CN-121998144-A - Drilling tool fault real-time prediction method and device based on deep learning

CN121998144ACN 121998144 ACN121998144 ACN 121998144ACN-121998144-A

Abstract

The invention provides a method and a device for predicting drilling tool faults in real time based on deep learning, which relate to the technical field of drilling tool fault real-time prediction, the invention monitors real-time parameter data of drilling tool work in real time, establishes a deep learning network model for drilling tool prediction, predicts fault time and fault characteristics through the deep learning network model, can discover the signs of possible faults of the drilling tool in advance, thereby timely taking maintenance measures, the method reduces the safety risk of the drilling tool, predicts the residual life of the drilling tool through predicting the fault time, improves the prediction precision, can help engineers to make a more effective drilling tool maintenance plan, avoids the situation of excessive maintenance or untimely maintenance, thereby optimizing the resource utilization efficiency, providing intelligent decision support based on a deep learning network model and real-time data analysis, and helping a management layer to make more accurate and effective decisions.

Inventors

CAI MINGJIE
LIU GANG
FU QIANG
DONG ZHONGJUN
HE MINGMIN
MAO DAN
LI YUYAO
TAN LEICHUAN
PENG HAO
LUO XIAOXUE

Assignees

中国石油天然气集团有限公司
中国石油集团川庆钻探工程有限公司

Dates

Publication Date: 20260508
Application Date: 20241105

Claims (19)

1. A drilling tool fault real-time prediction method based on deep learning is characterized by comprising the following steps: s1, acquiring attribute data of a target drilling tool, and acquiring environment data of the target drilling tool during working; S2, acquiring historical failure data and historical environment data of the drilling tool which are the same as or similar to the attribute data of the target drilling tool according to the attribute data of the target drilling tool, and analyzing failure characteristic data according to the historical failure data and the historical environment data so as to establish a deep learning network model; S3, monitoring real-time parameter data of the target drilling tool, performing fault correlation analysis on the real-time parameter data, and extracting change time course data from the real-time parameter data according to a fault correlation analysis result; S4, inputting the attribute data and the environment data of the target drilling tool obtained in the step S1 into the deep learning network model established in the step S2 for offline training, obtaining predicted fault time and fault characteristic data for drilling tool fault simulation, and then inputting the change time course data preprocessed in the step S3 into the deep learning network model obtained by offline training for data correction, so as to obtain new predicted fault time and fault characteristic data; S5, setting a fault risk threshold according to the new fault characteristic data according to the step S4, and comparing the fault risk threshold with the new fault characteristic data to judge the fault risk; s6, predicting the residual life of the target drilling tool according to the change time course data preprocessed in the step S3 and the new predicted fault time in the step S4.
2. The method for predicting drilling tool faults in real time based on deep learning as set forth in claim 1, wherein the step S1 is characterized in that an information collecting module is established at a drilling tool working end, the type and the size of a target drilling tool are collected through the information collecting module to be used as attribute data, and geological data of drilling tool work is input through the drilling tool working end to be used as environment data.
3. The method of claim 2, wherein the drill tool type comprises drill pipe, heavy drill pipe, drill collar and drill bit, and the geological data comprises geological blocks, borehole size and mud property.
4. The method for predicting the fault of the drilling tool based on deep learning in real time according to claim 2, wherein the step S2 is as follows: s2.1, collecting working data of drilling tools with the same attribute in a drilling tool working end according to the attribute data, extracting historical failure data from the working data, and extracting historical environment data corresponding to the same time record according to the historical failure data; S2.2, performing failure characteristic data analysis on the historical failure data combined with the historical environmental data to obtain failure characteristic data representing that the drilling tool cannot be continuously used under each environmental data, and then taking the historical failure data, the historical environmental data and the failure characteristic data as basic data, so that a deep learning network model for predicting the drilling tool is established.
5. The method for predicting drilling tool faults based on deep learning as claimed in claim 4, wherein in the step S2.1, working data of drilling tools with the same attribute are collected, including model and materials of the drilling tools.
6. The method of claim 5, wherein the historical failure data of the drilling tool is extracted from the collected working data, namely failure events which lead to the drilling tool not being used continuously, wherein the failure comprises fracture, abrasion or plastic deformation of the drilling tool.
7. The method for predicting drilling tool faults based on deep learning in real time as claimed in claim 6, wherein the historical environment data is environment data extracted when the drilling tool is in occurrence for each failure event which causes the drilling tool to be incapable of being used continuously.
8. The method of claim 7, wherein the failure characteristic data comprises failure mode characteristics, environment factor characteristics, operation parameter characteristics and use history characteristics, the failure mode characteristics comprise failure types and failure mechanisms, the environment factor characteristics comprise temperature, pressure, borehole size and geological characteristics, the operation parameter characteristics comprise rotating speed, weight on bit, displacement and vibration characteristics, and the use history characteristics comprise operation period and maintenance records.
9. The method for predicting drilling tool faults in real time based on deep learning, as set forth in claim 8, wherein the step S2.2 of establishing a deep learning network model comprises the steps of combining historical environment data and failure feature data, constructing a feature vector or a feature tensor for inputting the deep learning network model, using the marked historical failure data as a target, using the feature vector or the feature tensor as an input, and training the deep learning network model according to the following formula: X i ＝[E i ,F i ]; Wherein, X i includes historical environmental data and failure feature data, E i is environmental data when the ith failure event occurs, and F i is failure feature data related to the failure event; Wherein, the Is a predictive value, model is a deep learning function, receives input X i and processes it.
10. The method for predicting drilling tool faults in real time based on deep learning according to claim 1, wherein the step S3 is characterized in that the drilling tool working end is provided with real-time parameter data of a drilling tool sensor for monitoring the working of a target drilling tool, and real-time parameter data fed back by the drilling tool sensor in real time are extracted, wherein the real-time parameter data comprise drilling weight, rotating speed, vertical weight, torque, displacement, inlet and outlet density and gas measurement values, and only the drilling weight, rotating speed, vertical weight and torque parameters are reserved as change time-course data through fault correlation analysis.
11. The method for predicting drilling tool faults based on deep learning as set forth in claim 10, wherein the fault correlation analysis specifically refers to performing fault correlation analysis on each parameter in the real-time parameter data to determine the influence degree of the fault correlation analysis on key indexes in the drilling process, and the formula is as follows: wherein ρ X,Y is the covariance of X and Y, σ X and σ Y are the standard deviations of X and Y, the key indexes comprise the running state of the drill bit and the well bore condition, the value range of ρ X,Y is between [ -1,1], ρ X,Y =1 represents complete positive correlation, X and Y are completely synchronous, ρ X,Y = -1 represents complete negative correlation, Y is reduced when X is increased, ρ X,Y =0 represents no linear correlation, when the absolute value of ρ X,Y is larger than 0.7, the parameters are determined to be closely related to the key indexes, and the parameters closely related to the key index change are selected as important change time course data according to the result of fault correlation analysis.
12. The method for predicting drilling tool faults in real time based on deep learning as claimed in claim 1, wherein the step S3 is to preprocess the change time course data, specifically to process the change time course data by a continuous wavelet transformation method, and mark and divide the processed change time course data as data samples.
13. The method for predicting drilling tool faults based on deep learning in real time as claimed in claim 12, wherein the method for preprocessing the time course data of change specifically comprises the following steps: Firstly, for each parameter in the change time-course data, applying continuous wavelet transformation, and extracting features from the results of the continuous wavelet transformation; for each selected a and b, the wavelet function is scaled and translated, and then inner product operation is carried out with the signal x (t), and the integral operation shows how the wavelet function is matched with the signal x (t) under different time b and scale a; And then, marking the processed change time course data as data samples, and dividing the data samples into a training set, a verification set and a test set according to the marking so as to carry out subsequent model training and evaluation.
14. The method for predicting drilling tool failure in real time based on deep learning of claim 1, wherein in the step S4, the formula is as follows: Wherein x t represents attribute data and environment data at time t, Is the predicted fault time and fault signature data; wherein x't represents real-time varying schedule data of time t, Is the predicted failure time and failure characteristic data of the last moment, Is new predicted fault time and fault characteristic data corrected according to the real-time variation time course data x't.
15. The method for predicting failure of drilling tool based on deep learning in real time as set forth in claim 1, wherein the step S5 of judging the failure risk comprises: the fault characteristic data of the drilling tool comprises information of multiple dimensions, namely Fault signature data representing dimension i, the fault risk Threshold is threshold=threshold (1) ,Threshold (2) ,...,Threshold (i) , and the formula is as follows: Wherein, the Representing the relative deviation between the i-th dimensional fault signature and the threshold value when And if the i-th dimension fault characteristic exceeds the set fault risk threshold, judging that the drilling tool is a dangerous drilling tool.
16. The method for predicting the failure of the drilling tool based on deep learning in real time as set forth in claim 1, wherein the step S6 predicts the remaining life of the target drilling tool by the following formula: Where x t is current change time-course data, T remaining is residual life, T is current time, T ' is predicted failure time, f (T ' |x t ) is probability density function of tool failure time given the change time-course data x t at current time, and the predicted residual life T remaining can be obtained by integrating T '.
17. A drilling tool fault real-time prediction device based on deep learning is characterized by comprising: The system comprises a first module, a second module, a third module, a fourth module, a fifth module, a sixth module and a seventh module, wherein the first module is used for acquiring attribute data of a target drilling tool and acquiring environment data of the target drilling tool during working; The second module is used for monitoring real-time parameter data of the work of the target drilling tool; The third module is used for extracting real-time parameter data in the second module, performing fault correlation analysis in the real-time parameter data, extracting change time course data from the real-time parameter data according to a fault correlation analysis result, and preprocessing the change time course data; The fourth module is used for analyzing failure characteristic data according to the historical failure data and the historical environment data so as to establish a deep learning network model, inputting the attribute data and the environment data of the target drilling tool obtained by the first module into the established deep learning network model for offline training, obtaining predicted failure time and failure characteristic data for drilling tool failure simulation, inputting the change time course data preprocessed by the third module into the deep learning network model obtained by offline training for data correction so as to obtain new predicted failure time and failure characteristic data, setting a failure risk threshold according to the new failure characteristic data, comparing the failure risk threshold with the new failure characteristic data, judging the failure risk, and predicting the residual life of the target drilling tool according to the change time course data preprocessed by the third module and combining the new predicted failure time.
18. A computer device comprising a processor, an input device, an output device and a memory, the processor, the input device, the output device and the memory being interconnected, wherein the memory is adapted to store a computer program comprising program instructions, the processor being configured to invoke the program instructions to perform the method according to any of claims 1-16.
19. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program comprising program instructions which, when executed by a processor, cause the processor to perform the method according to any of claims 1-16.

Description

Drilling tool fault real-time prediction method and device based on deep learning Technical Field The invention relates to the technical field of drilling tool fault real-time prediction, in particular to a drilling tool fault real-time prediction method and device based on deep learning. Background Drilling is an important link of oil and natural gas resource exploration and development, and according to statistics, the drilling engineering cost accounts for 50% -80% of the whole oil and gas exploration and development cost. Along with the continuous increase of the oil and gas resource demands, the drilling engineering is gradually oriented to deep and ultra-deep oil and gas resources, and the drilling cost is also continuously increased, so how to reduce the drilling cost is a perpetual theme of the contemporary drilling engineering. However, the occurrence and handling of downhole accidents significantly reduces drilling efficiency, extends drilling cycle time, increases drilling costs, and is particularly costly. One of the main functions of the drilling tool is to transmit weight and torque, and therefore, the drilling tool is subjected to complex compressive, tensile, bending and torsional loads. These loads are dynamic and can cause the downhole drilling tool to vibrate in the longitudinal, transverse and torsional directions, the drill string is subjected to the composite stress under the action of dynamic and periodic loads, fatigue cracks can start at stress concentration points, along with the continuous action of the loads, the cracks propagate along the radial direction and the circumferential direction perpendicular to the axis of the drill rod, fatigue failure of the drill string is caused, once the drilling tool breaks down, the drilling tool is biased against the well wall and is difficult to center, the salvage difficulty is extremely high, great obstruction is brought to drilling construction, and great loss is brought to oil field companies. In the prior art, a life model is built through data of monitored equipment, and the state and the life of the equipment are judged through a model result. For example, chinese patent literature, publication No. CN114462662A, publication No. 2022, publication No. 5 and 10, entitled a method for predicting and analyzing life of drilling tool, which uses existing drilling report construction database, uses fully interconnected feedforward hidden layer network, i.e. back propagation learning rule, to build drilling tool life big data prediction model based on drilling parameters, uses drilling tool life big data prediction model based on drilling parameters to automatically track the parameters concerned by drilling tool, fits existing theoretical model and existing field experience data, and predicts tool life based on reliable actual data. However, the prior art with the publication number CN114462662a still has the defects that the prior art utilizes big data to predict information, so that the approximate service lives of tools under different conditions are obtained, the accuracy of the predicted service lives is insufficient, and the real-time prediction of the faults and the service lives of drilling tools cannot be realized. Disclosure of Invention In order to solve the problems in the prior art, the invention monitors the real-time parameter data of the drilling tool work in real time, establishes a deep learning network model for predicting the drilling tool, predicts the fault time and the fault characteristics through the deep learning network model, can discover the possible fault signs of the drilling tool in advance, thereby timely taking maintenance measures, reducing the safety risk of the drilling tool, and then predicts the residual life of the drilling tool through predicting the fault time, the prediction precision is improved, and can help engineers to make a more effective drilling tool maintenance plan, avoid the situation of excessive maintenance or untimely maintenance, thereby optimizing the resource utilization efficiency, providing intelligent decision support based on the deep learning network model and real-time data analysis, and helping a management layer to make more accurate and effective decisions. The invention is realized by the following technical scheme: The invention provides a drilling tool fault real-time prediction method based on deep learning, which comprises the following steps: s1, acquiring attribute data of a target drilling tool, and acquiring environment data of the target drilling tool during working; S2, acquiring historical failure data and historical environment data of the drilling tool which are the same as or similar to the attribute data of the target drilling tool according to the attribute data of the target drilling tool, and analyzing failure characteristic data according to the historical failure data and the historical environment data so as to establish a deep learning network model; S3