APPLICATION OF SURVEYING AND
MAPPING TECHNOLOGY BASED ON DEEP
LEARNING MODEL IN PETROLEUM
GEOLOGICAL EXPLORATION
Sheng Sun*
School of Modern Service Management, Shandong Youth University of Political
Science, Jinan, Shandong, 250103, China
sunsdjn2013@163.com
Ping Shu
School of Philosophy and Social Development, Shandong University, Jinan,
Shandong, 250103, China
Reception: 21/11/2022 Acceptance: 17/01/2023 Publication: 04/02/2023
Suggested citation:
S., Shang and S., Ping (2023). Application of surveying and mapping
technology based on deep learning model in petroleum geological
exploration. 3C Tecnología. Glosas de innovación aplicada a la pyme, 12(1),
159-174. https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
159
ABSTRACT
Surveying and mapping technology is one of the key technologies used in petroleum
geological exploration and has made significant contributions to geological
exploration. However, with the development of science and technology, traditional
surveying and mapping technology has low work efficiency and poor information
accuracy, which limits its application. This study proposes a surveying and mapping
technology based on the 1DCNN-LSTM deep learning model. Through feature
selection and feature optimization, the important features extracted by 1DCNN are
predicted through LSTM, and the development direction of surveying and mapping
technology is optimized and predicted to promote the development of new surveying
and mapping technologies. application. By using the orthogonal test to optimize the
input factors, determine the relative order of the influence of the factors, and use the
1DCNN-LSTM and BP neural network to train and verify the input factors respectively.
The research results show that 1DCNN-LSTM has higher prediction accuracy, and the
prediction accuracy is The results show that the 1DCNN-LSTM deep learning model
used in the optimization of petroleum geological exploration and mapping technology
in this study has strong practical significance.
KEYWORDS
1DCNN-LSTM; Mapping technology; Deep learning model; Neural network;
Optimization.
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
160
ABSTRACT
Surveying and mapping technology is one of the key technologies used in petroleum
geological exploration and has made significant contributions to geological
exploration. However, with the development of science and technology, traditional
surveying and mapping technology has low work efficiency and poor information
accuracy, which limits its application. This study proposes a surveying and mapping
technology based on the 1DCNN-LSTM deep learning model. Through feature
selection and feature optimization, the important features extracted by 1DCNN are
predicted through LSTM, and the development direction of surveying and mapping
technology is optimized and predicted to promote the development of new surveying
and mapping technologies. application. By using the orthogonal test to optimize the
input factors, determine the relative order of the influence of the factors, and use the
1DCNN-LSTM and BP neural network to train and verify the input factors respectively.
The research results show that 1DCNN-LSTM has higher prediction accuracy, and the
prediction accuracy is The results show that the 1DCNN-LSTM deep learning model
used in the optimization of petroleum geological exploration and mapping technology
in this study has strong practical significance.
KEYWORDS
1DCNN-LSTM; Mapping technology; Deep learning model; Neural network;
Optimization.
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
PAPER INDEX
ABSTRACT
KEYWORDS
1. INTRODUCTION
2. CORRELATION MODEL THEORY
2.1. Long Short Term Memory (LSTM)
2.2. One-Dimensional Convolutional Neural Network (1DCNN)
2.3. DCNN-LSTM network model construction
2.3.1. Model building
2.3.2. Model parameter determination and data preprocessing
3. TEST AND RESULT ANALYSIS
3.1. Orthogonal test to optimize input factors
3.2. DCNN-LSTM model prediction and result analysis
3.3. Application of New Technology of Surveying and Mapping in Petroleum
Geological Exploration
3.3.1. Field surveying and mapping
3.3.2. Cloth net
3.3.3. Dynamic real-time mapping
3.3.4. Geodetic control network
4. CONCLUSION
5. CONFLICT OF INTEREST
REFERENCES
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
161
1. INTRODUCTION
The traditional surveying and mapping technology has low work efficiency and
requires a lot of human resources to complete the survey work, but the final
information accuracy is poor [1]. Under the rapid development of modern social
science and technology, new surveying and mapping technologies with better
performance have been developed. Digital surveying and mapping are the main
forms, and they have been applied to the fields of hydrogeology, petroleum
engineering, and other fields, and played their due role in the field of geological
exploration. It greatly promotes the development of the exploration field, especially the
geological exploration, and the accuracy requirements of the surveying and mapping
technology are increasing day by day [2-3]. Therefore, it is extremely necessary to
optimize the surveying and mapping technology.
At present, there is little research on the optimization of surveying and mapping
technology, and it is of great significance to use deep learning models to optimize
surveying and mapping technology. In recent years, some scholars have done a lot of
work. Lu X H [4] established a prediction model of surveying and mapping technology
through regression analysis using cutting parameters as independent variables.
Beruvides [5] used the vibration signal sent out during the training process to establish
the prediction model of surveying and mapping technology by using the adaptive
neuro-fuzzy inference system and obtained a higher fitting index and better
generalization ability. Some researchers [6] used the improved particle swarm
algorithm to optimize the node selection of the hidden layer of the BP network and
established the prediction model of the surveying and mapping technology. In
addition, some researchers [7] proposed a parameter synchronization optimization
algorithm for GA signal feature recognition and mapping prediction, established a GA-
WPT-ELM prediction model, and obtained high prediction accuracy. With the
development of artificial intelligence, deep learning makes data processing and results
in prediction more efficient and accurate [8]. Its long short-term memory (LSTM)
neural network algorithm improves the gradient disappearance problem of traditional
recurrent neural networks (RNN) and provides a new method for the prediction of
sequence data. Prediction problems are applied in the field of new technologies [9].
Wang M W et al. [10] established a long short-term memory model and realized the
prediction of the wear of surveying and mapping tools by taking advantage of its
advantages of solving the accumulation effect. Some researchers [11] proposed a
traditional surveying and mapping stage identification model based on a deep LSTM
neural network, which can more accurately reflect the wear state of surveying and
mapping compared with traditional machine learning methods. Yu [12] proposed a
state recognition method based on LSTM, which has higher recognition accuracy than
BP neural network algorithm and SVM algorithm..
Although the LSTM network solves the problem of vanishing gradients, it has poor
performance for batch sequence data processing, resulting in lower accuracy of the
model in result prediction [13-14]. In this paper, taking petroleum surveying and
mapping technology as the research object, a prediction model based on the
combination of one-dimensional convolution and long short-term memory (1DCNN-
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
162
1. INTRODUCTION
The traditional surveying and mapping technology has low work efficiency and
requires a lot of human resources to complete the survey work, but the final
information accuracy is poor [1]. Under the rapid development of modern social
science and technology, new surveying and mapping technologies with better
performance have been developed. Digital surveying and mapping are the main
forms, and they have been applied to the fields of hydrogeology, petroleum
engineering, and other fields, and played their due role in the field of geological
exploration. It greatly promotes the development of the exploration field, especially the
geological exploration, and the accuracy requirements of the surveying and mapping
technology are increasing day by day [2-3]. Therefore, it is extremely necessary to
optimize the surveying and mapping technology.
At present, there is little research on the optimization of surveying and mapping
technology, and it is of great significance to use deep learning models to optimize
surveying and mapping technology. In recent years, some scholars have done a lot of
work. Lu X H [4] established a prediction model of surveying and mapping technology
through regression analysis using cutting parameters as independent variables.
Beruvides [5] used the vibration signal sent out during the training process to establish
the prediction model of surveying and mapping technology by using the adaptive
neuro-fuzzy inference system and obtained a higher fitting index and better
generalization ability. Some researchers [6] used the improved particle swarm
algorithm to optimize the node selection of the hidden layer of the BP network and
established the prediction model of the surveying and mapping technology. In
addition, some researchers [7] proposed a parameter synchronization optimization
algorithm for GA signal feature recognition and mapping prediction, established a GA-
WPT-ELM prediction model, and obtained high prediction accuracy. With the
development of artificial intelligence, deep learning makes data processing and results
in prediction more efficient and accurate [8]. Its long short-term memory (LSTM)
neural network algorithm improves the gradient disappearance problem of traditional
recurrent neural networks (RNN) and provides a new method for the prediction of
sequence data. Prediction problems are applied in the field of new technologies [9].
Wang M W et al. [10] established a long short-term memory model and realized the
prediction of the wear of surveying and mapping tools by taking advantage of its
advantages of solving the accumulation effect. Some researchers [11] proposed a
traditional surveying and mapping stage identification model based on a deep LSTM
neural network, which can more accurately reflect the wear state of surveying and
mapping compared with traditional machine learning methods. Yu [12] proposed a
state recognition method based on LSTM, which has higher recognition accuracy than
BP neural network algorithm and SVM algorithm..
Although the LSTM network solves the problem of vanishing gradients, it has poor
performance for batch sequence data processing, resulting in lower accuracy of the
model in result prediction [13-14]. In this paper, taking petroleum surveying and
mapping technology as the research object, a prediction model based on the
combination of one-dimensional convolution and long short-term memory (1DCNN-
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
LSTM) neural network is established to solve the problem of batch sequence data
processing, sample key feature learning, and small sample data processing. Mapping
technology optimizes precision problems. Through examples and experiments, the
effectiveness of the 1DCNN-LSTM prediction model for the prediction of the
development of mapping technology is verified.
2. CORRELATION MODEL THEORY
2.1. LONG SHORT TERM MEMORY (LSTM)
Compared with the traditional RNN, the core idea of the LSTM neural network is to
introduce "three gates" in each memory unit, use the three gates to interact with the
unit state, and change the information borne by the united state. The retention of
information is selectively determined within neurons. The most widely used LSTM
network structure is shown in Fig.1.
Figure 1. LSTM network structure.
As shown in Figure 1, the "three gates" of the LSTM network are the input gate
which determines the retention of new information; the output gate Ot determines the
output degree of information; the forgetting gate ft determines the retention of the
original information state [15]. Its mathematical expression is as follows:
In the formula: σ is the sigmoid activation function, the output range is 0~1; ht-1
is
the input at the previous moment; xt is the input at the current moment; W and b are
the weight coefficients and bias terms corresponding to the three gates, respectively
[16].
C
t-1
h
t-1
σtanhσ
i
t
tanh
σ
O
t
h
t
C
t
(1)
(2)
( )
tanh
t t t
h o C=∗
[ ]
( )
1
,
o o
t t t
o w h x bσ
−
=⋅+
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
163
The LSTM network reduces the number of network layers and the sequence length
through three gated structures effectively solves the problem of gradient
disappearance and realizes the prediction of sequence data [17]. However, due to the
poor processing of batch sequence data in the LSTM network itself, this paper
introduces a 1DCNN network structure to make up for this deficiency.
2.2.
ONE-DIMENSIONAL CONVOLUTIONAL NEURAL NETWORK
(1DCNN)
Convolutional neural network (CNN) is one of the most perfect algorithms in the
field of deep learning [18], which is divided into one-dimensional, two-dimensional,
and three-dimensional. Among them, 1DCNN is good at processing sequence data,
so this paper selects the 1DCNN network for data processing, the structure is shown
in Figure 2. As shown in Figure 2, the sequence data is input into 1DCNN for
preliminary feature extraction [19], and the sub-sequences composed of high-level
features are effectively extracted, and the interference information is removed as the
input node of the LSTM layer. At the same time, the network can directly identify local
simple patterns in the data and apply them to higher-level networks to form more
complex network patterns [20].
Figure 2. The structure of the 1DCNN convolutional layer.
Let the ith input data of the convolution layer be Ii, the convolution kernels are Wi,
each with n pieces, the bias is Bi, the activation function is f
, and the downsampling
operation is to further reduce the dimension of the features of the convolution output.,
and input the corresponding output to the fully connected layer, the fully connected
layer obtains the classification result of this round after weight transformation and
activation, and the corresponding classification error is obtained by comparing with
the true value of the classification [21]. Let the input feature of the fully connected
layer be T, the corresponding weight is W, the bias is B
, and the activation function is
f
, the output of the convolutional layer and the pooling method formula is as follows
[22]:
(3)
(4)
( )
t
O f WX b= +
1
0
t i t i
i
O ReLU W x b
λ−
+
=
⎡ ⎤
⎛ ⎞
= +
⎢ ⎥
⎜ ⎟
⎝ ⎠
⎣ ⎦
∑
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
164
The LSTM network reduces the number of network layers and the sequence length
through three gated structures effectively solves the problem of gradient
disappearance and realizes the prediction of sequence data [17]. However, due to the
poor processing of batch sequence data in the LSTM network itself, this paper
introduces a 1DCNN network structure to make up for this deficiency.
2.2. ONE-DIMENSIONAL CONVOLUTIONAL NEURAL NETWORK
(1DCNN)
Convolutional neural network (CNN) is one of the most perfect algorithms in the
field of deep learning [18], which is divided into one-dimensional, two-dimensional,
and three-dimensional. Among them, 1DCNN is good at processing sequence data,
so this paper selects the 1DCNN network for data processing, the structure is shown
in Figure 2. As shown in Figure 2, the sequence data is input into 1DCNN for
preliminary feature extraction [19], and the sub-sequences composed of high-level
features are effectively extracted, and the interference information is removed as the
input node of the LSTM layer. At the same time, the network can directly identify local
simple patterns in the data and apply them to higher-level networks to form more
complex network patterns [20].
Figure 2. The structure of the 1DCNN convolutional layer.
Let the ith input data of the convolution layer be Ii, the convolution kernels are Wi,
each with n pieces, the bias is Bi, the activation function is f, and the downsampling
operation is to further reduce the dimension of the features of the convolution output.,
and input the corresponding output to the fully connected layer, the fully connected
layer obtains the classification result of this round after weight transformation and
activation, and the corresponding classification error is obtained by comparing with
the true value of the classification [21]. Let the input feature of the fully connected
layer be T, the corresponding weight is W, the bias is B, and the activation function is
f, the output of the convolutional layer and the pooling method formula is as follows
[22]:
Input
1
2
3
4
5
6
1
2
3
4
5
6
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2
3
1
2 1
2 1
2 1
2
1
2
Output
(3)
(4)
( )
t
O f WX b= +
1
0
t i t i
i
O ReLU W x b
λ−
+
=
⎡ ⎤
⎛ ⎞
= +
⎢ ⎥
⎜ ⎟
⎝ ⎠
⎣ ⎦
∑
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
Among them, W is the convolution kernel, X∈RT × n is the input word vector matrix,
and parameter b
is the bias term. Commonly used nonlinear activation functions are
Sigmoid or ReLU. x is the output vector of the previous neural network [23].
2.3. DCNN-LSTM NETWORK MODEL CONSTRUCTION
2.3.1. MODEL BUILDING
Based on the ability of 1DCNN layer data processing and LSTM layer data
prediction, a 1DCNN-LSTM surface roughness prediction model was established [24].
Its structure is shown in Fig.3, including a one-dimensional convolutional layer, a
Batch-Normalize layer, an LSTM layer, and a fully connected layer [25]. After the data
is input, through the Conv algorithm, the features are identified and then entered into
the LSTM layer and the fully connected layer.
Figure 3. 1DCNN-LSTM prediction model structure.
2.3.2. MODEL PARAMETER DETERMINATION AND DATA
PREPROCESSING
There are many influencing factors of surveying and mapping technology, mainly
including field surveying and mapping n, network layout vf
, dynamic real-time
surveying and mapping ap, and geodetic surveying control network point β
. There is a
complex nonlinear relationship between these four parameters and new surveying
and mapping technology. Therefore, based on the establishment of the prediction
model framework, the prediction model is optimized through parameter selection [26],
and the specific steps are as follows:
(1) Input layer and output layer. The four milling parameters n, vf, ap and β are used
as the input node of the prediction model, and the surface roughness Ra is used as
the output node of the prediction model.
(2) Hidden layer. The hidden layer plays a key role in the network architecture. The
number of filters (filters) of the one-dimensional convolutional layer of the model is 1,
the size of the convolution kernel (kernel_size) is 3, and the stride (stride) is 1, and the
padding (padding) is 1. The LSTM layer node is 2, and the fully connected layer node
is 2.
(5)
( )
1
;
j x
j x
T i
i i
KT i
K
e
P y j x
e
θ
θ
θ
=
= =
∑
Conv1D
Input
n
vf
ap
β
Batch
Normalize LSTM Dense
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
165
(3) Learning rate. To adjust the appropriate learning rate parameters and avoid
going over the optimal solution or the model falling into the local optimal solution,
through continuous testing and adjustment, the Adam algorithm model is used, and
the learning rate parameter is finally selected as 0.001.
(4) Data preprocessing. In this dataset, n, vf, ap and β of each group of experiments
constitute a set of input parameters, the output parameters are the micro-milling
surface roughness of each group, and the input parameters and output parameters
constitute a set of sample data.
The purpose of preprocessing the sample data is to normalize the data features of
each dimension to the same value range so that the model training effect is better and
the generalization ability is stronger. For this purpose, all input data are normalized to
[0, 1] in this paper, and the normalization formula is:
In the formula: xi is the original data; xmin is the minimum value in the original data;
ximax is the maximum value in the original data; yi is the normalized value, and yi∈[0,1].
After data preprocessing, the data can be used as an input layer node for Ra
prediction.
3. TEST AND RESULT ANALYSIS
3.1. ORTHOGONAL TEST TO OPTIMIZE INPUT FACTORS
In the micro-slot milling experiment, this paper mainly considers the influence of
spindle speed n, feed rate vf, milling depth ap, and micro-milling cutter helix angle β
on the surface roughness. To fully consider the influence of the above four factors on
the surface roughness in a small number of experiments, a four-factor and five-level
orthogonal experiment were carried out, and the parameters are shown in Table 1.
Table 1. Orthogonal parameter factor level table.
The range analysis method was used to process the experimental results to obtain
the relationship between each factor and the surface roughness Ra, as shown in
Figure 4.
(15)
min
max min
i i
i
i i
x x
yx x
−
=−
Level Field surveying and mapping
coefficient
Network speed Dynamic real-time
mapping depth
Geodetic Control
Dot Spiral Angle
15 1.5 10 25
215 3.0 15 30
325 4.5 20 35
435 6.0 25 40
45 7.5 30 45
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
166
(3) Learning rate. To adjust the appropriate learning rate parameters and avoid
going over the optimal solution or the model falling into the local optimal solution,
through continuous testing and adjustment, the Adam algorithm model is used, and
the learning rate parameter is finally selected as 0.001.
(4) Data preprocessing. In this dataset, n, vf, ap and β of each group of experiments
constitute a set of input parameters, the output parameters are the micro-milling
surface roughness of each group, and the input parameters and output parameters
constitute a set of sample data.
The purpose of preprocessing the sample data is to normalize the data features of
each dimension to the same value range so that the model training effect is better and
the generalization ability is stronger. For this purpose, all input data are normalized to
[0, 1] in this paper, and the normalization formula is:
In the formula: xi is the original data; xmin is the minimum value in the original data;
ximax is the maximum value in the original data; yi is the normalized value, and yi∈[0,1].
After data preprocessing, the data can be used as an input layer node for Ra
prediction.
3. TEST AND RESULT ANALYSIS
3.1. ORTHOGONAL TEST TO OPTIMIZE INPUT FACTORS
In the micro-slot milling experiment, this paper mainly considers the influence of
spindle speed n, feed rate vf, milling depth ap, and micro-milling cutter helix angle β
on the surface roughness. To fully consider the influence of the above four factors on
the surface roughness in a small number of experiments, a four-factor and five-level
orthogonal experiment were carried out, and the parameters are shown in Table 1.
Table 1. Orthogonal parameter factor level table.
The range analysis method was used to process the experimental results to obtain
the relationship between each factor and the surface roughness Ra, as shown in
Figure 4.
(15)
min
max min
i i
i
i i
x x
yx x
−
=−
Level
Field surveying and mapping
coefficient
Network speed
Dynamic real-time
mapping depth
Geodetic Control
Dot Spiral Angle
1
5
1.5
10
25
2
15
3.0
15
30
3
25
4.5
20
35
4
35
6.0
25
40
45
7.5
30
45
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
Figure 4. The relationship between the experimental factors and the Ra value of the
surveying and mapping technology, a) field surveying and mapping coefficient; b) network
speed; c) Dynamic real-time mapping depth d) Geodetic control network point helix angle.
It can be seen from the figure that the dynamic real-time surveying and mapping
depth and the geodetic control network point helix angle have similar effects on the
surveying and mapping technology, the correlation coefficient increases with the
increase of its value, and the geodetic control network point helix angle has the most
significant impact on the surveying and mapping technology; surveying and mapping
The technical correlation coefficient decreases with the increase of the field surveying
and mapping coefficient; however, the influence of the network deployment speed is
not significant, so a slightly larger network deployment speed can be adopted to
improve the correlation coefficient of the surveying and mapping technology. In
addition, for the field surveying and mapping coefficient, with the increase of the
abscissa, the correlation coefficient shows a downward trend, while the distribution
speed has almost no change. However, for the dynamic real-time mapping depth and
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
0 10 20 30 40 50
0.020
0.025
0.030
0.035
0.040
0.045
Correlation coefficient
Field surveying and mapping coefficient
1 2 3 4 5 6 7 8
0.025
0.030
0.035
0.040
0.045
0.050
Correlation coefficient
Network speed
10 15 20 25 30
0.025
0.030
0.035
0.040
0.045
0.050
Correlation coefficient
Dynamic real-time mapping depth
25 30 35 40 45
0.030
0.035
0.040
0.045
0.050
0.055
Correlation coefficient
Geodetic Control Dot Spiral Angle
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
167
the geodetic control mesh point helix angle, both show an upward trend with
increasing depth and angle.
3.2. DCNN-LSTM MODEL PREDICTION AND RESULT ANALYSIS
To obtain accurate and credible batch training data sets, 139 groups of random
experiments were performed, and 164 groups of training data sets were obtained.
Combined with various influencing factors, the main parameters of the random
experiment are shown in Table 2. To make the model training more balanced, the
dataset is randomly distributed before training, and then the normalized dataset is
input into the model to start training [27]. The epochs are chosen to be 5000 times,
during the training process of each epoch, all training datasets will be trained once,
and the network automatically calculates the gradient of the batch loss concerning the
weights and updates the weights accordingly.
Table 2. Parameter range
The 164 datasets are divided into two groups, of which 150 are training sets and
the remaining 14 are validation sets. The training process of the 1DCNN-LSTM model
is shown in Figure 5. It can be seen from the figure that when the training round
reaches 3000 times, the training accuracy has reached about 95%, and the
verification accuracy has reached about 91%, indicating that the accuracy is high.
Very stable and meets forecast requirements. To verify the accuracy of the 1DCNN-
LSTM prediction model, 15 sets of experimental parameters were designed for
testing, and the experimental data were normalized and input into the two prediction
models of the 1DCNN-LSTM neural network and BP neural network, respectively.
prediction results.
Table 3. Prediction results
Field surveying and
mapping coefficient
Network speed Dynamic real-time
mapping depth
Geodetic Control Dot
Spiral Angle
5000-75000 1.5-100 6-100 25,30,35,40,45
Number 1DCNN-LSTM Neural Network BP neural network
Ra predicted
value
Error percentage
Ra predicted
value
Error percentage
1 0.1681 4.2 0.2181 11.09
2 0.0861 3.1 0.2132 12.15
3 0.0192 5.1 0.1055 16.33
4 0.0523 6.4 0.0755 21.36
5 0.0490 0.8 0.0489 18.55
6 0.0207 5.1 0.0230 20.91
7 0.0568 3.2 0.0110 4.64
8 0.0772 2.1 0.0225 9.24
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
168
the geodetic control mesh point helix angle, both show an upward trend with
increasing depth and angle.
3.2. DCNN-LSTM MODEL PREDICTION AND RESULT ANALYSIS
To obtain accurate and credible batch training data sets, 139 groups of random
experiments were performed, and 164 groups of training data sets were obtained.
Combined with various influencing factors, the main parameters of the random
experiment are shown in Table 2. To make the model training more balanced, the
dataset is randomly distributed before training, and then the normalized dataset is
input into the model to start training [27]. The epochs are chosen to be 5000 times,
during the training process of each epoch, all training datasets will be trained once,
and the network automatically calculates the gradient of the batch loss concerning the
weights and updates the weights accordingly.
Table 2. Parameter range
The 164 datasets are divided into two groups, of which 150 are training sets and
the remaining 14 are validation sets. The training process of the 1DCNN-LSTM model
is shown in Figure 5. It can be seen from the figure that when the training round
reaches 3000 times, the training accuracy has reached about 95%, and the
verification accuracy has reached about 91%, indicating that the accuracy is high.
Very stable and meets forecast requirements. To verify the accuracy of the 1DCNN-
LSTM prediction model, 15 sets of experimental parameters were designed for
testing, and the experimental data were normalized and input into the two prediction
models of the 1DCNN-LSTM neural network and BP neural network, respectively.
prediction results.
Table 3. Prediction results
Field surveying and
mapping coefficient
Network speed
Dynamic real-time
mapping depth
Geodetic Control Dot
Spiral Angle
5000-75000
1.5-100
6-100
25,30,35,40,45
Number
1DCNN-LSTM Neural Network
BP neural network
Ra predicted
value
Error percentage
Ra predicted
value
Error percentage
1
0.1681
4.2
0.2181
11.09
2
0.0861
3.1
0.2132
12.15
3
0.0192
5.1
0.1055
16.33
4
0.0523
6.4
0.0755
21.36
5
0.0490
0.8
0.0489
18.55
6
0.0207
5.1
0.0230
20.91
7
0.0568
3.2
0.0110
4.64
8
0.0772
2.1
0.0225
9.24
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
As shown in Table 3, to compare the prediction accuracy of the two models, the
average relative prediction error is used as the evaluation index, and the formula is as
follows:
In the formula: δ is the average relative prediction error; Rai
is the predicted value
of each model; Rai- is the measured value of the milling test.
Figure 5. 1DCNN-LSTM prediction model training process
According to the evaluation indicators, the 1DCNN-LSTM model is 5.90%, while the
BP model is 14.92%, and the evaluation effect of the 1DCNN-LSTM model is much
higher than that of the BP model. It shows that the sample features adaptively
extracted by the 1DCNN layer can better reflect the efficient data processing
capability of the network layer than the artificial extraction features, and the short
sequence samples composed of high-level features can effectively improve the
prediction accuracy for the data extraction and analysis of the LSTM layer. Based on
this, the prediction model established by the 1DCNN-LSTM network can accurately
predict the improvement direction of the mapping technology under different
parameters, which fully proves that the prediction model has strong applicability and
9 0.1191 9.1 0.0762 6.22
10 0.0875 7.8 0.0868 3.81
11 0.0542 6.1 0.0871 25.1
12 0.1121 5.3 0.0940 13.84
13 0.0258 2.2 0.0123 11.2
14 0.1391 10.1 0.1012 13.6
15 0.0913 7.1 0.0847 22.2
(16)
1
1
n
ai ai
i
n
ai
i
R R
R
δ=
=
−
=
∑
∑
0 500 1000 1500 2000 2500
50
60
70
80
90
100
Accuracy
Data
Verification accuracy
training curve
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
169
high prediction accuracy. According to the above results, it can be concluded that the
model meets the requirements of accurate prediction [28].
3.3. APPLICATION OF NEW TECHNOLOGY OF SURVEYING
AND MAPPING IN PETROLEUM GEOLOGICAL
EXPLORATION
3.3.1. FIELD SURVEYING AND MAPPING
When using new surveying and mapping technology for field surveying and
mapping work, it is necessary to select an accurate measurement point to ensure the
accuracy of the measurement results[29-30]. Since this measurement point has a
decisive impact on the accuracy of the entire measurement Make preparations such
as the frame to ensure that the technology can be effectively used.
3.3.2. CLOTH NET
In the work of network layout, it is necessary to use connection points or line
connections to achieve it. When performing network layout work for different
measurement areas, it is necessary to do a good job of understanding the local terrain
and formulate a reasonable network layout strategy according to the situation of the
measurement area. For example, in the process of work, two different ways of
construction network and information network are formulated according to the needs.
At the same time, reasonable network distribution can also ensure the network
strength during work, to ensure that the system can make full use of the network for
efficient data measurement and storage, and at the same time make the
measurement results more accurate.
3.3.3. DYNAMIC REAL-TIME MAPPING
The dynamic real-time surveying and mapping work requires a base station, and at
the same time surveying and mapping, it is ensured that each device is used
reasonably to improve the accuracy of the surveying and mapping work. In the survey
work, it is necessary to use a large number of wireless transmission technology, and
the obtained surveying and mapping results are sent to the information receiving
station. When observing whether the rover at the scene can receive the information
sent from different sending stations, it can also rely on the data transmitted by the
base station. to locate. The base station and the mobile station can use the data
observed by themselves and the difference value transmitted by themselves to
calculate to obtain the relative positions of different stations, to output and store the
three-dimensional coordinates.
3.3.4. GEODETIC CONTROL NETWORK
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
170
high prediction accuracy. According to the above results, it can be concluded that the
model meets the requirements of accurate prediction [28].
3.3. APPLICATION OF NEW TECHNOLOGY OF SURVEYING
AND MAPPING IN PETROLEUM GEOLOGICAL
EXPLORATION
3.3.1. FIELD SURVEYING AND MAPPING
When using new surveying and mapping technology for field surveying and
mapping work, it is necessary to select an accurate measurement point to ensure the
accuracy of the measurement results[29-30]. Since this measurement point has a
decisive impact on the accuracy of the entire measurement Make preparations such
as the frame to ensure that the technology can be effectively used.
3.3.2. CLOTH NET
In the work of network layout, it is necessary to use connection points or line
connections to achieve it. When performing network layout work for different
measurement areas, it is necessary to do a good job of understanding the local terrain
and formulate a reasonable network layout strategy according to the situation of the
measurement area. For example, in the process of work, two different ways of
construction network and information network are formulated according to the needs.
At the same time, reasonable network distribution can also ensure the network
strength during work, to ensure that the system can make full use of the network for
efficient data measurement and storage, and at the same time make the
measurement results more accurate.
3.3.3. DYNAMIC REAL-TIME MAPPING
The dynamic real-time surveying and mapping work requires a base station, and at
the same time surveying and mapping, it is ensured that each device is used
reasonably to improve the accuracy of the surveying and mapping work. In the survey
work, it is necessary to use a large number of wireless transmission technology, and
the obtained surveying and mapping results are sent to the information receiving
station. When observing whether the rover at the scene can receive the information
sent from different sending stations, it can also rely on the data transmitted by the
base station. to locate. The base station and the mobile station can use the data
observed by themselves and the difference value transmitted by themselves to
calculate to obtain the relative positions of different stations, to output and store the
three-dimensional coordinates.
3.3.4. GEODETIC CONTROL NETWORK
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
The new surveying and mapping technology in the geodetic control network is to
use satellite positioning technology to complete the measurement of the basic control
network. Since my country has a very large land area, the distance between each
geodetic control network Measuring tool does not perform effective distance
measurements. In the measurement of the urban control network, measurement tools
need to be used frequently, and the measurement tools need to cover a larger area
and have higher accuracy. The new technology of surveying and mapping has the
above characteristics, can meet the requirements of related surveying work, and has
the advantage of simple operation, which can solve the above surveying problems.
4. CONCLUSION
The traditional surveying and mapping technology has low work efficiency and
requires a lot of human resources to complete the survey work, but the final
information accuracy is poor. This study, this paper takes petroleum surveying and
mapping technology as the research object, and establishes a prediction model based
on the combination of one-dimensional convolution and long short-term memory
(1DCNN-LSTM) neural network, using orthogonal optimization to optimize input
parameters, increase prediction accuracy, and at the same time with BP The accuracy
of the neural network is compared, and the following conclusions are obtained: (1)
The input parameters of the 1DCNN-LSTM optimized by the orthogonal test
optimization method, the results predicted by the model have high prediction
accuracy, high prediction effectiveness, and correlation. The influencing factors are
the dynamic real-time surveying and mapping depth and the geodetic control network
point helix angle, field surveying, and network layout; (2) The prediction accuracy of
1DCNN-LSTM for oil exploration surveying and mapping technology is significantly
higher than that of BP neural network, and the errors of the two are the highest,
respectively. 10.1% and 25.1%; (3) The sample features adaptively extracted by the
DCNN layer can better reflect the efficient data processing capability of the network
layer than the artificially extracted features, and the short sequence samples
composed of high-level features are for the data of the LSTM layer. Extraction
analysis effectively improves prediction accuracy.
5. CONFLICT OF INTEREST
The authors declared that there is no conflict of interest.
REFERENCES
(1) Lu, W., Zhao, L. J., & Xu, R. (2021). Remote sensing image processing
technology based on mobile augmented reality technology in surveying
and mapping engineering. Soft Computing, (4). https://doi.org/10.1007/
s00500-021-05650-3
(2) Shan, X., Wang, Y., Dong, M., & Xia, J. (2021). Application Research and
Analysis of Geographic Information System in Intelligent City Surveying
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
171
and Mappinge. Journal of Physics: Conference Series, 1881(4), 042071. https://
doi.org/10.1088/1742-6596/1881/4/042071
(3) Xc, A., Bta, B., Yan, L. A., Xx, A., Lei, L. C., Hk, A., & Jl, A. (2019). Dynamic
recrystallization behavior of the Ti–48Al–2Cr–2Nb alloy during isothermal
hot deformation. Progress in Natural Science: Materials International, 29(5),
587-594. https://doi.org/10.1016/j.pnsc.2019.08.004
(4) Lu, X., Wang, F. R., Xue, L., Feng, Y., & Liang, S. Y. (2019). Investigation of
material removal rate and surface roughness using multi-objective
optimization for micro-milling of inconel 718. Industrial Lubrication and
Tribology, 71(6). https://doi.org/10.1108/ILT-07-2018-0259
(5) Beruvides, G., Casta?O, F., Quiza, R., & Haber, R. E. (2016). Surface
roughness modeling and optimization of tungsten–copper alloys in micro-
milling processes. Measurement, 246-252. https://doi.org/10.1016/
j.measurement.2016.03.002
(6) Shi, Z. L., Gong, Y., Cao, M., & Xiao, S. (2010). Discussion on the Application
of Surveying and Mapping Technology in the Internet of Things Times.
Modern Surveying and Mapping, 65(4), 503-515. https://doi.org/10.1016/
j.neuron.2010.01.035
(7) Shi, X., & Wang, B. (2021). Application of New Surveying and Mapping
Technology in the Construction of Smart City. E3S Web of Conferences,
236, 04031. https://doi.org/10.1051/e3sconf/202123604031
(8) Wang, J., Ran, R., Song, Z., & Sun, J. (2017). Short-Term Photovoltaic Power
Generation Forecasting Based on Environmental Factors and GA-SVM.
Journal of Electrical Engineering & Technology, 12(1), 64-71. https://doi.org/
10.5370/JEET.2017.12.1.064
(9) Zhang, X. Y., Deng, C. H., & Chen, T. T. (2011). Ultra-Short-Term Ahead
Generating Power Forecasting for PV System Based on Markov Chain for
Error Series. Advanced Materials Research, 347-353, 1498-1505. https://
doi.org/10.4028/www.scientific.net/AMR.347-353.1498
(10) Wang, M., Zhou, J., Gao, J., Li, Z., & Li, E. (2020). Milling Tool Wear
Prediction Method Based on Deep Learning under Variable Working
Conditions. IEEE Access, 8, 140726-140735. https://doi.org/10.1109/
ACCESS.2020.3010378
(11) Yi, S., Wang, X., & Tang, X. (2014). Deep Learning Face Representation by
Joint Identification-Verification. Advances in neural information processing
systems, 27. https://doi.org/10.48550/arXiv.1406.4773
(12) Yu, S., Zeng, W., Zhao, Y., Shao, Y., & Zhou, Y. (2012). Modeling the
Correlation of Composition-Processing-Property for TC11 Titanium Alloy
Based on Principal Component Analysis and Artificial Neural Network.
Journal of Materials Engineering & Performance, 21(11), 2231-2237. https://
doi.org/10.1007/s11665-012-0162-y
(13) Huang, Z., Wei, X., & Kai, Y. (2015). Bidirectional LSTM-CRF Models for
Sequence Tagging. Computer Science. https://doi.org/10.48550/
arXiv.1508.01991
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
172
and Mappinge. Journal of Physics: Conference Series, 1881(4), 042071. https://
doi.org/10.1088/1742-6596/1881/4/042071
(3) Xc, A., Bta, B., Yan, L. A., Xx, A., Lei, L. C., Hk, A., & Jl, A. (2019). Dynamic
recrystallization behavior of the Ti–48Al–2Cr–2Nb alloy during isothermal
hot deformation. Progress in Natural Science: Materials International, 29(5),
587-594. https://doi.org/10.1016/j.pnsc.2019.08.004
(4) Lu, X., Wang, F. R., Xue, L., Feng, Y., & Liang, S. Y. (2019). Investigation of
material removal rate and surface roughness using multi-objective
optimization for micro-milling of inconel 718. Industrial Lubrication and
Tribology, 71(6). https://doi.org/10.1108/ILT-07-2018-0259
(5) Beruvides, G., Casta?O, F., Quiza, R., & Haber, R. E. (2016). Surface
roughness modeling and optimization of tungsten–copper alloys in micro-
milling processes. Measurement, 246-252. https://doi.org/10.1016/
j.measurement.2016.03.002
(6) Shi, Z. L., Gong, Y., Cao, M., & Xiao, S. (2010). Discussion on the Application
of Surveying and Mapping Technology in the Internet of Things Times.
Modern Surveying and Mapping, 65(4), 503-515. https://doi.org/10.1016/
j.neuron.2010.01.035
(7) Shi, X., & Wang, B. (2021). Application of New Surveying and Mapping
Technology in the Construction of Smart City. E3S Web of Conferences,
236, 04031. https://doi.org/10.1051/e3sconf/202123604031
(8) Wang, J., Ran, R., Song, Z., & Sun, J. (2017). Short-Term Photovoltaic Power
Generation Forecasting Based on Environmental Factors and GA-SVM.
Journal of Electrical Engineering & Technology, 12(1), 64-71. https://doi.org/
10.5370/JEET.2017.12.1.064
(9) Zhang, X. Y., Deng, C. H., & Chen, T. T. (2011). Ultra-Short-Term Ahead
Generating Power Forecasting for PV System Based on Markov Chain for
Error Series. Advanced Materials Research, 347-353, 1498-1505. https://
doi.org/10.4028/www.scientific.net/AMR.347-353.1498
(10) Wang, M., Zhou, J., Gao, J., Li, Z., & Li, E. (2020). Milling Tool Wear
Prediction Method Based on Deep Learning under Variable Working
Conditions. IEEE Access, 8, 140726-140735. https://doi.org/10.1109/
ACCESS.2020.3010378
(11) Yi, S., Wang, X., & Tang, X. (2014). Deep Learning Face Representation by
Joint Identification-Verification. Advances in neural information processing
systems, 27. https://doi.org/10.48550/arXiv.1406.4773
(12) Yu, S., Zeng, W., Zhao, Y., Shao, Y., & Zhou, Y. (2012). Modeling the
Correlation of Composition-Processing-Property for TC11 Titanium Alloy
Based on Principal Component Analysis and Artificial Neural Network.
Journal of Materials Engineering & Performance, 21(11), 2231-2237. https://
doi.org/10.1007/s11665-012-0162-y
(13) Huang, Z., Wei, X., & Kai, Y. (2015). Bidirectional LSTM-CRF Models for
Sequence Tagging. Computer Science. https://doi.org/10.48550/
arXiv.1508.01991
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
(14) Li, M., Liu, X., & Xiong, A. (2002). Prediction of the mechanical properties of
forged TC11 titanium alloy by ANN. Journal of Materials Processing
Technology, 121(1), 1-4. https://doi.org/10.1016/S0924-0136(01)01006-8
(15) Cai, W., Zhai, B., Liu, Y., Liu, R., & Ning, X. (2021). Quadratic polynomial
guided fuzzy C-means and dual attention mechanism for medical image
segmentation. Displays, 70, 102106. https://doi.org/10.1016/
j.displa.2021.102106
(16) Miao, J., Wang, Z., Ning, X., Xiao, N., Cai, W., & Liu, R. (2022). Practical and
secure multifactor authentication protocol for autonomous vehicles in 5G.
Software: Practice and Experience. https://doi.org/10.1002/SPE.3087
(17) Liu, Z., Zhou, W., & Li, H. (2019). AB-LSTM: Attention-based Bidirectional
LSTM Model for Scene Text Detection. ACM Transactions on Multimedia
Computing Communications and Applications, 15(4), 1-23. https://doi.org/
10.1145/3356728
(18) Ning, X., Duan, P., Li, W., & Zhang, S. (2020). Real-time 3D face alignment
using an encoder-decoder network with an efficient deconvolution layer.
IEEE Signal Processing Letters, 27, 1944-1948. https://doi.org/10.1109/
LSP.2020.3032277
(19) Sagnika, S., Mishra, B., & Meher, S. K. An attention-based CNN-LSTM model
for subjectivity detection in opinion-mining. Neural Computing and
Applications, 1-14. https://doi.org/10.1007/s00521-021-06328-5
(20) Ning, X., Gong, K., Li, W., & Zhang, L. (2021). JWSAA: joint weak saliency
and attention aware for person re-identification. Neurocomputing, 453,
801-811. https://doi.org/10.1016/j.neucom.2020.05.106
(21) Shan, W. (2022). Digital streaming media distribution and transmission
process optimisation based on adaptive recurrent neural network.
Connection Science, 34(1), 1169-1180. https://doi.org/
10.1080/09540091.2022.2052264
(22) Chen, Y., Wang, L., Hu, J., & Ye, M. (2020). Vision-Based Fall Event Detection
in Complex Background Using Attention Guided Bi-directional LSTM.
https://doi.org/10.1109/ACCESS.2020.3021795
(23) Qi, F., Gao, C., Lan, W., Yue, Z., Song, T., & Li, Q. (2018). Spatio-temporal fall
event detection in complex scenes using attention guided LSTM. Pattern
Recognition Letters, S016786551830504X-. https://doi.org/10.1016/
j.patrec.2018.08.031
(24) Yan, C., Pang, G., Bai, X., Liu, C., Xin, N., Gu, L., & Zhou, J. (2021). Beyond
triplet loss: person re-identification with fine-grained difference-aware
pairwise loss. IEEE Transactions on Multimedia. https://doi.org/10.1109/
TMM.2021.3069562
(25) Nguyen, T. (2019). Spatiotemporal Tile-based Attention-guided LSTMs for
Traffic Video Prediction. https://doi.org/10.48550/arXiv.1910.11030
(26) Ying, L., Nan, Z. Q., Ping, W. F., Kiang, C. T., Pang, L. K., Chang, Z. H., Nam, L.
Adaptive weights learning in CNN feature fusion for crime scene
investigation image classification. Connection Science. https://doi.org/
10.1080/09540091.2021.1875987
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
173
(27) Li, D., & Lasenby, J. (2022). Spatiotemporal Attention-Based Graph
Convolution Network for Segment-Level Traffic Prediction. IEEE
Transactions on Intelligent Transportation Systems, 23(7), 8337-8345. https://
doi.org/10.1109/TITS.2021.3078187
(28) Hu, X., Liu, T., Hao, X., & Lin, C. (2022). Attention-based Conv-LSTM and Bi-
LSTM networks for large-scale traffic speed prediction. The Journal of
Supercomputing, 1-24. https://doi.org/10.1007/s11227-022-04386-7
(29) Juárez, F. F., Esenarro, D., Díaz, M., y Frayssinet, M. (2021). Model based on
balanced scorecard applied to the strategic plan of a peruvian public
entity. 3C Empresa. Investigación y pensamiento crítico, 10(4), 127-147. https://
doi.org/10.17993/3cemp.2021.100448.127-147
(30) Xu Bo, Wang Ran, Peng Baoying, Alqurashi Fahad Abdullah & Salama
Mohamad. (2021). Automatic parameter selection ZVD shaping algorithm
for crane vibration suppression based on particle swarm optimisation.
Applied Mathematics and Nonlinear Sciences, 7(1), 73-82. https://doi.org/
10.2478/AMNS.2021.1.00075
https://doi.org/10.17993/3ctecno.2023.v12n1e43.159-174
3C Tecnología. Glosas de innovación aplicadas a la pyme. ISSN: 2254-4143
Ed.43 | Iss.12 | N.1 January - March 2023
174