Patent Application
20250077748
The present invention belongs to the field of underground engineering safety, especially to the field of prevention and control of underground impact pressure in coal mines using an intelligent prediction model combining with-drill signal detection technology and intelligent prediction technology. Specifically, it relates to a method of intelligent prediction of coal stress and different diameters pressure relief based on optimization neural network.
Coal stress with the advancement of the working face is dynamic development and change, underground impact pressure and other dynamic disasters occur is the result of mining stress development and accumulation. At present, the drill hole pressure-relief borehole is one of the more commonly used methods to reduce the risk of coal impact, generally using equal diameter pressure-relief borehole. However, if the diameter of the pressure-relief borehole is too small, the expected pressure-relief effect cannot be achieved, and if the diameter of the hole is too large, it will cause damage to the surrounding rock and affect the stability of the surrounding rock. In order to realize accurate pressure relief, different diameters pressure relief technology has been developed at this stage. The different diameters pressure relief refers to the pressure relief from the surface of the surrounding rock to the depth of the surrounding rock drilling, the pressure relief drilling along the depth of the hole in the high stress areas need to be reamed to artificially change the distribution of stress in the surrounding rock. Compared with the traditional equal diameter pressure relief, different diameters pressure relief is more effective and can better protect the integrity of the surrounding rock. However, in order to achieve better decompression effect, the location of reaming needs to be decided according to the distribution of coal stress. The measure taken by the existing technology is to select a few representative locations within a certain distance to drill stress monitoring holes to monitor the stress at different depths. According to the monitoring results to determine the reaming parameters, so that all the pressure-relief borehole within the range of the fixed reaming parameters in accordance with the construction, which is essentially a rough method of reaming pressure relief. Therefore, how to measure the distribution of coal stress conveniently and quickly has become a bottleneck problem restricting the development of different diameters pressure relief technology.
The traditional drilling pressure relief method uses the amount of coal dust discharged from the drilling construction to indirectly reflect the coal stress, and there is an error in the numerical conversion process; the discharged coal dust needs to be weighed manually, which can not realize the rapid measurement of coal stress; the weighing of the coal dust needs to pause the drilling construction, which leads to the discontinuous monitoring of the parameters of the drilling process, and the frequent starting of the drilling rig may easily lead to the monitoring of the parameters of the drilling process, which are mostly anomalous; the data volume of the drilling parameters is huge, and the processing efficiency of the traditional method is low. Due to the above reasons, the traditional method of pressure relieving in drilling holes lacks precise guidance measures, and it is impossible to achieve precise and efficient pressure relieving effect.
The drilling parameters are characterized by a large number and variety of characteristics, and have a complex nonlinear relationship with coal stress, and the traditional linear and nonlinear methods are unable to establish a correlation model between the two. Artificial neural network is a mathematical model that applies a structure similar to the synaptic connections of the brain for information processing, and it is a kind of arithmetic model, which has been widely used in pattern recognition, signal processing, knowledge engineering, expert system, optimization combination, robot control and so on. However, artificial neural network technology is prone to overfitting problems and is difficult to ensure prediction accuracy. Genetic algorithm is an optimization algorithm that simulates natural selection and genetic mechanism to search for the optimal solution by simulating the evolutionary process, which can be used to automatically adjust the network structure and weight parameters to improve the prediction performance of neural networks. However, the use of neural networks and optimized neural network prediction model to solve the technical difficulties of intelligent prediction of rock mechanical parameters with drilling in coal mine underground has not been realized yet, and the key technologies that plague the realization of this technology are: Firstly, hardware testing equipment is needed to detect the drilling signals. Secondly, the collected drilling signals should be processed by computer system to establish the corresponding intelligent prediction model. Thirdly, the key premise of the above hardware detection equipment and the establishment of intelligent prediction model is also to coordinate the drilling construction of underground coal mine peripheral rock drilling holes, the collection of drilling parameters, and the reasonable allocation of time and space for the collection of drilling parameters used for the training of computational modeling and prediction design.
For this reason, it is necessary to study an operationally accurate pressure relief method that combines drill-following signal detection and intelligent prediction techniques.
In order to conveniently and quickly measure the distribution of coal stress and realize accurate pressure relief for mines with impact tendency, the present invention provides an intelligent prediction of coal stress and a different diameters pressure relief method based on an optimization neural network.
The technical solutions adopted in the present invention to achieve the above objectives are:
A method for intelligent prediction of coal stress and different diameters pressure relief based on an optimization neural network, comprising a computer readable medium operable on a computer with memory for the method of intelligent prediction of coal stress and different diameters pressure relief based on optimization neural network, and comprising program instructions for executing the following steps:
S1.1: Select a sampling working face in the direction of advancement of the mining working face, design pressure-relief boreholes at the sampling working face according to the degree of risk of underground impact pressure, and design at least one accompanying borehole near each pressure-relief borehole for stress monitoring;
S1.2.1: First, the accompanying borehole is drilled, and the method of drilling the accompanying borehole is divided into the following two construction scenarios:
In the first case, when the coal has a Proctor's coefficient value <3.0, an accompanying borehole is drilled parallel to it near each pressure-relief borehole. The accompanying borehole and the pressure-relief borehole are at the same level and at the same depth, and a plurality of hole stress gauges are set up at equal intervals along the depth in the accompanying borehole;
In the second case, when the coal has a Proctor's coefficient value <3.0, a number of accompanying boreholes of equal increasing depth are drilled at equal intervals around the week of the pressure-relief borehole. The depth of the deepest accompanying borehole is equal to the depth of the pressure-relief borehole, and a borehole stress gauge is placed at the bottom of each accompanying borehole;
S1.2.2: After the accompanying borehole is drilled, drilling of the pressure-relief borehole is started. Each time the pressure-relief borehole is drilled to a position at the same depth as the accompanying borehole's drilled stress gauge, the driller's accompanying drilling parameters are collected as a, and the stress value of that stress gauge is also collected as b;
The collation of all the pressure-relief borehole drilling collected during the drilling of the pressure-relief borehole a and the stress values b from the accompanying borehole stress gauges form the training set;
Adopting the coefficient of determination R2 as the evaluation index of model prediction accuracy, setting different numbers of training samples, hidden layers, hidden layer nodes, and different combinations of random drill parameters as independent variables, and the model prediction accuracy as the dependent variable, the comparison experiments of the corresponding artificial neural network models are established. The effects of the number of training samples, hidden layers, hidden layer nodes and different combinations of input feature parameters on the model prediction accuracy are examined. The optimal number of training samples, hidden layers, hidden layer nodes, and the optimal combination of feature parameters are preferred based on the impact results. The neural network prediction model is established based on the optimal number of training samples, implicit layers, implicit layer nodes, and the optimal combination of feature parameters that are preferred;
Genetic algorithm and particle swarm algorithm were used to optimize the neural network prediction model established in S2.2, respectively;
S2.4: The best neural network prediction model is preferred by comparing the coefficients of determination of a total of three neural network prediction models obtained in steps S2.2 and S2.3;
The working face adjacent to the sampling working face was selected as the prediction working face, and the parameters of the accompanying drilling were collected at the same time when the pressure-relief borehole construction was carried out at the prediction working face. The optimal neural network prediction model that has been constructed in step S2 is used to predict coal stress in real time;
Drilling pressure relief is a conventional measure to prevent and control underground impact pressure, but there are many problems with the conventional equal-diameter pressure relief technology: 1) Conventional equal-diameter pressure relief technology adopts a single borehole diameter for pressure relief, and the degree of pressure relief can only rely on adjusting the density of pressure relief holes (the density of pressure holes is the number of pressure relief holes per unit area), so the effect of pressure relief can't be accurately controlled. When the drilling density is small, the pressure relief effect is not obvious, and the safety of anti-punching is affected; When the drilling density is large, the construction speed is slow, and the integrity of the surrounding rock structure is affected to some extent; 2) The working face mining is a dynamic process, and the coal stress is dynamically changing with the advancement of the working face. At present, the real-time dynamic perception of coal stress has been a problem in the industry. Most of the monitoring of coal stress relies on pre-installed stress sensors, with limited monitoring range and lagging monitoring time, which can't be obtained in time and feedback regulation.
The invention starts from solving the practical problem that it is difficult to realize accurate pressure relief by conventional equal-diameter pressure relief, and based on the monitoring of rotary speed, feed pressure, motor torque and other accompanying drilling parameters during the construction of pressure relief drilling holes, it adopts artificial intelligence technology to intelligently predict the change of the coal stress in the drilling area. According to the pressure relief theory and diameter change technology, the pressure relief parameters such as drilling diameter and drilling speed are dynamically adjusted and optimized: Based on the predicted value of coal stress, analyze the stress distribution law of the coal and distinguish the high stress area and low stress area. Determine the reaming position and section length of the pressure relief drilling according to the distribution of the high stress area, and determine the reaming diameter according to the stress size of the high stress area. When the pressure relief drilling construction reaches the reaming position, the reaming construction is carried out according to the determined reaming section length and hole size.
Based on the predicted value of coal stress, analyze the coal stress distribution law and distinguish the high stress area and low stress area. Determine the reaming position and the length of the hole diameter section of the pressure relief drill hole according to the distribution of the high stress area. Determine the hole diameter of the reaming hole according to the stress size of the high stress zone. When the pressure-relief borehole drilling reaches the reaming position, the reaming construction is carried out according to the determined reaming section length and aperture size, and the purpose of accurate pressure relief is finally realized;
The technology of the invention has the following advantages in comparison with the conventional technology: 1) The effect of pressure relief is precise and controllable, the range of pressure relief hole diameter is optimized from single hole diameter to continuous variable hole diameter, and the number of pressure relief holes can be reduced by 50%; 2) The collection and analysis of drilling parameters can be realized, and the dynamic control of pressure relief in the drilling holes in the high-risk area of impact can be realized. The construction efficiency of pressure relief holes is increased from 25 m/day to 40 m/day, and the drilling construction is changed from the original three persons working in close proximity to one person working remotely from a distance, which shortens the exposure time of personnel in dangerous areas.
Therefore, the adoption of the above technology can not only improve the construction efficiency, but also ensure the safe mining of impact high-risk coal resources, and ultimately achieve the purpose of accurate pressure relief.
The best neural network prediction model is corrected at regular intervals when pressure relief drilling is performed at the prediction face. The corrected neural network prediction model is utilized to predict the value of stress at the time of coal stress.
Further, the specific process of step S3.3 predictive model revision is:
When pressure-relief boreholes are drilled at the prediction working face, a number of pressure-relief boreholes are selected at intervals to be drilled with stress-monitoring accompanying boreholes as verification accompanying boreholes. Construct the accompanying boreholes according to the method of step S1.2 and collect the stress values and the parameters of the pressure-relief boreholes at the corresponding locations, forming a validation dataset, and based on this validation dataset, verify the error of the optimal neural network prediction model of step S2.4. If the error exceeds the set range (the error value can be set according to the actual situation, such as setting the coefficient of determination greater than 0.8), the validation dataset will be added to the training sample set in step S1. The model is then corrected, and the corrected neural network prediction model is used to predict the value of coal stress at the time of mining, and if the error is within the set range, the model does not need to be corrected.
Further, the drilling-following parameter a described in step S1.2.2 includes drilling parameters and vibration parameters. Said drilling parameters include rotational speed, feed pressure, motor torque, current voltage and real-time power of the motor. Said vibration parameters include mean value, standard deviation, mean square deviation and center of gravity frequency.
The Proctor's coefficient value f described in the present invention is also known as the rock solidity coefficient, fastening coefficient, and is calculated by the formula f=R/10, where R is the uniaxial compressive strength of the rock in MPa. The critical value of the Proctor's coefficient value for distinguishing between the softness and hardness of the coal quality can be selected according to the specific conditions of the site.
The coefficient of determination described in the present invention is a commonly used evaluation index for evaluating the performance of neural network models, and its calculation method is public knowledge and common sense, so it will not be described in detail here.
The following describes the positive effects of the present invention in relation to the existing technology.
For impact-prone mines, drilling pressure relief boreholes to the working face is a common anti-impact measure. The existing technology is generally based on the working conditions and experience on-site drilling of equal-diameter decompression holes to decompress, which is equivalent to taking the surrounding rock stress as a constant value. In fact, with the increase of depth, the stress difference between deep and shallow perimeter rock is large, and the use of equal-diameter pressure relief borehole will easily lead to too small or too large pressure-relief borehole. If the diameter of the pressure-relief borehole is too small, the expected pressure-relief effect cannot be achieved, and if the diameter of the hole is too large, it will cause damage to the surrounding rock and affect the stability of the surrounding rock. In the present invention, when drilling pressure-relief boreholes, the drilling parameters are collected when drilling pressure-relief boreholes in one working face, and the accompanying borehole of the pressure-relief boreholes is used to obtain the stress value to construct a prediction model. In this way, the constructed prediction model can be used to predict the stress value of each pressure-relief borehole in real time, quickly and accurately when drilling pressure-relief boreholes in other working faces. At the same time, an accompanying borehole is set at a certain distance to correct the prediction model and improve the accuracy of coal stress prediction, so as to predict in advance the location of the pressure relief boreholes to be reamed and the size of the reamed holes to realize accurate pressure relief.
In order to more clearly illustrate the technical solutions in the embodiments or prior art of the present invention, the accompanying drawings to be used in the description of the embodiments or prior art will be briefly described below. Obviously, the accompanying drawings in the following description are only some of the embodiments of the present invention, and other accompanying drawings may be obtained based on these drawings by a person of ordinary skill in the art without creative labor.
In the picture:
1—pressure-relief borehole, 2—accompanying borehole, 3—stressometer, 4—different diameters pressure relief borehole, 5—verification of accompanying borehole.
2
1, 22, 23 . . . 2i represents the first accompanying borehole, the second accompanying borehole, the third accompanying borehole . . . the ith accompanying borehole, in that order.
The construction process of the embodiments of the present invention is described in detail below in conjunction with the accompanying drawings, so that the advantages and features of the present invention can be more easily understood by those skilled in the art. Thus, the scope of protection of the present invention is more clearly defined.
The technical solution of the present invention is described in detail below using two mines with underground impact pressure risk as examples.
As shown in
Before drilling the pressure-relief borehole 1, the accompanying borehole 2 is first drilled, in which stress gauges 3 are set at one meter intervals along the depth of the borehole. then the pressure-relief borehole 1 is constructed, and the pressure-relief borehole 1 is an equal-diameter drilled hole. During the drilling of the pressure-relief borehole 1, the accompanying drilling parameters a are collected at intervals of one meter, including drilling parameters (rotational speed, feed pressure, motor torque, current and voltage, and real-time power of the motor, etc.) and vibration parameters (mean, standard deviation, mean square deviation, and frequency of the center of gravity, etc.). Simultaneous acquisition of stress values b from the stress gauges in the accompanying borehole 2 corresponding to this pressure-relief borehole 1.
The collation of all the pressure-relief borehole 1 drilling with the drilling parameters a and the stress values b in the accompanying borehole 2 were collected to form the training set.
Adoption of the coefficient of determination R2 as an evaluation index of model prediction accuracy.
Different numbers of training samples, hidden layers, hidden layer nodes and different combinations of input feature parameters are set as independent variables. The model prediction accuracy, which is also known as the coefficient of determination R2, is used as the dependent variable to set up a comparison experiment of the corresponding artificial neural network model. The effects of the number of training samples, hidden layers, hidden layer nodes and different combinations of input feature parameters on the model recognition accuracy are examined to obtain the best neural network model;
The described combinations of different input characteristic parameters can be a full combination of the drill-following parameters. For example, the rotary speed, feed pressure, motor torque, current and voltage, motor real-time power, average value, standard deviation, mean square deviation, center of gravity frequency and other 9 with the drilling parameters of the full combination of the way a total of 511 combination samples. When there are more drilling parameters, the number of combination samples is very large, and the model calculation is large, wasting computational resources. At this time, the first use of principal component analysis to reduce the dimensionality of the drilling parameters, such as down to 5 dimensions, that is, 5 types of characteristic parameters. The 5 types of feature parameters as independent variables into the neural network model for training. After training, the best neural network prediction model corresponding to the combination of samples in the feature parameters will be the subsequent prediction of the need to collect with the drill parameters.
The algorithms such as genetic algorithm and particle swarm algorithm are used to optimize the best neural network model established in step 2.2 respectively. The best neural network model optimized by genetic algorithm and the best neural network model optimized by particle swarm algorithm are obtained respectively;
S2.4: The best neural network prediction model is preferred by comparing the best neural network model established in step 2.2 and the coefficients of determination of the two best neural network models obtained by optimization of the two algorithms in step S2.3;
As shown in
According to the coal stress prediction value, analyze the coal stress distribution law, distinguish the high stress area and low stress area, and determine the expansion position of the different diameters pressure relief borehole 4 (see the expansion section of the different diameters pressure relief borehole 4 in
When pressure-relief boreholes are drilled in the predicted working face, a number of pressure-relief boreholes are drilled at intervals (which can be adjusted according to the actual situation in the site) and the stress monitoring accompanying boreholes are called validation accompanying boreholes 5. The validation accompanying boreholes 5 are equipped with the stress gauges 3 in accordance with the manner of step S1.2, and when the pressure-relief boreholes corresponding to the validation accompanying boreholes 5 are drilled. Simultaneously, the drilling parameters and the stress values of the stress gauges in the verification accompanying borehole 5 are collected to form a verification data set. Based on this validation dataset, the optimal neural network prediction model of step S2.4 is validated for errors.
If the error exceeds a certain range (the error value can be set by itself according to the actual situation, such as setting the coefficient of determination to be greater than 0.8), the validation dataset will be added to the training sample set of step S1, and thus the optimal neural network prediction model will be corrected. If the error is within a certain range, the optimal neural network prediction model does not need to be corrected.
Example 2 is an example of a softer coal mine with a Proctor's coefficient value <3.0
Example 2 differs from Example 1 mainly in: first, the design and construction of the accompanying borehole is different at step S1 training set data collection; second, the drilling method for verifying the accompanying borehole 5 is different at step S3.3 prediction model correction. Specifically embodied in.
S1.1: In conjunction with
As shown in
As shown in
First, the first accompanying borehole 21, the second accompanying borehole 22, the third accompanying borehole 23 . . . the i-th accompanying borehole 2; is constructed according to the designed number of accompanying boreholes 2 and the depth of each accompanying borehole 2, and a stress gauge is installed at the bottom of each accompanying borehole 2. After the construction of the accompanying boreholes 2 is completed, the drilling of the pressure-relief borehole 1 is started, and when the pressure-relief borehole 1 is drilled to a depth of 1 meter, 2 meters . . . L meters (10 meters), the parameter a of the accompanying borehole is collected once, and the stress value b of the corresponding stress gauges 3 in the first accompanying borehole 21, the second accompanying borehole 22, the third accompanying borehole 23 . . . the i-th accompanying borehole 2i is collected at the same time.
In the correction of the neural network prediction model, a number of stress-monitoring accompanying boreholes of the pressure-relief borehole 1 are also drilled at intervals called validation accompanying boreholes 5 in accordance with the method of Example 1, and the method of drilling the validation accompanying boreholes 5 is carried out in accordance with the method of designing and constructing the neural network prediction model of Step S1. The different diameters accurate pressure relief prediction scheme implemented in the neighboring working face of the sampling working face in accordance with the prediction method of the present invention in Example 2 is shown in
The present invention has the beneficial effect of also including: (1) based on the intelligent prediction of coal stress change by following drilling parameters, combined with the existing pressure unloading theory, dynamic adjustment and optimization of pressure unloading drilling parameters such as drilling hole diameter and depth are carried out according to the needs of pressure unloading to realize the purpose of accurate pressure unloading; (2) accurately perceive the degree of coal stress concentration, prejudge the degree of impact disaster danger, and provide support for disaster prevention and control; (3) mapping the coal stress distribution law, to provide support for the working face roadway deployment, so that mining working face roadway deployment in avoiding the stress peak location.
It is to be noted that the innovation of the present invention mainly lies in setting up an accompanying borehole at the sampling working face utilizing each pressure-relief borehole. The pressure-relief borehole is used to obtain the stress at different depths of each pressure-relief borehole, and at the same time, the drilling parameters of the pressure-relief borehole are collected, and the one-to-one correspondence between the drilling parameters and the stress value is used for training through a neural network, so as to obtain the optimal neural network prediction model of coal stress. With this model, there is no need to drill a large number of stress monitoring holes in the future face, and only need to calibrate the model every once in a while, so that the stress values at different depths of each pressure-relief borehole can be collected quickly and accurately in real time, so as to obtain the parameters of the different diameters relief boreholes quickly and accurately. As to how to use the neural network to train on the drilling parameters is not the focus of the present invention, such as how to examine the impact of the number of training samples, implied layers, implied layer nodes, and different combinations of input feature parameters on the recognition accuracy of the model, as well as how to set up different numbers of training samples, implied layers, implied layer nodes, and different input feature parameters are all well known in the field of the technology, and the present invention does not make too much statement on this. In addition, the number and depth of accompanying boreholes 2 are determined in the embodiment on the basis of collecting the stress value and the accompanying drilling parameter at one meter per drilling, which is only a scale selected to facilitate operation and data processing under the conditions of considering the cost and the accuracy of the prediction results, and does not serve as a limitation on the technical solution of the present invention. In practice, as long as it can ensure that the acquisition of the accompanying drilling parameters and the stress values are at the same depth, any one of the scales is within the scope of protection of the present invention, and therefore a number of improvements made without departing from the principles described herein should also be regarded as within the scope of protection of the present invention.
| Number | Date | Country | Kind |
|---|---|---|---|
| 2023111218926 | Sep 2023 | CN | national |
This application is a continuation-in-part application of US 18/510,177 filed on 15 Nov. 2023 that claims priority to Chinese Patent Application Ser. No. CN2023111218926 filed on 1 Sep. 2023.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 18510177 | Nov 2023 | US |
| Child | 18814750 | US |
