Rock drillability prediction method based on multi-source information fusion

Abstract

The invention discloses a rock drillability prediction method based on multi-source information fusion, which comprises the following steps of: firstly, fusing drilling data and vibration data to predict the rock drillability, and unifying time scales by adopting a segmented cubic Hermite interpolation method aiming at the problem that the time scales of the drilling data and the vibration data are inconsistent; then, a mutual information method is adopted to carry out data correlation analysis on the input variables, and appropriate input variables are obtained; and finally, establishing a rock drillability prediction model by adopting an extreme learning machine method to obtain a uniaxial compressive strength value. The rock drillability is predicted through the drilling data and the vibration data, the drilling efficiency is improved, meanwhile, the stability of the drilling process is guaranteed, safety accidents are effectively reduced, guidance is provided for research on intelligent control of the drilling process, and practicability and applicability are achieved.

Description

Technical Field

[0001] The present invention relates to the field of coal mine drilling engineering, and in particular to a rock drillability prediction method based on multi-source information fusion. Background Art

[0002] The supply of energy resources is a key constraint on the sustainable development of the national economy and an important part of national security.

[0003] my country currently faces multiple challenges in coal exploration and development, including complex geological conditions, low resource density, high exploration difficulty, and environmental protection. Currently, China's drilling technology is relatively low and does not fully meet the needs of exploration and development. Therefore, my country needs to accelerate the development of a new generation of transformative drilling technologies to meet the current demands of complex coal resource exploration and development. Intelligent drilling technology can improve drilling success rates and reservoir penetration rates in coal exploration and development by controlling the drill bit's progress and direction in real time. Therefore, the development of intelligent drilling technology has important strategic significance and broad application prospects. Lithology identification is a key technology within intelligent drilling. Multi-source information fusion uses diverse geological data and information to determine the type, distribution, and physical characteristics of underground rocks, effectively identifying formation hardness and reducing drill bit wear and accidents.

[0004] In practical applications, the fusion of drilling and vibration data to predict rock drillability has achieved considerable success. Research has shown that by monitoring vibration signals during drilling in real time, combined with changes in drilling parameters, it is possible to predict the hardness, cracking, and heterogeneity of the rock formation in advance, and thus anticipate potential difficulties encountered during drilling operations. This method has been widely used in the exploration of resources such as coal, oil, and natural gas, and has demonstrated particularly promising results in drilling operations in coal mining areas. Summary of the Invention

[0005] The present invention provides a rock drillability prediction method based on multi-source information fusion, aiming to solve the problem of rock drillability prediction during the drilling process.

[0006] In order to solve this technical problem, the technical solution adopted by the present invention is:

[0007] A rock drillability prediction method based on multi-source information fusion includes the following steps:

[0008] S1: Fusion of drilling data and vibration data for rock drillability prediction, using piecewise cubic Hermite interpolation to unify the time scales of drilling data and vibration data;

[0009] S2: Use the mutual information method to perform correlation analysis on drilling data and vibration data, and obtain variables with adjacent correlations as input to the rock drillability prediction model;

[0010] S3: The extreme learning machine method is used to establish a rock drillability prediction model to obtain the uniaxial compressive strength value.

[0011] Optionally, in step S1, unifying the time scales of drilling data and vibration data using a piecewise cubic Hermite interpolation method includes:

[0012] G(r) is the piecewise cubic Hermite interpolation polynomial of f(r) at n+1 nodes, and its specific expression is as follows:

[0013]

[0014] In the formula, r represents the current interpolation point, r k and r k+1 Indicates the left and right endpoints of the kth segment, t k and t k+1 represents the known value of the function at the two end points, t k =f(r k ), t' k and t' k+1 Represents the first derivative (slope) of the function at the two end points, s k (r) and s k+1 (r) represents the interpolation basis function (multiplied by t k and t k+1 In the front, the influence of the control function value), S k (r) and S k+1 (r) interpolation basis function, G(r) represents the Hermite interpolation function, the estimated value at point x, k is the number of the current interval, in piecewise interpolation, the kth segment.

[0015] Optionally, in step S2, the data correlation analysis using the mutual information method includes:

[0016]

[0017] Where ρ(x, y), ρ(x), and ρ(y) represent the joint probability of x and y, the marginal probability of x, and the marginal probability of y, respectively; X represents the variables such as feed rate, feed pressure, rotation speed, main pump pressure, auxiliary pump pressure, frequency amplitude, instantaneous frequency, mean, and kurtosis; Y represents the uniaxial compressive strength; x is a specific value in X, and y is a specific value in Y; the larger the value of I(X; Y), the stronger the correlation between x and y; I(X; Y) = 0 indicates no correlation, and I(X; Y) = I(Y; X).

[0018] Optionally, in step S3, the use of the extreme learning machine method to establish a rock drillability prediction model to obtain a uniaxial compressive strength value includes:

[0019] The input of the rock drillability prediction model is the feed rate V ROP , mm / s, feed pressure F p , MPa, speed R, r / min, main pump pressure M p , MPa, auxiliary pump pressure S p , MPa, passband amplitude A B , m / s 2 、Instantaneous passband B I , m / s 2 , mean M e , m / s 2 And kurtosis K, unitless, output is U cs , MPa, the relationship between input and output is expressed as:

[0020] U cs =f ucs (V ROP ,F p ,R,M p ,S p ,A B ,B I ,M e ,K);

[0021] The extreme learning machine method is selected to construct a rock drillability prediction model, which is expressed as:

[0022] f ucs =ELM(n,ω,b);

[0023] Where n is the number of neurons in the hidden layer, ω is the weight connecting the input neurons and the hidden layer neurons, and b is the bias of the hidden layer neurons;

[0024] In the extreme learning machine, the output weight matrix β can be calculated according to the following formula;

[0025]

[0026] Where N is the length of the selected dataset, L is the number of hidden layer nodes; T represents matrix transpose, C represents a constant, and I represents the identity matrix;

[0027] The output matrix H can be calculated as follows:

[0028]

[0029] Where, is the activation function, is the Tanh activation function; ω and b are randomly generated.

[0030] Optionally, the randomly generated ω and b include:

[0031] A two-layer search strategy is adopted for selection. The inner search is as follows: first, a reasonable number of searches is set. According to the characteristics of the extreme learning machine that ω and b are randomly generated, the root mean square error of the ten-fold crossover is calculated, and ω and b with the smallest root mean square error are selected; the outer search is as follows: based on the inner search, a reasonable range of the number of hidden layer nodes and the number of searches are set. When the final root mean square error is minimized, the optimal number of hidden layer nodes n, ω and b are obtained.

[0032] Optionally, in S1, the drilling data includes feed speed, feed pressure, rotation speed, main pump pressure and auxiliary pump pressure.

[0033] Optionally, in S1, the vibration data includes the full-frequency amplitude, instantaneous full-frequency, mean value and kurtosis.

[0034] Optionally, in S2, the variables with adjacent correlations include feed speed, feed pressure, rotation speed, main pump pressure, auxiliary pump pressure, passband amplitude, instantaneous passband, mean value and kurtosis.

[0035] The technical solution provided by the present invention has the following beneficial effects:

[0036] First, drilling data and vibration data are integrated to predict rock drillability. To address the inconsistency between the time scales of drilling and vibration data, a piecewise cubic Hermite interpolation method is used to unify the time scales. Then, the mutual information method is used to analyze data correlation. Finally, an extreme learning machine method is used to establish a rock drillability prediction model to obtain uniaxial compressive strength values. This method uses drilling and vibration data to predict rock drillability, improving drilling efficiency while ensuring drilling stability. This method effectively reduces safety incidents and provides guidance for research on intelligent control of drilling processes. It is both practical and applicable. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0038] Figure 1 This is a step for implementing a rock drillability prediction method based on multi-source information fusion according to the present invention;

[0039] Figure 2 is the result of the piecewise cubic Hermite interpolation in the real-time example of the present invention;

[0040] Figure 3 This is a graph showing the results of predicting uniaxial compressive strength by fusing drilling data and vibration data in an embodiment of the present invention;

[0041] Figure 4is an error diagram of uniaxial compressive strength prediction by fusing drilling data and vibration data in an embodiment of the present invention;

[0042] Figure 5 is a graph of uniaxial compressive strength predicted using only drilling data in an embodiment of the present invention;

[0043] Figure 6 This is a prediction error diagram of uniaxial compressive strength prediction using only drilling data in an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0045] refer to Figure 1 , Figure 1 The rock drillability prediction method based on multi-source information fusion of the present invention is implemented in the following steps:

[0046] S1: Fusion of drilling data and vibration data for rock drillability prediction, using piecewise cubic Hermite interpolation to unify the time scales of drilling data and vibration data;

[0047] S2: Use the mutual information method to perform correlation analysis on drilling data and vibration data, and obtain variables with adjacent correlations as input to the rock drillability prediction model;

[0048] S3: The extreme learning machine method is used to establish a rock drillability prediction model to obtain the uniaxial compressive strength value.

[0049] In step S1, the piecewise cubic Hermite interpolation method is used to unify the time scales of drilling data and vibration data, including the interpolation principle: assuming that the known function f(r) satisfies f(r) at n+1 different nodes (i=0,1,…,n) in the interpolation interval [p,q] i )-f i and f'(r i )-f i '(i=0,1,…,n), if the function G(x) exists and satisfies the following conditions:

[0050] (1) The polynomial degree of G(r) in each cell is 3;

[0051] (2)G(r)∈C 1 [a,b];

[0052] (3)G(r i )=f(r i ),G'(r i )=f'(r i ),i=(0,1,...,n);

[0053] Then G(r) is called the piecewise cubic Hermite interpolation polynomial of f(x) at n+1 nodes, as shown in the following formula.

[0054]

[0055] In the formula, r represents the current interpolation point, r k and r k+1 Indicates the left and right endpoints of the kth segment, t k and t k+1 represents the known value of the function at the two end points, t k =f(r k ), t' k and t' k+1 Represents the first derivative (slope) of the function at the two end points, s k (r) and s k+1 (r) represents the interpolation basis function (multiplied by t k and t k+1 In the front, the influence of the control function value), S k (r) and S k+1 (r) interpolation basis function (multiplied by t' k and t' k+1 Earlier, the influence of the control derivative), G(r) represents the Hermite interpolation function, the estimated value at point x, and k is the number of the current interval (in piecewise interpolation, the kth segment).

[0056] In step S2, the data correlation analysis using the mutual information method includes:

[0057] In order to reduce model redundancy and improve model accuracy, the mutual information analysis method is used for correlation analysis. The formula is as follows:

[0058]

[0059] Where ρ(x, y), ρ(x), and ρ(y) represent the joint probability of x and y and the marginal probability of x and y, respectively; X represents the variables such as feed rate, feed pressure, rotation speed, main pump pressure, auxiliary pump pressure, frequency amplitude, instantaneous frequency, mean, and kurtosis; Y represents the uniaxial compressive strength; x is a specific value in X, and y is a specific value in Y; the larger the value of I(X; Y), the stronger the correlation between x and y; I(X; Y) = 0 indicates no correlation, and I(X; Y) = I(Y; X).

[0060] In step S3, the use of the extreme learning machine method to establish a rock drillability prediction model to obtain a uniaxial compressive strength value includes:

[0061] (1) The input of the rock drillability prediction model is the feed rate V ROP (mm / s), feed pressure F p (MPa), speed R (r / min), main pump pressure M p (MPa), auxiliary pump pressure S p (MPa), passband amplitude A B (m / s 2 ), instantaneous pass frequency B I (m / s 2 ), mean M e (m / s 2 ) and kurtosis K (unitless), the output is U cs (MPa). The relationship between input and output is expressed as

[0062] U cs =f ucs (V ROP ,F p ,R,M p ,S p ,A B ,B I ,M e ,K);

[0063] The Extreme Learning Machine (ELM) method is selected to construct a rock drillability prediction model. On the one hand, it has a simple structure and fast learning speed, and on the other hand, it has been widely used in the drilling process. The above can also be expressed as

[0064] f ucs =ELM(n,ω,b);

[0065] Where n is the number of hidden layer neurons, ω is the weight connecting the input neurons and the hidden layer neurons, and b is the bias of the hidden layer neurons.

[0066] In the extreme learning machine, the output weight matrix β can be calculated according to the following formula.

[0067]

[0068] Where N is the length of the selected dataset and L is the number of hidden layer nodes. The output matrix H can be calculated as follows.

[0069]

[0070] Where, Is the activation function, the activation function selected in this paper is the Tanh activation function.

[0071] In the extreme learning machine algorithm, the number of hidden layer nodes requires careful design, as ω and b are randomly generated. To select the optimal number of hidden layer nodes and biases, we employ a two-tiered search strategy. Inner search: First, we set a reasonable number of searches. Given the random nature of ω and b in the extreme learning machine, we calculate the root mean square error (RMSE) of a ten-fold crossover and select the ω and b that minimize the RMSE. Outer search: Based on the inner search, we set a reasonable range for the number of hidden layer nodes and the number of searches. When the final RMSE is minimized, we obtain the optimal number of hidden layer nodes, n, ω, and b.

[0072] Example 1:

[0073] This example uses 18,852 sets of drilling data and 504 sets of vibration data from a drilling site as the specific objects. These include key operating variables such as feed rate, feed pressure, power head speed, main pump pressure, and auxiliary pump pressure, as well as vibration data such as frequency amplitude, instantaneous frequency, mean, and kurtosis. First, the time scale of the drilling and vibration data is unified, using a 7-second time scale. Second, the mutual information method is used to select appropriate input variables. Finally, an extreme learning machine method is used to establish a uniaxial compressive strength prediction model.

[0074] The specific steps are as follows:

[0075] (1) Using the segmented cubic Hermite interpolation method to unify the time scales of drilling data and vibration data

[0076] The characteristic data representing vibration include the full-frequency amplitude, instantaneous full-frequency, mean value, effective value and kurtosis. The results of their cubic Hermite interpolation polynomials are as follows: Figure 2 shown.

[0077] (2) Using mutual information method to analyze data correlation

[0078] When data comes from different systems or sensors, their timestamps may be different. By unifying the time scale, we can ensure that the data is aligned on the same time axis, making it easier to perform data correlation analysis. ROP 、F p , R, M p and S p There is a strong correlation with uniaxial compressive strength. The drill bit primarily breaks the formation by cutting, shearing, squeezing, or grinding, rotating and feeding forward under the influence of feed pressure and rotation speed. Feed speed is sensitive to changes in the formation and indirectly reflects the difficulty of the drill bit penetrating the rock mass. The main pump pressure and auxiliary pump pressure control the drilling rig and feed pressure adjustment. From a data perspective, the mutual information between uniaxial compressive strength and drilling variables is shown in Table 1. ROP 、F p , R, M p and Sp The mutual information values ​​with the uniaxial compressive strength also show that they have a strong correlation.

[0079] Table 1 Mutual information between uniaxial compressive strength and drilling variables

[0080]

[0081] After unifying the time scale, the mutual information method was used to measure the correlation between the uniaxial compressive strength and vibration data. The results are shown in Table 2. B ), instantaneous frequency passing (B I ), mean(M e ) and kurtosis (K) have a large mutual information value and have a great influence on the uniaxial compressive strength. These four variables are selected as input. M ) has a small mutual information value and is not used as input.

[0082] Table 2 Mutual information between uniaxial compressive strength and vibration variables

[0083]

[0084] In summary, the input of the rock drillability prediction model is the feed rate V ROP (mm / s), feed pressure F p (MPa), speed R (r / min), main pump pressure M p (MPa), auxiliary pump pressure S p (MPa), passband amplitude A B (m / s 2 ), instantaneous pass frequency B I (m / s 2 ), mean M e (m / s 2 ) and kurtosis K (unitless), the output is U cs (MPa).

[0085] (3) Using the extreme learning machine method to establish a rock drillability prediction model to obtain the uniaxial compressive strength value

[0086] The predicted uniaxial compressive strength results are as follows Figure 3 As shown, the error results are as follows Figure 4 shown.

[0087] The comparative experimental results are based on the prediction using only drilling data. Figure 5 and Figure 6 shown.

[0088] In order to evaluate the model more comprehensively, three performance indicators are used for measurement, which are represented as follows:

[0089] (1) Root mean square error:

[0090]

[0091] (2) Mean absolute error:

[0092]

[0093] (3) Mean absolute percentage error:

[0094]

[0095] Where, is the predicted value of uniaxial compressive strength, is the actual value of uniaxial compressive strength.

[0096] Table 3 Comparison results of indicators from different data sources

[0097]

[0098] Table 3 compares the performance of metrics derived from different data sources. As can be seen, the results from integrating drilling and vibration data are superior to those from drilling data alone, in terms of root mean square error, mean absolute error, and mean absolute percentage error. This demonstrates that combining multiple data sources can effectively improve model performance. The combination of vibration and drilling data provides richer information, enabling the model to capture the complex characteristics of uniaxial compressive strength, thereby improving prediction accuracy.

[0099] The present invention has the following beneficial effects: First, drilling data and vibration data are integrated to predict rock drillability. To address the inconsistent time scales of drilling and vibration data, the piecewise cubic Hermite interpolation method is used to unify the time scales. Then, the mutual information method is used to analyze data correlation. Finally, an extreme learning machine method is used to establish a rock drillability prediction model to obtain uniaxial compressive strength values. This method uses drilling and vibration data to predict rock drillability, improving drilling efficiency while ensuring the stability of the drilling process, effectively reducing safety incidents, and providing guidance for research on intelligent control of the drilling process. It is practical and applicable.

[0100] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or system comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or system. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or system comprising the element.

[0101] The serial numbers of the embodiments of the present invention are for descriptive purposes only and do not represent superiority or inferiority of the embodiments. In a unit claim that lists several means, several of these means may be embodied by the same item of hardware. The use of the terms first, second, and third, etc., does not denote any order and should be construed as identifiers.

[0102] The above are only preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the scope of protection of the present invention.

Claims

1. A rock drillability prediction method based on multi-source information fusion, characterized in that: The following steps are involved: S1: Fusion of drilling data and vibration data for rock drillability prediction, using piecewise cubic Hermite interpolation to unify the time scales of drilling data and vibration data; S2: Use the mutual information method to perform correlation analysis on drilling data and vibration data, and obtain variables with adjacent correlations as input to the rock drillability prediction model; S3: The extreme learning machine method is used to establish a rock drillability prediction model to obtain the uniaxial compressive strength value.

2. The rock drillability prediction method based on multi-source information fusion according to claim 1 is characterized in that: In step S1, the use of the piecewise cubic Hermite interpolation method to unify the time scales of drilling data and vibration data includes: G(r) is the piecewise cubic Hermite interpolation polynomial of f(r) at n+1 nodes, and its specific expression is as follows: Where r represents the current interpolation point, r k and r k+1 Indicates the left and right endpoints of the kth segment, t k and t k+1 represents the known value of the function at the two end points, t k =f(r k ), t' k and t' k+1 Represents the first derivative (slope) of the function at the two end points, s k (r) and s k+1 (r) represents the interpolation basis function (multiplied by t k and t k+1 In the front, the influence of the control function value), S k (r) and S k+1 (r) interpolation basis function, G(r) represents the Hermite interpolation function, the estimated value at point x, k is the number of the current interval, in piecewise interpolation, the kth segment.

3. The rock drillability prediction method based on multi-source information fusion according to claim 1 or 2, characterized in that: In step S2, the data correlation analysis using the mutual information method includes: Where ρ(x, y), ρ(x), and ρ(y) represent the joint probability of x and y, the marginal probability of x, and the marginal probability of y, respectively; X represents the variables such as feed rate, feed pressure, rotation speed, main pump pressure, auxiliary pump pressure, frequency amplitude, instantaneous frequency, mean, and kurtosis; Y represents the uniaxial compressive strength; x is a specific value in X, and y is a specific value in Y; the larger the value of I(X; Y), the stronger the correlation between x and y; I(X; Y) = 0 indicates no correlation, and I(X; Y) = I(Y; X).

4. The rock drillability prediction method based on multi-source information fusion according to claim 1 or 2, characterized in that: In step S3, the use of the extreme learning machine method to establish a rock drillability prediction model to obtain a uniaxial compressive strength value includes: The input of the rock drillability prediction model is the feed rate V ROP , mm / s, feed pressure F p , MPa, speed R, r / min, main pump pressure M p , MPa, auxiliary pump pressure S p , MPa, passband amplitude A B , m / s 2 、Instantaneous passband B I , m / s 2 , mean M e , m / s 2 And kurtosis K, unitless, output is U cs , MPa, the relationship between input and output is expressed as: U cs =f ucs (V ROP ,F p ,R,M p ,S p ,A B ,B I ,M e ,K); The extreme learning machine method is selected to construct a rock drillability prediction model, which is expressed as: f ucs =ELM(n,ω,b); Where n is the number of neurons in the hidden layer, ω is the weight connecting the input neurons and the hidden layer neurons, and b is the bias of the hidden layer neurons; In the extreme learning machine, the output weight matrix β can be calculated according to the following formula; Where N is the length of the selected dataset, L is the number of hidden layer nodes; T represents matrix transpose, C represents a constant, and I represents the identity matrix; The output matrix H can be calculated as follows: Where, is the activation function, is the Tanh activation function; ω and b are randomly generated.

5. The rock drillability prediction method based on multi-source information fusion according to claim 4 is characterized in that: The randomly generated ω and b include: A two-layer search strategy is adopted for selection. The inner search is as follows: first, a reasonable number of searches is set. According to the characteristics of the extreme learning machine that ω and b are randomly generated, the root mean square error of the ten-fold crossover is calculated, and ω and b with the smallest root mean square error are selected; the outer search is as follows: based on the inner search, a reasonable range of the number of hidden layer nodes and the number of searches are set. When the final root mean square error is minimized, the optimal number of hidden layer nodes n, ω and b are obtained.

6. The rock drillability prediction method based on multi-source information fusion according to claim 1 or 2, characterized in that: In the aforementioned S1, the drilling data includes feed speed, feed pressure, rotation speed, main pump pressure and auxiliary pump pressure.

7. The rock drillability prediction method based on multi-source information fusion according to claim 1 or 2, characterized in that: In the above-mentioned S1, the vibration data includes the full-frequency amplitude, instantaneous full-frequency, mean value and kurtosis.

8. The rock drillability prediction method based on multi-source information fusion according to claim 1 or 2, characterized in that: In the aforementioned S2, the variables with adjacent correlations include feed speed, feed pressure, rotation speed, main pump pressure, auxiliary pump pressure, passband amplitude, instantaneous passband, mean value and kurtosis.

Images