Search

CN-122024409-A - Early warning method, medium, equipment and product for ground collapse

CN122024409ACN 122024409 ACN122024409 ACN 122024409ACN-122024409-A

Abstract

The invention discloses a ground collapse early warning method, medium, equipment and a product, which relate to the technical field of ground collapse early warning, and the method comprises the steps of carrying out data space-time alignment on multisource monitoring data of ground collapse to obtain observation vectors after space-time alignment, constructing a ground collapse causal graph according to geological priori data, carrying out causal graph structure learning, establishing a dynamic credibility state variable based on each data source, calculating multisource dynamic weights, extracting premonitory features from the observation vectors, calculating consistency scores in a preset time window based on the premonitory features and the dynamic weights, constructing a residual instability time prediction model based on causal graph reasoning results, the dynamic weights and the consistency scores, outputting residual instability time prediction of a target area, carrying out hierarchical early warning according to residual instability time threshold values and risk threshold values, taking on-site checking results as observation feedback, and updating credibility state variables, threshold values and prediction model parameters. The invention can effectively reduce the early warning false alarm rate.

Inventors

  • TAN FEI
  • LV JIAHE

Assignees

  • 中国地质大学（武汉）

Dates

Publication Date
20260512
Application Date
20251229

Claims (10)

  1. 1. The early warning method for the ground collapse is characterized by comprising the following steps of: s1, obtaining geological priori data of ground subsidence and multisource monitoring data, and performing data space-time alignment on the multisource monitoring data to obtain an observation vector after space-time alignment; S2, constructing a ground collapse causal graph according to geological priori data, and performing causal graph structure learning; S3, establishing a dynamic credibility state variable based on the monitoring data of each data source, and carrying out online estimation by adopting Bayesian updating and Kalman filtering according to the prior and recent observation residual errors of the causal graph to obtain the dynamic weight of the multi-source data; S4, extracting a premonitory feature from the observation vector after space-time alignment, and calculating a consistency score in a preset time window based on the premonitory feature and the dynamic weight; S5, constructing a residual destabilization time prediction model based on causal graph reasoning results, dynamic weights and consistency scores, and outputting residual destabilization time predictions of a target area; S6, comparing the residual destabilizing time prediction result with a preset residual destabilizing time threshold value and a preset risk threshold value to obtain an early warning grade of the target area, and updating the dynamic credibility state variables, the residual destabilizing time threshold value, the risk threshold value and the parameters of the residual destabilizing time prediction model of each data source by taking the field checking result as observation feedback.
  2. 2. The method for early warning of ground subsidence according to claim 1, wherein the multisource monitoring data comprises InSAR deformation sequences, GNSS displacements, GPR radar echoes, ground sound/vibration sensor signals, urban pipe network flow/pressure, rainfall and groundwater level.
  3. 3. The method for early warning of ground collapse according to claim 1, wherein prior constraints of irreversible edges and forbidden edges are set during the structure learning of the causal graph.
  4. 4. The method of claim 1, wherein the pre-characterization includes deformation curvature abrupt change, phase envelope drift, acceleration spectral peak up-shift, pipe pressure/flow micro-abrupt change, and abnormal energy windows of the earth sound.
  5. 5. The method for early warning of ground collapse according to claim 1, wherein the multisource dynamic weighting is formulated as: Wherein, the Representing the dynamic weight of the ith data source, And Representing the dynamic reliability state variables of the ith and kth data sources, respectively.
  6. 6. The method of claim 5, wherein the uniformity score is formulated as: wherein S represents a consistency score, Is shown in a preset time window The precursor characteristics of the ith data source in the database.
  7. 7. The early warning method for ground collapse according to claim 1, wherein the construction of the residual destabilizing time prediction model is specifically as follows: reasoning each monitoring data in the target area in a causal graph main control mechanism chain to form causal graph reasoning characteristics; weighting each monitoring data according to the dynamic weight to obtain a weighted monitoring characteristic; Splicing causal graph reasoning features, weighted monitoring features and consistency scores to form feature vectors; And constructing a model based on a machine learning method, and training the model constructed by the machine learning method by using the feature vector to obtain a residual instability time prediction model.
  8. 8. A computer-readable storage medium, in which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the method according to any of the claims 1-7.
  9. 9. An electronic device comprising a processor and a memory, the processor being interconnected with the memory, wherein the memory is configured to store a computer program comprising computer readable instructions, the processor being configured to invoke the computer readable instructions to perform the method of any of claims 1-7.
  10. 10. A computer program product comprising computer programs/instructions which, when executed by a processor, implement the steps of the method of any of claims 1-7.

Description

Early warning method, medium, equipment and product for ground collapse Technical Field The invention relates to the technical field of ground collapse early warning, in particular to a ground collapse early warning method, medium, equipment and a product. Background Ground subsidence is a common geological disaster, and forms a serious threat to urban infrastructure safety, traffic operation and resident life and property safety. Currently, there are many limitations to ground collapse monitoring that rely on a single data source or simple multi-source data superposition. The partial monitoring method only adopts single deformation monitoring means such as InSAR (interferometric synthetic aperture radar) or GNSS (global navigation satellite system), and the like, is difficult to comprehensively capture complex precursor information before ground collapse occurs, and is easy to cause early warning lag or false alarm due to the one-sidedness of the data. In addition, some multisource monitoring schemes do not consider the credibility difference of different data sources under different situations, all data are weighted equally, and when a certain data source is abnormally affected by environmental interference (such as InSAR observation under bad weather influence and flow data influenced by pipe network instantaneous fluctuation), the accuracy of an early warning result can be seriously affected. Meanwhile, the existing early warning method lacks of systematic modeling of the ground subsidence cause relation, and cannot clearly comb the internal association of rainfall, seepage, leakage loss, stratum state, sedimentation/cavity, so that early warning logic is ambiguous, and residual instability time is difficult to accurately predict. In addition, most schemes do not establish an effective feedback mechanism, can not dynamically optimize model parameters and thresholds according to field checking results, and can easily cause early warning performance degradation and degradation after long-term use, and are difficult to adapt to complex and changeable geology and environmental conditions. Disclosure of Invention The invention aims to solve the problem that the existing ground collapse monitoring is dependent on a single data source or simple multi-source data superposition and has various limitations, and provides a ground collapse early warning method which comprises the following steps: s1, obtaining geological priori data of ground subsidence and multisource monitoring data, and performing data space-time alignment on the multisource monitoring data to obtain an observation vector after space-time alignment; S2, constructing a ground collapse causal graph according to geological priori data, and performing causal graph structure learning; S3, establishing a dynamic credibility state variable based on the monitoring data of each data source, and carrying out online estimation by adopting Bayesian updating and Kalman filtering according to the prior and recent observation residual errors of the causal graph to obtain the dynamic weight of the multi-source data; S4, extracting a premonitory feature from the observation vector after space-time alignment, and calculating a consistency score in a preset time window based on the premonitory feature and the dynamic weight; S5, constructing a residual destabilization time prediction model based on causal graph reasoning results, dynamic weights and consistency scores, and outputting residual destabilization time predictions of a target area; S6, comparing the residual destabilizing time prediction result with a preset residual destabilizing time threshold value and a preset risk threshold value to obtain an early warning grade of the target area, and updating the dynamic credibility state variables, the residual destabilizing time threshold value, the risk threshold value and the parameters of the residual destabilizing time prediction model of each data source by taking the field checking result as observation feedback. Further, the multisource monitoring data comprises InSAR deformation sequences, GNSS displacement, GPR radar echoes, ground sound/vibration sensor signals, urban pipe network flow/pressure, rainfall and groundwater level. Further, when the causal graph structure is learned, prior constraints of irreversible edges and forbidden edges are set. Further, the precursor features include deformation curvature abrupt changes, phase envelope shifts, acceleration spectral peak shifts up, pipe pressure/flow micro-abrupt changes, and earth sound anomaly energy windows. Further, the multisource dynamic weight reuse formula is expressed as: Wherein, the Representing the dynamic weight of the ith data source,AndRepresenting the dynamic reliability state variables of the ith and kth data sources, respectively. Further, the consistency score is formulated as: wherein S represents a consistency score, Is shown in a preset time windowThe precursor characteristics