of the dynamic risk warning integral from the dynamic safety benchmark integral. This factor is then used to output a degradation attenuation coefficient, which is applied to the initial upper limit of equipment operating parameters to generate intermediate degradation parameter values ​​through attenuation multiplication. This achieves adaptive and smooth modulation of the degradation depth by risk fluctuations. On the other hand, it breaks through the limitations of conventional single-layer attenuation by extracting the attenuation difference between the initial upper limit and the intermediate degradation parameters. The joint penalty weight of the attenuation difference and the dynamic risk warning integral is calculated, and a weighted penalty suppression amount is output based on this weighted attenuation difference. Finally, the intermediate degradation parameters and the penalty suppression amount are used to perform a difference operation to output candidate values ​​for the degradation control permission benchmark.

[0034] This process effectively solves the technical problem that traditional degradation strategies are prone to control jitter and excessive degradation blocking operations at the risk critical boundary. Through the dynamic offsetting mechanism of penalty inhibition, the degradation curve achieves a smooth transition in the low-risk area and a sharp contraction in the high-risk area. This ensures that the generated degradation control authority benchmark candidate value can not only accurately follow the implicit fluctuations of risk to carry out fine-grained authority deprivation, but also avoid the step collapse of the physical operating parameters of the equipment. This greatly improves the operational continuity and control robustness of the leased property under risk intervention.

[0035] In this embodiment, S5 specifically includes: S51. Extract the spatial constraint deviation feature, calculate the absolute value of the backward difference between adjacent time steps of the spatial constraint deviation feature as the temporal mutation distribution feature, calculate the mean and standard deviation of the temporal mutation distribution feature, add 3 times the standard deviation of the temporal mutation distribution feature to the mean of the temporal mutation distribution feature as the mutation judgment threshold, divide the mean of the temporal mutation distribution feature by the reference frequency of 0.5 Hz to calculate the phase transition triggering coefficient, output the value 1 when the temporal mutation distribution feature of the spatial constraint deviation feature is greater than the mutation judgment threshold, otherwise output the value 0, and generate a phase transition signal; S52. When the phase transition signal is 1, the phase transition trigger coefficient is multiplied by the dynamic risk warning integral, and then the product of the dynamic risk warning integral and the bias coefficient 0.2 is added to perform joint mapping, output the off-chain final state settlement data, and perform SHA256 hash operation on the off-chain final state settlement data to output the hash digest. S53. Extract the baseline candidate value for downgraded control permissions, concatenate the baseline candidate value for downgraded control permissions with the hash digest and encapsulate it into a binary structure, and submit the binary structure to the blockchain for evidence storage.

[0036] In this embodiment, S6 specifically includes: S61. Extract the binary structure on the blockchain and parse it to obtain the off-chain final settlement data, hash digest and downgrade control permission baseline candidate value. Perform SHA256 hash operation on the off-chain final settlement data to output the reconstructed hash digest. Perform bit-by-bit comparison between the reconstructed hash digest and the hash digest. When they are completely consistent, output the value 1; otherwise, output the value 0 as the integrity verification result. S62. Extract the degradation boundary threshold of the blockchain evidence storage, calculate the difference between the upper limit of the initial device operating parameters and the candidate value of the degradation control authority benchmark, divide the difference by the upper limit of the initial device operating parameters to calculate the degradation attenuation degree, compare the degradation attenuation degree with the degradation boundary threshold, output the value 1 when the degradation attenuation degree is less than or equal to the degradation boundary threshold, otherwise output the value 0 as the out-of-bounds verification result. S63. Perform a logical AND operation on the integrity check result and the out-of-bounds check result. If the result of the logical AND operation is 1, it passes. Perform an elliptic curve digital signature algorithm to sign the hash digest and output the downgrade execution credential.

[0037] In its implementation, this invention deeply inherits and transforms the traditional blockchain data verification and device control authorization mechanism. It inherits the basic trust framework of blockchain hash anti-tampering verification and digital signature confirmation, and fundamentally reconstructs the pain points of traditional on-chain verification that only focuses on data consistency while ignoring the security of control logic, and that edge control commands are easily subject to malicious replay or unauthorized tampering.

[0038] Traditional methods typically only perform a single hash comparison verification on on-chain data, which cannot prevent the downgrade parameters generated off-chain from being maliciously tampered with into unauthorized instructions that exceed physical security boundaries. This invention, however, firstly analyzes the binary structure on the blockchain, performs a hash operation to reconstruct the off-chain final state settlement data, and compares it with the original hash digest to output the integrity verification result, ensuring the absolute consistency of on-chain and off-chain state data. The more core innovation lies in the pioneering dual-dimensional joint constraint mechanism of "integrity verification and out-of-bounds logic verification," breaking through the blind spot of traditional one-dimensional verification. By extracting the downgrade boundary threshold of blockchain evidence, calculating the downgrade attenuation degree of the downgrade control authority benchmark candidate value relative to the upper limit of the initial device operating parameters, and performing an out-of-bounds comparison with the downgrade boundary threshold to output the out-of-bounds verification result, and finally performing a logical AND operation based on the integrity verification result and the out-of-bounds verification result, the hash digest is signed and the downgrade execution certificate is output only when both dimensions pass simultaneously.

[0039] This process completely solves the technical problem of device damage or asset loss caused by malicious unauthorized downgrade commands executed after the edge is compromised under the traditional mechanism. By transforming the security boundary threshold of risk control into an immutable hard logical constraint on the blockchain, it achieves a leap from "post-event data storage and traceability" to "in-event control logic strong verification". It ensures that the issued downgrade execution certificate is not only tamper-proof at the data layer, but also prevents overstepping at the physical security boundary layer, providing a double endorsement with extremely high trust strength and security bottom line for the final execution of the edge control terminal.

[0040] In this embodiment, S7 specifically includes: S71. Extract the downgrade execution credential and hash digest. Based on the blockchain public key, perform a decryption operation on the downgrade execution credential and output the decrypted hash digest. Compare the decrypted hash digest with the hash digest bit by bit. If they are completely consistent, output the value 1; otherwise, output the value 0 as the credential verification result. S72. When the voucher verification result is 1, it passes. Extract the candidate value of the downgraded control permission benchmark from the local cache of the edge control terminal, and establish the candidate value of the downgraded control permission benchmark as the target downgraded control permission benchmark. S73. Extract the initial device operating parameter upper limit anchored to the local security register of the edge control terminal, divide the target downgrade control permission benchmark by the initial device operating parameter upper limit to calculate the normalized downgrade ratio, multiply the normalized downgrade ratio by the initial device operating parameter upper limit, and output the device physical operating parameters.

[0041] Example 1: To verify the feasibility of this invention in the supervision of leased assets, the method of this invention was applied to the intelligent risk control and asset supervision system of a large construction machinery leasing company (hereinafter referred to as "Company L"). In traditional leased asset monitoring systems, rule-based judgment based on a single geofence or simple trajectory deviation alarms are typically used. These methods not only struggle to accurately capture the progressive trend of non-compliant operations in complex construction environments, but also fail to quantify the non-linear implicit fluctuations in the risk accumulation process, easily leading to delays in intervention to downgrade high-risk operations or misjudgments of normal operations. To solve the above problems, Company L decided to adopt a blockchain-based risk warning method for leasing platforms proposed in this invention.

[0042] During implementation, Company L first utilizes multi-source fusion positioning sensor terminals deployed on the leased assets to acquire real-time spatiotemporal trajectory data streams. It also collects the spatial geofence coordinates of the leased assets and the initial upper limit of the device operating parameters at the edge control terminal. After preprocessing operations such as spatiotemporal registration and coordinate alignment, it constructs the input data base for off-chain monitoring. Simultaneously, Company L packages the spatial geofence coordinates, the historical spatiotemporal trajectory of the leased assets, and the initial upper limit of the device operating parameters into an initial block and uploads it to the blockchain for evidence storage. Furthermore, it writes the initial upper limit of the device operating parameters into the local security register of the edge control terminal to complete anchoring, serving as the benchmark for subsequent risk calculation and degradation control.

[0043] Company L extracts the spatial geofence coordinates stored in the blockchain, aligns real-time spatiotemporal trajectory data with these coordinates, calculates the shortest distance, and adaptively extracts a dynamic deviation metric based on evolutionary distribution characteristics, outputting spatial constraint deviation features. Next, the spatial constraint deviation features and real-time spatiotemporal trajectory data are input into an improved Flowformer model, sequentially passing through a spatiotemporal flow feature embedding layer and a nonlinear manifold mapping layer to extract a nonlinear risk feature tensor. Subsequently, a fluctuation evolution mechanism is introduced, performing exponential decay memory field sliding convolution extraction and state-dependent gated selective aggregation on the nonlinear risk feature tensor to generate a dynamic state sequence. Feature dimension extreme value comparison is performed to extract the risk critical boundary, and the logarithmic elasticity deviation and feature dynamic divergence of the dynamic state sequence relative to the risk critical boundary are calculated. The feature dynamic divergence drives a random walk with mean-recovery constraints to generate temporal evolution features. After dual-segment heterogeneous phase transition activation and fusion, the risk-implied fluctuation tensor is output, and finally, the dynamic risk warning integral is calculated and output.

[0044] In the core degradation and rights confirmation execution phase, this invention uses dynamic risk warning integrals as modulation parameters. It performs attenuation product and joint penalty weighted operations with reference to the anchored initial device operating parameter upper limit to generate candidate values ​​for degradation control authority. Based on the temporal abrupt change distribution characteristics of spatial constraint deviation features, it outputs phase transition signals to drive the generation of off-chain final state settlement data. The hash digests of the candidate values ​​for degradation control authority and the off-chain final state settlement data are encapsulated into a binary structure and submitted to the blockchain. The blockchain parses the binary structure to verify the integrity of the off-chain final state settlement data and calculates the degradation attenuation degree of the candidate values ​​for degradation control authority and performs an out-of-bounds comparison with the degradation boundary threshold. After dual-dimensional verification, the hash digest is signed to generate a degradation execution credential. Finally, the edge control terminal decrypts and compares the degradation execution credential, establishing the candidate values ​​for degradation control authority as the target degradation control authority benchmark. It calculates the normalized degradation ratio mapping and outputs the device physical operating parameters, achieving a closed-loop transition from spatiotemporal deviation perception to on-chain rights confirmation and then to edge security execution.

[0045] During implementation, Company L's technical team discovered that, compared to traditional fence alerts and static degradation methods, the method of this invention significantly improves the accuracy of risk warnings for leased assets and the adaptability of degradation intervention. Traditional methods cannot quantify the implicit fluctuations in risk evolution, and degradation controls lacking on-chain verification are easily tampered with or excessively restrict equipment performance. In contrast, the method of this invention, through nonlinear risk feature extraction, fluctuation evolution deduction, and blockchain dual-dimensional verification, effectively achieves accurate quantification of leased asset violation risks and secure adaptive degradation of equipment permissions.

[0046] To further verify the actual performance of the method of the present invention, Company L conducted a detailed comparative test between the method of the present invention and the traditional method. The specific performance data is shown in Table 1: Table 1. Performance Comparison of Risk Warning Methods on Company L Leasing Platform

[0047] As shown in Table 1, the performance of the risk warning system of the leasing platform was comprehensively improved after applying the method of this invention. The accuracy of risk warning increased from 79.5% of the traditional method to 96.2%, and the false trigger rate of degradation strategy decreased from 15.2% to 1.5%, significantly improving the accuracy of risk status perception and intervention and avoiding unnecessary interruptions to normal operations. The implicit risk underreporting rate decreased from 18.6% to 2.1%, effectively avoiding the risk of asset loss caused by progressive violations. The rate of degradation control command tampering decreased to 0, and the rate of unauthorized degradation execution decreased sharply from 8.3% to 0.2%, thoroughly ensuring the anti-tampering and hard constraint security of edge control. The single on-chain confirmation response time was shortened from 45 seconds to 6 seconds, and the equipment physical parameter mapping deviation decreased from 12.0% to 0.8%, significantly enhancing the system's timeliness and mapping accuracy. In addition, the success rate of recovering defaulted assets increased from 65.0% to 94.5%, and the risk control and maintenance costs decreased from RMB 1.2 million / year to RMB 550,000 / year, significantly reducing operating costs.

[0048] Through the method of this invention, Company L has successfully achieved precise quantitative early warning of the risk of non-compliance of leased assets and secure adaptive downgrading of equipment control permissions. This effectively curbs the unauthorized use and malicious escape of assets, ensures the safety of leased assets, significantly improves the intelligence and credibility of risk control in leasing business, significantly reduces the monitoring burden of risk control personnel, enhances the robustness and anti-attack capability of the regulatory system, and provides strong technical support for the construction of a trustworthy leasing Internet of Things.

[0049] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A risk warning method for a blockchain-based rental platform, characterized in that, Includes the following steps: S1. Store the spatial compliance constraint data of the leased property and the initial control authority benchmark of the edge control terminal on the blockchain, and simultaneously anchor the initial control authority benchmark to the edge control terminal. Generate a downgrade boundary threshold based on the initial control authority benchmark and store it on the blockchain. S2. Obtain real-time spatiotemporal trajectory data of the leased object and input it into the offline monitoring chain. Perform feature mapping on the real-time spatiotemporal trajectory data based on spatial compliance constraint data, and output spatial constraint deviation features. S3. Based on spatial constraint deviation characteristics and real-time spatiotemporal trajectory data, nonlinear risk characteristics are extracted using an improved Flowformer model; Introduction The fluctuation evolution mechanism sequentially executes critical deviation measurement, temporal evolution diffusion, and nonlinear phase transition activation to deduce the implicit fluctuations of risk; finally, the integral solution outputs the dynamic risk warning integral. S4. The downgrade strategy is activated by the dynamic risk warning score, and the dynamic risk warning score is used as a modulation parameter to perform downgrade mapping with reference to the anchored initial control authority benchmark, generating candidate values ​​for the downgrade control authority benchmark. S5. Driven by real-time spatiotemporal trajectory data, the on-chain intervention of critical phase transition is generated, and the off-chain final state settlement data is generated based on dynamic risk warning points. The candidate value of the downgrade control authority benchmark and the hash digest of the off-chain final state settlement data are submitted to the blockchain in parallel. S6. The blockchain verifies the integrity of the off-chain final settlement data based on the hash digest and verifies whether the candidate value of the downgrade control authority benchmark exceeds the downgrade boundary threshold. After the two-dimensional verification is confirmed, the hash digest is signed to generate a downgrade execution certificate. S7. The edge control terminal responds to the trust verification of the off-chain final settlement data and the corresponding degradation execution certificate. Based on the endorsement of the degradation execution certificate, the candidate value of the degradation control authority benchmark is established as the target degradation control authority benchmark and mapped to the device physical operating parameters.

2. The risk warning method for a blockchain-based leasing platform according to claim 1, characterized in that, S1 specifically includes: S11. Collect the spatial geofence coordinates of the leased object, the historical spatiotemporal trajectory of the leased object, and the upper limit of the initial device operating parameters of the edge control terminal. Package the spatial geofence coordinates, the historical spatiotemporal trajectory of the leased object, and the upper limit of the initial device operating parameters into an initial block and upload it to the blockchain for evidence storage. Write the upper limit of the initial device operating parameters into the local security register of the edge control terminal to complete the anchoring. S12. Extract the upper limit of the initial device operating parameters anchored in the local security register, extract the historical spatiotemporal trajectory of the leased item stored in the blockchain, statistically analyze the boundary crossing distribution characteristics of the historical spatiotemporal trajectory of the leased item relative to the spatial geofence coordinates, adaptively calculate the benchmark tolerance deviation ratio based on the boundary crossing distribution characteristics, and output the degradation boundary threshold by combining the product operation of the benchmark tolerance deviation ratio and the upper limit of the initial device operating parameters. S13. Encapsulate the degradation boundary threshold into a boundary constraint block and upload it to the blockchain for evidence storage, and establish a hash mapping index on the blockchain between the initial upper limit of device operating parameters and the degradation boundary threshold.

3. The risk warning method for a blockchain-based leasing platform according to claim 1, characterized in that, S2 specifically includes: S21. Obtain the real-time spatiotemporal trajectory data of the leased object and input it into the off-chain monitoring. Extract the spatial geofence coordinates stored in the blockchain and align the real-time spatiotemporal trajectory data with the spatial geofence coordinates to output the aligned trajectory sequence. S22. Calculate the shortest distance from each trajectory point in the alignment trajectory sequence to the coordinate boundary of the spatial geographic fence, statistically analyze the evolutionary distribution characteristics of the shortest distance, and adaptively extract the dynamic deviation metric benchmark based on the evolutionary distribution characteristics. S23. Perform nonlinear mapping on the shortest distance of each trajectory point based on the dynamic deviation metric benchmark, and output the spatial constraint deviation characteristics.

4. The risk warning method for a blockchain-based leasing platform according to claim 1, characterized in that, The improved Flowformer model includes a spatiotemporal flow feature embedding layer, a nonlinear manifold mapping layer, and a risk integral calculation layer: The spatiotemporal flow feature embedding layer is used to perform spatiotemporal alignment and feature fusion of spatial constraint deviation features and real-time spatiotemporal trajectory data, and inject position awareness and local temporal evolution information to output a spatiotemporal flow embedding tensor. The nonlinear manifold mapping layer is used to map the spatiotemporal flow embedding tensor to a high-dimensional latent space to generate a basic latent space feature tensor, and simultaneously generate a residual tensor of the same dimension; a temporal causal constraint is applied to the basic latent space feature tensor to perform global feature aggregation, thereby generating an enhanced latent space feature tensor; The enhanced latent space feature tensor is fused with the same-dimensional residual tensor to output a nonlinear risk feature tensor; The implied volatility extrapolation layer is used to introduce a volatility evolution mechanism, decouple and extract the risk evolution state and the risk critical boundary from the nonlinear risk feature tensor, and calculate the feature dynamic divergence based on the deviation of the risk evolution state from the risk critical boundary. The temporal diffusion evolution is driven by the aforementioned dynamic divergence to generate temporal evolution features; The temporal evolution features are activated and fused through a dual-segment heterogeneous phase transition to output a risk-implicit fluctuation tensor. The risk integral solution layer is used to perform adaptive aggregation on the risk implicit volatility tensor along the time dimension to output a risk evolution feature vector, and to perform dimensionality reduction mapping and interval normalization on the risk evolution feature vector to output a dynamic risk warning integral.

5. The risk warning method for a blockchain-based leasing platform according to claim 4, characterized in that, The aforementioned fluctuation evolution mechanism specifically includes: The nonlinear risk feature tensor is extracted by exponential decay memory field sliding convolution along the time dimension, and state-dependent gated selective aggregation is applied to the convolution extraction result. Then, hyperbolic tangent activation and scaling affine mapping are performed on the aggregation result to generate a dynamic state sequence. The nonlinear risk feature tensor is sequentially subjected to feature dimension extreme value comparison extraction, hard-limited truncation amplification and additive offset and scaling mapping to generate the risk critical boundary; Calculate the element-by-element comparison and division of the dynamic state sequence relative to the risk critical boundary on the corresponding feature channel to obtain the feature space approximation ratio; perform a logarithmic transformation on the approximation ratio to generate the logarithmic elasticity deviation. Based on the logarithmic elasticity deviation, backward difference is performed to extract the momentum deviation rate of change in the time dimension. The statistical mean of the momentum deviation rate of change in the time dimension is extracted and used to perform centering detrending. Then, quadratic moment accumulation with decay memory is performed on the detrending result to calculate the characteristic dynamic divergence. The adaptive early warning time window span is calculated based on the dynamic divergence of the features. The dynamic divergence of the features is used as the evolutionary variance parameter. The inverse step size operator is constructed in combination with the adaptive early warning time window span to generate the time deviation step size. Using the evolution variance parameter as the fluctuation scale and the time aberration step size as the walk interval, the long-term equilibrium state and recovery decay coefficient are calculated based on the statistical mean and detrending results. A stepwise random walk with mean recovery constraint and nonlinear scaling mapping are performed to generate time-series evolution features. The temporal evolution features are activated by a dual-segment heterogeneous phase transition of critical threshold gating and overcritical exponential surge, respectively generating subcritical gated response tensors and overcritical exponential surge response tensors, and the routing mask generated by threshold decision is extracted. Based on the routing mask, bitwise selective fusion is performed on the subcritical gated response tensor and the overcritical exponential surge response tensor to calculate the future risk uncertainty spillover value, and the risk implied volatility tensor is aggregated and output according to multi-scale time windows.

6. The risk warning method for a blockchain-based leasing platform according to claim 5, characterized in that, The exponentially decaying memory field sliding convolution extraction specifically includes: Extract the feature channel dimension span of the nonlinear risk feature tensor and calculate the local variance of the nonlinear risk feature tensor in the time dimension. Based on the feature channel dimension span and the local variance in the time dimension, dynamically calculate the receptive field window length to construct the initial convolution kernel. Based on the time distance of each element in the initial convolution kernel relative to the current calculation time, an exponential decay function is applied to generate exponential decay constraint weights. The exponential decay constraint weights are then multiplied element-wise with the internal weights of the initial convolution kernel to generate a time decay convolution kernel. The time decaying convolution kernel is trunculated along the time dimension of the nonlinear risk feature tensor with a step size of 1. The truncated local time window features are multiplied element-wise with the time decaying convolution kernel and summed to obtain the local time decaying response value. The center time feature aligned with the current calculation time in the local time window features is extracted as the residual memory component. The local time decay response value and the residual memory component are added bitwise to generate the memory field convolution compensation result. The convolution extraction result is generated by traversing and sliding along the time dimension and splicing the memory field convolution compensation results corresponding to each time.

7. The risk warning method for a blockchain-based leasing platform according to claim 1, characterized in that, S4 specifically includes: S41. Extract the dynamic risk warning integral and the upper limit of the initial equipment operating parameters anchored to the local security register of the edge control terminal, and extract the historical time series distribution characteristics of the nonlinear risk feature tensor to adaptively extract the dynamic security benchmark integral. S42. Calculate the relative deviation of the dynamic risk warning integral from the dynamic safety benchmark integral, calculate the dynamic scaling factor based on the relative deviation, multiply the dynamic risk warning integral and the dynamic scaling factor, and output the degradation attenuation coefficient. S43. Using the degradation attenuation coefficient as a modulation parameter, perform attenuation product operation on the upper limit of the initial equipment operating parameters, and output the intermediate degradation parameter value. S44. Calculate the attenuation difference between the initial upper limit of equipment operating parameters and the intermediate degradation parameter values, and calculate the joint penalty weight of the attenuation difference and the dynamic risk warning score; S45. Perform a weighted operation on the attenuation difference based on the joint penalty weight, output the penalty suppression amount, perform a difference operation on the intermediate degradation parameter value and the penalty suppression amount, and output the candidate value of the degradation control authority baseline.

8. The risk warning method for a blockchain-based leasing platform according to claim 1, characterized in that, S5 specifically includes: S51. Extract spatial constraint deviation features, statistically analyze the temporal abrupt change distribution features of spatial constraint deviation features, calculate the phase transition triggering coefficient based on the temporal abrupt change distribution features, perform adaptive boundary judgment based on the temporal abrupt change distribution features, and output the phase transition signal. S52. Based on phase transition signal triggering, the phase transition triggering coefficient is jointly mapped with the dynamic risk warning integral to output off-chain final state settlement data, and a hash operation is performed on the off-chain final state settlement data to output a hash digest. S53. Extract the baseline candidate value for downgraded control permissions, encapsulate the baseline candidate value for downgraded control permissions and hash digest into a binary structure, and submit the binary structure to the blockchain for evidence storage.

9. A risk warning method for a blockchain-based leasing platform according to claim 1, characterized in that, S6 specifically includes: S61. Extract and parse the binary structure on the blockchain to obtain the off-chain final settlement data, hash digest and downgrade control permission baseline candidate value. Perform hash operation on the off-chain final settlement data to output the reconstructed hash digest. Compare the reconstructed hash digest with the hash digest and output the integrity verification result. S62. Extract the degradation boundary threshold of the blockchain evidence storage, calculate the degradation attenuation degree of the baseline candidate value of degradation control authority relative to the upper limit of the initial device operating parameters, perform an out-of-bounds comparison between the degradation attenuation degree and the degradation boundary threshold, and output the out-of-bounds verification result. S63. Perform a logical AND operation based on the integrity check result and the out-of-bounds check result. If the logical AND operation passes, sign the hash digest and output the downgrade execution credential.

10. A risk warning method for a blockchain-based leasing platform according to claim 1, characterized in that, Specifically, S7 includes: S71. Extract the downgrade execution credential and hash digest. Perform a decryption operation on the downgrade execution credential based on the blockchain public key and output the decrypted hash digest. Perform a consistency comparison between the decrypted hash digest and the hash digest and output the credential verification result. S72. Based on the successful verification of credentials, extract the candidate values ​​of the downgraded control permission benchmark cache in the local cache of the edge control terminal, and establish the candidate values ​​of the downgraded control permission benchmark as the target downgraded control permission benchmark. S73. Extract the initial device operating parameter upper limit anchored to the local security register of the edge control terminal, calculate the normalized degradation ratio of the target degradation control permission baseline candidate value relative to the initial device operating parameter upper limit, perform a product operation on the normalized degradation ratio and the initial device operating parameter upper limit, and output the device physical operating parameters.

Images