Search

CN-121978768-A - Geophysical prospecting earth deep detection system and method based on artificial intelligence

CN121978768ACN 121978768 ACN121978768 ACN 121978768ACN-121978768-A

Abstract

The invention discloses a geophysical prospecting earth deep detection system and method based on artificial intelligence, the system comprises a multi-source data acquisition subsystem, an intelligent preprocessing subsystem, an AI inversion decision subsystem, a dynamic interpretation subsystem and a visual output subsystem, wherein the subsystems realize data interaction and collaborative work through a high-speed communication network. According to the invention, through constructing a full-flow intelligent architecture of acquisition, pretreatment, inversion, explanation and visualization, a distributed geophysical prospecting sensor array, edge computing nodes, a multi-model AI inversion engine and a three-dimensional visualization terminal are integrated, multi-mode data of seismic waves, electromagnetism, gravity and magnetic force are synchronously acquired, and the high-precision extraction of deep geological parameters is realized through enhanced pretreatment and multi-objective optimized inversion, so that the technical pain points with low data processing efficiency and strong inversion multi-solution property in the traditional method are solved, and the detection precision is improved.

Inventors

  • LUO ZHONGQIN
  • JIN XUELIANG
  • TIAN XUEFENG
  • LI NING
  • HE YAZHOU

Assignees

  • 中国煤炭地质总局地球物理勘探研究院

Dates

Publication Date
20260505
Application Date
20260121

Claims (9)

  1. 1. The geophysical prospecting earth deep detection system based on artificial intelligence is characterized by comprising a multi-source data acquisition subsystem, an intelligent preprocessing subsystem, an AI inversion decision subsystem, a dynamic interpretation subsystem and a visual output subsystem, wherein the subsystems realize data interaction and collaborative work through a high-speed communication network; The multi-source data acquisition subsystem consists of a distributed geophysical prospecting sensor array, a data transmission module and a synchronous control unit, wherein the geophysical prospecting sensor array comprises a seismic wave sensor, an electromagnetic sensor, a gravity sensor and a magnetic sensor and is used for synchronously acquiring multi-mode geophysical prospecting data of the deep part of the earth, the synchronous control unit realizes synchronous acquisition time sequence of each sensor, and the data transmission module transmits original data through a 5G/optical fiber network; the intelligent preprocessing subsystem is deployed at the edge computing node and comprises a data cleaning module, an outlier removing module, a multi-mode characteristic enhancing module and a data standardization module, and is used for enhancing and preprocessing original geophysical prospecting data and outputting high-reliability structured data; The AI inversion decision-making subsystem comprises a pre-training inversion model library, an increment learning module and a multi-target optimization module, wherein the pre-training inversion model library comprises a seismic wave inversion model, an electromagnetic parameter inversion model and a gravity magnetic joint inversion model based on deep learning, the increment learning module continuously absorbs new detection data optimization model parameters, and the multi-target optimization module generates an optimal inversion result by taking the maximization of detection precision and the minimization of data fitting errors as targets; The dynamic interpretation subsystem comprises a geologic body identification module, a parameter association analysis module and a dynamic correction module, and based on the inversion result and known geologic data, deep geologic structures, mineral distribution and disaster hidden danger are intelligently identified, and interpretation conclusion is dynamically corrected; The visual output subsystem comprises a three-dimensional modeling module, a result display module and a data export module, converts inversion results and interpretation conclusions into a three-dimensional geological model, and supports multidimensional visual display and standardized data export.
  2. 2. The geophysical prospecting earth deep detection method based on artificial intelligence is characterized by comprising the following steps of: S1, synchronously acquiring multisource geophysical prospecting data, namely synchronously acquiring seismic wave data, electromagnetic data, gravity data and magnetic force data through a distributed geophysical prospecting sensor array, calibrating acquisition time sequences through a synchronous control unit, and uploading the acquisition time sequences to an intelligent preprocessing subsystem through a data transmission module; S2, multi-mode data enhancement preprocessing, namely cleaning, outlier rejection, feature enhancement and standardization processing are carried out on original geophysical prospecting data to obtain structured data; S3, AI model training and optimization, namely, based on historical geophysical prospecting data and known geological data, performing initial training on a deep learning model in an inversion model library, and absorbing new scene data through an incremental learning module to optimize model parameters so as to ensure that the model is adapted to geological conditions of different detection areas; S4, intelligent inversion calculation, namely inputting the preprocessed structured data into a trained AI inversion model, and inverting to obtain deep geological parameters (stratum thickness, rock density, resistivity and wave velocity) of the earth by combining a multi-objective optimization algorithm; s5, dynamically geologic interpretation, namely intelligently identifying the type of the deep geologic body, mineral resource distribution and geologic hazard hidden danger based on geologic parameters obtained by inversion and combined with regional geologic background information, and dynamically correcting interpretation results; And S6, three-dimensional visual output, namely constructing a three-dimensional geological model by the inversion result and the interpretation conclusion, and carrying out multidimensional display through a visual terminal to support data export and result sharing.
  3. 3. The artificial intelligence based geophysical prospecting earth depth detecting method according to claim 2, wherein the method comprises the steps of: s1-1, seismic wave data acquisition, namely acquiring longitudinal wave and transverse wave propagation signals through a detector array, setting the sampling frequency to be 250-1000Hz, and recording parameters such as signal amplitude, phase, propagation time and the like; S1-2, electromagnetic data acquisition, namely adopting a controllable source electromagnetic method and a natural field electromagnetic method to acquire electromagnetic signals with the frequency range of 1Hz-10kHz, and recording electric field intensity, magnetic field intensity and impedance parameters; S1-3, acquiring gravity data, namely acquiring gravity acceleration abnormal data through a gravity meter, and synchronously recording longitude and latitude and elevation information of an acquisition point at 1-5 seconds of sampling intervals; s1-4, magnetic force data acquisition, namely acquiring geomagnetic field total intensity and component abnormal data through a magnetometer, wherein the resolution is less than or equal to 0.1nT, and synchronously calibrating instrument errors; And S1-5, synchronous control, namely ensuring that the acquisition time sequence error of each sensor is less than or equal to 1ms by using a GPS time service and clock synchronization technology, and ensuring the data security by adopting an encryption protocol for data transmission.
  4. 4. The artificial intelligence based geophysical prospecting earth depth detecting method according to claim 2, wherein the method comprises the steps of: S2-1, data cleaning, namely adopting a sliding window method to remove random noise in data, and deleting invalid data (such as zero value and extreme value caused by sensor fault) through data consistency check; S2-2, outlier rejection, namely adopting an improved isolated forest algorithm to combine with statistical test (Grubbs test) to identify and reject systematic outliers caused by geological anomalies and random outliers caused by instrument errors; S2-3, multi-mode feature enhancement, namely extracting time domain features (peak value, energy and main frequency) and frequency domain features (frequency spectrum distribution and phase spectrum) from seismic wave data, extracting impedance features and polarization features from electromagnetic data, extracting gradient features and abnormal boundary features from heavy magnetic data, and constructing a multi-dimensional feature set; and S2-4, data standardization, namely mapping the geophysical prospecting data with different dimensions to the same distribution interval by adopting a Z-score normalization algorithm, and eliminating dimensional differences to obtain structured data.
  5. 5. The artificial intelligence based geophysical prospecting earth depth detecting method according to claim 2, wherein the method comprises the steps of: s3-1, constructing a model, namely adopting a U-Net improved network (an attention adding mechanism) for a seismic wave inversion model, adopting a Deep Belief Network (DBN) for an electromagnetic parameter inversion model, and adopting a multi-task convolution neural network (MT-CNN) for a heavy magnetic joint inversion model; s3-2, initial training, namely training a model until a loss function converges (the loss value is less than or equal to 0.03) by utilizing historical geophysical prospecting data (comprising known drilling verification data) and geological parameter labels; S3-3, incremental optimization, namely adopting an Elastic Weight Consolidation (EWC) algorithm to continuously absorb new detection region data and verification results, and finely adjusting model parameters to avoid catastrophic forgetting; and S3-4, evaluating model inversion accuracy through Root Mean Square Error (RMSE) and correlation coefficient (R2), wherein the RMSE is less than or equal to 5%, and the R2 is more than or equal to 0.9.
  6. 6. The artificial intelligence based geophysical prospecting earth depth detecting method according to claim 2, wherein the method comprises the steps of: S4-1, inputting the structured data into a corresponding inversion model according to modal classification, and improving the calculation efficiency by adopting a batch processing mechanism; S4-2, multi-objective optimization, namely constructing an optimization function Wherein 、 、 Is a weight coefficient (satisfy + + =1), T is the inversion time-consuming, Time consuming for maximum allowance; S4-3, inversion calculation, namely accelerating the inversion process through GPU parallel calculation, and outputting core geological parameters such as stratum thickness, rock density (error is less than or equal to 3%), resistivity (error is less than or equal to 8%), wave speed (error is less than or equal to 4%), and the like; and S4-4, checking results, namely cross checking inversion results by combining known drilling data and a regional geological map, and eliminating invalid inversion results.
  7. 7. The artificial intelligence based geophysical prospecting earth depth detecting method according to claim 2, wherein the method comprises the steps of: S5-1, identifying a geologic body, namely identifying rock types such as sedimentary rock, magma rock and metamorphic rock through a pre-trained geologic body classification model (ResNet-50) based on inversion parameters, and delineating a mineral body distribution range; s5-2, parameter association analysis, namely adopting a Bayesian network to analyze association relation between geological parameters and mineral resources and disaster hidden danger, and quantifying resource reserve probability and disaster occurrence risk level; S5-3, dynamically correcting, namely dynamically adjusting interpretation model parameters by combining real-time detection data and newly acquired geological data, and updating geological interpretation conclusion; and S5-4, verifying the result by means of field geological investigation, drilling verification and the like, and ensuring that the interpretation accuracy is more than or equal to 85%.
  8. 8. The artificial intelligence based geophysical prospecting earth depth detecting method according to claim 2, wherein the method comprises the steps of: S6-1, three-dimensional modeling, namely adopting VTK (VisualizationToolkit) to construct a three-dimensional geological model and restoring the spatial distribution characteristics of a deep geological structure; S6-2, visually displaying, namely supporting slice viewing, transparency adjustment and multi-view switching, and displaying geological parameter distribution, geologic body boundaries, resource distribution ranges and disaster hidden danger areas; S6-3, data export, namely supporting export of a standardized data format (SEGY, ASCII, shapefile), and being compatible with mainstream geological exploration software; and S6-4, carrying out achievement sharing, namely realizing authority sharing of detection achievement through a cloud platform, and supporting multi-terminal access and viewing.
  9. 9. The geophysical prospecting earth deep detecting system based on artificial intelligence according to claim 1, further comprising a fault diagnosis and self-adaptive adjustment module, wherein the fault diagnosis and self-adaptive adjustment module monitors the working state of the sensor, a data transmission link and the running condition of a model in real time, automatically switches a standby sensor when the sensor is in fault, starts an edge node local cache when the data transmission is interrupted, and triggers an incremental learning process when the inversion accuracy of the model is reduced, so that the stable running of the system is ensured.

Description

Geophysical prospecting earth deep detection system and method based on artificial intelligence Technical Field The invention relates to the technical field of geophysical prospecting earth deep detection, in particular to an artificial intelligence-based geophysical prospecting earth deep detection system and method. Background Earth deep exploration is a core foundation for mineral resource exploration, geological disaster early warning and underground space development, but the traditional detection technology still faces a plurality of bottlenecks under complex geological conditions. On the one hand, the traditional detection is dependent on a single geophysical prospecting means (such as single seismic prospecting and electrical prospecting), so that the multi-element characteristics of the deep geologic body are difficult to comprehensively capture, the precise synchronization mechanism is lacked in multi-source data acquisition, the time sequence error is large, the data integrity and consistency are insufficient, the problems of deep signal attenuation, noise interference and the like are superposed, and the quality of the original data is further reduced. Meanwhile, the traditional inversion model is mostly a single fixed framework, lacks the adaptation capability to different geological scenes, cannot continuously absorb new detection data to optimize the model, is easy to forget disastrous, and has strong inversion polynomials, low precision and difficult to accurately acquire core geological parameters such as stratum thickness, rock density and the like. In addition, the geological interpretation process is highly dependent on manual experience, analysis subjectivity on inversion results is high, a dynamic correction mechanism is lacked, real-time detection data and new geological data are difficult to integrate rapidly, interpretation conclusion accuracy is insufficient, updating is lagged, meanwhile, detection results are presented in a two-dimensional report or a simple model, visualization degree is low, a data format is not compatible with mainstream exploration software, sharing efficiency is low, and popularization and application of results are restricted. Furthermore, the traditional system lacks a perfect fault diagnosis and self-adaptive adjustment mechanism, so that problems such as sensor faults, data transmission interruption and the like easily cause interruption of a detection flow, and further influence the detection efficiency and stability. The problems commonly lead to the defects of low precision, poor efficiency, weak scene suitability, insufficient result practicality and the like of the traditional earth deep detection, and the requirements of the fields of current deep mineral exploration, geological disaster early warning and the like on high-precision and intelligent detection are difficult to meet, so that an innovative detection technical scheme integrating multi-source data fusion, intelligent processing, dynamic interpretation and stable operation is needed. Disclosure of Invention In order to solve the problems in the background art, the invention aims to provide a geophysical prospecting earth deep detection system and method based on artificial intelligence, which have the advantages of multi-source data fusion, intelligent processing, dynamic interpretation and stable operation, and solve the problems of insufficient data integrity, poor preprocessing effect, low inversion precision, interpretation dependence on artificial experience, weak result practicality and insufficient system stability of the traditional detection method. In order to achieve the aim, the invention provides the technical scheme that the geophysical prospecting earth deep detection system based on artificial intelligence comprises a multi-source data acquisition subsystem, an intelligent preprocessing subsystem, an AI inversion decision subsystem, a dynamic interpretation subsystem and a visual output subsystem, wherein the subsystems realize data interaction and collaborative work through a high-speed communication network; The multi-source data acquisition subsystem consists of a distributed geophysical prospecting sensor array, a data transmission module and a synchronous control unit, wherein the geophysical prospecting sensor array comprises a seismic wave sensor, an electromagnetic sensor, a gravity sensor and a magnetic sensor and is used for synchronously acquiring multi-mode geophysical prospecting data of the deep part of the earth, the synchronous control unit realizes synchronous acquisition time sequence of each sensor, and the data transmission module transmits original data through a 5G/optical fiber network; the intelligent preprocessing subsystem is deployed at the edge computing node and comprises a data cleaning module, an outlier removing module, a multi-mode characteristic enhancing module and a data standardization module, and is used for enhancing and preprocessing