CN115680645A - Rock mass characteristic real-time prediction method and system based on multi-source information fusion while drilling - Google Patents
Rock mass characteristic real-time prediction method and system based on multi-source information fusion while drilling Download PDFInfo
- Publication number
- CN115680645A CN115680645A CN202211182857.0A CN202211182857A CN115680645A CN 115680645 A CN115680645 A CN 115680645A CN 202211182857 A CN202211182857 A CN 202211182857A CN 115680645 A CN115680645 A CN 115680645A
- Authority
- CN
- China
- Prior art keywords
- drilling
- rock mass
- vibration
- prediction
- fusion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 239000011435 rock Substances 0.000 title claims abstract description 131
- 238000005553 drilling Methods 0.000 title claims abstract description 111
- 238000000034 method Methods 0.000 title claims abstract description 57
- 230000004927 fusion Effects 0.000 title claims abstract description 42
- 238000010801 machine learning Methods 0.000 claims abstract description 18
- 230000008569 process Effects 0.000 claims abstract description 14
- 238000012549 training Methods 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 230000003190 augmentative effect Effects 0.000 claims description 2
- 239000011159 matrix material Substances 0.000 description 18
- 238000011160 research Methods 0.000 description 9
- 230000001133 acceleration Effects 0.000 description 8
- 238000012544 monitoring process Methods 0.000 description 8
- 238000012360 testing method Methods 0.000 description 7
- 238000005070 sampling Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 238000010276 construction Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000006872 improvement Effects 0.000 description 4
- 238000007500 overflow downdraw method Methods 0.000 description 4
- 238000011161 development Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000007246 mechanism Effects 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 230000035515 penetration Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000011065 in-situ storage Methods 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000009527 percussion Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000005086 pumping Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Earth Drilling (AREA)
Abstract
The invention belongs to the technical field of drilling, and particularly relates to a rock mass characteristic real-time prediction method and system based on multi-source information fusion while drilling. The method comprises the following steps: step 1, acquiring drilling speed, torque, vibration Y and vibration Z information while drilling in a drilling process in real time; step 2, respectively inputting the drilling speed, the torque, the vibration Y and the vibration Z into a machine learning model to obtain classification prediction results of four single signal sources; and 3, fusing the classification prediction results of the four single signal sources obtained in the step 2 based on a conditional probability and minimum risk decision level fusion theory to obtain a final rock mass characteristic prediction result. The invention also provides a system for realizing the method. The method realizes more accurate real-time prediction of rock mass characteristics by selecting four appropriate signal sources and by multi-source data fusion based on decision level, and has good application prospect in the drilling field.
Description
Technical Field
The invention belongs to the technical field of drilling, and particularly relates to a rock mass characteristic real-time prediction method and system based on multi-source information fusion while drilling.
Background
With the requirement of human beings on underground space development, a plurality of new engineering technical problems appear, wherein the real-time prediction of rock mechanical characteristics based on characteristic signals of a drilling process is a problem which needs to be solved urgently and is very challenging. The method can effectively ensure the drilling efficiency and the coring quality by acquiring the rock mass characteristics of the stratum to be drilled in real time based on the drilling means, can reduce the number of the cores, and has great significance for the development of the whole geological exploration technology. Some scholars have conducted a series of studies on this section and its related contents.
Some researchers studied some relationships of drilling parameters to rock mass properties from the perspective of the bit cutting mechanism. Based on the theory of equilibrium of axial force and torque work and the principle of conservation of energy for rock cutting during drilling, karasawa et al derived a formula for rock drillability strength and rig torque and axial force, describing the relationship between uniaxial compressive strength and drilling parameters. Song et al introduced rock cutting mechanism to analyze stress on the cutting edge of angle plate drill bit and established mathematical models of axial load, torque and rock-soil mechanical parameters of drill bit. Wang Qi et al, by establishing an indoor mechanical test model, establish a correlation mechanical model between drilling parameters and the cohesion and internal friction angle of the rock, and verify through indoor test data.
With the development of artificial intelligence technology, some students have achieved a series of research results in the aspect of developing research on relevant characteristics of rocks and rocks by using a machine learning method. For example inverting rock mechanics parameters based on drilling rate, rotational speed and torque. Yue Zhongqi et al developed and deployed a Drilling Process Monitor (DPM) in the weathered volcanic rock of hong kong, china, and the results showed that the rate of penetration of the rotary percussion drill was constant for the same uniform continuous rock mass. Tan Zhuoying and the like find that monitoring parameters such as effective axial pressure, drilling tool rotating speed, drilling rate and the like have good reaction on rock strength change at a boundary through field drilling tests. Ahmed Gowida et al established a relationship of drilling parameters to uniaxial compressive strength of rock based on Artificial Neural Networks (ANN), adaptive Neural Fuzzy Inference Systems (ANFIS), support Vector Machines (SVM), and the like by collecting rate of penetration (ROP), mud pumping rate (GPM), riser pressure (SPP), revolutions Per Minute (RPM), torque (T), and Weight On Bit (WOB). Besides predicting uniaxial compressive strength and lithology of complete rock based on drilling parameters, some students also begin to pay attention to the prediction of rock mass quality by using the drilling parameters, and the rock mass quality is inverted mainly by using real-time signals in the working process of a TBM (tunnel boring machine) based on a machine learning method. But is limited by the miniaturization and light weight of geological drilling equipment, and the rock mass inversion research directly based on drilling parameters is hardly available in the field of geological exploration.
In conclusion, the research developed from the angle of cutting rock mass by the drill bit can make a certain explanation on the relevant mechanism and rule of drilling parameters, but the research is difficult to completely and accurately express the boundary condition of a complex geological object, and the boundary condition is greatly different from the real geological background. However, inversion of a machine learning method based on measured data often ignores deep mining of multiple types of sensing signal sources due to limitation of a single signal source or limitation of fusion and mining of multi-source signals, and thus great contingency and limitation exist in a prediction result. Therefore, a machine learning method which integrates multiple signal sources is needed in the field, so that rock mass characteristics can be predicted more accurately.
Disclosure of Invention
The invention belongs to the technical field of drilling, and particularly relates to a rock mass characteristic real-time prediction method and system based on multi-source information fusion while drilling, aiming at more accurately predicting rock mass characteristics according to multi-source information while drilling.
A rock mass characteristic real-time prediction method based on multi-source information fusion while drilling comprises the following steps:
and 3, fusing the classification prediction results of the four single signal sources obtained in the step 2 based on a conditional probability and minimum risk decision level fusion theory to obtain a final rock mass characteristic prediction result.
Preferably, in step 2, the machine learning model is a KNN algorithm-based model.
Preferably, the sample size of the training data used for training the machine learning model in step 2 is 250.
Preferably, the training data used to train the machine learning model in step 2 is augmented in the following manner: the next cycle is set every 0.06m from the starting point.
Preferably, in the training data used for training the machine learning model in step 2, a smote interpolation algorithm is used to process the unbalanced samples.
Preferably, the classification prediction result or the rock mass characteristic prediction result is used for classifying the II, III, IV and V rock masses in the rock mass basic quality classification.
Preferably, in step 3, the decision-level fusion theory based on conditional probability and minimum risk adopts one of the following three types of risk matrices:
or the like, or, alternatively,
| Ⅱ | Ⅲ | Ⅳ | Ⅴ | |
| Ⅱ | 0 | 0.7 | 1.2 | 1.2 |
| Ⅲ | 0.9 | 0 | 0.7 | 0.5 |
| Ⅳ | 1 | 0.9 | 0 | 0.4 |
| Ⅴ | 1.5 | 1.5 | 1.2 | 0 |
or the like, or, alternatively,
| Ⅱ | Ⅲ | Ⅳ | Ⅴ | |
| Ⅱ | 0 | 1 | 1.2 | 1.2 |
| Ⅲ | 0.9 | 0 | 0.7 | 0.5 |
| Ⅳ | 1 | 0.8 | 0 | 0.4 |
| Ⅴ | 1.5 | 1.5 | 1.2 | 0 |
the invention also provides a system for realizing the rock mass characteristic real-time prediction method based on the multi-source information fusion while drilling, which comprises the following steps:
the input module is used for inputting drilling speed, torque, vibration Y and vibration Z information while drilling;
the calculation module is used for calculating to obtain a final rock mass characteristic prediction result according to the information while drilling;
and the output module is used for outputting the final rock mass characteristic prediction result.
The invention also provides a computer-readable storage medium, wherein a computer program is stored on the computer-readable storage medium, and the computer program is used for realizing the rock mass characteristic real-time prediction method based on the multi-source information fusion while drilling.
In the invention, the drilling speed reflects the difficulty of drilling, and for the same set of drilling equipment, different drilling speeds exist when drilling different stratums in the actual construction process, and the drilling speed can reflect the change of the lithology of the stratums and comprehensively reflect the quality characteristics of underground space rock masses. The torque is the rotary force provided by the drilling machine for rotary drilling of the drilling equipment, the rotary torque enables the equipment to effectively cut into the rock mass, and the torque can better reflect the hardness degree and the crushing degree of the underground space rock mass. Compared with the direct monitoring sensor, the three-axis vibration acceleration sensor has the advantages that the information quantity reflected by the vibration sensor is richer, but the information monitoring is not direct enough, the vibration signal is formed by superposition of vibration of drilling equipment and a rock excitation process and additional vibration of carrying machinery, and due to the fact that the frequency of the vibration sensor is high, the extremely high sampling frequency brings great difficulty to an algorithm of inversion while drilling while ensuring rich information monitoring. Wherein, the vibration Y is the vibration acceleration in the drilling direction, and the directions of the vibration X and the vibration Z are determined according to the direction of the vibration Y and a right-hand coordinate system.
The method is based on four kinds of information while drilling, namely drilling speed, torque, vibration Y and vibration Z, and respectively adopts a machine learning model to obtain the classification prediction result of a single signal source, and the prediction result matrix formed based on the four kinds of signal sources is based on the conditional probability minimum risk theory, and fusion is realized from decision-level fusion on the classification results of heterogeneous signals by setting different risk levels, so that the quality levels of several kinds of sample rocks are effectively predicted, and the final overall classification precision and the classification precision of a small amount of samples are effectively improved. Experiments show that the recognition accuracy of several types of rock masses after decision-level fusion can reach 97% at most, and the method is obviously improved compared with an inversion result based on a single signal source.
Obviously, many modifications, substitutions, and variations are possible in light of the above teachings of the invention, without departing from the basic technical spirit of the invention, as defined by the following claims.
The present invention will be described in further detail with reference to the following examples. This should not be understood as limiting the scope of the above-described subject matter of the present invention to the following examples. All the technologies realized based on the above contents of the present invention belong to the scope of the present invention.
Drawings
FIG. 1 shows a method for grading the mass of a rock mass;
FIG. 2 is a schematic diagram of a drill-back overlap sampling method;
FIG. 3 is a data source correlation coefficient;
FIG. 4 is a ratio of sample numbers before and after smote interpolation;
FIG. 5 is a graph of the effect of k on the model prediction accuracy Pre;
FIG. 6 is a sample result of optimal prediction based on various data source characteristics;
FIG. 7 shows the improvement in prediction accuracy (pre for example) due to the increase in the number of samples;
FIG. 8 is the conditional probabilities of various rock masses in combination with the predictions;
FIG. 9 is a prediction result based on conditional probability;
FIG. 10 is a risk coefficient matrix;
FIG. 11 is a prediction of a sample after decision-level fusion;
FIG. 12 is a prediction result based on the minimum risk principle;
FIG. 13 shows key parameters of prediction models of several types of rock masses under different fusion conditions.
Detailed Description
It should be noted that, in the embodiment, the algorithm of the steps of data acquisition, transmission, storage, processing, etc. which are not specifically described, as well as the hardware structure, circuit connection, etc. which are not specifically described, can be implemented by the contents disclosed in the prior art.
The system of the embodiment comprises:
the input module is used for inputting drilling speed, torque, vibration Y and vibration Z information while drilling;
the calculation module is used for calculating to obtain a final rock mass characteristic prediction result according to the information while drilling;
and the output module is used for outputting the final rock mass characteristic prediction result.
The method for predicting rock mass characteristics in real time by adopting the system comprises the following steps:
and 3, fusing the classification prediction results of the four single signal sources obtained in the step 2 based on a conditional probability and minimum risk decision level fusion theory to obtain a final rock mass characteristic prediction result.
And classifying the II, III, IV and V rock masses in the basic quality classification of the rock masses according to the classification prediction result or the rock mass characteristic prediction result.
The conditional probability and minimum risk decision level fusion theory belongs to the prior art, and the specific principle is as follows:
the conditional probability refers to the probability of occurrence of an event ω under the condition that another event X has occurred. Expressed as: p (ω | X). For the prediction of the rock mass quality grade, P (ω | X) refers to the probability of finally deciding a certain type of rock mass under the combination of the rock mass quality grades respectively predicted by several types of sensors.
The minimum risk decision theory is a basic method in statistical pattern recognition, and the method considers the loss possibly brought by wrong classification of different classes on the basis of carrying out conditional probability analysis on data. The decision theory considers the probability of each feature combination relation to each decision category, does not need to consider the feature quantity or the dependency relation of the fusion information, and only needs each feature combination to be independent. The risk of rock mass misclassification of different projects is not the same, for example, the risk of rock mass misclassification from class iii to class v is higher than the risk of rock mass misclassification from class v to class iii, the former may cause the cost of later projects to be increased, and the latter may directly cause project design to be less rigorous so as to cause the quality of projects to be reduced, which causes serious project accidents, so that several different classes have different risk coefficients. According to the investigation and prediction requirements of actual engineering, the most appropriate expected prediction result can be selected by setting different risk coefficients, and further rock mass quality grading based on input while-drilling signal source fusion is realized. The decision-level fusion adopted in the embodiment is to consider the misclassification matrix and calculate the minimum risk on the basis of the calculation of the classification probability of each type of decision combination, so as to complete the final decision.
R(α i |X)＝minR(α i | x), then x ∈ w k (2)
Of the above formula, R (alpha) i | X) represents a conditional risk, which is embodied in the embodiment as a risk determined to be a certain type of rock mass under a certain prediction combination, P (w) j | x) represents a conditional probability, embodied in the present embodiment as a probability of being determined to be a certain type of rock mass under a certain prediction combination. Lambda (. Alpha.) (alpha.) i ，w j ) The corresponding risk factor for each prediction.
In this embodiment, according to the requirement for accuracy of classification results of different types of rock masses, the decision-level fusion theory based on conditional probability and minimum risk adopts one of the following three types of risk matrices:
| Ⅱ | Ⅲ | | Ⅴ | |
| Ⅱ | ||||
| 0 | 0.6 | 0.8 | 1 | |
| Ⅲ | 0.6 | 0 | 0.6 | 0.8 |
| Ⅳ | 0.8 | 0.6 | 0 | 0.6 |
| |
1 | 0.8 | 0.6 | 0 |
or the like, or, alternatively,
| Ⅱ | Ⅲ | | Ⅴ | |
| Ⅱ | ||||
| 0 | 0.7 | 1.2 | 1.2 | |
| Ⅲ | 0.9 | 0 | 0.7 | 0.5 |
| |
1 | 0.9 | 0 | 0.4 |
| Ⅴ | 1.5 | 1.5 | 1.2 | 0 |
or the like, or a combination thereof,
| Ⅱ | Ⅲ | | Ⅴ | |
| Ⅱ | ||||
| 0 | 1 | 1.2 | 1.2 | |
| Ⅲ | 0.9 | 0 | 0.7 | 0.5 |
| |
1 | 0.8 | 0 | 0.4 |
| Ⅴ | 1.5 | 1.5 | 1.2 | 0 |
the technical solution of the present invention will be further described below by specific experimental examples. In the following experimental examples, the method steps not specifically described can be set up with reference to example 1.
Experimental example 1 selection of input data
1. Construction of a database
The data adopted in the experimental example comes from a certain project, the crossed stratum is T3xj, the lithology of the crossed stratum mainly comprises siltstone, argillaceous siltstone and thin coal seam, and the geological conditions are very complex. The main drilling equipment is a ROCK-600 type full-hydraulic portable drilling machine, coring drilling is carried out in a rope coring mode, the effective data acquisition length is 80m, and 9.6m covering layers (not used as main research data of the experimental example) are included, so that 70.4m total whole-process signal characteristics while drilling and whole-section ROCK body characteristic information are acquired in the experimental example.
In the experimental example, 4 sensors are arranged on a drilling machine, and 6 types of signal sources including bit pressure, torque, drilling speed, vibration-X, vibration-Y and vibration-Z are monitored. The first three are obtained by instrument panel information recording and interpolation of the drilling machine, and the vibration signal is directly measured by a triaxial vibration acceleration arranged at the front end of the drilling machine.
Drilling pressure: the pressure of the direct contact position between the drill bit and the rock body can directly reflect the acting force acting on the drill bit, is directly related to the strength of the rock body and is influenced by the feeding total pressure.
Drilling speed: the difficulty degree of drilling is reflected, for the same set of drilling equipment, in the actual construction process, when drilling different stratums, different drilling speeds exist, the drilling speed can reflect the change of the lithology of the stratums, and meanwhile, the quality characteristics of underground space rock masses can be comprehensively reflected.
Torque: the drilling machine provides rotary force for rotary drilling of drilling equipment, the rotary torque enables the equipment to effectively cut into a rock mass, and the torque can better reflect the hardness degree and the crushing degree of the underground space rock mass.
Triaxial vibration acceleration: compared with the direct monitoring sensor, the vibration sensor has richer reflected information quantity, but the information monitoring is not direct enough, the vibration signal is formed by superposing the vibration of the drilling equipment and the rock excitation process and the additional vibration of the carried machinery, and because the frequency of the vibration sensor is higher, the extremely high sampling frequency brings great difficulty to the algorithm of inversion while drilling while ensuring rich information monitoring. Wherein, the vibration Y is the vibration acceleration in the drilling direction, and the directions of the vibration X and the vibration Z are determined according to the direction of the vibration Y and a right-hand coordinate system.
2. Construction of machine learning sample libraries
1. Construction of rock mass tag
For the whole collected 70.4m drilling process in the field, based on the grade division method of RMR quality, according to the field in-situ test and rock mass logging, the data fine description and the rock mass quality grading of the whole research section are realized, and because underground water is not disclosed in the drilling process, the influences of several factors of uniaxial compressive strength, RQD, joint condition and joint interval are mainly considered, and fig. 2 is a method schematic diagram of the rock mass logging.
The method is characterized in that 47 times of secondary units can be extracted from a single secondary period by taking the initial point as a secondary period every 1.5m, and meanwhile, new secondary periods are arranged at intervals of 0.06m (the overlapping rate of adjacent secondary periods is 96%), the total amount of data is expanded by 26 times compared with a single secondary period without considering the overlapping rate, and the method fully and effectively excavates the drilling information. The drill-back overlap sampling method is illustrated in fig. 3.
2. Pre-processing of monitoring signals
Regarding data source feature extraction, because the sampling frequency and the signal feature of each type of signal have great difference, different features are respectively extracted according to the signal feature difference of different data sources. The torque reflects the magnitude of the rotary cutting force, and the numerical value collected by the sensor in the experimental example comprises the specific magnitude of each rotary force in the whole turn, so that the numerical characteristic of the rotary cutting force is directly extracted, and for a single research turn, the maximum value, the minimum value, the average value and the variance in the turn reflect the characteristics of the torque in the turn; the data recorded on the drilling rate is relatively less, and is calculated indirectly by recording the time in each fixed interval distance, so that the data of the drilling rate represents the drilling speed v selected in each round and the change of the speed
The sampling frequency of the vibration signal is highest, but too many characteristic points bring great difficulty to inversion, time domain information of the vibration signal is firstly acquired by a sensor, the abscissa is time(s), and the ordinate is vibration acceleration (g); this is generated as vibration frequency domain information based on fourier transform, with frequency (HZ) on the abscissa and amplitude intensity on the ordinate. Based on the generated time-frequency domain information, the peak frequency and the peak intensity are respectively extracted once every 2s, and the maximum value, the minimum value, the average value and the variance of the peak frequency value and the peak intensity in the time are respectively calculated and finally used as the vibration signal characteristics in the time. The extracted features of several types of signals are shown in the table below.
Based on the correlation test result of the mean value and the label of each element, the correlation test result is shown in fig. 4, the correlation between the weight on bit and the vibration acceleration in the X direction and the final rock mass specific score value is relatively low, the correlation between the rest types of signals and the label value is relatively high, and the correlation between signal sources is relatively low, so that the drilling speed, the torque, the vibration Y and the vibration Z are finally selected as input signal sources to participate in machine learning prediction.
3. Method for processing unbalanced samples
The rock mass types processed by the method have high imbalance, so that reasonable interpolation of small sample data is required based on an effective data interpolation method, the idea of a Smote algorithm (Synthetic Minority updating Technique) is to synthesize a new Minority sample, select b from the nearest neighbor (according to Euclidean distance) of each Minority sample a, and then select a new point from a and b as a new Minority sample.
Based on the smote interpolation algorithm, the finally obtained sample has good performance on balance. The specific sample ratios before and after interpolation are shown in fig. 5.
3. Single source prediction
Respectively carrying out 8:2, randomly dividing the ratio into a training set and a test set, wherein 1222 samples are used as original samples, 2580 samples are obtained after interpolation according to smote data, and 240 samples and 539 samples are obtained after the original samples and the data are enhanced respectively. For various signal sources, the extracted features are trained and predicted respectively based on a KNN algorithm, the minimum distance is measured based on Euclidean distance, the accuracy of K =3, 4, 5, 6 and 7 under different conditions is tested respectively, according to experimental results, the selection of the K value has different adaptability to different signal source prediction results, and a sample subjected to data enhancement has better performance on prediction accuracy. As can be seen from fig. 6, the choice of the k value has a significant impact on the data results for the same sample of the same class of signal. The prediction accuracy of II-class and V-class rock masses is relatively large along with the change of k values, the prediction accuracy of III-class rock masses and V-class rock masses is relatively stable, and a model trained by data reaching class balance after expansion is relatively less influenced by k.
Fig. 7 reflects the optimal prediction result based on each sensing signal, and visually represents whether each verification sample of each sensing signal source is predicted accurately under the optimal prediction result based on the KNN algorithm. As shown in the following table:
in all single signal sources, the prediction accuracy of several types of signal sources to various rock masses has obvious complementation and difference. In terms of overall prediction accuracy. The prediction accuracy of the torque signal to the original data is 70.3%, and the prediction accuracy after sample expansion is 71.1%. The prediction accuracy of the drilling rate signal on the original data is 78.9%, and the prediction accuracy after sample expansion is 73.5%. The vibration-Y signal has a prediction accuracy of 78.1% for the original data and 78.4% after sample expansion. The prediction accuracy of the vibration-Z signal for the original data is 65.6%, and the prediction accuracy after sample expansion is 71.2%. In general, the prediction accuracy of several types of signal sources has a remarkable difference, and the other types of signal sources have small difference except for the vibration-Z signal. Specifically, for the specific classification conditions of several types of rock masses, the inversion model based on the vibration signal (Y, Z) has higher classification accuracy for the II-type rock masses and the V-type rock masses, taking the vibration-Y signal as an example, the values of PRC, REC and F1 of the expansion samples of the inversion model after classification by the KNN algorithm are respectively 0.85, 0.69 and 0.76, and the values of PRC, REC and F1 of the prediction results of the V-type rock masses are respectively 0.89, 0.74 and 0.81. Compared with two types of drilling parameters, the prediction accuracy is remarkably improved. Taking the inversion result of the torque signal as an example, after the torque signal based on the expansion sample is classified by the KNN algorithm, the PRC, REC and F1 values of the II-class rock mass are predicted to be 0.67, 0.56 and 0.61, and the PRC, REC and F1 values of the V-class rock mass are predicted to be 0.66, 0.7 and 0.68 respectively. However, from the evaluation results of the type iii and type iv rock masses, the inversion results based on the vibration signal source do not show more excellent characteristics than the other two types of signal sources, for example, the PRC, REC and F1 values of the prediction results of the type iv rock mass based on the drilling rate signal source are 0.75, 0.793 and 0.771, respectively, and the PRC, REC and F1 values of the prediction results of the type iv rock mass based on the vibration-Y signal source are 0.63, 0.75 and 0.69, respectively.
From the overall training results, the accuracy of the added samples is improved to a certain extent compared with the accuracy of the original data, but the accuracy is not absolute, as shown in fig. 8, when the number of the training samples is less than 250, the prediction accuracy can be remarkably improved by increasing the number of the training samples, when the number of the training samples exceeds 500, the prediction accuracy of some classes is rather reduced, for example, after the classification effect of the iv-class rock mass is nearly doubled, the classification accuracy is still unsatisfactory for the PRC of the information sources of the types of torque, drilling speed and vibration-Y, and for the original small sample classes of the types of the ii-class rock mass and the v-class rock mass, the new samples generated after the smote interpolation have better classification prediction effects on the training results of the types of the ii-class rock mass and the v-class rock mass, but have reverse inhibition on the prediction effects of the types of the iii-class rock mass and the iv-class rock mass. The sensitivity of the sensing information of several types to the data samples is different, for example, the torque signal presents a situation of increasing and then decreasing for the increasing prediction accuracy of the sample size, while the vibration-Y signal is small for the sample size involved in the experiment, and the trend that the accuracy increases with the increasing sample size is basically maintained. Therefore, the accurate classification of several types of rock masses is difficult to be considered simultaneously in the prediction of a single signal source, the classification accuracy of certain types of small samples can be improved to a certain extent by a sample data enhancement method, the improvement on the accuracy is limited, and the classification accuracy of multiple samples in original data can be reduced.
Experimental example 2 decision-level fusion results
The experimental example uses the input data selected in experimental example 1 to study the decision-level fusion results.
Based on the conditional probabilities, the maximum conditional probability in each combination is used as a final decision category, and classification results predicted by the KNN classifier are displayed, a prediction matrix a is formed according to final results obtained by four sensors, theoretically, the prediction matrix a has 4 × 4 × 4 × 4=256 different types, the final results show that the prediction data in the research has 127 prediction combinations, and the corresponding conditional probabilities in each prediction combination are shown in fig. 9. The decision-level fusion results of several types of rock masses based on the maximum conditional probability are shown in fig. 10.
And respectively setting a symmetric risk matrix and an asymmetric risk matrix, and setting a risk coefficient matrix for the wrong cost of different types of rocks, so that the minimum risk under the judgment matrix is selected as the final decision under the condition. In the experimental example, four signal sources of drilling speed, torque, vibration-Y and vibration-Z are respectively and independently subjected to rock mass quality grading prediction by adopting a KNN algorithm, risk coefficients under different decisions are determined, and the risk coefficient matrixes selected in the experimental example comprise three types, wherein the a type matrix is a symmetric risk matrix (lambda) ij ＝λ ji ) Class b and class c are asymmetric risk matrices (λ) ij ≠λ ji ). And selecting a risk coefficient matrix corresponding to the misclassification of the II, III, IV and V rock masses as shown in figure 11. The symmetrical risk matrix is set according to the assumption that the cost of every two rock masses in wrong division is consistent, and the asymmetric risk matrix considers that the cost of the rock masses in wrong division from the rock masses with higher quality grades to the rock masses with lower quality grades is lower than that of the rock masses in wrong division from the rock masses with lower quality grades to the rock masses with higher quality grades. And setting three different risk matrixes by combining field work experience. The predicted results of the three different types of risk matrices are shown in fig. 12.
From the final prediction result, compared with the fusion method directly determined by conditional probability, the fusion method based on the minimum risk has more obvious improvement on the prediction correctness of several types of rock masses. As shown in fig. 13, the distribution of the mispredicted samples under the concentrated fusion condition is visually displayed, and fusion1, fusion2, fusion3, and fusion4 respectively represent fusion directly based on the conditional probability, fusion of the first-type risk matrix, fusion of the second-type risk matrix, and fusion of the third-type risk matrix.
Compared with the prediction result of a single signal source, the multi-classification prediction result after decision-level fusion is better improved. Wherein the accuracy of the prediction of II, III, IV and V rock masses based on the decision level fusion of the conditional probability can reach 0.8, 0.88, 0.96 and 0.9 respectively; the accuracy of the symmetric risk matrix for predicting the II, III, IV and V rock masses is respectively 0.9, 0.85, 0.98 and 0.84 based on the consideration of the mutual mispartition influence; in consideration of two different types of risk matrixes with mutually wrong cost, the prediction accuracy of the II, III, IV and V types of rock masses is respectively 0.8, 0.9, 0.93, 0.96, 0.9, 0.81, 0.93 and 0.97, so that the different risk matrixes set according to different experience judgment have larger influence on the classification prediction accuracy of several different types of rock masses. Compared with a fusion method model based on conditional probability, the fusion method considering the minimum risk has certain improvement on indexes in all aspects, and particularly shows more outstanding characteristics on small samples and important sample problems. Different risk judgment matrixes are set according to different engineering requirements, so that the prediction result can adapt to different engineering requirements on the premise of keeping certain accuracy.
Through the embodiment and the experimental example, the method realizes more accurate real-time prediction of rock mass characteristics by selecting four appropriate signal sources of drilling speed, torque and vibration acceleration (Y axis and Z axis) and by multi-source data fusion based on decision level, and has good application prospect in the drilling field.
Claims (9)
1. A rock mass characteristic real-time prediction method based on multi-source information fusion while drilling is characterized by comprising the following steps:
step 1, acquiring drilling speed, torque, vibration Y and vibration Z information while drilling in a drilling process in real time;
step 2, respectively inputting the drilling speed, the torque, the vibration Y and the vibration Z into a machine learning model to obtain classification prediction results of four single signal sources;
and 3, fusing the classification prediction results of the four single signal sources obtained in the step 2 based on a conditional probability and minimum risk decision level fusion theory to obtain a final rock mass characteristic prediction result.
2. The method for predicting rock mass characteristics in real time according to claim 1, characterized in that: in step 2, the machine learning model is a KNN algorithm-based model.
3. The method for predicting rock mass characteristics in real time according to claim 1, characterized in that: the sample size of the training data used to train the machine learning model in step 2 is 250.
4. The method for predicting rock mass characteristics in real time according to claim 1, characterized in that: the training data used to train the machine learning model in step 2 is augmented in the following way: the next cycle is set at intervals of 0.06m from the starting point.
5. The method for predicting rock mass characteristics in real time according to claim 1, characterized in that: and (3) in training data used for training the machine learning model in the step 2, processing the unbalanced sample by adopting a smote interpolation algorithm.
6. The method for predicting rock mass characteristics in real time according to claim 1, wherein: and the classification prediction result or the rock mass characteristic prediction result is used for classifying the II, III, IV and V rock masses in the rock mass basic quality classification.
7. The method for predicting rock mass characteristics in real time according to claim 6, characterized in that: in step 3, the decision-level fusion theory based on the conditional probability and the minimum risk adopts one of the following three types of risk matrices:
or the like, or, alternatively,
or the like, or, alternatively,
。
8. a system for realizing the rock mass characteristic real-time prediction method based on multi-source information-while-drilling fusion as claimed in any one of claims 1 to 7, is characterized by comprising the following steps:
the input module is used for inputting drilling speed, torque, vibration Y and vibration Z information while drilling;
the calculation module is used for calculating to obtain a final rock mass characteristic prediction result according to the information while drilling;
and the output module is used for outputting the final rock mass characteristic prediction result.
9. A computer-readable storage medium characterized by: the rock mass real-time characteristic prediction method based on multi-source information while drilling fusion is characterized by comprising the steps of storing a computer program for realizing the rock mass characteristic real-time prediction method based on multi-source information while drilling fusion according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211182857.0A CN115680645A (en) | 2022-09-27 | 2022-09-27 | Rock mass characteristic real-time prediction method and system based on multi-source information fusion while drilling |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211182857.0A CN115680645A (en) | 2022-09-27 | 2022-09-27 | Rock mass characteristic real-time prediction method and system based on multi-source information fusion while drilling |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115680645A true CN115680645A (en) | 2023-02-03 |
Family
ID=85062764
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211182857.0A Pending CN115680645A (en) | 2022-09-27 | 2022-09-27 | Rock mass characteristic real-time prediction method and system based on multi-source information fusion while drilling |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115680645A (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110231170A1 (en) * | 2008-11-26 | 2011-09-22 | Total Sa | Estimation of lithological properties of a geological zone |
| US20180171774A1 (en) * | 2016-12-21 | 2018-06-21 | Schlumberger Technology Corporation | Drillstring sticking management framework |
| AU2020101353A4 (en) * | 2019-09-22 | 2020-08-20 | China University Of Geosciences (Wuhan) | Indoor drilling perception test system for mechanical properties of rock masses |
| CN112529341A (en) * | 2021-02-09 | 2021-03-19 | 西南石油大学 | Drilling well leakage probability prediction method based on naive Bayesian algorithm |
| CN112949202A (en) * | 2021-03-19 | 2021-06-11 | 交通运输部科学研究院 | Bayesian network-based rockburst probability prediction method |
| CN113338894A (en) * | 2021-07-15 | 2021-09-03 | 西安石油大学 | Control method of small intelligent drilling machine |
| US20210285316A1 (en) * | 2017-06-27 | 2021-09-16 | Schlumberger Technology Corporation | Real-Time Well Construction Process Inference Through Probabilistic Data Fusion |
| CN114429563A (en) * | 2021-12-16 | 2022-05-03 | 山东大学 | Surrounding rock integrity identification method and system based on while-drilling test and TBM rock slag image |
| CN114492521A (en) * | 2022-01-21 | 2022-05-13 | 成都理工大学 | Intelligent lithology while drilling identification method and system based on acoustic vibration signals |
-
2022
- 2022-09-27 CN CN202211182857.0A patent/CN115680645A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110231170A1 (en) * | 2008-11-26 | 2011-09-22 | Total Sa | Estimation of lithological properties of a geological zone |
| US20180171774A1 (en) * | 2016-12-21 | 2018-06-21 | Schlumberger Technology Corporation | Drillstring sticking management framework |
| US20210285316A1 (en) * | 2017-06-27 | 2021-09-16 | Schlumberger Technology Corporation | Real-Time Well Construction Process Inference Through Probabilistic Data Fusion |
| AU2020101353A4 (en) * | 2019-09-22 | 2020-08-20 | China University Of Geosciences (Wuhan) | Indoor drilling perception test system for mechanical properties of rock masses |
| CN112529341A (en) * | 2021-02-09 | 2021-03-19 | 西南石油大学 | Drilling well leakage probability prediction method based on naive Bayesian algorithm |
| CN112949202A (en) * | 2021-03-19 | 2021-06-11 | 交通运输部科学研究院 | Bayesian network-based rockburst probability prediction method |
| CN113338894A (en) * | 2021-07-15 | 2021-09-03 | 西安石油大学 | Control method of small intelligent drilling machine |
| CN114429563A (en) * | 2021-12-16 | 2022-05-03 | 山东大学 | Surrounding rock integrity identification method and system based on while-drilling test and TBM rock slag image |
| CN114492521A (en) * | 2022-01-21 | 2022-05-13 | 成都理工大学 | Intelligent lithology while drilling identification method and system based on acoustic vibration signals |
Non-Patent Citations (5)
| Title |
|---|
| 刘正好: "基于多源信息融合的隧道岩体参数钻机感知***及识别方法研", 中国优秀硕士学位论文全文数据库（工程科技Ⅱ辑）, 15 November 2021 (2021-11-15), pages 57 * |
| 张瑞新等: "节理岩体关键块体稳定的概率分析", 岩土力学, vol. 35, no. 5, 31 May 2014 (2014-05-31), pages 1399 - 1405 * |
| ***等: "城市地下空间施工风险因素耦合效应研究", 土木工程学报, 31 December 2021 (2021-12-31), pages 76 - 86 * |
| 李科杰: "新编传感器技术手册", 31 January 2002, 国防工业出版社, pages: 180 - 181 * |
| 谢毅等: "山区铁路隧道超长定向钻探关键技术研究", 铁道工程学报, 31 August 2022 (2022-08-31), pages 1 - 5 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109635461B (en) | Method and system for automatically identifying surrounding rock grade by using while-drilling parameters | |
| Wu et al. | Real-time rock mass condition prediction with TBM tunneling big data using a novel rock–machine mutual feedback perception method | |
| US11899154B2 (en) | DAS same-well monitoring real-time microseismic effective event identification method based on deep learning | |
| CN110609320B (en) | Pre-stack seismic reflection pattern recognition method based on multi-scale feature fusion | |
| CN111520123A (en) | Mechanical drilling speed prediction method, device and equipment | |
| Basarir et al. | The use of soft computing methods for the prediction of rock properties based on measurement while drilling data | |
| CN115329657A (en) | Drilling parameter optimization method and device | |
| US20220276143A1 (en) | Method and system for automatic evaluation of cutting element during wear test | |
| Abbas et al. | Drill bit selection optimization based on rate of penetration: application of artificial neural networks and genetic algorithms | |
| NO20101482A1 (en) | Prediction of production zone | |
| Bhatawdekar et al. | Building information model for drilling and blasting for tropically weathered rock. | |
| Vivas et al. | Real-Time Model for Thermal Conductivity Prediction in Geothermal Wells Using Surface Drilling Data: A Machine Learning Approach | |
| Silversides et al. | Boundary identification and surface updates using MWD | |
| Li et al. | Pattern recognition of stick-slip vibration in combined signals of DrillString vibration | |
| CN115680645A (en) | Rock mass characteristic real-time prediction method and system based on multi-source information fusion while drilling | |
| Bai et al. | Intelligent real-time predicting method for rock characterization based on multi-source information integration while drilling | |
| LaBelle | Lithological classification by drilling | |
| Siva Prasad et al. | Influence of static and dynamic rock properties on drillability prognosis for mining and tunnelling projects | |
| CN215256164U (en) | Device for monitoring formation change of drilling machine in drilling process | |
| CN115046516A (en) | Sliding surface position accurate determination method based on single-sliding-surface r-type deep hole inclination measuring curve | |
| US11434757B2 (en) | Direct hydrocarbon indicators analysis informed by machine learning processes | |
| Serapião et al. | Artificial immune systems for classification of petroleum well drilling operations | |
| Li et al. | Completion Design Improvement Using a Deep Convolutional Network | |
| JPH09317372A (en) | Method for searching rock mass | |
| You et al. | Stratigraphic identification using real-time drilling data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |