Dry-method precise positioning construction control system and method based on machine learning

Abstract

The invention belongs to the technical field of construction control, and particularly discloses a dry-method precise positioning construction control system and method based on machine learning. Multi-dimensional data such as drilling resistance, vibration frequency, shovel head angle deviation, position and soil layer hardness are acquired in real time through a multi-source data acquisition module; the deviation calculation module carries out dynamic comparison analysis based on data corresponding to a preset sampling area, the dynamic adjustment module dynamically optimizes the drilling speed and angle through a multi-condition triggering mechanism and a transfer adjustment rule, and finally the control execution module generates a precise control instruction to drive the Luoyang shovel to execute deviation correction action. Through multi-dimensional data fusion analysis and a real-time dynamic adjustment strategy, the path precision and the stratum adaptability of the drilling operation are remarkably improved, meanwhile, the adaptability of equipment to stratum sudden change is improved, the construction risk under the complex geological condition is reduced, and collaborative optimization of the drilling efficiency and the operation safety is achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of construction control. Specifically, it relates to a dry-method precise positioning construction control system and method based on machine learning. Background Art

[0002] In the field of building construction, traditional pile foundation construction methods have many drawbacks. In this context, the dry-method construction method using a Luoyang shovel, due to its own characteristics, brings new opportunities for dry-method precise positioning construction control. In order to ensure the quality of pile foundation construction, it is necessary to accurately position its construction process.

[0003] The prior art, such as a pile foundation construction control method, system, terminal and storage medium disclosed in a Chinese invention patent application with the application number 202410770492.6, includes obtaining the current position and the preset position, leveling the pile driver through control algorithm matching, obtaining parameters for excavation, and performing pressure grouting after detecting the stop instruction, and then improving the accuracy through preset algorithms and parameter adjustment.

[0004] The prior art, such as a full-automatic static pile driver positioning control system based on GPS technology disclosed in a Chinese invention patent application with the application number 201610441593.4, realizes the automatic positioning and horizontal adjustment of the pile driver by combining GPS technology, especially carrier phase differential technology, emphasizing high-precision positioning and automatic adjustment.

[0005] For the above technical solutions, obviously, there are still the following deficiencies in the current pile foundation construction control: 1. When predicting the soil layer, less attention is paid to the evaluation of the sampling area, mainly combined with past experience, resulting in certain deviations in the effectiveness and accuracy of soil layer prediction, and the reference is not strong.

[0006] 2. It depends on preset pile foundation parameters for excavation and deviation correction, but does not integrate the real-time soil layer data and the predicted soil layer data to verify the comprehensive deviation situation in the depth sequence, resulting in certain deviations in the guarantee of deviation correction.

[0007] 3. It only depends on external positioning data and does not combine the self-attitude of the Luoyang shovel and the change of soil layer resistance for multi-parameter collaborative adjustment, which is prone to cumulative errors in complex strata. Summary of the Invention

[0008] In view of this, in order to solve the problems raised in the above background art, a dry-method precise positioning construction control system and method based on machine learning are proposed.

[0009] The object of the present invention can be achieved by the following technical solutions: The present invention provides a dry method precise positioning construction control system based on machine learning. The system includes: a multi-source data acquisition module that collects drilling resistance, vibration frequency, deviation of the shovel head angle, and position in real time, and synchronously collects the hardness data of the current soil layer.

[0010] A deviation calculation module that calculates the deviation of the resistance, vibration frequency, and hardness of the soil layer corresponding to the preset sampling area in combination with the drilling resistance, vibration frequency, and hardness data.

[0011] A dynamic adjustment module that dynamically adjusts the drilling speed and drilling angle based on the drilling adjustment rules in combination with the deviation value of the shovel head angle and the difference in the actual path position.

[0012] A control execution module that generates a control instruction based on the adjusted drilling speed and drilling angle, and drives the Luoyang shovel to perform a deviation correction action.

[0013] Among them, the triggering conditions of the dynamic adjustment module include: any deviation amount exceeds the set threshold, or the difference in the actual path position exceeds the limit, or the deviation of the shovel head angle continues to exceed the time limit.

[0014] The present invention also provides a dry method precise positioning construction control method based on machine learning. The method includes: S1. Multi-source data acquisition: Collect drilling resistance, vibration frequency, deviation value of the shovel head angle, and position data in real time, and synchronously collect the hardness data of the current soil layer.

[0015] S2. Construction deviation calculation: Calculate the deviation of the resistance, vibration frequency, and hardness of the soil layer corresponding to the preset sampling area in combination with the drilling resistance, vibration frequency, and hardness data.

[0016] S3. Construction dynamic adjustment: Dynamically adjust the drilling speed and drilling angle based on the drilling adjustment rules in combination with the deviation value of the shovel head angle and the difference in the actual path position.

[0017] S4. Drilling control execution: Generate a control instruction based on the adjusted drilling speed and drilling angle, and drive the Luoyang shovel to perform a deviation correction action.

[0018] Compared with the prior art, the beneficial effects of the present invention are as follows: (1) The present invention obtains parameters such as drilling resistance, vibration frequency, angle deviation, position, and soil layer hardness in real time through multi-source data acquisition. After the deviation calculation module dynamically compares with the preset parameters, the dynamic adjustment module is triggered to optimize the drilling speed and angle based on multiple conditions. Finally, the control execution module generates a precise instruction to drive the deviation correction, realizing multi-dimensional data fusion and real-time dynamic adjustment, significantly improving the accuracy of the drilling path and the adaptability to the formation, and reducing the construction risk in complex geological conditions.

[0019] (2) Based on the planned drilling hole diameter and the critical safety distance calculated by the elastic mechanics formula, the present invention scientifically calibrates the preselected area, breaks through the dependence on traditional experience, and integrates soil parameters such as spectral reflectance and moisture content to calculate the Jaccard similarity coefficient, realizing the quantitative comparison of the soil characteristics between the planned drilling area and the preselected area, enhancing the effectiveness, accuracy, and reference of soil layer prediction, and providing reliable data support for the dynamic adjustment of drilling parameters.

[0020] (3) The present invention quantifies and evaluates the deviation of drilling parameters by constructing the depth sequence comparison of the actual resistance gradient / distribution curve and the preset sampling area control curve, and combines the weight compensation of the historical accident database, realizing the dynamic verification of the rectification action and ensuring the accuracy of the construction index adjustment under complex strata.

[0021] (4) By integrating drilling resistance, vibration frequency, angle deviation, position, and soil layer hardness, the dynamic adjustment module fuses angle deviation, position difference, and multi-parameter deviation amount, and combines weighted summation and Sigmoid function mapping to solve the cumulative error problem caused by single external positioning, realizing the differential precision and coordination control of drilling speed and angle under complex strata, and significantly improving the construction path accuracy and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for describing the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0023] Figure 1 It is a schematic diagram of the structural connection of each module of the system of the present invention.

[0024] Figure 2 It is a schematic diagram of the implementation steps flow of the method of the present invention.

[0025] Figure 3 It is a schematic diagram of the selection process of the preset sampling area of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] Please refer to Figure 1As shown, the present invention provides a dry method precise positioning construction control system based on machine learning, and the system includes: a multi-source data acquisition module, a deviation calculation module, a dynamic adjustment module and a control execution module.

[0028] In the above, the deviation calculation module is connected to the multi-source data acquisition module and the dynamic adjustment module respectively, and the dynamic adjustment module is connected to the control execution module.

[0029] The multi-source data acquisition module collects the drilling resistance, vibration frequency, shovel head angle deviation and position in real time, and simultaneously collects the hardness data of the current soil layer.

[0030] Specifically, the multi-source data acquisition module includes: obtaining the three-dimensional coordinates of the Luoyang shovel in real time through the RTK positioning module installed on the shovel body, and generating the real-time position of the Luoyang shovel.

[0031] A dual-axis inclination sensor is used to detect the deviation between the shovel head and the set angle as the shovel head angle deviation.

[0032] The pressure sensor array and vibration sensor array are integrated on the blade of the shovel head to measure the drilling resistance value and drilling vibration frequency value of each soil layer in real time.

[0033] The deviation calculation module calculates the resistance, vibration frequency and hardness deviation of the soil layer corresponding to the preset sampling area in combination with the drilling resistance, vibration frequency and hardness data.

[0034] Specifically, see Figure 3 As shown, the selection process of the preset sampling area is as follows: R1, import the planned aperture of the planned drilling area in the construction site, and calculate the critical safety distance by the elastic mechanics formula , each pre-selected area is calibrated at the construction site based on the critical safety distance.

[0035] Understandably, the reference formula for the critical safety distance is: ,in, represents the critical safety distance, represents the elastic modulus, Represents Poisson's ratio, the default value is 0.3, represents the planned aperture, It represents the tensile strength of soil. The elastic modulus and Poisson's ratio can be inverted by measuring the stress-strain relationship of soil with an expansion probe, and the tensile strength of soil can be obtained by direct tensile test.

[0036] It should be added that the critical safety distance will increase with the increase of the planned aperture. For example, assuming that the elastic modulus of a site is , Poisson's ratio , Poisson's ratio , when the planned aperture d = 50cm, the critical safety distance calculated by the formula is approximately , if the aperture is increased to d=80cm, , in order to avoid soil damage caused by the superposition of inter-hole stress.

[0037] It should also be added that the pre-selected areas are marked with To determine the interval distance, the diameter of the planned drilling area is used as the calibration diameter to calibrate each pre-selected area.

[0038] R2. Use the Voronoi diagram algorithm to divide the honeycomb grid and determine the sampling points in the planned drilling area and the pre-selected area.

[0039] It can be understood that each sampling point is composed of a grid center point and each edge vertex, such as a vertex of a hexagonal grid.

[0040] R3. Collect the spectral reflectance, water content and penetration index of each sampling point, standardize them respectively, and output the standardized data.

[0041] It should be added that the spectral reflectance is collected using a portable spectrometer, the moisture content is collected using a TDR or FDR probe, and the penetration index needs to be collected using a static penetration instrument.

[0042] R4. Construct a standardized processed data set for the planned drilling area and each pre-selected area, and use the Jaccard similarity coefficient to calculate the similarity between the soil in each pre-selected area and the planned drilling area.

[0043] R5. For the same pre-selected area, the standardized processed data of each sampling point are mapped into weights, which are recorded as soil parameter consistency weights.

[0044] Exemplarily, the specific mapping formula for mapping to weight is: , represents the weight, Indicates standardized processing data.

[0045] It should be noted that when mapping to weights, the standard deviations of the standardized spectral reflectance, moisture content and micro-penetration index need to be mapped separately to obtain the mapping weights of the spectral reflectance, moisture content and micro-penetration index, and then the weights of the spectral reflectance, moisture content and micro-penetration index are fused to obtain the soil parameter consistency weights.

[0046] It should be added that the weights of the fused spectral reflectance, water content, and mini - penetration index include the fusion ratios for setting the weights of the fused spectral reflectance, water content, and mini - penetration index. Among them, the mini - penetration index directly reflects the soil mechanical strength and density, and plays a key role in the assessment of the stability of pile - hole formation and bearing capacity of pile foundations. The water content significantly affects the physical state of the soil such as plasticity, permeability, and mechanical properties such as shear strength and hole - wall stability, and is a parameter that needs to be key - monitored during pile - foundation construction. The spectral reflectance is mainly used to analyze soil components such as organic matter and minerals, and assist in judging soil types. Therefore, in the scenario of pile - foundation excavation, the attention to mechanical properties is relatively high. Thus, the mini - penetration index has the highest influence, the water content has the second - highest influence, and the spectral reflectance has the lowest influence. Correspondingly, the fusion ratio of the mini - penetration index is higher, which can be taken as 45%, the fusion ratio of the water content is the second - highest, which can be taken as 35%, and the fusion ratio of the spectral reflectance is the lowest, which can be taken as 20%.

[0047] R6. Perform similarity correction based on the consistency weights of the soil parameters, and use the pre - selected area corresponding to the corrected maximum similarity as the preset sampling area.

[0048] In the embodiment of the present invention, the critical safety distance is calculated based on the planned drilling hole diameter and the elastic - mechanics formula to scientifically calibrate the pre - selected area, breaking through the dependence on traditional experience. And soil parameters such as spectral reflectance and water content are integrated to calculate the Jaccard similarity coefficient, realizing the quantitative comparison of the soil characteristics between the planned drilling area and the pre - selected area, enhancing the effectiveness, accuracy, and reference of soil - layer prediction, and providing reliable data support for the dynamic adjustment of drilling parameters.

[0049] It can be understood that performing similarity correction based on the consistency weights of the soil parameters specifically means multiplying the consistency weights of the soil parameters by the corresponding similarities.

[0050] Specifically, calculate the resistance, vibration frequency, and hardness deviation of the soil layer corresponding to the preset sampling area, including: A1. For the current soil layer, take the historical peak value of the drilling resistance collected in real - time as the target resistance value, and calculate and output the resistance gradient through the difference of the target resistance values of adjacent soil layers.

[0051] A2. For each currently accumulated drilled soil layer, construct the resistance gradient under the current depth sequence, denoted as the actual resistance gradient, and intercept the resistance gradient corresponding to the current soil layer from the preset sampling area, denoted as the control resistance gradient.

[0052] A3. Compare the actual and control resistance gradients and output the deviation of the soil - layer resistance gradient.

[0053] A4. For each soil layer, fit the resistance-time curve based on the time series to generate the actual resistance distribution curve at the current depth sequence. Meanwhile, intercept the resistance distribution curve at the current depth sequence from the preset sampling area, which is denoted as the control resistance distribution curve.

[0054] A5. Compare the actual resistance distribution curve with the control resistance distribution curve, output the deviation amount of the soil layer resistance distribution, and integrate the deviation amount of the soil layer resistance gradient and the deviation amount of the soil layer resistance distribution to obtain the resistance deviation amount of the current soil layer.

[0055] A6. For the vibration frequency and hardness, repeat steps A1 to A5, and sequentially output the deviation amounts of the vibration frequency and hardness of the current soil layer.

[0056] In the embodiment of the present invention, by constructing the depth sequence comparison of the actual resistance gradient / distribution curve and the control curve of the preset sampling area, combined with the weight compensation of the historical accident database, the deviation of the drilling parameters is quantitatively evaluated, and the dynamic verification of the rectification action is realized, so as to ensure the accuracy of the construction index adjustment under complex strata.

[0057] It can be understood that the drilling resistance deviation directly reflects the reaction force of the current soil layer on the drill bit and is related to the formation density and integrity. When the resistance differs greatly from the collected resistance in the preset sampling area, it is necessary to timely adjust the drilling speed. The vibration frequency deviation characterizes the operation stability of the equipment and is related to the drill bit wear and formation heterogeneity such as fractures and interlayers. When the vibration frequency differs greatly from the collected vibration frequency in the preset sampling area, it indicates that the equipment state or the formation condition differs greatly from the expectation. At this time, in order to ensure the accuracy of the drilling path, it is necessary to timely adjust the drilling speed. The soil layer hardness deviation reflects the physical strength of the formation material and is related to the rock and soil type and structure. Compared with the corresponding collected hardness in the preset sampling area, if the hardness is too high or too low, it is necessary to adjust the drilling angle or the drilling speed to ensure smooth drilling and ensure the drilling accuracy.

[0058] In summary, comprehensively considering these three parameters can comprehensively evaluate the formation condition and the equipment state, so as to adjust the angle and speed, avoid equipment failures, and improve the efficiency.

[0059] It should also be added that the drilling resistance, vibration frequency, and the hardness data of each soil layer collected in real time corresponding to the sampling area are all pre-synchronized and processed in the same way as the currently collected drilling resistance, vibration frequency, and the hardness data of the current soil layer.

[0060] It should be noted that the formation parameters have spatial correlation and temporal evolution, and it is necessary to analyze in combination with the spatial and temporal dimensions. In actual engineering, the formation heterogeneity and the equipment state are coupled to affect the parameters, and it is difficult to distinguish the source of the abnormality with a single index. Therefore, a comprehensive analysis is carried out by combining both space and time, that is, an analysis is carried out from two dimensions of gradient and distribution.

[0061] Understandably, the gradient deviation focuses on abrupt changes in adjacent soil layer parameters such as rock fractures and hard-soft interlayers to identify local anomalies, and the distribution deviation captures the overall trend deviation of parameters over time or depth such as progressive soil layer softening or equipment performance degradation. At the same time, the gradient deviation is vulnerable to single-point noise such as instantaneous resistance fluctuations, and the distribution deviation depends on the overall distribution pattern of the data, such as the filtered trend. By combining the gradient deviation and the distribution deviation to cover the composite anomaly scenario of "local mutation + global trend", the false alarm rate can be reduced. For example, local noise will not significantly change the overall distribution deviation, avoiding misjudgment.

[0062] Furthermore, the output of the soil layer resistance gradient deviation amount in step A3 includes: A31. For the depth of each soil layer, set the influence weight of the resistance deviation corresponding to each soil layer in combination with the historical accident database.

[0063] Understandably, through the integration of depth weight differentiation and gradient accumulation, the quantitative assessment ability of the formation risk is significantly improved, especially suitable for deep complex formation engineering.

[0064] A32. Calculate the difference between the actual resistance gradient and the reference resistance gradient corresponding to each soil layer, and based on the influence weight of the resistance deviation, perform weighted summation to obtain the cumulative resistance gradient difference, which is used as the soil layer resistance gradient deviation amount.

[0065] Understandably, the specific setting process of setting the influence weight of the resistance deviation corresponding to each soil layer in combination with the historical accident database is as follows: Extract the occurrence frequencies of each accident type corresponding to the depth of each soil layer from the historical accident database, and select the highest occurrence frequency as the reference occurrence frequency for the depth of each soil layer.

[0066] Statistically count the number of accident types with occurrence frequencies exceeding the set threshold corresponding to the depth of each soil layer, and divide it by the total number of accident types to output the triggering ratio of effective accident types corresponding to the depth of each soil layer , indicating the th soil layer depth, .

[0067] Normalize the reference occurrence frequencies of each soil layer depth, and mark the processing result as the accident occurrence ratio, denoted as .

[0068] Set the influence weight of the resistance deviation , and select the influence weight of the resistance deviation of each soil layer from the influence weights of the resistance deviation of each soil layer depth. Among them, the influence weight of the resistance deviation The specific formula is: , indicating the pre-set The comprehensive construction risk weight corresponding to each soil layer depth is obtained by weighted summation of the engineering importance weight, geological risk weight, and construction feedback weight. Among them, the engineering importance weight measures the key role of the soil layer in the engineering structure and reflects its direct contribution to the structural safety. Its specific value is comprehensively determined by combining the building foundation design code and the pile foundation form design requirements formulated by structural engineers. The geological risk weight quantifies the geological risk attributes of the soil layer itself, and its value can be comprehensively determined based on exploration data and regional disasters. The construction feedback weight reflects the abnormal feedback monitored in real time during the drilling process, such as a sudden drop in drilling speed and a sudden change in torque, and is a dynamic weight. represents the accident risk weight. and respectively represent the proportion coefficients of the effective accident type trigger ratio and the accident occurrence ratio. and respectively represent the comprehensive construction risk weight item and the proportion coefficient corresponding to the construction risk weight item.

[0069] It should be added that the effective accident type trigger ratio is analyzed from the aspect of the coverage breadth of accident types, and the accident occurrence ratio is analyzed from the aspect of the frequency of accident occurrence. They can be respectively valued at 0.6 and 0.4, following the principle of giving priority to coverage breadth to avoid omission of a single accident type. And and can be respectively valued at 0.55 and 0.45.

[0070] It also should be added that the construction feedback weight is obtained by combining the pre-set reference feedback weight and the risk trigger weight. Exemplarily, the reference feedback weight can be valued at 0.1. If the drilling speed decline rate exceeds the trigger threshold, the drilling speed decline rate and the corresponding trigger threshold are respectively denoted as and , and is used as the risk trigger weight.

[0071] It should be noted that when screening out the resistance deviation influence weights of each soil layer from the resistance deviation influence weights of each soil layer depth, when a certain soil layer matches the resistance deviation influence weights of multiple soil layer depths, the maximum value is taken as the influence weight of the corresponding soil layer.

[0072] Furthermore, the output of the soil layer resistance distribution deviation amount in step A5 includes: A51. Project the actual resistance distribution curve to the position of the control resistance distribution curve, and count the total length of the overlapping area curve of the projection.

[0073] A52. Extract the slope, the number of peak points, and the number of valley points from the actual resistance distribution curve and the control resistance distribution curve respectively. Combining the total length of the overlapping area curve of the projection, the slope, the number of peak points, and the number of valley points, set the curve shape deviation compensation factor for each soil layer.

[0074] A53. For the same soil layer, extract the amplitudes of the actual resistance distribution curve and the control resistance distribution curve respectively, take the difference between the two to obtain the resistance change amplitude difference, and at the same time extract the maximum resistance respectively, and take the difference to obtain the maximum resistance difference.

[0075] A54. Based on the curve shape deviation compensation factor, correct the resistance change amplitude difference of each soil layer, and output the corrected resistance change amplitude difference of each current soil layer.

[0076] A55. Obtain the cumulative maximum resistance difference by synthesizing the maximum resistance differences of each soil layer, calculate the average value of the corrected resistance change amplitude differences of each soil layer to obtain the average change amplitude difference, and use the cumulative maximum resistance difference and the average change amplitude difference as the soil layer resistance distribution deviation amount.

[0077] Regarding step A52, it can be understood that the specific setting process of setting the curve shape deviation compensation factor for each soil layer is as follows: If the signs of the corresponding slopes of the actual resistance distribution curve and the control resistance distribution curve are inconsistent, assign the curve shape deviation compensation factor as 1, otherwise, record the slope difference between the two as , divide the total length of the curve in the projection overlapping area by the total length of the control resistance distribution curve to obtain the projection overlapping ratio and record it as , record the differences between the number of peak points and the number of valley points of the actual resistance distribution curve and the control resistance distribution curve as and , and select the maximum value from and as the target feature point number difference .

[0078] Set the curve shape deviation compensation factor, denoted as , , , and respectively represent the compensation ratio coefficients corresponding to the slope deviation, the projection overlapping ratio deviation and the target feature point number difference deviation, and are respectively the set reference slope difference and feature point number difference.

[0079] In a specific embodiment, , and can be respectively taken as 0.5, 0.3 and 0.2. At the same time, the set reference slope difference can be taken as 0.1, and the feature point number difference can be taken as 20% of the number of peak points of the control resistance distribution curve.

[0080] The said dynamic adjustment module, in combination with the bucket angle deviation value and the actual path position difference, dynamically adjusts the drilling speed and drilling angle based on the drilling adjustment rule.

[0081] Specifically, the specific confirmation process of the actual path position difference is as follows: F1. Vertically project the real-time collected positions onto the planned path, and calculate the perpendicular distances from each projection point to the planned path.

[0082] It can be understood that the perpendicular distances from each projection point to the planned path can be calculated by the distance formula between two points, and the specific calculation formula will not be elaborated here.

[0083] F2. If the number of projection points with non-zero consecutive perpendicular distances is within the set threshold, extract the maximum perpendicular distance, and at the same time calculate the average perpendicular distance of the projection points. Output the actual path position difference by synthesizing the maximum perpendicular distance and the average perpendicular distance.

[0084] It can be understood that the set threshold for the number of projection points with non-zero consecutive perpendicular distances can be taken as times the total number of projection points, and output the actual path position difference by synthesizing the maximum perpendicular distance and the average perpendicular distance. The actual path position difference can be obtained by performing a weighted sum of the maximum perpendicular distance and the average perpendicular distance. Among them, the weights of the maximum perpendicular distance and the average perpendicular distance can be taken as 0.7 and 0.3 respectively.

[0085] F3. If the number of consecutive projection points with non-zero perpendicular distances is not within the set threshold, take the sum of the perpendicular distances from each projection point to the planned path as the actual path position difference.

[0086] It should be noted that when traditional path position deviation is carried out, on the one hand, it focuses on the current position, which belongs to static deviation. On the other hand, when dynamic tracking is carried out, the maximum position deviation is often considered. By distinguishing accidental fluctuations and systematic offsets through the threshold of consecutive points, it avoids misjudgment caused by a single extreme value. At the same time, the global cumulative mode strengthens the sensitivity to persistent small deviations, reduces the risk of missed reports, and in the local mode, combining extreme values and means can better reflect the deviation intensity than simply the maximum value.

[0087] In a specific embodiment, the maximum position deviation selection method is compared with the existing method. At the same time, the position deviation limit value for triggering deviation correction is set to 5 mm, the total number of projection points is 15, the set threshold for the number of consecutive projection points is 5, and three scenarios of local accidental deviation, systematic continuous offset, and mixed offset are set for comparison. The comparison results are shown in Table 1.

[0088] Table 1 Schematic table of comparison data for position deviation selection methods

[0089]

[0090] As shown in Table 1, the existing method has false negatives in the scenario of systematic continuous deviation (the true deviation value is 20 mm > the threshold, but it is not triggered because the maximum value of a single point is 3 mm). However, the present invention accurately identifies through global accumulation, and the triggering rate is significantly improved. In the scenario of local accidental deviation, the existing method triggers deviation correction due to an extreme value of 5 mm, which may actually be an instantaneous interference. The present invention determines it as accidental fluctuation, greatly reducing ineffective adjustments. Generally speaking, compared with the scenario of local accidental deviation, the present invention can avoid over-response to instantaneous interference. Compared with the scenario of systematic continuous deviation, it can solve the problem of gradual path deviation caused by the insensitivity of the existing method to continuous small deviations. Compared with the scenario of mixed deviation, it maintains a high detection rate and has both extreme value response and trend judgment capabilities. That is, through multi-dimensional error analysis, the present invention can significantly improve the judgment rationality while ensuring sensitivity.

[0091] Further, before dynamically adjusting the drilling speed and drilling angle, a drilling adjustment trigger assessment is performed, including: if the deviation amount of the shovel head angle exceeds the set threshold, or the duration of the deviation amount of the shovel head angle exceeds the set time, or the ratio of the actual path position difference to the planned path is exceeded, the drilling angle adjustment is triggered.

[0092] If any of the deviation amounts of resistance, vibration frequency, and hardness exceeds the set threshold, the drilling speed adjustment is triggered.

[0093] It can be understood that the angle deviation directly reflects the loss of direction control, and the attitude needs to be corrected immediately to avoid path deviation. The lateral path deviation needs to adjust the angle to return to the planned trajectory first. When the deviation amount of soil layer resistance / hardness / vibration frequency exceeds the set threshold, it indicates a formation mutation, and reducing the speed can relieve the load. The present invention sets the trigger assessment conditions for drilling adjustment in accordance with this principle.

[0094] Further, dynamically adjusting the drilling speed and drilling angle includes confirming the adjustment ratio. The specific confirmation process is as follows: 1) Standardize the deviation amount of the shovel head angle, the duration of the deviation amount of the shovel head angle, and the actual path position difference respectively, sum the processed results after weighting, and use them as the input variables of the Sigmoid function, and output the adjustment ratio of the drilling angle.

[0095] It can be understood that the deviation amount of the shovel head angle and the actual path position difference can be processed by Min-Max standardization, and the duration of the deviation amount of the shovel head angle can be processed by logarithmic standardization to cope with the long-tail distribution, that is, to avoid negative infinity when t = 0 and compress the high-value range at the same time. And the Min-Max standardization process and the logarithmic standardization process are common existing processing methods, and their specific representation formulas are not shown here.

[0096] It should be added that the deviation amount of the shovel head angle represents the urgency of the current deviation and belongs to the instantaneous state quantity that dominates the influence. The deviation duration reflects the severity of the cumulative deviation and has a secondary influence. The path position difference reflects the correction requirement of the overall offset and belongs to the long-term effect of the position error. They can be respectively valued at 0.5, 0.3, and 0.2.

[0097] 2) Standardize the deviation amounts of resistance, vibration frequency, and hardness, and after weighted summation of the processing results, input them into the Sigmoid function to output the adjustment ratio of the drilling speed.

[0098] It is understandable that the deviation amounts of resistance, vibration frequency, and hardness are standardized.

[0099] It should be added that the set threshold of the deviation amount, the set limit value of the actual path position difference, and the exceeding time threshold of the deviation amount of the shovel head angle are determined through comprehensive experiments based on historical experience data.

[0100] It also should be added that the weights of the deviation amount of resistance, the deviation amount of vibration frequency, and the deviation amount of hardness can be respectively valued at 0.4, 0.2, and 0.4.

[0101] In the embodiment of the present invention, by integrating the drilling resistance, vibration frequency, angle deviation, position, and soil layer hardness, the dynamic adjustment module fuses the angle deviation, position difference, and multi-parameter deviation amounts, combines weighted summation with Sigmoid function mapping, solves the problem of cumulative error caused by single external positioning, realizes the differential precision and coordination control of the drilling speed and angle under complex strata, significantly improves the construction path accuracy and safety. At the same time, it also improves the smoothness of the adjustment of drilling parameters under complex strata and ensures the construction accuracy.

[0102] In a specific embodiment, dynamically adjusting the drilling speed and drilling angle further includes screening out the corresponding drilling speed adjustment range and drilling angle adjustment range from the pre-set adjustment rules based on the adjustment ratio, and taking the product of the adjustment ratio and the adjustment range corresponding to the adjustment ratio as the specific adjustment value.

[0103] Exemplarily, when the adjustment ratio < 0.3, the adjustment range of the drilling angle is ±2° of the current drilling angle, and the adjustment range of the drilling speed is ±5% of the current drilling speed. When 0.3 ≤ adjustment ratio < 0.7, the adjustment range of the drilling angle is ±5° of the current drilling angle, and the adjustment range of the drilling speed is ±15% of the current drilling speed. When the adjustment ratio ≥ 0.7, shutdown processing is performed.

[0104] It should be added that the positive and negative signs of the drilling angle adjustment direction are determined by the deviation direction of the actual parameters relative to the target state. Exemplarily, when the deviation amount of the actual drilling angle relative to the planned drilling angle is greater than 0 and the actual direction is to the right, a left correction is required. The current adjustment coefficient is 0.3, that is, a negative adjustment is performed. For example, 。

[0105] It should be added that when the deviation amount of the vibration frequency is greater than 0 and exceeds the set threshold, the speed is forced to decrease to prevent equipment damage or accidents. When the deviation amount of the vibration frequency is less than or equal to 0, the resistance and hardness deviation amounts are respectively marked as ΔM and ΔH, and the positive and negative signs of the drilling speed adjustment direction are confirmed according to the following table.

[0106] Table 2 Schematic table of the drilling speed adjustment direction.

[0107]

[0108] It should be noted that when ΔM>0 and ΔM>0, it indicates high resistance but soft formation, which may be a temporary obstacle. At this time, the speed reduction is a conservative speed reduction. The fixed speed reduction range is, for example, -5% to -10%, and it is confirmed in combination with the adjustment ratio. For example, is used as the specific speed reduction value, where, represents the adjustment ratio.

[0109] The control execution module generates a control command based on the adjusted drilling speed and drilling angle, and drives the Luoyang shovel to perform a deviation correction action.

[0110] Among them, the triggering conditions of the dynamic adjustment module include: any deviation amount exceeds the set threshold, or the actual path position difference exceeds the limit, or the deviation amount of the shovel head angle continues to exceed the time limit.

[0111] In the embodiment of the present invention, parameters such as drilling resistance, vibration frequency, angle deviation, position, and soil layer hardness are obtained in real time through multi-source data acquisition. After the deviation calculation module dynamically compares with the preset parameters, the dynamic adjustment module is triggered to optimize the drilling speed and angle based on multiple conditions. Finally, the control execution module generates a precise command to drive the deviation correction, realizing multi-dimensional data fusion and real-time dynamic adjustment, significantly improving the accuracy of the drilling path and the formation adaptability, and reducing the construction risk in complex geological conditions.

[0112] Please refer to Figure 2 As shown, the present invention also provides a dry method precise positioning construction control method based on machine learning. The method includes: S1. Multi-source data acquisition: Real-time acquisition of drilling resistance, vibration frequency, deviation value of the shovel head angle, and position data, and simultaneous acquisition of the current soil layer hardness data.

[0113] S2. Construction deviation calculation: Combine the drilling resistance, vibration frequency, and hardness data to calculate the deviation amounts of the resistance, vibration frequency, and hardness of the soil layer corresponding to the preset sampling area.

[0114] S3. Construction dynamic adjustment: Combine the deviation value of the shovel head angle and the difference in the actual path position, and dynamically adjust the drilling speed and drilling angle based on the drilling adjustment rules.

[0115] S4. Drilling control execution: Generate control instructions based on the adjusted drilling speed and drilling angle, and drive the Luoyang shovel to perform the deviation correction action.

[0116] The above content is only an example and explanation of the concept of the present invention. Those skilled in the art of this technology can make various modifications or supplements to the described specific embodiments or use similar methods for substitution. As long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, they should fall within the protection scope of the present invention.

Claims

1. A dry method precise positioning construction control system based on machine learning, characterized in that, The system includes: Multi-source data acquisition module collects drilling resistance, vibration frequency, shovel head angle deviation and position in real time, and simultaneously collects hardness data of the current soil layer; The deviation calculation module combines the drilling resistance, vibration frequency and hardness data to calculate the resistance, vibration frequency and hardness deviation of the soil layer corresponding to the preset sampling area; The dynamic adjustment module dynamically adjusts the drilling speed and drilling angle based on the drilling adjustment rules by combining the shovel head angle deviation value and the actual path position difference; The control execution module generates control instructions based on the adjusted drilling speed and drilling angle, and drives the Luoyang shovel to perform the deviation correction action; The triggering conditions of the dynamic adjustment module include: any deviation exceeds a set threshold, or the actual path position difference exceeds a limit, or the shovel head angle deviation continues to time out.

2. The dry method precise positioning construction control system based on machine learning according to claim 1, characterized in that: The multi-source data acquisition module comprises: The RTK positioning module installed on the shovel body obtains the three-dimensional coordinates of the Luoyang shovel in real time and generates the real-time position of the Luoyang shovel; A dual-axis inclination sensor is used to detect the deviation between the shovel head and the set angle as the shovel head angle deviation; The pressure sensor array and vibration sensor array are integrated on the blade of the shovel head to measure the drilling resistance value and drilling vibration frequency value of each soil layer in real time.

3. The dry method precise positioning construction control system based on machine learning according to claim 1, characterized in that: The process of selecting the preset sampling area is as follows: Import the planned hole diameter of the planned drilling area within the construction site, and calculate the critical safety distance through the elastic mechanics formula , and calibrate each preselected area on the construction site based on the critical safety distance; Use the Voronoi diagram algorithm to divide the honeycomb grid and determine the sampling points in the planned drilling area and the pre-selected area; Collect the spectral reflectance, water content and penetration index of each sampling point, perform standardization processing on them respectively, and output the standardized processing data; A standardized processed data set of the planned drilling area and each pre-selected area is constructed, and the similarity between the soil in each pre-selected area and the planned drilling area is calculated using the Jaccard similarity coefficient; For the same pre-selected area, the standardized processed data of each sampling point are mapped into weights, which are recorded as soil parameter consistency weights; The similarity correction is performed based on the soil parameter consistency weight, and the preselected area corresponding to the corrected maximum similarity is used as the preset sampling area.

4. The dry method precise positioning construction control system based on machine learning according to claim 1, characterized in that: The calculation of the resistance, vibration frequency and hardness deviation of the soil layer corresponding to the preset sampling area includes: A1. For the current soil layer, the historical peak value of the drilling resistance collected in real time is taken as the target resistance value, and the output resistance gradient is calculated by the difference of the target resistance values ​​of adjacent soil layers; A2. For each soil layer currently drilled cumulatively, a resistance gradient under the current depth sequence is constructed, recorded as the actual resistance gradient, and the resistance gradient corresponding to the current soil layer is intercepted from the preset sampling area, recorded as the reference resistance gradient; A3. Compare the actual resistance gradient with the control resistance gradient and output the soil resistance gradient deviation; A4. Fit the resistance-time curve based on the time series for each soil layer to generate the actual resistance distribution curve under the current depth series, and at the same time intercept the resistance distribution curve under the current depth series from the preset sampling area, which is recorded as the control resistance distribution curve; A5. Compare the actual resistance distribution curve with the control resistance distribution curve, output the soil layer resistance distribution deviation, integrate the soil layer resistance gradient deviation and the soil layer resistance distribution deviation to obtain the resistance deviation of the current soil layer; A6. For the vibration frequency and hardness, repeat steps A1 to A5, and sequentially output the vibration frequency and hardness deviation of the current soil layer.

5. The dry method precise positioning construction control system based on machine learning according to claim 4, characterized in that: The output of the soil layer resistance gradient deviation amount includes: For each soil layer corresponding to the depth of the soil layer, set the resistance deviation influence weight corresponding to each soil layer in combination with the historical accident database; Calculate the difference between the actual resistance gradient and the control resistance gradient corresponding to each soil layer, and based on the resistance deviation influence weight, perform weighted summation to obtain the cumulative resistance gradient difference, which is used as the soil layer resistance gradient deviation amount.

6. The dry-method precise positioning construction control system based on machine learning according to claim 4, characterized in that: The output of the soil layer resistance distribution deviation amount includes: Project the actual resistance distribution curve to the position of the control resistance distribution curve, and count the total length of the overlapping area curve of the projection; Extract the slope, the number of peak points, and the number of valley points from the actual resistance distribution curve and the control resistance distribution curve respectively. Combine the total length of the overlapping area curve of the projection, the slope, the number of peak points, and the number of valley points, and set the curve shape deviation compensation factor for each soil layer; For the same soil layer, extract the amplitudes of the actual resistance distribution curve and the control resistance distribution curve respectively, and calculate the difference between the two to obtain the resistance change amplitude difference. At the same time, extract the maximum resistance respectively and calculate the difference to obtain the maximum resistance difference; Based on the curve shape deviation compensation factor, correct the resistance change amplitude difference of each soil layer, and output the corrected resistance change amplitude difference of the current soil layers; Combine the maximum resistance differences of each soil layer to obtain the cumulative maximum resistance difference. Calculate the average value of the corrected resistance change amplitude differences of each soil layer to obtain the average change amplitude difference. Use the cumulative maximum resistance difference and the average change amplitude difference as the soil layer resistance distribution deviation amount.

7. The dry method precise positioning construction control system based on machine learning according to claim 1, characterized in that: The specific confirmation process of the actual path position difference is as follows: Vertically project the real-time collected position onto the planned path, and calculate the perpendicular distance from each projection point to the planned path; If the number of projection points with non-zero consecutive perpendicular distances is within the set threshold, extract the maximum perpendicular distance, and at the same time calculate the average perpendicular distance of the projection points. Combine the maximum perpendicular distance and the average perpendicular distance to output the actual path position difference; If the number of consecutive projection points with non-zero perpendicular distances is not within the set threshold, use the sum of the perpendicular distances from each projection point to the planned path as the actual path position difference.

8. The dry method precise positioning construction control system based on machine learning according to claim 1, characterized in that: Before dynamically adjusting the drilling speed and drilling angle, a drilling adjustment trigger assessment is performed, including: If the deviation amount of the shovel head angle exceeds the set threshold, or the duration of the deviation amount of the shovel head angle exceeds the set time, or the ratio of the actual path position difference to the planned path is exceeded, trigger the adjustment of the drilling angle; If any one of the resistance, vibration frequency, and hardness deviation amounts exceeds the set threshold, trigger the adjustment of the drilling speed.

9. The dry method precise positioning construction control system based on machine learning according to claim 1, wherein: The dynamic adjustment of the drilling speed and drilling angle includes confirming the adjustment ratio. The specific confirmation process is: Perform standardization processing on the deviation amount of the shovel head angle, the duration of the deviation amount of the shovel head angle, and the actual path position difference respectively. Perform weighted summation on the processing results and use them as the input variables of the Sigmoid function, and output the adjustment ratio of the drilling angle; Perform standardization processing on the resistance, vibration frequency, and hardness deviation amounts. After weighted summation of the processing results, input them into the Sigmoid function, and output the adjustment ratio of the drilling speed.

10. A dry-method precise positioning construction control method based on machine learning, characterized in that, This method includes: S1. Multi-source data collection: Real-time collect the drilling resistance, vibration frequency, deviation value of the shovel head angle and position data, and synchronously collect the current soil layer hardness data; S2. Construction deviation calculation: Combine the drilling resistance, vibration frequency and hardness data to calculate the resistance, vibration frequency and hardness deviation of the soil layer corresponding to the preset sampling area; S3. Construction dynamic adjustment: Combine the deviation value of the shovel head angle and the actual path position difference, and dynamically adjust the drilling speed and drilling angle based on the drilling adjustment rules; S4. Drilling control execution: Generate control instructions based on the adjusted drilling speed and drilling angle, and drive the Luoyang shovel to perform deviation correction actions.

Images