Automatic data separation method for each stage of shield tunneling process and computer readable medium

Abstract

The present invention relates to a method for automatically separating data at each stage of a shield tunneling process and a computer-readable medium, wherein the method for automatically separating data includes: obtaining tunneling parameter data; extracting tunneling status data; judging whether the time nodes of adjacent data are continuous, and dividing each initial tunneling cycle segment; smoothing the data; calculating the minimum peak point of each tunneling stage of the data, and dividing the data into a single tunneling cycle segment; calculating the maximum peak point of the single tunneling cycle segment, and dividing the data into ascending segment data and steady tunneling segment data segmentation points according to the first maximum peak point; performing piecewise linear fitting on the data to calculate the slope, and using the positive and negative change points of the slope as the segmentation points of the steady tunneling segment and the descending segment to achieve separation of the steady segment and the descending segment; obtaining the separated ascending segment, steady tunneling segment and descending segment data. Compared with the prior art, the present invention has the advantages of being able to effectively avoid misjudgment and complete data recognition.

Description

Technical Field

[0001] The invention relates to the technical field of unmanned shield construction, and in particular to a method for automatically separating data at various stages of a shield tunneling process and a computer-readable medium. Background Art

[0002] Shield construction is moving towards intelligentization from manual control to unmanned driving, in which autonomous decision-making of tunneling parameters is an extremely important part. For this reason, a large number of scholars have used machine learning algorithms to train models based on the collected data and try to achieve autonomous recommendation of tunneling parameters. Among them, the data used for machine learning model training is based on the shield machine's stable stage data, so the separation and extraction of the ascending section, stable section, and descending section data is extremely important.

[0003] In the prior art, research on the separation of shield tunneling parameters is ongoing. For example, Chinese patent CN110110419A discloses a TBM tunneling parameter prediction method based on multi-objective learning. This method obtains the original data of TBM tunneling parameters through the TBM intelligent data acquisition system, performs noise reduction and enhancement processing on the collected TBM tunneling parameters, obtains the data of the rising section and the stable section of the cycle, and uses artificial intelligence algorithms to effectively characterize the nonlinear mapping relationship between the TBM characteristic indicators of the rising section and the multi-objective variables of the stable section, and establishes a rock-machine interaction model based on multi-objective learning. Chinese patent CN110852423A discloses a tunneling performance and control parameter prediction method based on transfer learning. This method collects tunneling data of the tunneling machine reflecting the operating status of the equipment online, pre-processes the tunneling data of the tunneling machine, extracts the rising section and stable section data of each tunneling cycle, and uses deep learning methods to establish a tunneling performance parameter prediction model and a control parameter prediction model for completed projects. Chinese patent CN110895730A discloses a method for predicting TBM excavation parameters based on the LSTM algorithm. This method selects key excavation parameters in the construction excavation process, automatically separates the rising stage and the stable stage in the TBM excavation process through the RSPRA algorithm, and uses the LSTM algorithm for model training to realize the prediction of the excavation parameters in the stable stage.

[0004] The above methods all separate the rising segment data and the stable segment data based on the time continuity of the excavation state or based on the continuous positive slope. However, due to the volatility of the data and the multiple adjustments of the rising segment, it is difficult to ensure that the slope of the rising segment data is always positive. Therefore, these methods can often only identify part of the rising segment and cannot accurately identify all the rising segment data. Summary of the invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a method and computer-readable medium for automatically separating data at each stage of the shield tunneling process, which can effectively avoid misjudgment and have complete data identification.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A method for automatically separating data at each stage of a shield tunneling process, the data separation method comprising:

[0008] Step 1: Obtain excavation parameter data;

[0009] Step 2: Extracting excavation status data from the data obtained in step 1 according to the core excavation parameters;

[0010] Step 3: Determine whether the time nodes of adjacent data are continuous in the data extracted in step 2, and divide the data into initial excavation cycle segments according to the discontinuous time nodes;

[0011] Step 4: Smooth the data in step 3;

[0012] Step 5: Calculate the minimum peak point for each excavation stage of the data in step 4, and divide the data into single excavation cycle segments according to the minimum peak point;

[0013] Step 6: Calculate the maximum peak point of a single excavation cycle segment, and divide the data into ascending segment data and steady excavation segment data division points according to the first maximum peak point;

[0014] Step 7: Perform piecewise linear fitting on the data to calculate the slope, and use the positive and negative change points of the slope as the dividing points between the steady excavation section and the descending section to separate the steady section from the descending section;

[0015] Step 8: Obtain the data of the separated ascending section, steady excavation section and descending section.

[0016] Preferably, the tunneling parameter data acquired in step 1 include the shield machine cutter head rotation speed, cutter head torque, propulsion speed and total propulsion force.

[0017] Preferably, the step 2 is specifically:

[0018] The data in the excavation state is identified by the fact that the four parameters of the shield machine cutter head speed, cutter head torque, propulsion speed and total propulsion force are simultaneously greater than zero, that is, the data when the shield machine cutter head speed, cutter head torque, propulsion speed and total propulsion force are simultaneously greater than zero are extracted.

[0019] Preferably, the step 4 uses a moving window smoothing polynomial to smooth the data, and the specific steps are:

[0020] Step 4-1: Given the window width w and the threshold of the fitting polynomial n;

[0021] Step 4-2: Delimit the data with a window width of w from the starting position, use the least squares method to fit the data with an n-order polynomial to obtain a fitting function, and use the fitting function to recalculate the value of the window center position to replace the original data;

[0022] Step 4-3: Move the window backward and repeat step 4-2, that is, continuously update the value at the center of the window until the window reaches the final position.

[0023] Preferably, the step 5 is specifically:

[0024] Step 5-1: Multiply the data processed in step 4 by -1, and then identify the maximum value point. The specific method is:

[0025] By comparing the sizes of the three adjacent points, if the value of a point is greater than the values ​​of the adjacent previous and next points at the same time, then the point is the maximum point. The judgment formula is:

[0026] T(i-1)<T(i)>T(i+1)

[0027] Among them, T(i) represents the i-th point, T(i-1) represents the i-1-th point, and T(i+1) represents the i+1-th point;

[0028] Step 5-2: Screen the maximum peak point. The specific method is as follows:

[0029] Preset the thresholds of peak prominence p, peak width W and distance d between peaks, calculate the peak prominence p, peak width W and distance d between peaks at each maximum point in step 5-1, and when the peak prominence p, peak width W and distance d between peaks all meet the threshold requirements, then judge the maximum point as the maximum peak point;

[0030] Step 5-3: The maximum peak point selected in step 5-2 is the minimum peak point of the original data, and the data is segmented according to the minimum peak point;

[0031] Step 5-4: Based on the segmented cyclic segment data, if the number of samples of the segment data is less than a preset threshold, the cyclic segment data is removed.

[0032] More preferably, the calculation method of the peak protrusion p and peak width W in 5-2 is:

[0033] p(i)=T(i)-max(minT l ,minT r )

[0034] W(i)=T(i)-p(i)×r

[0035] Where p(i) is the peak prominence of the i-th maximum point; W(i) is the peak width of the i-th maximum point; T(i) is the value of the i-th maximum point; T l and T r They are the intersection areas of the maximum point contour lines and the first larger extreme points on the left and right; r is the calculation position of the peak width.

[0036] Preferably, the step 6 is specifically:

[0037] Step 6-1: By comparing the values ​​of the three adjacent points, if the value of a point is greater than the values ​​of the adjacent previous and next points at the same time, then the point is the maximum value point. The judgment formula is:

[0038] T(i-1)<T(i)>T(i+1)

[0039] Among them, T(i) represents the i-th point, T(i-1) represents the i-1-th point, and T(i+1) represents the i+1-th point;

[0040] Step 6-2: Screen the maximum peak point. The specific method is as follows:

[0041] Preset the thresholds of peak prominence p, peak width W and distance d between peaks, calculate the peak prominence p, peak width W and distance d between peaks at each maximum point in step 5-1, and when the peak prominence p, peak width W and distance d between peaks all meet the threshold requirements, then judge the maximum point as the maximum peak point;

[0042] Step 6-3: Divide the data into ascending segment data and steady excavation segment data according to the first maximum peak point.

[0043] Preferably, the method for calculating the slope by piecewise linear fitting in step 7 is specifically:

[0044] First, the allowable fitting error threshold is given;

[0045] Secondly, determine the segmentation point. Since the linear fitting error between two points must be zero, we continue to add data points until the fitting error reaches the threshold, which means the segmentation is completed.

[0046] Finally, the current fitted slope of each segment represents the slope of that local area in the original data.

[0047] Preferably, the specific identification method of the dividing point between the steady excavation section and the descending section in step 7 is:

[0048] The slope of the descending section is always negative. Starting from the end of the data, look forward for the first data point with a positive slope. This point is the dividing point between the steady excavation section and the descending section, that is, the end point of the steady excavation section and the starting point of the descending section.

[0049] A computer-readable medium stores a method for automatically separating data at each stage of a shield tunneling process as described in any one of the above items.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] 1. Avoiding misjudgment: The method for automatically separating data at each stage of the shield tunneling process in the present invention can conveniently divide the data into independent single-cycle tunneling segments by identifying the minimum peak point of the initial tunneling cycle segment, thereby avoiding the appearance of data from multiple tunneling cycle segments. At the same time, the identification of the starting point of the rising segment is relatively accurate, thereby avoiding misjudgment as an empty push segment.

[0052] 2. Complete data identification: The automatic data separation method for each stage of the shield excavation process in the present invention can identify the complete rising segment data by identifying the maximum peak point of a single excavation cycle, avoiding the situation where only part of the rising segment data can be identified due to the continuous positive slope; at the same time, the data segmentation points of the stable segment and the descending segment are identified by identifying the positive and negative change points of the slope, avoiding the situation where the stable segment data contains the descending segment data. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flow chart of the method for automatically separating data at each stage of the shield tunneling process in the present invention;

[0054] Figure 2 This is a cutter head torque variation trend diagram of raw data of a certain ring in a certain area in an embodiment of the present invention;

[0055] Figure 3 A diagram showing a trend of cutter head torque changes after identification of a tunneling state of a certain ring in a certain area in an embodiment of the present invention;

[0056] Figure 4 In the embodiment of the present invention, Figure 3 A change trend diagram of the cutter head torque in the first initial cycle segment identified in;

[0057] Figure 5 In the embodiment of the present invention, Figure 3 A change trend diagram of the cutter head torque in the second initial cycle segment identified in;

[0058] Figure 6 Schematic diagram of the flow of a moving window smoothing polynomial algorithm in an embodiment of the present invention;

[0059] Figure 7 In the embodiment of the present invention Figure 4 Smoothing effect diagram;

[0060] Figure 8 In the embodiment of the present invention Figure 5 Smoothing effect diagram;

[0061] Fig. 9 Schematic diagram of peak protrusion p and peak width W in an embodiment of the present invention;

[0062] Fig.10 In the embodiment of the present invention Figure 7 The minimum peak point identification result diagram;

[0063] Fig.11 It is a schematic diagram of judging the rising section and the stable section of the existing method in an embodiment of the present invention;

[0064] Fig.12 In the embodiment of the present invention Fig.10 The separation result diagram of Fig.12 (a) Fig.12 (b) and Fig.12 (c) Three independent single-cycle segments identified respectively;

[0065] Fig.13 In the embodiment of the present invention Fig.12 The maximum peak point identification result diagram of Fig.13 (a) Fig.13 (b) and Fig.13 (c) respectively Fig.12 (a) Fig.12 (b) and Fig.12 (c) Maximum peak point identification result;

[0066] Fig.14 In this embodiment Fig.12 The result diagram of the identification of the stationary segment end point of Fig.14 (a) Fig.14 (b) and Fig.14 (c) respectively Fig.12 (a) Fig.12 (b) and Fig.12 (c) the identification result;

[0067] Fig.15 In this embodiment Fig.12 The identification results of the rising section, steady excavation section and descending section of Fig.15 (a) Fig.15 (b) and Fig.15 (c) respectively Fig.12 (a) Fig.12 (b) and Fig.12 (c) The recognition result. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0069] According to the search results and related papers, in order to predict the stable stage data, it is necessary to separate the rising segment data from the stable segment data before model training. The existing method is based on the separation method of the rising segment and the stable segment data of the shield machine excavation parameters. The steps are as follows:

[0070] (1) Extraction of excavation cycle: the total thrust and cutter torque of the shield machine are greater than zero, and the penetration is less than 20 mm / rev. The data is the excavation state data. According to the excavation time record of the shield machine, the continuous excavation time of more than 600 s is the excavation cycle segment;

[0071] (2) Outlier judgment and processing: Calculate the mean μ and standard deviation σ of the penetration and excavation speed of the entire excavation process. Based on the 3σ criterion, when the excavation speed and penetration are greater than μ+3σ, treat them as outliers and take the average of the five data points adjacent to the outlier to replace the value;

[0072] (3) Data smoothing: In order to eliminate the influence of white noise, the average value of 15 adjacent data is taken to filter the data;

[0073] (4) The rising segment and the stable segment data are separated. The torque curve is divided into 10 segments using piecewise linear fitting. The segmentation method with the smallest error is the segmentation point. The slope of each segment is calculated separately. The segment with the largest slope is the starting position of the rising segment. The segment with continuous positive slope is the entire rising segment data. The data point with a smaller slope after the end of the rising segment is the stable segment.

[0074] The main disadvantage of the above method is that the excavation cycle segment extraction will include multiple excavation cycle segments rather than a single excavation cycle segment extracted. This is because when the adjacent geological environment does not change much, the next excavation cycle of the shield machine will not start from the static state of the previous excavation cycle segment, but will reduce the excavation parameters of the previous excavation cycle to a certain extent (core excavation parameters such as cutter head torque and propulsion speed are still greater than 0), and slowly rise to directly form the next excavation cycle segment. Therefore, the above method uses the data of total propulsion force and cutter head torque greater than zero and penetration less than 20mm / rev to identify the excavation state. It will include multiple excavation cycle segments. Secondly, taking the point with the largest slope as the starting point of the rising segment is prone to misjudgment. The slope of the empty push segment when the shield machine is started is large, so judging the starting point of the rising segment by the slope is prone to misjudgment as an empty push segment. Thirdly, when separating the rising segment and the stable segment data, taking the segment with a continuous positive slope will result in only identifying part of the excavation segment, but not all of the rising segment data. This is because, on the one hand, when the geological environment is complex, the rising segment data will be adjusted twice or even multiple times, and on the other hand, the volatility of the data will cause the slope to be negative locally. Taking the segment with a continuous positive slope can only identify the first adjusted rising segment or part of the rising segment data, but not all of the rising segment data. Finally, the stable segment data includes the descending segment data, and there is no accurate separation of the stable segment and the descending segment data.

[0075] This embodiment relates to a method for automatically separating data at each stage of a shield tunneling process, and its process is as follows: Figure 1 As shown, including:

[0076] Step 1: According to the historical tunneling records of the shield machine, tunneling parameter data is obtained, including but not limited to the cutter head speed, cutter head torque, propulsion speed and total propulsion force of the shield machine;

[0077] Step 2: Extracting excavation status data from the data obtained in step 1 according to the core excavation parameters;

[0078] The shield machine collects data in real time, including a large amount of data when it is in the shutdown state. The data in the excavation state is identified by the fact that the four parameters of excavation speed, cutter head speed, total propulsion force, and cutter head torque are all greater than zero at the same time; the original data and excavation state data identification results of a certain ring in a certain area are as follows Figure 2 and Figure 3 shown.

[0079] Step 3: Determine whether the time nodes of adjacent data are continuous in the data extracted in step 2, and divide the data into initial excavation cycle segments according to the discontinuous time nodes;

[0080] Determine whether the time nodes of adjacent data of each excavation cycle segment after processing in step 2 are continuous, wherein the discontinuous time nodes are the segmentation points, i.e., the critical points between the previous cycle and the next cycle, and divide the data into initial excavation cycle segments according to this point;

[0081] Figure 4 and Figure 5 Given Figure 3 The identification results of the initial excavation cycle segment show that Figure 4 For data containing multiple driving cycle segments, Figure 5 The data is for a single driving cycle section;

[0082] Step 4: Smooth the data in step 3. The process is as follows: Figure 6 As shown;

[0083] Step 4-1: Given a window width w and a threshold of a fitting polynomial n, in this embodiment, w=31, n=2, where the window width w must be an odd number;

[0084] Step 4-2: Delimit the data with a window width of w from the starting position, use the least squares method to fit the data with an n-order polynomial to obtain a fitting function, and use the fitting function to recalculate the value of the window center position to replace the original data;

[0085] Step 4-3: Move the window backward and repeat step 4-2, that is, continuously update the value at the center of the window until the window reaches the final position;

[0086] Figure 4 , Figure 5 The smoothing result is Figure 7 , Figure 8 As shown;

[0087] It should be noted that the smoothing method in this embodiment may also use a wavelet transform denoising algorithm to replace the moving window smoothing polynomial algorithm.

[0088] Step 5: Calculate the minimum peak point for each excavation stage of the data in step 4, and divide the data into single excavation cycle segments according to the minimum peak point;

[0089] The identification of the minimum peak point can be solved by multiplying the cyclic excavation section data by -1 and identifying the maximum peak point. The maximum peak point must be at the maximum value point. Therefore, the maximum value point is first identified, and the maximum value is screened according to the characteristics of the peak. Finally, the maximum value point that meets the conditions is the maximum peak point. The specific steps for maximum peak point identification are as follows:

[0090] Step 5-1: Multiply the data processed in step 4 by -1, and then identify the maximum value point. The specific method is:

[0091] By comparing the sizes of the three adjacent points, if the value of a point is greater than the values ​​of the adjacent previous and next points at the same time, then the point is the maximum point. The judgment formula is:

[0092] T(i-1)<T(i)>T(i+1)

[0093] Among them, T(i) represents the i-th point, T(i-1) represents the i-1-th point, and T(i+1) represents the i+1-th point;

[0094] Step 5-2: Screen the maximum peak point. The specific method is as follows:

[0095] Preset the thresholds of peak prominence p, peak width W and distance d between peaks, calculate the peak prominence p, peak width W and distance d between peaks at each maximum point in step 5-1, and when the peak prominence p, peak width W and distance d between peaks all meet the threshold requirements, then judge the maximum point as the maximum peak point;

[0096] The peak prominence p and peak width W are shown as follows: Fig. 9 As shown, the calculation methods are:

[0097] p(i)=T(i)-max(minT l ,minT r )

[0098] W(i)=T(i)-p(i)×r

[0099] Where p(i) is the peak prominence of the i-th maximum point; W(i) is the peak width of the i-th maximum point; T(i) is the value of the i-th maximum point; T l and T r are the intersection areas of the maximum point contour lines and the first larger extreme points on the left and right; r is the calculation position of the peak width;

[0100] In this embodiment, the peak prominence threshold is 200, the peak width threshold is 30, the minimum distance threshold between peaks is 300, and the peak width calculation position is 0.5, that is, the half-peak position. After the above processing, the maximum peak point caused by abnormality or noise can be eliminated. The maximum peak point corresponds to the minimum peak point of the original data. Figure 7 The recognition results are shown below Fig.10 As shown in the figure, this point is the starting point of the ascending section in each excavation cycle section. The existing method does not separate multiple excavation cycle sections, and takes the point with the largest slope as the starting point of the ascending section, which is prone to misjudgment. Fig.11 As shown in the figure, in fact, the misjudged shield machine has not yet advanced, but is only in the startup stage, also known as the empty push stage, which often has a large slope. Therefore, it is easy to make a misjudgment, resulting in an error in the identification of the starting position of the ascending section. The misjudgment of the ascending section will further lead to an error in the identification of the stable section.

[0101] Step 5-3: The maximum peak point selected in step 5-2 is the minimum peak point of the original data, and the data is segmented according to the minimum peak point; Fig.10 The separation results are as follows Fig.12 As shown, Fig.12 (a) Fig.12 (b) and Fig.12 (c) Three independent single-cycle segments identified respectively;

[0102] Step 5-4: The shield machine will continue to excavate for a period of time during normal excavation, so the samples of each section of normal excavation should be greater than a certain predetermined number. Therefore, it is necessary to exclude the data of each cycle segment after segmentation if the number of samples of the segment is less than the preset threshold. In this embodiment, the duration of normal excavation of the shield machine is at least 300s, so the preset threshold is set to 300, that is, the samples with a number less than 300 are excluded;

[0103] Step 6: Calculate the maximum peak point of a single excavation cycle segment, and divide the data into ascending segment data and steady excavation segment data division points according to the first maximum peak point;

[0104] The calculation method of the maximum peak point is:

[0105] Step 6-1: By comparing the values ​​of the three adjacent points, if the value of a point is greater than the values ​​of the adjacent previous and next points at the same time, then the point is the maximum value point. The judgment formula is:

[0106] T(i-1)<T(i)>T(i+1)

[0107] Among them, T(i) represents the i-th point, T(i-1) represents the i-1-th point, and T(i+1) represents the i+1-th point;

[0108] Step 6-2: Screen the maximum peak point. The specific method is as follows:

[0109] Preset the thresholds of peak prominence p, peak width W and distance d between peaks, calculate the peak prominence p, peak width W and distance d between peaks at each maximum point in step 5-1, and when the peak prominence p, peak width W and distance d between peaks all meet the threshold requirements, then judge the maximum point as the maximum peak point;

[0110] The calculation methods of peak prominence p and peak width W are:

[0111] p(i)=T(i)-max(minT l ,minT r )

[0112] W(i)=T(i)-p(i)×r

[0113] Where p(i) is the peak prominence of the i-th maximum point; W(i) is the peak width of the i-th maximum point; T(i) is the value of the i-th maximum point; T l and T r are the intersection areas of the maximum point contour lines and the first larger extreme points on the left and right; r is the calculation position of the peak width;

[0114] In this embodiment, the threshold of the peak prominence is set to 100 when identifying the maximum peak point, and the other parameters are the same as the thresholds when identifying the minimum peak point in step 5. Fig.12 The maximum peak point identification result is as follows Fig.13 As shown, Fig.13 (a) Fig.13 (b) and Fig.13 (c) respectively Fig.12 (a) Fig.12 (b) and Fig.12 (c) the identification result;

[0115] Step 6-3: Divide the data into ascending segment data and steady excavation segment data division points according to the first maximum peak point;

[0116] Step 7: Perform piecewise linear fitting on the data to calculate the slope, and use the positive and negative change points of the slope as the dividing points between the steady excavation section and the descending section to separate the steady section from the descending section;

[0117] The slope is calculated by piecewise linear fitting. This is because discrete data is still fluctuating data after noise reduction. Replacing the differential slope of the data with the slope of local linear fitting is more accurate. The core of segmentation is the calculation of the segmentation point. The segmentation point can be found by fitting error, and the local slope can be calculated according to the segmentation point. The specific steps are as follows:

[0118] First, the allowable fitting error threshold is given;

[0119] Secondly, determine the segmentation point. Since the linear fitting error between two points must be zero, we continue to add data points until the fitting error reaches the threshold, which means the segmentation is completed.

[0120] Finally, the current fitted slope of each segment represents the slope of that local area in the original data.

[0121] The specific identification method of the dividing point between the steady excavation section and the descending section is:

[0122] The slope of the descending section is always negative. Start from the end of the data and look forward for the first data point with a positive slope. This point is the dividing point between the steady excavation section and the descending section, that is, the end point of the steady excavation section and the start point of the descending section. Fig.12 The recognition results are as follows Fig.14As shown, Fig.14 (a) Fig.14 (b) and Fig.14 (c) respectively Fig.12 (a) Fig.12 (b) and Fig.12 (c) the identification result;

[0123] Step 8: Obtain the data of the separated ascending section, steady excavation section and descending section. Fig.12 The recognition results are as follows Fig.15 As shown, Fig.15 (a) Fig.15 (b) and Fig.15 (c) respectively Fig.12 (a) Fig.12 (b) and Fig.12 (c) The recognition result.

[0124] This embodiment also relates to a computer-readable medium, in which any of the above-mentioned methods for automatically separating data at various stages of a shield tunneling process is stored.

[0125] The core points of the above-mentioned automatic data separation method at each stage of shield tunneling process are:

[0126] (1) Using the minimum peak point to divide the excavation cycle segment into single excavation cycle segment data

[0127] When the geological environment does not change much, the shield machine will not stop in the previous excavation cycle, but the excavation parameters will slowly decrease to a certain value and then directly transition to the next cycle excavation segment. In traditional methods, this situation cannot be solved only by the excavation state and time interval. Therefore, in order to obtain the data of a single excavation cycle segment, the initial excavation cycle segment data must be segmented. This method proposes to use the minimum peak point to conveniently segment the data into independent excavation cycle segment data, which is also the starting point of the rising segment in each excavation cycle segment.

[0128] (2) Using the maximum peak point to divide the single excavation cycle into the rising segment and the stable excavation segment data

[0129] In the existing methods, the slope is continuously positive to judge the rising segment data. In fact, on the one hand, due to the existence of noise (although the noise can only be weakened but not eliminated after noise reduction), the local slope may be positive. On the other hand, sometimes the rising segment data will be adjusted twice or even multiple times. Therefore, only by continuously positive slopes can only part of the excavation rising segment data be identified, and the complete excavation rising segment data cannot be identified. Since the end of the excavation rising segment to the stable segment will form a maximum peak point, this method proposes to divide the excavation cycle segment into the rising segment and the stable excavation segment data through the first maximum peak point.

[0130] (3) Separate the stable segment and the descending segment by using the positive and negative change points of the slope

[0131] The data of the descending section changes smoothly because the excavation data does not need to be adjusted continuously according to the rock-machine interaction like the ascending section. Therefore, the slope change of the descending section data can obtain accurate results. Since the slope of the descending section data is negative, the data is calculated from back to front, and the point where the slope changes from negative to positive is the dividing point between the stable section and the descending section.

[0132] This method separates the ascending segment and the stable segment data based on the identification of the maximum peak point, and can easily identify all the ascending segment data, thereby achieving effective separation of the ascending segment and the stable segment data. On the other hand, none of the current methods considers splitting the fused multi-excavation cycle segment data into single-excavation cycle segment data, and using the point with the largest slope as the starting point of the ascending segment is easy to misjudge as an empty push segment, and no method for separating the stable segment and the descending segment data is given, resulting in a certain deviation in the extracted training data. Therefore, this patent is significantly different from the currently disclosed patents and papers, and has obvious innovation.

[0133] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A method for automatically separating data at each stage of shield tunneling process, characterized in that: The automatic data separation method comprises: Step 1: Obtain excavation parameter data; Step 2: Extracting excavation status data from the data obtained in step 1 according to the core excavation parameters; Step 3: Determine whether the time nodes of adjacent data are continuous in the data extracted in step 2, and divide the data into initial excavation cycle segments according to the discontinuous time nodes; Step 4: Smooth the data in step 3; Step 5: Calculate the minimum peak point for each excavation stage of the data in step 4, and divide the data into single excavation cycle segments according to the minimum peak point; Step 6: Calculate the maximum peak point of a single excavation cycle segment, and divide the data into ascending segment data and steady excavation segment data division points according to the first maximum peak point; Step 7: Perform piecewise linear fitting on the data to calculate the slope, and use the positive and negative change points of the slope as the dividing points between the steady excavation section and the descending section to separate the steady section from the descending section; Step 8: Obtain the data of the separated ascending section, steady excavation section and descending section.

2. The method for automatically separating data at each stage of shield tunneling process according to claim 1 is characterized in that: The excavation parameter data obtained in step 1 include the shield machine cutter head rotation speed, cutter head torque, propulsion speed and total propulsion force.

3. The method for automatically separating data at each stage of shield tunneling process according to claim 1 is characterized in that: The step 2 is specifically as follows: The data in the excavation state is identified by the fact that the four parameters of the shield machine cutter head speed, cutter head torque, propulsion speed and total propulsion force are simultaneously greater than zero, that is, the data when the shield machine cutter head speed, cutter head torque, propulsion speed and total propulsion force are simultaneously greater than zero are extracted.

4. The method for automatically separating data at each stage of shield tunneling process according to claim 1 is characterized in that: The step 4 uses a moving window smoothing polynomial to smooth the data, and the specific steps are: Step 4-1: Given the window width w and the threshold of the fitting polynomial n; Step 4-2: Delimit the data with a window width of w from the starting position, use the least squares method to fit the data with an n-order polynomial to obtain a fitting function, and use the fitting function to recalculate the value of the window center position to replace the original data; Step 4-3: Move the window backward and repeat step 4-2, that is, continuously update the value at the center of the window until the window reaches the final position.

5. The method for automatically separating data at each stage of shield tunneling process according to claim 1 is characterized in that: The step 5 is specifically as follows: Step 5-1: Multiply the data processed in step 4 by -1, and then identify the maximum value point. The specific method is: By comparing the sizes of the three adjacent points, if the value of a point is greater than the values ​​of the adjacent previous and next points at the same time, then the point is the maximum point. The judgment formula is: T(i-1)<T(i)>T(i+1) Among them, T(i) represents the i-th point, T(i-1) represents the i-1-th point, and T(i+1) represents the i+1-th point; Step 5-2: Screen the maximum peak point. The specific method is as follows: Preset the thresholds of peak prominence p, peak width W and distance d between peaks, calculate the peak prominence p, peak width W and distance d between peaks at each maximum point in step 5-1, and when the peak prominence p, peak width W and distance d between peaks all meet the threshold requirements, then judge the maximum point as the maximum peak point; Step 5-3: The maximum peak point selected in step 5-2 is the minimum peak point of the original data, and the data is segmented according to the minimum peak point; Step 5-4: Based on the segmented cyclic segment data, if the number of samples of the segment data is less than a preset threshold, the cyclic segment data is removed.

6. The method for automatically separating data at each stage of shield tunneling process according to claim 5 is characterized in that: The calculation method of the peak protrusion p and peak width W in 5-2 is: p(i)=T(i)-max(minT l ,minT r ) W(i)=T(i)-p(i)×r Where p(i) is the peak prominence of the i-th maximum point; W(i) is the peak width of the i-th maximum point; T(i) is the value of the i-th maximum point; T l and T r They are the intersection areas of the maximum point contour lines and the first larger extreme points on the left and right; r is the calculation position of the peak width.

7. The method for automatically separating data at each stage of shield tunneling process according to claim 1 is characterized in that: The step 6 is specifically as follows: Step 6-1: By comparing the values ​​of the three adjacent points, if the value of a point is greater than the values ​​of the adjacent previous and next points at the same time, then the point is the maximum value point. The judgment formula is: T(i-1)<T(i)>T(i+1) Among them, T(i) represents the i-th point, T(i-1) represents the i-1-th point, and T(i+1) represents the i+1-th point; Step 6-2: Screen the maximum peak point. The specific method is as follows: Preset the thresholds of peak prominence p, peak width W and distance d between peaks, calculate the peak prominence p, peak width W and distance d between peaks at each maximum point in step 5-1, and when the peak prominence p, peak width W and distance d between peaks all meet the threshold requirements, then judge the maximum point as the maximum peak point; Step 6-3: Divide the data into ascending segment data and steady excavation segment data according to the first maximum peak point.

8. The method for automatically separating data at each stage of shield tunneling process according to claim 1 is characterized in that: The method for calculating the slope by piecewise linear fitting in step 7 is specifically: First, the allowable fitting error threshold is given; Secondly, determine the segmentation point. Since the linear fitting error between two points must be zero, we continue to add data points until the fitting error reaches the threshold, which means the segmentation is completed. Finally, the slope of the linear fit for each segment represents the local slope in the original data.

9. The method for automatically separating data at each stage of shield tunneling process according to claim 1 is characterized in that: The specific identification method of the dividing point between the steady excavation section and the descending section in step 7 is: The slope of the descending section is always negative. Starting from the end of the data, look forward for the first data point with a positive slope. This point is the dividing point between the steady excavation section and the descending section, that is, the end point of the steady excavation section and the starting point of the descending section.

10. A computer-readable medium, characterized in that The computer-readable medium stores a method for automatically separating data at each stage of a shield tunneling process according to any one of claims 1 to 9.

Images