CN113033676B

CN113033676B - Stratum dividing method, device, equipment and storage medium for combined error loss of data in hole

Info

Publication number: CN113033676B
Application number: CN202110339981.2A
Authority: CN
Inventors: 吕小宁; 郭建湖; 刘铁; 孙红林; 廖进星; 张占荣; 石碧波; 张凯翔; 刘庆辉; 王卫国; 谢百义; 李慈航; 李萍; 冯光胜; 张曦; 程昊
Original assignee: China Railway Siyuan Survey and Design Group Co Ltd
Current assignee: China Railway Siyuan Survey and Design Group Co Ltd
Priority date: 2021-03-30
Filing date: 2021-03-30
Publication date: 2022-09-09
Anticipated expiration: 2041-03-30
Also published as: CN113033676A

Abstract

The embodiment of the invention discloses a method, a device, equipment and a storage medium for stratigraphic division of in-hole data joint error loss, wherein the method comprises the following steps: obtaining first picture data, first waveform data and a first stratum lithologic name corresponding to a first preset depth position in the hole; determining a discrimination error representing the combined error loss of the data in the hole according to the first picture data, the first waveform data and the first formation lithology name; performing learning training based on the discrimination error and the first stratum lithology name, and establishing a neural network model; and obtaining second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model.

Description

Stratum dividing method, device, equipment and storage medium for combined error loss of data in hole

Technical Field

The invention relates to the technical field of geological exploration, in particular to a method, a device, equipment and a storage medium for stratigraphic division of in-hole data joint error loss.

Background

In the existing engineering geological exploration, in order to find out the structure, lithology and physical mechanical parameters of the foundation, on the basis of surface engineering geological mapping, the engineering geological exploration is developed, an engineering geological drilling machine is used for drilling holes underground, rock samples or rock cores are extracted, representative soil samples or rock samples are selected for indoor test and assay, in-situ tests such as penetration in the holes or dynamic exploration and geophysical exploration are carried out if necessary, lithology identification is carried out on site, and the stratum structure is divided Complicated stratums such as fault fracture zones and the like are limited by drilling technology, and the quality of stratum division cannot be guaranteed.

Meanwhile, in engineering geological exploration, various rock and soil testing devices and methods such as in-hole optical imaging, acoustic testing, resistivity imaging and the like exist, in-hole rock and soil imaging, rock and soil physical mechanical parameters and the like can be obtained and used for core description, stratum division and the like, however, the achievement dimensions provided by each testing method are different, and the single testing method is used for data interpretation alone, so that the improvement of the stratum division quality is limited. However, no effective solution is available for this problem.

Disclosure of Invention

In view of this, embodiments of the present invention desirably provide a method, an apparatus, a device, and a storage medium for stratigraphic division with intra-pore data joint error loss.

The technical embodiment of the invention is realized as follows:

the embodiment of the invention provides a stratum dividing method for combining hole data with error loss, which comprises the following steps:

obtaining first picture data, first waveform data and a first stratum lithologic name corresponding to a first preset depth position in the hole;

determining a discrimination error representing the combined error loss of the data in the hole according to the first picture data, the first waveform data and the first formation lithology name;

performing learning training based on the discrimination error and the first stratum lithology name, and establishing a neural network model;

and obtaining second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model.

In the above solution, the obtaining of the first picture data, the first waveform data, and the first formation lithology name corresponding to the first preset depth position in the hole includes:

obtaining a plurality of waveform characteristic curves, imaging images and first stratum lithology names corresponding to a first preset depth position in the hole;

carrying out graying processing on the imaging image, and determining the first picture data corresponding to the imaging image;

presetting the plurality of waveform characteristic curves, and determining the first waveform data corresponding to the plurality of waveform characteristic curves; the dimension of the first picture data is different from the dimension of the first waveform data.

In the above scheme, the performing the preset processing on the multiple waveform characteristic curves to determine the first waveform data corresponding to the multiple waveform characteristic curves includes:

sequentially connecting the plurality of waveform characteristic curves in series to obtain one-dimensional data corresponding to the plurality of waveform characteristic curves;

and carrying out normalization processing on the one-dimensional data, and determining first waveform data corresponding to the one-dimensional data.

In the foregoing solution, the determining, according to the first picture data, the first waveform data, and the first formation lithology name, a discrimination error representing a joint error loss of the data in the borehole includes:

obtaining a first discrimination error of the first picture data on the stratum at a first preset depth position in the hole according to the first picture data and the first stratum lithology name;

obtaining a second discrimination error of the first waveform data to the stratum at a first preset depth position in the hole according to the first waveform data and the first stratum lithology name;

and determining a discrimination error representing the intra-hole data joint error loss according to the first discrimination error and the second discrimination error.

In the foregoing solution, the determining a discriminant error characterizing a joint error loss of the data in the hole according to the first discriminant error and the second discriminant error includes:

obtaining a first weight coefficient and a second weight coefficient; the first weight coefficient represents the influence degree of the first picture data on the lithological name of the first stratum; the second weight coefficient represents the influence degree of the first waveform data on the lithology name of the first stratum;

determining a discrimination error characterizing a joint error loss of the in-hole data based on the first weight coefficient, the second weight coefficient, the first discrimination error, and the second discrimination error.

In the above scheme, the performing learning training based on the discrimination error and the first formation lithology name to establish a neural network model includes:

optimizing an initial neural network model based on the discrimination error and the first formation lithology name;

the optimized initial neural network model is called a neural network model.

In the foregoing scheme, the determining, according to the second picture data, the second waveform data, and the neural network model, a second lithology name of the stratum corresponding to the second preset depth position in the hole to be identified includes:

determining lithology type probability values corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model;

and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified based on the lithology type probability value.

The embodiment of the invention provides a stratum dividing device for combining error loss of data in a hole, which comprises: the device comprises an obtaining unit, a first determining unit, a establishing unit and a second determining unit, wherein:

the obtaining unit is used for obtaining first picture data, first waveform data and a first formation lithology name corresponding to a first preset depth position in the hole;

the first determining unit is configured to determine a discrimination error representing a joint error loss of the data in the borehole according to the first picture data, the first waveform data, and the first formation lithology name obtained by the obtaining unit;

the establishing unit is used for learning and training based on the discrimination error determined by the first determining unit and the lithological name of the first stratum, and establishing a neural network model;

the second determining unit is configured to obtain second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determine a second lithology name of the stratum corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model obtained by the establishing unit.

In the above scheme, the obtaining unit is further configured to obtain a plurality of waveform characteristic curves, an imaging image, and a first formation lithology name corresponding to a first preset depth position in the hole; carrying out graying processing on the imaging image, and determining the first picture data corresponding to the imaging image; presetting the plurality of waveform characteristic curves, and determining the first waveform data corresponding to the plurality of waveform characteristic curves; the dimension of the first picture data is different from the dimension of the first waveform data.

In the above scheme, the obtaining unit is further configured to sequentially connect the multiple waveform characteristic curves in series to obtain one-dimensional data corresponding to the multiple waveform characteristic curves; and carrying out normalization processing on the one-dimensional data, and determining first waveform data corresponding to the one-dimensional data.

In the above scheme, the first determining unit is further configured to obtain, according to the first picture data and the first formation lithology name, a first discrimination error of the first picture data with respect to a formation at a first preset depth position in the hole; obtaining a second discrimination error of the first waveform data to the stratum at a first preset depth position in the hole according to the first waveform data and the first stratum lithology name; and determining a discrimination error representing the intra-hole data joint error loss according to the first discrimination error and the second discrimination error.

In the above solution, the first determining unit is further configured to obtain a first weight coefficient and a second weight coefficient; the first weight coefficient represents the influence degree of the first picture data on the lithological name of the first stratum; the second weight coefficient represents the influence degree of the first waveform data on the lithological name of the first stratum; determining a discrimination error characterizing a joint error loss of the in-hole data based on the first weight coefficient, the second weight coefficient, the first discrimination error, and the second discrimination error.

In the above solution, the establishing unit is further configured to optimize an initial neural network model based on the discrimination error and the formation name; the optimized initial neural network model is called a neural network model.

In the above scheme, the second determining unit is further configured to determine a lithologic type probability value corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model; and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified based on the lithology type probability value.

The embodiment of the invention provides a formation partitioning device for combined error loss of data in a hole, which comprises a memory and a processor, wherein the memory stores a computer program capable of running on the processor, and the processor realizes any step of the method when executing the program.

Embodiments of the present invention provide a computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements any of the steps of the above-mentioned method.

The embodiment of the invention provides a method, a device, equipment and a storage medium for stratigraphic division by combining borehole data with error loss, wherein the method comprises the following steps: obtaining first picture data, first waveform data and a first stratum lithologic name corresponding to a first preset depth position in the hole; determining a discrimination error representing the combined error loss of the data in the hole according to the first picture data, the first waveform data and the first formation lithology name; performing learning training based on the discrimination error and the lithological name of the first stratum, and establishing a neural network model; and obtaining second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model. By adopting the technical scheme of the embodiment of the invention, the discrimination error representing the combined error loss of the data in the hole is determined according to the first picture data, the first waveform data and the first stratum lithology name; performing learning training based on the discrimination error and the lithological name of the first stratum, and establishing a neural network model; obtaining second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model; the method realizes the fusion of the data in the holes with different dimensions and the error between the output probability of the fused data optimization model and the real lithology category of the stratum, is used for predicting the lithology, greatly improves the stratum division precision, reduces the artificial identification risk, and makes up the defect of single test interpretation.

Drawings

FIG. 1 is a schematic diagram of a flow chart of an implementation of a method for stratigraphic division with in-hole data joint error loss according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a plurality of waveform characteristic curves in a method for stratigraphic division with in-hole data joint error loss according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of an imaging image in a method for stratigraphic division with intra-pore data joint error loss according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of first picture data in a method for stratigraphic division with intra-borehole data joint error loss according to an embodiment of the present invention;

FIG. 5 is a schematic view of an application scenario of feature extraction of waveform data in a method for stratigraphic division with intra-borehole data joint error loss according to an embodiment of the present invention;

FIG. 6 is a schematic diagram of a neural network model established in a method for stratigraphic division with in-pore data joint error loss according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of the structure of the formation partitioning apparatus for data joint error loss in a borehole according to an embodiment of the present invention;

FIG. 8 is a diagram of a hardware entity structure of formation partitioning equipment for joint error loss of data in a borehole according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the following describes specific technical solutions of the present invention in further detail with reference to the accompanying drawings in the embodiments of the present invention. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.

The embodiment provides a formation partitioning method for data joint error loss in a hole, which is applied to formation partitioning equipment for data joint error loss in a hole, and the functions implemented by the method can be implemented by calling a program code through a processor in design equipment, where of course, the program code can be stored in a computer storage medium, and thus, the computing equipment at least comprises a processor and a storage medium.

The embodiment provides a formation partitioning method for intra-hole data joint error loss, and fig. 1 is a schematic flow chart of an implementation of the formation partitioning method for intra-hole data joint error loss according to the embodiment of the present invention, as shown in fig. 1, the method includes:

step S101: and obtaining first picture data, first waveform data and a first formation lithology name corresponding to a first preset depth position in the hole.

It should be noted that the formation partitioning method with the intra-hole data joint error loss may be a formation partitioning method with multiple intra-hole data joint error losses and/or a formation partitioning method with multiple intra-hole data joint error losses, for example, a formation partitioning method with multiple intra-hole test data joint error losses and/or a formation partitioning method with multiple intra-hole test data joint error losses. The hole may be a borehole.

Obtaining first picture data, first waveform data and a first formation lithology name corresponding to a first preset depth position in the hole can be used for obtaining a sample data set in the hole, wherein the sample data set comprises the first picture data, the first waveform data and the first formation lithology name corresponding to the first preset depth position in the hole; wherein, the hole can be one hole or a plurality of holes; the first preset depth may be any depth in the hole, and is not limited herein. As an example, the first preset depth may be any depth, for example, a certain depth. In practical application, the obtaining of the first picture data, the first waveform data and the first formation lithology name corresponding to the first preset depth position in the hole may be to obtain the first picture data, the first waveform data and the first formation lithology name corresponding to different preset depth positions in one hole respectively; or obtaining first picture data, first waveform data and first stratum lithologic names which respectively correspond to the same preset depth positions in the holes; or the obtained part is first picture data, first waveform data and first stratum lithology names which respectively correspond to different preset depth positions in one hole, and the obtained part is first picture data, first waveform data and first stratum lithology names which respectively correspond to the same preset depth positions in a plurality of holes. The first picture data may be a plurality of pieces of picture data; the first waveform data may be a plurality of waveform data; the first formation lithology name may be a plurality of formation lithology names. One preset depth position in each hole corresponds to one picture data, one waveform data and one stratum lithology name; the preset depth position is in one-to-one correspondence with one picture data, one waveform data and one stratum lithology name. The first picture data, the first waveform data and the first formation lithology name may be determined according to actual conditions, which is not limited herein. As an example, the first picture data may be grayed-out picture data; the first waveform data may be normalized vector data; the lithological name of the first stratum can be granite, limestone, sedimentary rock, magma rock and the like.

Step S102: and determining a discrimination error representing the combined error loss of the data in the hole according to the first picture data, the first waveform data and the first formation lithology name.

It should be noted that the determining, according to the first picture data, the first waveform data, and the first formation lithology name, a discrimination error representing a loss of a joint error of the in-hole data may be a first discrimination error corresponding to the in-hole data obtained according to the first picture data and the first formation lithology name, a second discrimination error corresponding to the in-hole data obtained according to the first waveform data and the first formation lithology name, and a discrimination error representing a loss of a joint error of the in-hole data determined based on the first discrimination error and the second discrimination error. Wherein, the discriminant error representing the intra-hole data joint error loss can be understood as a total discriminant error, and for convenience of understanding, the first discriminant error can be recorded as loss ₁ (ii) a The second determination error may be denoted as loss ₂ (ii) a The discriminant error representing the intra-pore data joint error loss can be recorded as loss.

Step S103: and performing learning training based on the discrimination error and the first stratum lithology name, and establishing a neural network model.

It should be noted that, the learning training is performed based on the discrimination error and the first formation lithologic name, and the establishing of the neural network model may be performed based on the discrimination error and the first formation lithologic name to perform convolutional neural network learning training, so as to establish the neural network model. The convolutional neural network comprises a convolutional layer, a pooling layer and a full-connection layer; the convolutional layer gradually performs abstract extraction of features on the data through kernel, stride and padding; the pooling layers comprise maximum pooling, average pooling, moving average pooling, L2 norm pooling and the like; and the full connection layer is mapped through formalization. In the embodiment, the cutting of the feature dimension can be completed through the convolution layer and the pooling layer, so that the features can be abstractly extracted; after the multi-layer convolution and pooling operation, features are mapped to lithology categories in the formation through fully connected layers.

Step S104: and obtaining second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model.

It should be noted that the hole to be identified may be a new drilled hole; the preset depth may be any depth, which is not limited herein, and as an example, the preset depth may be a certain depth. Obtaining second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified can be used for obtaining second picture data and second waveform data of a certain depth position of a new drilled hole; wherein the second picture data is similar in format and type to the first picture data; the second waveform data is similar in format and type to the first waveform data.

Determining a second lithology name corresponding to the second preset depth position in the hole to be recognized according to the second picture data, the second waveform data and the neural network model, wherein the second lithology name corresponding to the second preset depth position in the hole to be recognized can be obtained by inputting the second picture data and the second waveform data corresponding to the second preset depth position in the hole to be recognized into the neural network model, outputting a vector based on the neural network model after processing, and determining the second lithology name corresponding to the second preset depth position in the hole to be recognized based on the vector; and the second stratum lithology name can be understood as a predicted stratum name corresponding to the second preset depth position in the hole to be identified.

According to the stratum dividing method for the combined error loss of the data in the hole, provided by the embodiment of the invention, the judgment error representing the combined error loss of the data in the hole is determined according to the first picture data, the first waveform data and the first stratum lithology name; performing learning training based on the discrimination error and the first stratum lithology name, and establishing a neural network model; obtaining second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model; the method realizes the fusion of the data in the holes with different dimensions and the error between the output probability of the fused data optimization model and the real lithology category of the stratum, is used for predicting the lithology, greatly improves the stratum division precision, reduces the risk of manual identification, and makes up the defect of single test interpretation.

In an optional embodiment of the present invention, the obtaining the first picture data, the first waveform data, and the first formation lithology name corresponding to the first preset depth position in the hole includes: obtaining a plurality of waveform characteristic curves, imaging images and first stratum lithology names corresponding to a first preset depth position in the hole; carrying out graying processing on the imaging image, and determining the first picture data corresponding to the imaging image; presetting the plurality of waveform characteristic curves, and determining the first waveform data corresponding to the plurality of waveform characteristic curves; the dimension of the first picture data is different from the dimension of the first waveform data.

In this embodiment, obtaining the plurality of waveform characteristic curves, the imaging image and the first formation lithology name corresponding to the first preset depth position in the hole may be obtaining the plurality of waveform characteristic curves, the imaging image and the formation lithology name in the hole at a certain depth position in each hole respectively; the plurality of waveform characteristic curves may be determined according to actual conditions, and are not limited herein. As an example, the plurality of waveform characteristic curves may be elastic wave characteristic curves, and specifically, the elastic wave characteristic curves may be shear wave and/or longitudinal wave characteristic curves; the imaging image may be determined according to actual conditions, and is not limited herein. As an example, the imaged image may be an optically imaged image. The formation lithology name may be determined according to actual conditions, and is not limited herein. As an example, the formation lithology designation may be granite, limestone, sedimentary rock, magmatic rock, and the like.

For convenience of understanding, the holes are drilled holes, and the first preset depth position is a certain depth, which is determined according to actual conditions, and is not limited herein, and as an example, the certain depth position may be a drilled hole 0.5 m. FIG. 2 is a schematic diagram of a plurality of waveform characteristic curves in a method for stratigraphic division with in-pore data joint error loss according to an embodiment of the present invention; as shown in fig. 2, the plurality of waveform characteristic curves are 4 waveform characteristic curves, which are 11, 12, 13, and 14, wherein 11 and 12 are both longitudinal wave waveform characteristic curves, and 13 and 14 are both transverse wave waveform characteristic curves. FIG. 3 is a schematic diagram of an imaging image in a method for stratigraphic division with intra-borehole data joint error loss according to an embodiment of the present invention; in practical applications, the image may be an in-hole optical image of the formation at a depth in the borehole, also referred to as a raw image, as shown in fig. 3.

Graying the imaging image, determining that the first picture data corresponding to the imaging image can be used for inlaying the imaging image, graying the inlaid imaging image, and determining the first picture data corresponding to the imaging image; the mosaicing of the imaging images can be mosaicing of the imaging images into images with uniform pixel number w in the image width direction and pixel number h in the image height direction; w and h are respectively the maximum values of the pixel numbers in the width direction and the height direction of all the sample images; the pixel value of the blank portion is set to 0. And graying the embedded imaging image, and determining the first picture data corresponding to the imaging image can be to graying the embedded imaging image according to a first preset formula and determine the first picture data corresponding to the imaging image. As an example, the first preset formula may be represented by the following formula (1):

Gray(i,j)＝R(i，j)*0.299+G(i，j)*0.587+B(i，j)*0.114； (1)

wherein i is the pixel position in the image width direction, j is the pixel position in the image height direction, and R (i, j) is the red channel intensity value in the image width direction i and in the image height direction j; g (i, j) is a green channel intensity value in the height direction j of the image width direction i; b (i, j) is a blue channel intensity value in the image width direction i and the height direction j; gray (i, j) is a Gray value in the image width direction i and the height direction j.

For convenience of understanding, fig. 4 is a schematic diagram of first picture data in a stratigraphic division method with intra-pore data joint error loss according to an embodiment of the present invention; as shown in fig. 4, the first picture data may be understood as grayscale picture data.

Presetting the plurality of waveform characteristic curves, and determining the first waveform data corresponding to the plurality of waveform characteristic curves; the preset processing may include a tandem processing, a normalization processing, and the like.

The dimension of the first picture data is different from the dimension of the first waveform data, which may be understood as the dimension of the first picture data is different from the dimension of the first waveform data. As an example, the dimension of the first picture data may be two-dimensional data; the dimension of the first waveform data may be one-dimensional data.

In an optional embodiment of the present invention, the performing a preset process on the plurality of waveform characteristic curves to determine the first waveform data corresponding to the plurality of waveform characteristic curves includes: sequentially connecting the plurality of waveform characteristic curves in series to obtain one-dimensional data corresponding to the plurality of waveform characteristic curves; and carrying out normalization processing on the one-dimensional data, and determining first waveform data corresponding to the one-dimensional data.

In this embodiment, the multiple waveform characteristic curves are sequentially connected in series, and the obtaining of the one-dimensional data corresponding to the multiple waveform characteristic curves may be that the multiple waveform characteristic curves are sequentially connected in series end to form the one-dimensional data corresponding to the multiple waveform characteristic curves. The one-dimensional data may be determined according to an actual situation, and is not limited herein. As an example, the one-dimensional data may be a one-dimensional vector. For convenience of understanding, the plurality of waveform characteristic curves may be 4, each waveform characteristic curve has a length of n, and the 4 waveform characteristic curves with the length of n are connected together in series end to end in sequence to form a one-dimensional vector of (1, n × 4).

Performing normalization processing on the one-dimensional data, wherein the step of determining the first waveform data corresponding to the one-dimensional data can be performed by performing normalization processing on the one-dimensional data according to a second preset formula, and the step of determining the first waveform data corresponding to the one-dimensional data; the second preset formula may be determined according to an actual situation, and is not limited herein. As an example, when the one-dimensional data is a one-dimensional vector, the element of the one-dimensional vector is denoted as x _i The second preset formula may be represented by the following formula (2):

in the formula (2), x _i Is any one element in the one-dimensional data, min (x) _i ) Is the smallest element in the one-dimensional data, max (x) _i ) Is the largest element in the one-dimensional data, x _i The first waveform data after normalization processing in the one-dimensional data is used for ensuring that each element of the first waveform data is within the interval of (0, 1), so that the convergence rate of the model in the subsequent accelerated training process is facilitated.

In an optional embodiment of the present invention, the determining, according to the first picture data, the first waveform data, and the first formation lithology name, a discriminant error characterizing a joint error loss of the data in the borehole includes: obtaining a first discrimination error of the first picture data on the stratum at a first preset depth position in the hole according to the first picture data and the first stratum lithology name; obtaining a second discrimination error of the first waveform data to the stratum at a first preset depth position in the hole according to the first waveform data and the first stratum lithology name; and determining a discrimination error representing the intra-hole data joint error loss according to the first discrimination error and the second discrimination error.

This implementationIn an example, the first discrimination error of the first picture data with respect to the stratum at the first preset depth position in the hole is obtained according to the first picture data and the first stratum lithology name, and may be that first feature extraction is performed on the first picture data with a first preset model under the first stratum lithology name, so as to obtain a first discrimination error of the first picture data with respect to the stratum at the first preset depth position in the hole; as an example, the first preset model may be an AlexNet neural network model. The AlexNet neural network model has a deeper layer number and is excellent in the aspects of image capturing characteristics, image classification and the like. The AlexNet neural network model comprises a convolutional layer, a pooling layer and a full-connection layer. The convolutional layer gradually performs abstract extraction of features on data through kernel, stride and padding. For the first picture data, n is data x m, n is width, and m is height, the dimension after convolution can be expressed by the following equations (3) and (4), and out is output _w Is the width of the abstracted feature, out _h Is the height of the abstracted feature.

In formulae (3) and (4), kernel _w For the width of the convolution kernel, kernel _h Padding, being the height of the convolution kernel _w To fill width, padding _h To the height of the filling, stride _w For step size in the width dimension, stride _h Is the step size in the height dimension, out _w Is the width of the abstracted feature, out _h Is the height of the abstracted feature.

The pooling layers include maximum pooling, average pooling, sliding average pooling, L2 norm pooling, and the like. For geological images, key information needs to be captured, the dimensionality of the features is gradually reduced by adopting maximum pooling, and meanwhile the overfitting resistance of the model is improved.

The fully-connected layer can finish cutting of feature dimensions through the convolution layer and the pooling layer, and features can be abstractly extracted. The features are then mapped to k lithology categories by fully connected layers. Formalized, the feature dimension after abstraction is (out) _w ,out _h ) Having a post-stretching dimension of (1, out) _w *out _h ) The classification problem of k lithologies is completed by the following formulas (6) and (7). Wherein the dimension of out is (1, out) _w *out _h ) The dimension of w is (out) _w *out _h N), the dimension of b is (1, n). In the softmax formula, k represents the dimension, i.e., the number of classes classified. w and b are parameters to be trained in CNN, and are initialized by the xavier algorithm.

prob＝softmax(out*w+b) (6)

For the convenience of understanding, the construction of the AlexNet network is completed through a tensoflow framework and is exemplified.

Assuming that the size of an image input by the convolutional layer 1 is 1024 × 256, the number of convolution kernels is 96, the upper part and the lower part are on two GPUs, and each GPU respectively calculates 48 kernels; the convolution kernel size is 11 × 3, the step size is 4, and padding is 0, indicating that the edge is not extended.

The convolutional layer 2 input is the characteristic output of the previous layer of convolution, the number of convolutions is 256, and two GPUs respectively have 128 convolution kernels. The size of the convolution kernel is 5 x 48; the padding size is 2 and the step size is 1. After convolution, Local Response Normalization (LRN) is performed, and the Local Normalization is performed according to equation (8), and finally maximum pooling operation is performed, wherein the pooling size is 3 × 3 and the pooling step size is 2.

In the formula (8), the first and second groups,

represented is the output of the ReLU activation function at the (x, y) position of the ith convolution kernel, ngr denotes

N represents the number of convolution kernels.

Representing the result of local normalization.

The convolutional layer 3 input is the output of the second layer, the number of convolutional kernels is 384, the convolutional kernel size is 3 × 256, and the padding size is 1.

The convolutional layer 4 input is the output of the third layer, the number of convolutional kernels is 384, the size of the convolutional kernels is 3 x 3, and the padding size is 1.

The convolution layer 5 input is the output of the fourth layer, the number of convolution kernels is 256, the convolution kernel size is 3 x 3, the filling size is 1, the maximum pooling operation is directly performed after convolution, the pooling size is 3 x 3, and the step size is 2.

Layers 6,7 and 8 are all connected layers, the number of neurons in each layer is 4096, and finally the output softmax is k. k is the total number of lithology classes that need to be classified. In the fully-connected layer, a dropout mechanism is adopted, and partial nerve units in the fully-connected layer are randomly discarded to prevent overfitting of the model. If the dropout mechanism is not adopted, the calculation formula of the network adopts (9) and (10), wherein l represents the current layer number and f represents an activation function.

After the dropout mechanism is adopted, the calculation formula of the network adopts (11), (12), (13) and (14), as follows, wherein the Bernoulli function randomly generates vectors of 0 and 1 with probability p, and y ^l The multiplication achieves the effect of discarding part of the neural units.

Through the operations of convolution, pooling and full-connection layers, the probability under k lithologies can be obtained, and the training optimization of the model is completed by continuously optimizing the error between the model output probability and the real lithology category through the following formula (15). Wherein ns is the number of samples in the training data set, f-function represents the AlexNet network structure, L2 regularization is used to prevent overfitting, γ is the weight,

representing the feature of the ith sample lithology picture.

For convenience of understanding, the feature extraction may be performed on the picture data by an AlexNet network, and the AlexNet network may be abstracted as y in consideration of the inclusion of a convolutional layer, a pooling layer, a nonlinear activation function, and the like ₁ ＝w ₁ *x ₁ +b ₁ Wherein w is ₁ And b ₁ Weights and biases, x, representing AlexNet network parameters ₁ Representing first picture data, y ₁ Is the output of the AlexNet network.

In practical application, the first picture data under the first stratum lithology name may fluctuate, and in order to make the AlexNet network more optimal, different training samples are required under the first stratum lithology name, that is, different training samples are required under the same lithology category, so that the "multi-knowledge" capability, that is, the capability of discriminating the lithology, is enhanced. Therefore, when data are collected for a certain rock, operations such as removing the data set are not performed, all data in the data set are all involved in training, and the robustness of the AlexNet network is enhanced. For example, the first picture data and the first stratigraphic lithology name may be input to an AlexNet neural network model to obtain a first discrimination error of the first picture data with respect to a stratigraphic at a first preset depth position in the hole, the process is a repeated iterative process, an early stopping method (early stopping) is used to determine an optimization termination criterion of the AlexNet neural network model, a training data set corresponding to the first picture data and the first stratigraphic lithology name may be divided into a training set validation set, training is performed only on the training set, and when the discrimination error of the AlexNet neural network model on the validation set is worse than that of the last training set in each period, the training is stopped, and a parameter in the last iteration result is used as a final parameter, so as to obtain a final AlexNet neural network model. Namely, the model which best performs on the verification set is the final AlexNet neural network model.

Obtaining a second discrimination error of the first waveform data on the stratum at the first preset depth position in the hole according to the first waveform data and the lithological name of the first stratum, wherein the second discrimination error can be obtained by performing second feature extraction on the waveform data by using a second preset model; the second predetermined model may be determined according to actual conditions, and as an example, the second predetermined model may be a fully-connected neural network (DNN) model. The fully-connected neural network model mainly comprises an input layer, an nl layer hidden layer and an output layer (nl is a hyper-parameter).

For convenience of understanding, an application scene schematic diagram for feature extraction of waveform data is illustrated here; FIG. 5 is a schematic view of an application scenario of feature extraction of waveform data in a method for stratigraphic division with in-pore data joint error loss according to an embodiment of the present invention; as shown in fig. 5, the plurality of waveform characteristic curves are 4, and 21 is 4 waveform characteristic curves; 22 is an input layer of the fully-connected neural network model, and is used for inputting corresponding first waveform data after 4 waveform characteristic curves are sequentially connected in series and subjected to normalization processing, namely, a (1, 4 × m) vector with 4 × m values in total serves as data of the input layer; 23. 24, 25 are hidden layers of the fully-connected neural network model, each neural unit in the hidden layer is a function, and the features are abstracted by the following formula (4), and the result is used as the input of the next layer. Where ω and b are both parameters that require training. 26 is the output layer of the fully-connected neural network model; the output layer can map the matrix to k lithology categories by specifying the size of the matrix omega and b, and the probability distribution calculation of k lithologies is completed. The lithology classification of the fully-connected neural network model is completed through the characteristic abstraction of the hidden layer and the output layer in the fully-connected neural network model, and the error of the current training turn can be obtained through the loss calculation of the cross entropy of the fully-connected neural network model and the existing label, so that the training aim is to continuously reduce the error of the model on a data set. The error calculation formula (16) is as follows:

wherein ns is the number of samples of the training data set, q function represents a fully-connected neural network, and q (x) is a judgment result vector of the model to the current sample i, namely the model is predicted to be a lithologic type;

representing the characteristic of the ith sample elastic wave;

to facilitate understanding, it is illustrated herein that for waveform data, feature extraction may be performed by a fully connected neural network. Given that the fully-connected neural network contains many neural units in the hidden layer, it can be abstracted as y ₂ ＝w ₂ *x ₂ +b ₂ Wherein w is ₂ And b ₂ Weights and offsets, x, representing parameters of a fully connected network ₂ Representing elastic wave curve data, y ₂ Is the output.

In practical application, the first waveform data under the first stratum lithology name may fluctuate, and in order to make the fully-connected neural network model more optimal, different training samples are required under the first stratum lithology name, that is, different training samples are required under the same lithology category, so that the "multi-knowledge" capability, that is, the capability of discriminating lithology, is enhanced. Therefore, when data are collected on a certain rock, operations such as removing the data set are not performed, all data under the type are all involved in training, and the robustness of the full-connection neural network model is enhanced. For example, the first waveform data and the first formation lithology name may be input to a fully-connected neural network model to obtain a second discrimination error of the first waveform data for a formation at a first preset depth position in the hole, the process is a repeated iteration process, an early-stopping method may be used to determine an optimization termination criterion of the fully-connected neural network model, a training data set corresponding to the first waveform data and the first formation lithology name may be divided into a training set validation set, training is performed only on the training set, training is stopped when the discrimination error of the fully-connected neural network model on the validation set is calculated in each cycle and is worse than the last training result, and a parameter in the last iteration result is used as a final parameter to obtain a final fully-connected neural network model. Namely, the model which best represents on the verification set is the final fully-connected neural network model.

For convenience of understanding, for example, the waveform data is taken as the elastic wave data, and the elastic wave data of granite may have various cases [0,2,1, …,1,0], [0,2,2, …,1,0],. and so on. Although different data exist, the neural network can be optimized aiming at each piece of data, firstly [0,2,1, …,1,0] is input into the fully-connected neural network model, the parameters of the fully-connected neural network model are updated through reverse transmission after errors are obtained, and then [0,2,2, …,1,0] is input into the network.

In an optional embodiment of the present invention, the determining a discriminant error characterizing a joint error loss of the data in the hole according to the first discriminant error and the second discriminant error comprises: obtaining a first weight coefficient and a second weight coefficient; the first weight coefficient represents the influence degree of the first picture data on the lithological name of the first stratum; the second weight coefficient represents the influence degree of the first waveform data on the lithology name of the first stratum; determining a discrimination error characterizing a joint error loss of the in-hole data based on the first weight coefficient, the second weight coefficient, the first discrimination error, and the second discrimination error.

In this embodiment, the first weight coefficient represents a degree of influence of the first picture data on the lithological name of the first formation; the second weight coefficient represents the influence degree of the first waveform data on the lithology name of the first stratum; the method mainly considers that under different scenes, the influence degrees of picture data and waveform data on the formation lithology judgment are different, for example, in the scene that the picture data can reflect the type of the formation lithology better, a first weight coefficient of the picture data for judging the formation lithology is properly adjusted to be larger, and a second weight coefficient of the waveform data for judging the formation lithology is adjusted to be smaller; in the scene that the waveform data can reflect the lithological character type of the stratum, the first weight coefficient of the lithological character of the stratum is judged by properly reducing the picture data, and the second weight coefficient of the lithological character of the stratum is judged by increasing the waveform data. In practical applications, the sum of the first weight coefficient and the second weight coefficient is one, and for convenience of understanding, the first weight coefficient may be represented as α, the second weight coefficient may be represented as β, and α and β satisfy α + β ═ 1.

Determining the discriminant error representing the joint error loss of the in-hole data based on the first weight coefficient, the second weight coefficient, the first discriminant error, and the second discriminant error may be determining the discriminant error representing the joint error loss of the in-hole data according to a third predetermined formula based on the first weight coefficient, the second weight coefficient, the first discriminant error, and the second discriminant error; for ease of understanding, the first weight coefficient may be denoted as α, the second weight coefficient may be denoted as β, and α and β satisfy α + β ═ 1; the first discriminant error can be recorded as loss ₁ (ii) a The second discrimination error may be denoted as loss ₂ (ii) a The discriminant error characterizing the joint error loss of the data in the hole can be denoted as loss. As an example, the third preset formula may be represented by the following formula (17):

loss＝α×loss ₁ +β×loss ₂ (17)

illustrating by a piece of data, the error of a piece of data during training can be represented by the following equation (18):

loss＝α×-y ⁽ⁱ⁾ log(w ₁ *x ₁ +b ₁ )+β×-y ⁽ⁱ⁾ log(w ₂ *x ₂ +b ₂ ) (18)

the formula is simplified to the formula (19):

loss＝-y ⁽ⁱ⁾ ×[α×log(w ₁ *x ₁ +b ₁ )+β×log(w ₂ *x ₂ +b ₂ )] (19)

during the training process, the variation of the parameters is expressed by the following equations (20), (21), (22), (23):

the learning rate (learning rate) during training is lr. It is calculated that after one piece of data is trained, the parameters of the model are expressed by the following formulas (24), (25), (26) and (27):

in practical application, it is assumed that the first picture data is marked as x ⁽ⁱ⁾ The first discrimination error is recorded as y ⁽ⁱ⁾ Then y is ⁽ⁱ⁾ Can be expressed by the following equation (28):

y ⁽ⁱ⁾ ＝AlexNet(x ⁽ⁱ⁾ ) (28)

in the formula (28), i represents any picture data, y ⁽ⁱ⁾ For a vector, each element in the vector may correspond to the probability of the formation lithology name, respectively. E.g. y ⁽ⁱ⁾ ＝[0.7,0.1,0.1,0.1]Respectively corresponding to [ granite, limestone, sedimentary rock, magmatic rock ]]Can be understood as the probability that the picture data corresponds to the granite0.7, corresponding to limestone 0.1, corresponding to sedimentary rock 0.1, corresponding to magmatic rock 0.1.

Assume that the waveform data is denoted as x ^(j) And the second judgment error is recorded as y ^(j) Then y is ^(j) Can be expressed by the following formula (29):

y ^(j) ＝DNN(x ^(j) ) (29)

in the formula (29), j represents arbitrary waveform data, y ^(j) Each element in the vector can respectively correspond to the probability of the formation lithology name, and the dimensionality is the number of formation lithology categories set during training. E.g. y ^(j) ＝[0.4,0.3,0.2,0.1]Respectively corresponding to [ granite, limestone, sedimentary rock, magmatic rock ]]It can be understood that the waveform data has a probability of 0.4 for granite, 0.3 for limestone, 0.2 for sedimentary rock, and 0.1 for magmatic rock.

The first discrimination error is recorded as y ⁽ⁱ⁾ And the second judgment error is recorded as y ^(j) And the discrimination error representing the intra-pore data joint error loss is denoted as Y, the third preset formula can be expressed by the following formula (30):

Y＝ay ⁽ⁱ⁾ +βy ^(j) (30)

for convenience of understanding, in a scene that the picture data can reflect the lithology type of the stratum, a first weight coefficient for judging the lithology of the stratum by the picture data should be appropriately increased, and a second weight coefficient for judging the lithology of the stratum by the waveform data should be decreased, wherein the first weight coefficient α is assumed to be 0.7; the second weight coefficient beta is 0.3, y ⁽ⁱ⁾ ＝[0.7,0.1,0.1,0.1]Respectively corresponding to [ granite, limestone, sedimentary rock, magma rock ]]；y ^(j) ＝[0.4,0.3,0.2,0.1]Respectively corresponding to [ granite, limestone, sedimentary rock, magmatic rock ]](ii) a Y may be represented by the following formula (31):

Y＝0.7×y ⁽ⁱ⁾ +0.3×y ^(j) ＝[0.7×0.7+0.4×0.3,0.1×0.7+0.3×0.3,0.1×0.7+0.2×0.3,0.1×0.7+0.1×0.3]＝[0.61,0.16,0.13,0.1]

(31)

in an optional embodiment of the present invention, the performing learning training based on the discrimination error and the lithology name of the first formation to establish a neural network model includes: optimizing an initial neural network model based on the discrimination error and the first formation lithology name; the optimized initial neural network model is called a neural network model.

In this embodiment, optimizing the initial neural network model based on the discrimination error and the first formation lithology name may be optimizing the initial neural network model by an optimizer based on the discrimination error and the first formation lithology name; wherein the optimizer may be an adam optimizer; the initial neural network model may be a convolutional neural network model. In practical application, the optimization process is a repeated iteration process, and an early stopping method (early stopping) can be used to determine an initial neural network model optimization termination standard, a training data set corresponding to the discriminant error and the first formation lithology name can be divided into a training set and a verification set, training is performed only on the training set, error loss of the initial neural network model on the verification set is calculated in each period, training is stopped when the joint error loss of the initial neural network model on the verification set is worse than that of the previous training result, and parameters in the previous iteration result are used as optimization parameters to optimize the initial neural network model. Namely, the model which best performs on the verification set is the optimized initial neural network model.

The optimized initial neural network model is called a neural network model, and the optimized initial neural network model can be called a neural network model; in particular, the optimized convolutional neural network model may be referred to as an available neural network model.

For convenience of understanding, fig. 6 is a schematic diagram illustrating a neural network model established in a stratigraphic division method for intra-pore data joint error loss according to an embodiment of the present invention; 31 is an imaged image; 32, performing mosaic imaging on the imaging image; performing graying processing on the embedded imaging image to determine the first picture data corresponding to the imaging image; 34 is a classification neural network for training the first picture data; 35 is a first discrimination error obtained by the first picture data through a classification neural network; 36 are four waveform characteristic curves; the method comprises the following steps that 37, four waveform characteristic curves are sequentially connected in series to obtain first waveform data corresponding to the waveform characteristic curves; 38 is a fully-connected neural network that trains the first waveform data; 39 is a second judgment error obtained by the first waveform data through a full-connection neural network; 40, joint learning, specifically, a process of obtaining a discrimination error through the joint learning of the first discrimination error and the second discrimination error, and then a process of optimizing the discrimination error and the first formation lithology name through an adam optimizer; 41 is a neural network model.

In an optional embodiment of the present invention, the determining, according to the second picture data, the second waveform data, and the neural network model, a second lithology name of the second stratum corresponding to the second preset depth position in the hole to be identified includes: determining a lithology type probability value corresponding to the second preset depth position in the hole to be recognized according to the second picture data, the second waveform data and the neural network model; and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified based on the lithology type probability value.

In this embodiment, determining the lithology type probability value corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model may be inputting the second picture data and the second waveform data into the neural network model, and by calculating forward transmission of the neural network model, a vector may be output, where a value at each position in the vector represents a lithology type probability belonging to a corresponding subscript.

The lithology type probability value is used for determining the lithology type probability value, the second stratum lithology name corresponding to the second preset depth position can be selected to be the largest lithology type probability value based on the lithology type probability value, and the stratum name corresponding to the largest lithology type probability value is called the lithology name of the second stratum corresponding to the second preset depth position in the hole to be identified.

For convenience of understanding, it is assumed that, according to the second picture data, the second waveform data and the neural network model, lithology type probability values corresponding to the second preset depth position in the hole to be identified are determined to be [0.61,0.16,0.13,0.1], which respectively correspond to [ granite, limestone, sedimentary rock, and magma ], and then the granite corresponding to 0.61 is a second formation lithology name corresponding to the second preset depth position in the hole to be identified.

According to the stratum dividing method for the combined error loss of the data in the hole, provided by the embodiment of the invention, the judgment error representing the combined error loss of the data in the hole is determined according to the first picture data, the first waveform data and the first stratum lithology name; performing learning training based on the discrimination error and the lithological name of the first stratum, and establishing a neural network model; obtaining second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model; the method realizes the fusion of the data in the holes with different dimensions and the error between the output probability of the fused data optimization model and the real lithology category of the stratum, is used for predicting the lithology, greatly improves the stratum division precision, reduces the artificial identification risk, and makes up the defect of single test interpretation.

In this embodiment, a formation partitioning device for intra-hole data joint error loss is provided, and fig. 7 is a schematic structural diagram of a formation partitioning device for intra-hole data joint error loss according to an embodiment of the present invention, as shown in fig. 7, the device 200 includes: an obtaining unit 201, a first determining unit 202, a establishing unit 203 and a second determining unit 204, wherein:

the obtaining unit 201 is configured to obtain first picture data, first waveform data, and a first formation lithology name corresponding to a first preset depth position in the hole;

the first determining unit 202 is configured to determine a discrimination error representing a joint error loss of the intra-pore data according to the first picture data and the first waveform data obtained by the obtaining unit 201;

the establishing unit 203 is configured to perform learning training based on the discrimination error determined by the first determining unit 202 and the first formation lithology name, and establish a neural network model;

the second determining unit 204 is configured to obtain second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determine a second lithology name of the stratum corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data, and the neural network model obtained by the establishing unit 203.

In other embodiments, the obtaining unit 201 is further configured to obtain a plurality of waveform characteristic curves, an imaging image, and a first formation lithology name corresponding to a first preset depth position in the hole; carrying out graying processing on the imaging image, and determining the first picture data corresponding to the imaging image; presetting the plurality of waveform characteristic curves, and determining the first waveform data corresponding to the plurality of waveform characteristic curves; the dimension of the first picture data is different from the dimension of the first waveform data.

In other embodiments, the obtaining unit 201 is further configured to sequentially connect the plurality of waveform characteristic curves in series to obtain one-dimensional data corresponding to the plurality of waveform characteristic curves; and carrying out normalization processing on the one-dimensional data, and determining first waveform data corresponding to the one-dimensional data.

In other embodiments, the first determining unit 202 is further configured to obtain a first discrimination error of the first picture data with respect to a stratum at a first preset depth position in the hole according to the first picture data and the first stratum lithology name; obtaining a second discrimination error of the first waveform data to the stratum at a first preset depth position in the hole according to the first waveform data and the first stratum lithology name; and determining a discrimination error representing the intra-hole data joint error loss according to the first discrimination error and the second discrimination error.

In other embodiments, the first determining unit 202 is further configured to obtain a first weighting factor and a second weighting factor; the first weight coefficient represents the influence degree of the first picture data on the lithological name of the first stratum; the second weight coefficient represents the influence degree of the first waveform data on the lithology name of the first stratum; determining a discrimination error characterizing a joint error loss of the in-hole data based on the first weight coefficient, the second weight coefficient, the first discrimination error, and the second discrimination error.

In other embodiments, the establishing unit 203 is further configured to optimize an initial neural network model based on the discrimination error and the formation name; the optimized initial neural network model is called a neural network model.

In other embodiments, the second determining unit 204 is further configured to determine a lithology type probability value corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model; and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified based on the lithology type probability value.

The above description of the apparatus embodiments, similar to the above description of the method embodiments, has similar beneficial effects as the method embodiments. For technical details not disclosed in the embodiments of the apparatus according to the invention, reference is made to the description of the embodiments of the method according to the invention for understanding.

It should be noted that, in the embodiment of the present invention, if the above stratigraphic division method with combined error loss of pore data is implemented in the form of a software functional module and is sold or used as a standalone product, it may also be stored in a computer readable storage medium. With this understanding, technical embodiments of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a stratigraphic division device (which may be a personal computer, a server, or a network device) to perform all or part of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read Only Memory (ROM), a magnetic disk, or an optical disk. Thus, embodiments of the invention are not limited to any specific combination of hardware and software.

Correspondingly, the stratum partitioning device for joint error loss of data in a hole provided by the embodiment of the invention comprises a memory and a processor, wherein the memory stores a computer program capable of running on the processor, and the processor executes the program to realize the steps in the stratum partitioning method for joint error loss of data in a hole provided by the embodiment.

Correspondingly, the embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps in the method for stratigraphic division with combined error loss of in-hole data provided by the above-mentioned embodiment.

Here, it should be noted that: the above description of the storage medium and device embodiments is similar to the description of the method embodiments above, with similar advantageous effects as the method embodiments. For technical details not disclosed in the embodiments of the storage medium and the apparatus according to the invention, reference is made to the description of the embodiments of the method according to the invention.

It should be noted that fig. 8 is a schematic structural diagram of a hardware entity of a formation partitioning apparatus for joint error loss of data in a hole in an embodiment of the present invention, as shown in fig. 8, a hardware entity of the formation partitioning apparatus 300 includes: a processor 301 and a memory 303, optionally, the stratigraphic division apparatus 300 may further comprise a communication interface 302.

It will be appreciated that the memory 303 can be either volatile memory or nonvolatile memory, and can include both volatile and nonvolatile memory. Among them, the nonvolatile Memory may be a Read Only Memory (ROM), a Programmable Read Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a magnetic random access Memory (FRAM), a Flash Memory (Flash Memory), a magnetic surface Memory, an optical disk, or a Compact Disc Read-Only Memory (CD-ROM); the magnetic surface storage may be disk storage or tape storage. Volatile Memory can be Random Access Memory (RAM), which acts as external cache Memory. By way of illustration and not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), Enhanced Synchronous Dynamic Random Access Memory (Enhanced DRAM), Synchronous Dynamic Random Access Memory (SLDRAM), Direct Memory (DRmb Access), and Random Access Memory (DRAM). The memory 303 described in connection with the embodiments of the invention is intended to comprise, without being limited to, these and any other suitable types of memory.

The method disclosed by the above embodiment of the present invention may be applied to the processor 301, or implemented by the processor 301. The processor 301 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware or instructions in the form of software in the processor 301. The Processor 301 may be a general purpose Processor, a Digital Signal Processor (DSP), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like. Processor 301 may implement or perform the methods, steps, and logic blocks disclosed in embodiments of the present invention. The general purpose processor may be a microprocessor or any conventional processor or the like. The steps of the method disclosed by the embodiment of the invention can be directly implemented by a hardware decoding processor, or can be implemented by combining hardware and software modules in the decoding processor. The software modules may be located in a storage medium located in the memory 303, and the processor 301 reads the information in the memory 303 and performs the steps of the aforementioned methods in conjunction with its hardware.

In an exemplary embodiment, the stratigraphic division Device may be implemented by one or more Application Specific Integrated Circuits (ASICs), DSPs, Programmable Logic Devices (PLDs), Complex Programmable Logic Devices (CPLDs), Field Programmable Gate Arrays (FPGAs), general purpose processors, controllers, Micro Controllers (MCUs), microprocessors (microprocessors), or other electronic components for performing the foregoing methods.

In the several embodiments provided in the present invention, it should be understood that the disclosed method and apparatus may be implemented in other manners. The above-described device embodiments are merely illustrative, for example, the division of the unit is only a logical functional division, and there may be other division ways in actual implementation, such as: multiple units or components may be combined, or may be integrated into another observation, or some features may be omitted, or not performed. In addition, the communication connections between the components shown or discussed may be through interfaces, indirect couplings or communication connections of devices or units, and may be electrical, mechanical or other.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, that is, may be located in one place, or may be distributed on a plurality of network units; some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

Those of ordinary skill in the art will understand that: all or part of the steps for realizing the method embodiments can be completed by hardware related to program instructions, the program can be stored in a computer readable storage medium, and the program executes the steps comprising the method embodiments when executed; and the aforementioned storage medium includes: various media that can store program codes, such as a removable Memory device, a Read-Only Memory (ROM), a magnetic disk, or an optical disk.

Alternatively, the integrated unit according to the embodiment of the present invention may be stored in a computer-readable storage medium if it is implemented in the form of a software functional unit and sold or used as a separate product. Based on such understanding, the technical embodiments of the present invention or portions thereof that contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a stratigraphic division device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the methods described in the embodiments of the present invention. And the aforementioned storage medium includes: various media that can store program code, such as removable storage devices, ROMs, magnetic or optical disks, etc.

The method, the apparatus, the device and the computer storage medium for stratigraphic division equipment with in-hole data joint error loss described in the embodiment of the present invention are only examples of the embodiments of the present invention, but are not limited thereto, and the method, the apparatus, the device and the computer storage medium for stratigraphic division equipment with in-hole data joint error loss are all within the scope of the present invention.

It should be appreciated that reference throughout this specification to "one embodiment" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. It should be understood that, in various embodiments of the present invention, the sequence numbers of the above-mentioned processes do not mean the execution sequence, and the execution sequence of each process should be determined by its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention. The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element identified by the phrase "comprising an … …" does not exclude the presence of other identical elements in the process, method, article, or apparatus that comprises the element.

The above description is only an embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of changes or substitutions within the technical scope of the present invention, and all such changes or substitutions are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the appended claims.

Claims

1. A method for stratigraphic division with in-hole data joint error loss, the method comprising:

determining a discrimination error representing the combined error loss of the data in the hole according to the first picture data, the first waveform data and the first stratum lithology name;

performing learning training based on the discrimination error and the lithological name of the first stratum, and establishing a neural network model;

obtaining second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model;

wherein determining a discrimination error characterizing a joint error loss of the borehole data according to the first picture data, the first waveform data, and the first formation lithology name comprises:

obtaining a second judgment error of the first waveform data to the stratum at a first preset depth position in the hole according to the first waveform data and the first stratum lithology name;

2. The method of claim 1, wherein the obtaining the first picture data, the first waveform data and the first formation lithology name corresponding to the first preset depth position in the hole comprises:

obtaining a plurality of waveform characteristic curves, imaging images and first stratum lithological names corresponding to a first preset depth position in the hole;

3. The method according to claim 2, wherein the pre-processing the plurality of waveform characteristic curves to determine the first waveform data corresponding to the plurality of waveform characteristic curves comprises:

4. The method of claim 1, wherein determining a discriminant error characterizing a joint error loss of the intra-hole data from the first discriminant error and the second discriminant error comprises:

and determining a discrimination error representing the intra-pore data joint error loss based on the first weight coefficient, the second weight coefficient, the first discrimination error and the second discrimination error.

5. The method of claim 1, wherein the learning training based on the discrimination error and the first formation lithology name, building a neural network model, comprises:

the optimized initial neural network model is called a neural network model.

6. The method according to claim 1, wherein the determining, according to the second picture data, the second waveform data and the neural network model, a second lithology name corresponding to the second preset depth position in the hole to be identified comprises:

7. An in-hole data joint error-loss stratigraphic division apparatus, said apparatus comprising: the device comprises an obtaining unit, a first determining unit, a establishing unit and a second determining unit, wherein:

the establishing unit is used for performing learning training based on the discrimination error determined by the first determining unit and the lithological name of the first stratum, and establishing a neural network model;

the second determining unit is configured to obtain second picture data and second waveform data corresponding to a second preset depth position in the hole to be identified, and determine a second lithology name of the stratum corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model obtained by the establishing unit;

the first determining unit is further configured to obtain a first discrimination error of the first picture data with respect to a formation at a first preset depth position in the hole according to the first picture data and the first formation lithology name; obtaining a second discrimination error of the first waveform data to the stratum at a first preset depth position in the hole according to the first waveform data and the first stratum lithology name; and determining a discrimination error representing the intra-hole data joint error loss according to the first discrimination error and the second discrimination error.

8. The apparatus according to claim 7, wherein the obtaining unit is further configured to obtain a plurality of waveform characteristic curves, an imaging image and a first formation lithology name corresponding to a first preset depth position in the hole; carrying out graying processing on the imaging image, and determining the first picture data corresponding to the imaging image; presetting the plurality of waveform characteristic curves, and determining the first waveform data corresponding to the plurality of waveform characteristic curves; the dimension of the first picture data is different from the dimension of the first waveform data.

9. The apparatus according to claim 8, wherein the obtaining unit is further configured to sequentially connect the plurality of waveform characteristic curves in series to obtain one-dimensional data corresponding to the plurality of waveform characteristic curves; and carrying out normalization processing on the one-dimensional data, and determining first waveform data corresponding to the one-dimensional data.

10. The apparatus according to claim 7, wherein the first determining unit is further configured to obtain a first weighting factor and a second weighting factor; the first weight coefficient represents the influence degree of the first picture data on the lithological name of the first stratum; the second weight coefficient represents the influence degree of the first waveform data on the lithology name of the first stratum; determining a discrimination error characterizing a joint error loss of the in-hole data based on the first weight coefficient, the second weight coefficient, the first discrimination error, and the second discrimination error.

11. The apparatus of claim 7, wherein the building unit is further configured to optimize an initial neural network model based on the discrimination error and the formation name; the optimized initial neural network model is called a neural network model.

12. The apparatus according to claim 7, wherein the second determining unit is further configured to determine a lithology type probability value corresponding to the second preset depth position in the hole to be identified according to the second picture data, the second waveform data and the neural network model; and determining a second stratum lithology name corresponding to the second preset depth position in the hole to be identified based on the lithology type probability value.

13. An in-hole data joint error-loss stratigraphic division apparatus comprising a memory and a processor, said memory storing a computer program executable on the processor, wherein said processor when executing said program implements the steps in the method of any of claims 1 to 6.

14. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.