CN-121834720-B - CPTU-based multi-source geological survey data fusion method

CN121834720BCN 121834720 BCN121834720 BCN 121834720BCN-121834720-B

Abstract

The invention belongs to the technical field of geotechnical engineering investigation and artificial intelligence intersection, and in particular relates to a CPTU-based multi-source geological investigation data fusion method, which comprises the steps of obtaining a CPTU data set and a drilling data set and preprocessing; the method comprises the steps of carrying out data set matching based on depth, extracting CPTU center point characteristics, window statistical characteristics and multi-scale characteristics extracted by wavelet transformation to form a characteristic set by utilizing a sliding window, constructing a standardized training sample by the characteristic set and a soil body parameter label, carrying out model training by adopting a multi-output random forest, and applying the trained model to full-depth continuous section prediction and uncertainty quantification. According to the invention, the discrete drilling data and the continuous CPTU data are fused, so that a fusion frame of multi-source heterogeneous data and soil parameter mapping is established, the difficult problems of data fracture and uncertainty quantification in the traditional method are solved, and reliable prediction and decision support are provided for ocean geotechnical engineering, infrastructure construction and geological disaster assessment.

Inventors

LIU TAO
CAI YUNPENG
WANG DONG
ZHAO JIAN
ZHENG TIANYUAN
GAO YANG
LIU LELE
CAI GUOJUN

Assignees

中国海洋大学

Dates

Publication Date: 20260512
Application Date: 20260313

Claims (7)

1. The multi-source geological survey data fusion method based on CPTU is characterized by comprising the following steps of: s1, acquiring soil parameters of a CPTU data set and drilling test data points, and performing fusion pretreatment; S2, searching a point with the smallest absolute depth difference from each drilling test data point in the CPTU depth sequence according to the depth value as a matching anchor point; S3, setting a dynamic sliding window along the depth direction by taking the matched anchor point as a center, and fusing and extracting three types of characteristics in the window, namely a) a CPTU data set corresponding to the anchor point as a center point characteristic, b) statistical characteristics of CPTU data in the window, c) CPTU multi-scale characteristics of the window; The CPTU multi-scale features are extracted through wavelet transformation, and the method is specifically as follows: s3c1, carrying out multi-layer wavelet decomposition on the cone tip resistance signal in the sliding window by using a db4 wavelet basis function to obtain detail coefficients of each layer, wherein the wavelet decomposition calculation formula is as follows: , Wherein J is the number of decomposition layers, A J is the approximation coefficient of the J layers, and D j is the detail coefficient of the J layers; s3c2, for each layer of detail coefficient array, when the detail coefficient array is valid, calculating the energy and standard deviation of the detail coefficient of the layer, wherein the standard deviation calculation formula is as follows: , Wherein, the Is the mean value of the detail coefficients of the j-th layer, Is the number of the layer coefficients; s3c3, calculating total energy and average standard deviation based on detail coefficients of all the effective layers; s3c4, the energy, standard deviation and average standard deviation of each layer of detail coefficient are combined to form the CPTU multi-scale feature; S4, constructing a feature set fusing local context information by the three types of features in the S3, associating the feature set with a soil body parameter label corresponding to an anchor point, and constructing a multidimensional feature vector to form a standardized training sample.
2. The method for merging multi-source geological survey data based on CPTU according to claim 1, further comprising step S5, wherein the standardized training sample is used for training a subsequent multi-output random forest, carrying out joint prediction on a plurality of soil parameters, and applying the trained model to a complete CPTU penetration curve to generate a complete prediction section comprising a prediction mean, a standard deviation, a confidence interval and an information entropy.
3. The method for merging multi-source geological survey data based on CPTU according to claim 1, wherein the CPTU data set comprises original parameters and derivative parameters, the original parameters comprise cone tip resistance, side friction resistance and pore water pressure, and the derivative parameters comprise normalized cone tip resistance, friction resistance ratio and pore pressure ratio.
4. The CPTU-based multi-source geological survey data fusion method of claim 1, wherein in step S3, the statistical features of the in-window CPTU data are a plurality of statistics of the calculated cone tip resistance signal in the window.
5. The method for merging multi-source geological survey data based on CPTU according to claim 1, wherein in step S3c1, before wavelet decomposition is performed, it is judged whether the length of the cone tip resistance signal meets the preset requirement of the number of decomposition layers, and if not, the number of decomposition layers is automatically adjusted to the maximum number of safety layers allowed by the current signal length.
6. The CPTU-based multi-source geological survey data fusion method of claim 1, wherein said sliding window is a fixed point number symmetric sliding window centered on said matching anchor point.
7. The method of claim 1, wherein in step S4, after the feature set is constructed for each drilling point, the percentage of the effective features in the feature set is calculated, and only when the percentage exceeds 50%, the data points are reserved for training sample construction.

Description

CPTU-based multi-source geological survey data fusion method Technical Field The invention belongs to the technical field of geotechnical engineering investigation and artificial intelligence intersection, and particularly relates to a CPTU-based multi-source geological investigation data fusion method. Background The marine rock soil investigation is used as a precondition for marine resource development and marine engineering construction, and the accuracy and reliability of the investigation result are related to the safety and stability of the marine engineering and the full life cycle cost. The accurate acquisition of physical and mechanical parameters (such as non-drainage shear strength, compression modulus, permeability coefficient, oversolidification ratio, fine grain content and the like) of the ocean soil body is a precondition for realizing the fine design, safety evaluation and long-term performance prediction of structures such as ocean platforms, submarine pipelines, offshore wind power foundations and the like. At present, a method for acquiring soil engineering property parameters mainly depends on drilling sampling and indoor geotechnical test. Although the method can directly obtain soil samples and provide more detailed physical and mechanical parameters, the method is regarded as 'gold standard' for a long time. However, in the marine environment, the inherent defects are greatly amplified, namely, firstly, the offshore drilling operation is influenced by severe environments such as stormy waves and currents, so that the economic cost is high, the operation period is long, and secondly, the soil sample is difficult to disturb in the sampling process, and the authenticity of the data is influenced. In order to overcome the limitation of drilling sampling, the pore-pressure static sounding (CPTU) technology has become an indispensable core means for marine geotechnical investigation due to the advantages of high efficiency, economy, in-situ and continuous testing. However, these two types of core data have fundamental differences in spatial distribution, physical meaning and data dimension, and there are significant shortcomings in practical applications, mainly in the following aspects: (1) CPTU continuous data and drilling discrete data are split, the CPTU continuous data and the drilling discrete data have fundamental differences in spatial distribution, physical significance and data dimension, are difficult to directly fuse and utilize, the current engineering practice mainly relies on subjective experience to carry out rough and qualitative manual comparison and stratum division on the CPTU continuous data and the drilling discrete data, and the CPTU continuous data and the drilling discrete data are low in efficiency and poor in consistency. (2) Each drilling test value represents the attribute of a certain volume of soil body with a certain depth, under the conditions of sparse drilling data and dense CPTU data, single-point or single-scale characteristics are difficult to effectively represent the actual soil body characteristics, so that the characteristic expression information quantity is limited, and a method for extracting the equivalent characteristic representing the volume of soil body from a high-density CPTU signal is lacking. (3) The existing soil body parameter interpretation method based on CPTU data generally only gives a deterministic prediction result, only outputs the deterministic result, cannot quantify the confidence level of the prediction result, and is difficult to meet the requirement of high-risk decision in engineering on the reliability of the result. Disclosure of Invention The invention can overcome the defects, and provides a multi-source geological survey data fusion method based on CPTU, which solves the problems that in the prior art, discrete data of drilling and continuous data of CPTU are split, direct fusion and utilization are difficult, single characteristics are difficult to effectively represent actual soil characteristics, so that intelligent interpretation sample basis is weak, and uncertainty of a prediction result cannot be quantized. In order to achieve the above purpose, the multi-source geological survey data fusion method based on CPTU of the invention comprises the following steps: s1, acquiring soil parameters of a CPTU data set and drilling test data points, and performing fusion pretreatment; S2, searching a point with the smallest absolute depth difference from each drilling test data point in the CPTU depth sequence according to the depth value as a matching anchor point; S3, setting a dynamic sliding window along the depth direction by taking the matched anchor point as a center, and fusing and extracting three types of characteristics in the window, namely a) a CPTU data set corresponding to the anchor point as a center point characteristic, b) statistical characteristics of CPTU data in the window, c) CPTU multi-scale char