Drill jumbo body positioning and pose compensation method

Abstract

The invention discloses a drill jumbo body positioning and pose compensation method which comprises the following steps: S1, arranging a laser target and a drill jumbo at proper positions based on the position of the drill jumbo relative to a working surface; s2, adjusting the imaging of the laser target on the working surface, and adjusting the angle of the depth camera; s3, extracting the existing position and posture information as the standard posture of the drill jumbo; s4, in the previous working process, the pose of the drill jumbo is recorded in real time, the deviation value is calculated, and the change rule of the pose and the time when the pose exceeds the deviation range are learned; and S5, when the work is officially carried out, carrying out pose adjustment once at set intervals according to the rock drilling platform stall pose prediction model. The intelligent drill jumbo is high in intelligent degree, the working stability, working precision and working efficiency of the drill jumbo are greatly improved, the drill rod clamping probability is reduced, the manual labor intensity is reduced, precision errors caused by manual participation in posture adjustment are reduced, and the intelligent drill jumbo has very high engineering practical value.

Description

Technical Field

[0001] The invention belongs to the technical field of rock drilling rig posture control, and in particular relates to a rock drilling rig body positioning and posture compensation method. Background Art

[0002] A drilling rig is a type of rock drilling equipment required for tunnel or underground engineering construction. The drilling rig is mainly composed of a rock drill, a drill arm, and some necessary mechanisms, and can support multiple rock drills to operate simultaneously. At present, the positioning of tunnel drilling rigs requires manual cooperation with instruments. In the past, when using a drilling rig for underground construction, the holes on the working surface would be manually arranged first, and then the hole sequence planning would be input into the drilling rig. During the work process, unless there is an unexpected situation such as drill jamming, the position of the drilling rig is generally not readjusted. When a drill jam occurs, it takes manpower and time to readjust the position of the drilling rig, and it may also damage the drill rod.

[0003] With the increasing difficulty of mining and the advancement of technology, the use of manual positioning and posture adjustment of the drilling rig does not meet the requirements of modern intelligent mining. Moreover, in current mining, it does not occur after the operation starts, but in the subsequent work process, due to various factors such as vibration of the drilling rig, uneven force, and complex working environment, the position of the drilling rig will deviate from the standard position, resulting in a decrease in work accuracy. Therefore, it is necessary to monitor the posture deviation of the drilling rig in real time during the work process and compensate the deviation to the drilling rig.

[0004] In the prior art, there is still a lack of an automatic posture adjustment method for a rock drilling rig with high intelligence and high precision. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method for positioning the body of a rock drilling trolley and compensating the posture, in view of the deficiencies in the above-mentioned prior art. The method has a high degree of intelligence, greatly improves the working stability, working accuracy and working efficiency of the rock drilling trolley, reduces the probability of drill sticking, reduces the labor intensity, and reduces the accuracy error caused by manual participation in posture adjustment, and has a high engineering practical value.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for positioning and compensating the body of a rock drilling trolley, the method comprising the following steps:

[0007] Step S1, based on the position of the drilling rig relative to the working surface, placing the laser target and the drilling rig at appropriate positions so that all the imaging points of the laser target fall within the working surface and can be recognized by the depth camera installed on the drilling rig;

[0008] Step S2, adjusting the imaging of the laser target on the working surface so that the three points projected by the laser target form an upright isosceles triangle and are evenly located in the middle of the working surface; adjusting the angle of the depth camera so that the working surface is evenly centered in the camera imaging;

[0009] Step S3, based on the position and posture adjusted in step S2, extracting the existing position and posture information as the standard posture of the drilling rig;

[0010] Step S4, during the previous working process, the position and posture of the drilling rig are recorded in real time to calculate the deviation, and the changing pattern of the position and posture and the time when the deviation range is exceeded are learned;

[0011] Step S5: When the work is officially carried out, the drilling rig posture adjustment is performed at regular intervals according to the drilling rig posture prediction model.

[0012] In the above-mentioned method for positioning and compensating the body of a drilling rig, step S3 extracts the existing position and posture information as the standard posture of the drilling rig based on the position and posture adjusted in step S2, and adopts the principal component analysis method and the convolutional neural network coupling algorithm.

[0013] In the above-mentioned method for positioning and compensating the body of a drilling rig, the specific process of using the principal component analysis method and the convolutional neural network coupling algorithm in step S3 to extract the existing position and posture information as the standard posture of the drilling rig based on the position and posture adjusted in step S2 is as follows:

[0014] Step S301, using a depth camera to capture working surface image information, and passing it into an image feature extraction module of a convolutional neural network;

[0015] Step S302: perform principal component analysis on the convolution kernel used for feature extraction in the convolutional neural network, and only retain the part whose cumulative contribution rate is higher than the cumulative contribution rate threshold; wherein the calculation formula of the cumulative contribution rate is:

[0016]

[0017] Among them, η represents the cumulative contribution rate, p refers to the number of eigenvalues ​​of the parameter matrix of the convolution kernel used for feature extraction in the convolutional neural network, and λ k Refers to the kth eigenvalue after arranging the eigenvalues ​​from large to small;

[0018] Step S303: Use the obtained reduced-dimensional feature image as a standard posture image of the drilling rig.

[0019] In the above-mentioned method for positioning and compensating a drilling rig body, the convolutional neural network is a VGG16 convolutional neural network, which includes 13 convolutional layers, 3 fully connected layers and 5 pooling layers. The convolutional layers and the pooling layers are used for feature extraction, and the fully connected layers are used for classification.

[0020] In step S301, the image information of the working surface is captured by the depth camera and transmitted to the image feature extraction module of the convolutional neural network, and then the following process is performed:

[0021] Step S3011, firstly, perform two convolution operations to extract the main feature information of the image; then activate it by the ReLU activation function so that the model can better fit the nonlinear function; finally, perform the maximum pooling layer to reduce the image size and the amount of calculation; repeat this process again for the obtained image and output the image;

[0022] Step S3012, the image processed in step S3011 is first subjected to three convolution operations, then activated by the ReLU activation function, and the output image is finally subjected to the maximum pooling layer to reduce the image size, and the processed image is repeated twice to obtain the result image;

[0023] In step S302, principal component analysis is performed on the convolution kernel used for feature extraction in the convolutional neural network. The original matrix of the standardized processing in the principal component analysis method is:

[0024]

[0025] in, m is the number of matrix rows, n is the number of matrix columns;

[0026] The normalized matrix is:

[0027] The process of principal component analysis is:

[0028] Step S3021, calculate the covariance matrix R of the standardized sample X:

[0029]

[0030] in,

[0031] Step S3022: Calculate the eigenvalues ​​of the correlation coefficient matrix λ1, λ2, ...λ n And the corresponding eigenvectors, and arrange the eigenvectors into a matrix row by row from top to bottom according to the corresponding eigenvalue size, and take the first k rows to form the matrix P;

[0032] Step S3023: multiply the original data set by the eigenvector matrix P to obtain a matrix after dimension reduction.

[0033] In the above-mentioned method for positioning and compensating the body of a rock drilling rig, the cumulative contribution rate threshold is 85%.

[0034] In the above-mentioned method for positioning and compensating the body of a rock drilling rig, in the VGG16 convolutional neural network, the convolution kernel in the convolution layer uses a small convolution kernel of 3×3, and the maximum pooling layer in the five pooling layers uses a 2×2 filter.

[0035] In the above-mentioned method for positioning and compensating the body of a drilling rig, in the previous working process described in step S4, the posture of the drilling rig is recorded in real time to calculate the deviation, and the changing law of the posture and the time when the deviation range is exceeded are learned; the specific process is:

[0036] Step S401, calculating the maximum deviation of the drilling trolley body when the drill bit is stuck, setting a threshold based on the deviation, and recording the time interval for reaching the threshold each time;

[0037] Step S402, during the working process, real-time recording of the displacement law of the vehicle body caused by the force generated by the drill arm in the working range, position and process; different displacement laws of the vehicle body posture caused by different situations will result in different times to reach the threshold;

[0038] Step S403: Use the collected data to improve the training set of the rock drilling rig and autonomously adjust the posture of the training rock drilling rig to improve the prediction accuracy of the rock drilling rig.

[0039] In the above-mentioned method for positioning and compensating the body of a drilling rig, when the work described in step S5 is formally carried out, according to the drilling rig posture prediction model, the posture adjustment is performed once at a certain period of time, and the time interval for each position adjustment of the drilling rig is set according to the time interval obtained in step S4. The principle is that the actually designed time interval is less than the time interval obtained in step S4.

[0040] Compared with the prior art, the present invention has the following advantages: first, the present invention uses a laser target in conjunction with a depth camera to determine the standard posture of the rock drilling trolley and data collection of the subsequent working process; then, a convolutional neural network is used to extract features from the collected data, and in order to reduce the amount of calculation and the speed of calculation, the principal component analysis method is used to analyze the convolution kernel of the convolutional neural network that may have a linear relationship, and extract the part with a cumulative contribution rate of 85%; finally, the posture information at this time is solved by the extracted feature information and compared with the standard posture information to obtain the deviation, and then the deviation is compensated to the vehicle body to achieve autonomous posture adjustment of the rock drilling trolley; the real-time recording of the vehicle body position data during the working process is realized, and then the real-time posture detection of the rock drilling trolley is realized, the degree of intelligence is high, the working stability, working accuracy and working efficiency of the rock drilling trolley are greatly improved, the probability of drill jamming is reduced, the labor intensity is reduced, and the precision error caused by manual participation in posture adjustment is reduced, and it has a high engineering practical value.

[0041] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flowchart of the method of the present invention;

[0043] Figure 2 Schematic diagram of the working position of the drilling trolley body in an embodiment of the present invention;

[0044] Figure 3 This is a structural diagram of the convolutional neural network used in the present invention;

[0045] Figure 4 Schematic diagram of posture solution used in a specific embodiment of the present invention;

[0046] Figure 5 This is a schematic diagram of the dimensionality reduction principle of the convolution kernel using the principal component analysis method of the present invention. DETAILED DESCRIPTION

[0047] like Figure 1 and Figure 2 As shown, the drilling trolley body positioning and posture compensation method of the present invention comprises the following steps:

[0048] Step S1, based on the position of the drilling rig relative to the working surface, placing the laser target and the drilling rig at appropriate positions so that all the imaging points of the laser target fall within the working surface and can be recognized by the depth camera installed on the drilling rig;

[0049] In specific implementation, the selected depth camera can extract the depth information of the target point, and the selected depth camera and laser target can still work normally under insufficient light conditions; based on the position of the drilling rig relative to the working surface, the laser target is placed at the top of the tunnel, above the drilling rig, and the projection point of the laser target is ensured to be unobstructed. The laser target can use a pan-tilt to manually operate the pan-tilt to adjust the position of the laser target; in principle, all the imaging points of the laser target fall within the working surface and can be recognized by the visual camera;

[0050] Step S2, adjusting the imaging of the laser target on the working surface so that the three points projected by the laser target form an upright isosceles triangle and are evenly located in the middle of the working surface; adjusting the depth camera angle so that the working surface is evenly centered in the camera imaging; so as to better determine the standard posture of the rock drilling rig relative to the working surface;

[0051] Step S3, based on the position and posture adjusted in step S2, extracting the existing position and posture information as the standard posture of the drilling rig;

[0052] Step S4, during the previous working process, the position and posture of the drilling rig are recorded in real time to calculate the deviation, and the changing pattern of the position and posture and the time when the deviation range is exceeded are learned;

[0053] Step S5: When the work is officially carried out, the drilling rig posture adjustment is performed at regular intervals according to the drilling rig posture prediction model.

[0054] In this embodiment, step S3 extracts the existing position and posture information as the standard posture of the drilling rig based on the position and posture adjusted in step S2, and adopts the principal component analysis method and convolutional neural network coupling algorithm.

[0055] In this embodiment, the specific process of using the principal component analysis method and the convolutional neural network coupling algorithm in step S3 to extract the existing position and posture information as the standard posture of the drilling rig based on the position and posture adjusted in step S2 is as follows:

[0056] Step S301, using a depth camera to capture working surface image information, and passing it into an image feature extraction module of a convolutional neural network;

[0057] Step S302: perform principal component analysis on the convolution kernel used for feature extraction in the convolutional neural network, and only retain the part whose cumulative contribution rate is higher than the cumulative contribution rate threshold; wherein the calculation formula of the cumulative contribution rate is:

[0058]

[0059] Among them, η represents the cumulative contribution rate, p refers to the number of eigenvalues ​​of the parameter matrix of the convolution kernel used for feature extraction in the convolutional neural network, and λ kRefers to the kth eigenvalue after arranging the eigenvalues ​​from large to small;

[0060] Step S303: Use the obtained reduced-dimensional feature image as a standard posture image of the drilling rig.

[0061] In this embodiment, the convolutional neural network is a VGG16 convolutional neural network with outstanding performance in the field of image feature extraction. The VGG16 convolutional neural network includes 13 convolutional layers, 3 fully connected layers and 5 pooling layers, wherein the pooling layer does not involve weights, the convolutional layer and the pooling layer are used for feature extraction, and the fully connected layer is used for classification;

[0062] like Figure 3 As shown, in step S301, the work surface image information is captured by the depth camera and transmitted to the image feature extraction module of the convolutional neural network, and then the following process is performed:

[0063] Step S3011, firstly, perform two convolution operations to extract the main feature information of the image; then activate it by the ReLU activation function so that the model can better fit the nonlinear function; finally, perform the maximum pooling layer to reduce the image size and the amount of calculation; repeat this process again for the obtained image and output the image;

[0064] Step S3012, the image processed in step S3011 is first subjected to three convolution operations, then activated by the ReLU activation function, and the output image is finally subjected to the maximum pooling layer to reduce the image size, and the processed image is repeated twice to obtain the result image;

[0065] In step S302, principal component analysis is performed on the convolution kernel used for feature extraction in the convolutional neural network. The original matrix of the standardized processing in the principal component analysis method is:

[0066]

[0067] in, m is the number of matrix rows, n is the number of matrix columns;

[0068] The normalized matrix is:

[0069] The process of principal component analysis is:

[0070] Step S3021, calculate the covariance matrix R of the standardized sample X:

[0071]

[0072] in,

[0073] Step S3022: Calculate the eigenvalues ​​of the correlation coefficient matrix λ1, λ2, ...λ n And the corresponding eigenvectors, and arrange the eigenvectors into a matrix row by row from top to bottom according to the corresponding eigenvalue size, and take the first k rows to form the matrix P;

[0074] Step S3023: multiply the original data set by the eigenvector matrix P to obtain a matrix after dimension reduction.

[0075] In this embodiment, the cumulative contribution rate threshold is 85%.

[0076] In this embodiment, in the VGG16 convolutional neural network, the convolution kernel in the convolution layer uses a small 3×3 convolution kernel, and the maximum pooling layer among the five pooling layers uses a 2×2 filter.

[0077] Generally, in order to obtain higher accuracy, the network will use dozens or even hundreds of convolution kernels. The increase in the number of convolution kernels will extract more detailed features of the target object, but it will also cause feature redundancy. Moreover, the more convolution kernels there are, the closer the relationship between different convolution kernels will be, or some convolution kernels will produce linear relationships. Therefore, this method is based on Python and uses the convolution neural network VGG16 to extract the features of the working surface. Then, the principal component analysis method is used to reduce the dimensionality of the convolution kernels used in the convolution neural network (such as Figure 5 As shown in Figure 2, selecting the part of the convolution kernel where the cumulative contribution rate of the main components reaches 85% can reduce the accuracy drop caused by feature redundancy and improve the operation speed.

[0078] The present invention uses principal component analysis and convolutional neural network coupling algorithm to adjust and monitor the drilling rig body posture, neural network extracts working surface features, uses principal component analysis to reduce the feature dimension, and retains the first n main features. It can reduce the number of features and the amount of calculation, and can also reduce the accuracy error caused by feature redundancy.

[0079] In this embodiment, during the previous working process described in step S4, the position and posture of the drilling rig are recorded in real time to calculate the deviation, and the changing law of the position and posture and the time when the deviation range is exceeded are learned; the specific process is:

[0080] Step S401, calculating the maximum deviation of the drilling trolley body when the drill bit is stuck, setting a threshold based on the deviation, and recording the time interval for reaching the threshold each time;

[0081] In specific implementation, the maximum deviation of the drilling rig body that may cause the drill to get stuck is calculated based on the drilling requirements and the drill arm size;

[0082] Step S402, during the working process, real-time recording of the displacement law of the vehicle body caused by the force generated by the drill arm in the working range, position and process; different displacement laws of the vehicle body posture caused by different situations will result in different times to reach the threshold;

[0083] Step S403: Use the collected data to improve the training set of the rock drilling rig and autonomously adjust the posture of the training rock drilling rig to improve the prediction accuracy of the rock drilling rig.

[0084] In this embodiment, when the work described in step S5 is formally carried out, according to the drilling rig posture prediction model, when the posture adjustment is performed once at a certain time interval, the time interval for each position adjustment of the drilling rig is set according to the time interval obtained in step S4. In principle, the actual designed time interval is less than the time interval obtained in step S4. Due to the complex environment, the deviation of the drilling rig must always be kept within an acceptable range, so the time interval set in step S5 is less than the time interval learned in step S4.

[0085] When implementing it, Figure 4 As shown in the figure, when the posture is solved at point m, the drilling rig is in the standard posture; when the posture deviates to point n, the distance offset of the vehicle body is calculated from the distance of the line segment mn, and the angle offset of the vehicle body is calculated from the midline of the triangle, thereby solving the posture offset of the drilling rig.

[0086] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent structural change made to the above embodiment based on the technical essence of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for positioning and compensating a drilling rig body, characterized in that: The method comprises the following steps: Step S1, based on the position of the drilling rig relative to the working surface, placing the laser target and the drilling rig at appropriate positions so that all the imaging points of the laser target fall within the working surface and can be recognized by the depth camera installed on the drilling rig; Step S2, adjusting the imaging of the laser target on the working surface so that the three points projected by the laser target form an upright isosceles triangle and are evenly located in the middle of the working surface; adjusting the angle of the depth camera so that the working surface is evenly centered in the camera imaging; Step S3, based on the position and posture adjusted in step S2, extracting the existing position and posture information as the standard posture of the drilling rig; Step S4, during the previous working process, the position and posture of the drilling rig are recorded in real time to calculate the deviation, and the changing pattern of the position and posture and the time when the deviation range is exceeded are learned; Step S5: When the work is officially carried out, the drilling rig posture adjustment is performed at regular intervals according to the drilling rig posture prediction model.

2. A method for positioning and compensating a drilling rig body according to claim 1, characterized in that: In step S3, based on the position and posture adjusted in step S2, the existing position and posture information is extracted as the standard posture of the drilling rig, and the principal component analysis method and the convolutional neural network coupling algorithm are adopted.

3. A method for positioning and compensating a drilling rig body according to claim 2, characterized in that: In step S3, the principal component analysis method and the convolutional neural network coupling algorithm are used to extract the existing position and posture information as the standard posture of the drilling rig based on the position and posture adjusted in step S2. The specific process is as follows: Step S301, using a depth camera to capture working surface image information, and passing it into an image feature extraction module of a convolutional neural network; Step S302: perform principal component analysis on the convolution kernel used for feature extraction in the convolutional neural network, and only retain the part whose cumulative contribution rate is higher than the cumulative contribution rate threshold; wherein the calculation formula of the cumulative contribution rate is: Among them, η represents the cumulative contribution rate, p refers to the number of eigenvalues ​​of the parameter matrix of the convolution kernel used for feature extraction in the convolutional neural network, and λ k Refers to the kth eigenvalue after arranging the eigenvalues ​​from large to small; Step S303: Use the obtained reduced-dimensional feature image as a standard posture image of the drilling rig.

4. A method for positioning and compensating a drilling trolley body according to claim 3, characterized in that: The convolutional neural network is a VGG16 convolutional neural network, which includes 13 convolutional layers, 3 fully connected layers and 5 pooling layers. The convolutional layers and pooling layers are used for feature extraction, and the fully connected layers are used for classification. In step S301, the image information of the working surface is captured by the depth camera and transmitted to the image feature extraction module of the convolutional neural network, and then the following process is performed: Step S3011, firstly, perform two convolution operations to extract the main feature information of the image; then activate it by the ReLU activation function so that the model can better fit the nonlinear function; finally, perform the maximum pooling layer to reduce the image size and the amount of calculation; repeat this process again for the obtained image and output the image; Step S3012, the image processed in step S3011 is first subjected to three convolution operations, then activated by the ReLU activation function, and the output image is finally subjected to the maximum pooling layer to reduce the image size, and the processed image is repeated twice to obtain the result image; In step S302, principal component analysis is performed on the convolution kernel used for feature extraction in the convolutional neural network. The original matrix of the standardized processing in the principal component analysis method is: in, m is the number of matrix rows, n is the number of matrix columns; The normalized matrix is: The process of principal component analysis is: Step S3021, calculate the covariance matrix R of the standardized sample X: in, ; Step S3022: Calculate the eigenvalues ​​of the correlation coefficient matrix λ1, λ2, ...λ n And the corresponding eigenvectors, and arrange the eigenvectors into a matrix row by row from top to bottom according to the corresponding eigenvalue size, and take the first k rows to form the matrix P; Step S3023: multiply the original data set by the eigenvector matrix P to obtain a matrix after dimension reduction.

5. A method for positioning and compensating a drilling trolley body according to claim 3 or 4, characterized in that: The cumulative contribution rate threshold is 85%.

6. A method for positioning and compensating a drilling trolley body according to claim 4, characterized in that: In the VGG16 convolutional neural network, the convolution kernel in the convolution layer uses a small 3×3 convolution kernel, and the maximum pooling layer in the five pooling layers uses a 2×2 filter.

7. A method for positioning and compensating a drilling trolley body according to claim 4, characterized in that: In the previous working process described in step S4, the position and posture of the drilling rig are recorded in real time to calculate the deviation, and the changing law of the position and posture and the time when the deviation range is exceeded are learned; the specific process is: Step S401, calculating the maximum deviation of the drilling trolley body when the drill bit is stuck, setting a threshold based on the deviation, and recording the time interval for reaching the threshold each time; Step S402, during the working process, real-time recording of the displacement law of the vehicle body caused by the force generated by the drill arm in the working range, position and process; different displacement laws of the vehicle body posture caused by different situations will result in different times to reach the threshold; Step S403: Use the collected data to improve the training set of the rock drilling rig and autonomously adjust the posture of the training rock drilling rig to improve the prediction accuracy of the rock drilling rig.

8. A method for positioning and compensating a drilling rig body according to claim 4, characterized in that: When the work described in step S5 is formally carried out, according to the drilling rig posture prediction model, the posture adjustment is performed once at a certain period of time. The time interval for each position adjustment of the drilling rig is set according to the time interval obtained in step S4. The principle is that the actual designed time interval is less than the time interval obtained in step S4.

Images