Intelligent control-based core accurate cutting method

Abstract

The application discloses a core accurate cutting method based on intelligent control, and particularly relates to the technical field of core processing, which comprises the following steps: collecting the surface topography, texture features and acoustic response signals of the core, fusing multi-source data to reconstruct a three-dimensional model and generate a potential distribution diagram of the sandwich structure; constructing a stable cutting path avoiding the sandwich through spatial curvature analysis; in the cutting process, the position offset of the cutter head is acquired in real time, and point-by-point fine adjustment is performed; the contact angle and the feed speed are dynamically adjusted in the cutting execution; after the cutting is completed, structural integrity analysis and quality scoring are performed, if the score is lower than a threshold, path reconstruction and re-cutting operation are triggered; the application avoids the sandwich structure through intelligent path planning, and improves the cutting stability; dynamic parameter adjustment and point-by-point fine adjustment mechanism are adopted to realize high-precision adaptive control of the cutting process; a cutting behavior atlas is constructed to support experience learning and feedback adjustment of path optimization, and the overall cutting quality and intelligent level are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of core processing, more particularly, the present application relates to a core accurate cutting method based on intelligent control. BACKGROUND

[0002] In the process of geotechnical engineering, oil and gas exploration and geological sample analysis, the core is an important original sample for structural analysis, physical and mechanical testing and mineral identification, and its integrity and cutting accuracy directly affect the reliability of the downstream analysis conclusion. At present, the core cutting mostly adopts the way of manual setting path, fixed speed and angle for straight line cutting. This kind of method has certain applicability for samples with uniform structure and no interlayer, but when facing core samples with interlayer structure, weak cementation zone or stiffness mutation area, there are problems such as cutting path misjudgment, local damage, interlayer exposure, which seriously affect the subsequent sample utilization value.

[0003] The traditional path setting method lacks the perception ability of the internal structure of the core, and cannot effectively avoid the interlayer or potential structural weakness. At the same time, in the cutting execution process, the current technology generally adopts fixed parameter setting, lacks real-time response control mechanism for different lithology changes, and is easy to cause the knife head to be stuck, the structure to be cracked or the cutting to be deviated. In addition, lacking of trajectory adaptive fine tuning mechanism and result quality feedback ability, once the cutting deviates, it cannot be adjusted and compensated in time. Therefore, the present application proposes a core accurate cutting method based on intelligent control to solve the above problems. SUMMARY

[0004] To achieve the above purpose, the present application provides the following technical scheme:

[0005] The core accurate cutting method based on intelligent control comprises the following steps:

[0006] Collect the surface topography, texture features and internal acoustic response signals of the core sample, reconstruct a three-dimensional space model through multi-source information fusion, and generate a potential interlayer structure distribution map;

[0007] Based on the interlayer distribution map and the core shape model, a cutting path that fits the real shape of the core is constructed by using spatial curvature analysis, and the cutting path is selected to have a continuous trend outside the interlayer area through the shortest structure stability evaluation;

[0008] In the cutting process, the spatial deviation data between the cutting knife head and the core is obtained in real time through position tracking, and the data is input into the trajectory deviation judgment process to fine tune the cutting path point by point;

[0009] When executing the cutting action, the contact angle and feed speed are dynamically adjusted to adapt to the local structural stiffness change of the core, so as to realize a stable, continuous and interlayer-unbroken cutting process;

[0010] After the cutting is completed, the target section is analyzed for structural integrity, and a quality score is calculated in combination with the trajectory fitting rate. If the score is lower than a preset threshold, a path reconstruction and re-cutting instruction is executed.

[0011] In a preferred embodiment, the construction process of the cutting path includes generating a plurality of candidate path sets, each path being initially fitted based on the spatial curvature features extracted from the three-dimensional structure model of the core, forming trajectory samples with cutting feasibility;

[0012] For each candidate path, the local stress concentration coefficient corresponding to the path is calculated, which is defined as the ratio of the average stress gradient of the area crossed by the path under the simulated cutting load to the core stiffness response. The top several paths sorted by stress concentration coefficient from small to large are selected as the final candidate set to reduce the risk of structural disturbance during cutting and improve overall stability.

[0013] The generation of the potential distribution map of the interlayer structure is based on joint modeling of multi-source data, which includes acoustic response signals, image texture changes, and surface microstructure scanning data. After boundary determination, the interlayer judgment function is synthesized. This function integrates the confidence of different source signals through logical integration and constructs an interlayer probability field in the overall core model.

[0014] The integral expression in the interlayer judgment function adopts one of the three typical combination methods: based on the weighted rule of logical product, the maximum confidence selection rule, or the Bayesian confidence update rule.

[0015] In a preferred embodiment, the final confirmation of the cutting path is based on the constructed candidate path set. A coordinate fitting mechanism is introduced in the candidate set. By matching the spatial attitude of the core end face and the long axis direction, a local reference coordinate frame is established to eliminate the influence of core deformation. In this coordinate space, the candidate path set is mapped to a feasible trajectory subset within the high-dimensional structure. The shortest stability of the path and the interlayer boundary obstacle avoidance ability are used as target parameters. Discrete point array construction is used for trajectory iterative optimization, and the point array is non-uniformly encrypted near the interlayer boundary.

[0016] The path optimization process is completed in a variable weight graph search manner. The path optimization process is based on the comprehensive evaluation of the stress distribution performance and spatial obstacle avoidance ability of each path in the core structure model and the candidate path set. The path risk score rule is adaptively adjusted according to the lithology input parameters, so that the final selected path has the minimum disturbance risk and the highest morphological fit in the overall structure.

[0017] In a preferred embodiment, the trajectory deviation judgment process is realized by comparing the spatial position data collected during the continuous cutting process with the preset trajectory, wherein the trajectory comparison process uses the minimum curvature change calculation method between continuous time discrete sampling points, and the deviation judgment process introduces a comparison mechanism for three different stages before, during and after cutting, automatically marks the error burst point and the expected error turning point position during execution, and determines whether to execute the trajectory correction instruction by combining the three-stage error clustering atlas and the multi-threshold judgment standard.

[0018] In a preferred embodiment, the dynamic adjustment process of the contact angle and the feed speed in the cutting action is combined with the core surface response characteristics and the current position of the cutter head to make a joint decision, and the calculation of the real-time adjustment parameters is completed in each time slice, and the possible contact form of the next step is judged through a limited prediction window.

[0019] The joint adjustment strategy does not use a fixed value switching method, but continuously adjusts based on a local target optimal control model, wherein the control quantity can be determined jointly according to three indicators:

[0020] The current position, the tangent direction of the interlayer predicted boundary, the instantaneous incremental change of the cutting force of the cutter head, and the core surface friction feedback signal form a local adjustment cost function in the form of a weighted function, realizing the continuous surface of the parameters.

[0021] In a preferred embodiment, if it is judged that the path reconstruction trigger state is triggered, the historical cutting trajectory, the error record, and the interlayer exposure position information will be automatically archived, and a time sequence cutting behavior trajectory atlas will be formed, which is used to generate the next round of path correction scheme; the atlas construction uses an event chain representation method, all key turning nodes are defined as structural state change points, which constitute a decision tree input data set, and are used as prior conditions in the next path planning.

[0022] In a preferred embodiment, the structural integrity analysis includes quantitative evaluation of the flatness of the cutting surface, the number of structural mutation edges, and the exposure degree of the interlayer, wherein the flatness evaluation uses the average value of the residual error of the fitting surface, the number of mutation edges is the number of regions with a first derivative change greater than a set threshold in the cutting surface, and the exposure degree of the interlayer is calculated by integrating the predicted interlayer density through the cutting surface. The quality score is output by a preset machine learning model based on the above calculation results and the trajectory fitting rate.

[0023] In a preferred embodiment, the point-by-point fine tuning process of the cutting path uses a dynamic optimization strategy based on a rolling window during execution, and in the execution interval continuously sliding along the cutting direction, whether to make a slight adjustment to the path point and its adjacent points is judged according to the stress change rate of the current path point and the cutting error trend at each time;

[0024] To ensure that the cutting trajectory realizes high-precision local correction under the premise of maintaining the overall structural stability, the amplitude of each adjustment should not exceed the standard deviation range of the path change curvature in the historical cutting record, which is generated according to the statistical characteristics of the trajectory variation of the same lithology type samples in the historical path set.

[0025] Technical effects and advantages of the present application:

[0026] The present application realizes the identification and path avoidance of the core internal interlayer, weak cementation zone and other invisible structures by introducing the interlayer structure potential distribution map and interlayer probability field, combining spatial curvature analysis and structural stress concentration judgment mechanism. In the path generation stage, a plurality of candidate paths are constructed and a stress disturbance factor is introduced for screening, and the finally determined cutting path can maximize the avoidance of high-risk areas, effectively reducing the structural disturbance and sample damage risk in the cutting process. Compared with the existing fixed path straight cutting method, the present application significantly improves the adaptability and path stability of complex lithology samples.

[0027] In the cutting execution process of the present application, a dynamic fine-tuning mechanism based on a rolling window is introduced, which collects stress change rate and cutting error trend information in each execution cycle to determine whether the current path point needs to be slightly adjusted. At the same time, the contact angle and the feed speed are dynamically adjusted by a local control model composed of three feedback indicators, ensuring that the cutting behavior remains continuous, smooth and high-precision in a changing structure environment. This mechanism enables the tool head to have adaptive ability to respond to complex structure feedback, effectively avoiding the problems of structural tearing, cutting shock or interlayer exposure caused by fixed parameter cutting in traditional methods.

[0028] The present application constructs a cutting behavior trajectory map, archives the historical trajectory, error record and interlayer exposure position in each cutting process, and establishes a behavior node sequence in the form of an event chain, which is used to train the next round of path correction model or provide prior knowledge to assist path screening. This mechanism not only supports intelligent optimization after path reconstruction, but also has behavior data accumulation and adaptive ability, so that the cutting strategy can "learn" from historical experience and gradually form the optimal behavior evolution logic. Compared with the traditional static path scheme, the present application has stronger stability, predictability and intelligent control ability. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to facilitate the understanding of those skilled in the art, the present application will be further described below with reference to the accompanying drawings;

[0030] Figure 1 The principle diagram of the intelligent control-based core precise cutting method in the present application. DETAILED DESCRIPTION

[0031] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described, obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0032] With reference to Figure 1 The following examples are obtained:

[0033] Embodiment 1: The present application proposes a core accurate cutting method based on intelligent control. First, a spatial model containing core geometric shape and internal interlayer structure is established by fusing multi-source data such as image texture, acoustic response and three-dimensional laser scanning, and on this basis, an interlayer probability field is constructed to represent the distribution and risk level of the interlayer area. According to the model, a plurality of feasible cutting paths are generated, each path is fitted based on the spatial curvature characteristics, and the stress concentration degree in the local area is evaluated, and the path set with the minimum stress disturbance is selected as the candidate path. Subsequently, in the established coordinate reference frame, the candidate path is spatially mapped and optimized, combined with the interlayer boundary avoidance ability and the path stability, through the variable weight graph search and non-uniform sampling strategy, the final cutting trajectory with the lowest structural interference risk is selected.

[0034] During the cutting execution process, by real-time acquisition of cutting error, stress change and cutter feedback data, it is judged whether the current path deviates from the safety window, and if necessary, the path fine tuning is triggered. The fine tuning process adopts a rolling window mechanism, in each execution cycle, the local response of the current trajectory point is modified point by point, and the adjustment amplitude is limited to not exceed the standard deviation range of the historical trajectory curvature fluctuation, so as to ensure the stability and reliability of the cutting process. At the same time, the cutting behavior parameters such as feed speed and contact angle will be dynamically adjusted combined with friction feedback and stress change, to realize the accurate adaptation to the complex core structure. The whole execution process will continuously record the cutting behavior atlas, which is used to optimize the path selection and control strategy of the same type of core in the subsequent process, and after the cutting is completed, the effect is evaluated according to the indexes such as cutting surface flatness and interlayer exposure degree, if the score does not reach the preset threshold, the path is automatically reconstructed and recut, so as to form a self-adaptive, learnable and high-robustness precise cutting closed-loop system.

[0035] Specifically includes the following steps:

[0036] The surface morphology, texture characteristics and internal acoustic response signals of the collected core sample are collected, a three-dimensional space model is reconstructed through multi-source information fusion, and a potential distribution map of the interlayer structure is generated; it is the basic link of the whole technical process, and the purpose is to fully master the external structure and internal structure of the core. Through the fusion of image, texture, acoustic and other sensing means, not only can the surface of the core be "seen", but also the hidden structure such as interlayer and weakly cemented zone in the core can be "perceived", which can provide reliable data support for subsequent path planning.

[0037] Based on the interlayer distribution map and the core shape model, a cutting path that fits the real shape of the core is constructed by using spatial curvature analysis, and the cutting path is evaluated by the shortest structural stability to select the continuous trend outside the interlayer area; through "intelligent path design", instead of using straight lines or manually planned cutting lines, a more fitting path is automatically generated based on the actual geometry and structural curvature of the core. At the same time, a "structural stability" evaluation mechanism is introduced to ensure that the path neither passes through the interlayer nor has mechanical safety, greatly reducing the risk of breakage.

[0038] During the cutting process, the spatial offset data between the cutting tool and the core is obtained in real time through position tracking, and the data is input into the trajectory deviation judgment process to make point-by-point adjustments to the cutting path; it is a real-time correction logic in the path execution process, that is, "cutting while correcting". By monitoring the offset between the tool and the path in real time, the deviation can be responded to at the first time, the trajectory is adjusted point by point, and the cutting precision is ensured.

[0039] When performing the cutting action, the contact angle and feed speed are dynamically adjusted to adapt to the local structural stiffness change of the core, so as to realize a stable, continuous and non-destructive cutting process; different rock properties have different stiffness and force feedback, so it is necessary to adjust the cutting angle and speed in real time. This mechanism automatically adjusts the working state when encountering uneven hard and soft structures, avoids damaging the interlayer or producing irregular cross sections, and ensures consistent cutting quality.

[0040] After cutting, the structural integrity of the target section is analyzed, and the quality score is calculated in combination with the trajectory fitting rate. If it is lower than the preset threshold, execute the path reconstruction and re-cutting instructions, form a "feedback outlet" of closed-loop control, not only perform cutting, but also actively evaluate the good and bad of the results. The scoring standard not only includes whether the section is flat, but also focuses on whether the interlayer and other high-risk structures are exposed. Once the score is not up to standard, automatically enter the next round of path adjustment and re-cutting to ensure high reliability and high fidelity of the output results.

[0041] The construction process of the cutting path includes generating a plurality of candidate path sets, each path is initially fitted based on the spatial curvature features extracted from the three-dimensional structure model of the core, and a trajectory sample with cutting feasibility is formed;

[0042] For each candidate path, the local stress concentration coefficient corresponding to the path is calculated, which is defined as the ratio of the average stress gradient of the area crossed by the path under the simulated cutting load to the stiffness response of the core. The top several paths sorted by the stress concentration coefficient from small to large are selected as the final candidate set to reduce the risk of structural disturbance during cutting and improve overall stability.

[0043] Before designing the core cutting trajectory, the complete three-dimensional structure information of the core is first obtained, which includes the surface profile, the spatial distribution of internal interlayers, the degree of geometric distortion, and the overall spatial coordinate configuration. Based on the three-dimensional structure, the spatial curvature extraction algorithm, such as Gaussian curvature or principal curvature curve analysis method, is used to model the structural curvature of the core overall shape, and then a plurality of cutting path samples are generated on the modeling results.

[0044] These sample paths can be generated in different strategies such as center axis fitting, end face translation cutting, and interlayer symmetric avoidance to form a set of "candidate path set" in three-dimensional space along different attitude directions. Each path is a spatial trajectory with geometric continuity and meets the basic cutting physical requirements, and has processing feasibility. For example, the starting point of the path can be set at the center of one end face of the core, and the ending point can be set at the opposite surface. The path can be a curved line that fits the core curvature while avoiding the interlayer boundary.

[0045] For each candidate path, the local structural response of the path during cutting is simulated based on the structural finite element method or regional stress estimation model, and the "local stress concentration coefficient" is calculated. The calculation method of the coefficient is as follows: first, a virtual equal cutting load is applied in each small spatial element covered by the path, and the stress gradient (i.e. the stress change rate per unit length) generated in the region is recorded. Then, the stress gradient of all path covered regions is averaged. Subsequently, the stiffness response values of the core material along the path, such as elastic modulus, Poisson's ratio, or shear modulus, are obtained by looking up the table or presetting the lithology parameters. The stress gradient average and the stiffness response value are divided to obtain the local stress concentration coefficient of the path.

[0046] The physical meaning of the coefficient is that under the same load, whether the path is prone to local stress concentration in the area it passes through, thereby causing structural rupture or micro-crack propagation. The smaller the value, the more uniform the stress distribution during cutting, and the less likely to cause damage.

[0047] After the coefficient calculation of all paths is completed, it is sorted from small to large, and a number of paths with high coefficients are selected as the "final candidate path set". This step is equivalent to selecting a subset of the actual structure disturbance risk from multiple theoretically feasible paths, providing a smaller and higher quality option set for subsequent path optimization and dynamic execution. This sorting mechanism takes into account the structural reliability and cutting stability, effectively improving the controllability of the overall cutting process and the integrity of the finished sample.

[0048] Taking a shale interlayer type core as an example, the core length is 120 mm, the diameter is 38 mm, and there are 3 irregular interlayers inside. After scanning and modeling, 10 cutting path samples are generated, and the path distribution is offset from the core center axis by about 0-5 mm. Local stress concentration simulation is performed on each path, and it is found that the stress gradient of path P3 is the smallest, and the rock stiffness parameter of the region it passes through is 28 GPa, and the corresponding stress concentration coefficient is 1.2; although path P8 is shorter in geometry, it passes through the interlayer, resulting in high stress gradient and low stiffness, and the stress concentration coefficient is 4.5. Finally, paths P1, P2, P3 and P6 with stress concentration coefficients below 2.0 are selected as the candidate set and enter the next trajectory optimization link.

[0049] The joint design of the candidate path set construction and the stress concentration coefficient screening mechanism can ensure that the paths entering the execution link are not only the shortest in geometry, but also the safest in structure physics. This method significantly improves the adaptability to complex interlayers and weakly cemented structures in core cutting tasks, reduces the damage rate, and improves the sample section quality, providing a better data foundation for subsequent analysis.

[0050] The generation of the interlayer structure potential distribution map is based on multi-source data joint modeling, which includes acoustic response signals, image texture changes and surface microstructure scanning data. After boundary determination, the interlayer judgment function is synthesized, which integrates the confidence of different source signals through logical integral method, and constructs the interlayer probability field in the core overall model.

[0051] The integral expression in the interlayer judgment function adopts one of the three typical combination methods: based on the logical product weighting rule, the maximum confidence selection rule or the Bayesian confidence updating rule.

[0052] Before performing the core precise cutting, the internal structural features of the core need to be accurately identified, especially the interlayer regions with potential cutting risks. Therefore, the present application proposes to generate an interlayer structure potential distribution map through multi-source data joint modeling, and to construct an interlayer probability field based on this, thereby providing a data basis for subsequent cutting path construction and avoidance judgment.

[0053] Multi-source data refers to core structure information obtained through different sensing technologies, including but not limited to acoustic response signals, image texture changes, and surface microstructure scanning data.

[0054] Acoustic response signals are used to reflect the acoustic impedance changes of different medium interfaces inside the core. Interlayer regions usually exhibit enhanced reflection or velocity changes.

[0055] Image texture changes refer to the analysis of parameters such as gray level gradient, direction consistency, and texture scale of core surface images. Interlayers are often accompanied by obvious texture mutations.

[0056] Surface microstructure scanning data obtains the three-dimensional structure of core surface micro-topography through laser or white light interference. Interlayer boundary regions exhibit micro-protrusions, cracks, or abnormal roughness.

[0057] The above three types of data sources are pre-processed and boundary determination operations are performed in their respective dimensions. For example, potential interlayer boundaries are determined by setting reflection intensity thresholds, image gradient mutation thresholds, or roughness jump values. Each data source forms an independent "interlayer preliminary judgment map" to establish a confidence region based on spatial position.

[0058] In order to form a unified interlayer identification result, an interlayer judgment function is constructed to fuse the results of the three types of signals according to the confidence level. The "logical integral method" here refers to integrating the judgment results of the existence of interlayers in different source signals at a certain spatial position to form a total confidence score.

[0059] The integral expression in the interlayer judgment function can adopt one of the three typical combination methods:

[0060] Based on the logical product weighting rule: the product of the interlayer judgment confidence values of each data source at the same point is taken as the joint judgment result. Each source confidence value has a weight, which can be set as an empirical value or automatically generated by a pre-trained model.

[0061] Maximum confidence selection rule: at each spatial point, only the highest confidence value among the three is selected as the final judgment value to enhance the dominance of sensitive source signals.

[0062] Bayesian confidence updating rule: set the prior probability as the statistical characteristics of the interlayer distribution of historical cores, and take the three sensing results as the likelihood term. Perform Bayesian update on the target point to obtain the posterior probability as the probability value of the point belonging to the interlayer.

[0063] The integral method can be selected according to different lithology types, data quality or identification accuracy requirements, and finally a interlayer probability field map covering the entire core space is formed. The map divides the core into multiple sub-regions, each sub-region is assigned a probability value (range 0-1) of the existence of interlayer, the higher the value, the more likely the region is an interlayer.

[0064] Taking a section of shale sample as an example, the core length is 100 mm, the image texture data (resolution 1 point per mm), acoustic echo data (sampling density 1 point per 2 mm) and surface micro scanning data (resolution 1 point per 0.5 mm) are collected. In the image texture, there is a continuous gray level mutation at the position of 30-40 mm, the reflection energy in the same interval of acoustic data is obviously increased, and the microstructure data shows that the roughness peak value is enhanced, and the boundary overlap rate of the three is about 85%. Set the initial weight as image 0.3, acoustic 0.5, and microstructure 0.2, use the logical product weighting rule, assign each interlayer a probability of 0.78-0.92 in the region, and finally construct the interlayer probability field map. On this basis, according to the avoidance principle, i.e. avoiding the area exceeding the preset probability value, the subsequent path construction stage can preferentially avoid this area, realizing the effective avoidance of interlayer structure.

[0065] By introducing multi-source perception fusion and confidence modeling, the accuracy and spatial resolution of interlayer identification can be significantly improved without relying on single image or manual annotation. Through the flexible expression structure of the interlayer judgment function, the method has self-adaptive ability under different lithology conditions, and can dynamically generate high adaptability interlayer probability field, providing accurate guidance for path planning, cutting avoidance and structure protection.

[0066] The final confirmation of the cutting path is based on the constructed candidate path set. In the candidate set, a coordinate fitting mechanism is introduced, which matches the spatial attitude of the core end face and the long axis direction, establishes a local reference coordinate frame to eliminate the influence of core deformation, and maps the candidate path set to a feasible trajectory subset in the high-dimensional structure in the coordinate space. The shortest stability of the path and the interlayer boundary obstacle avoidance ability are taken as the target parameters, and the discrete point array construction method is used for trajectory iterative optimization, and the point array is non-uniformly encrypted near the interlayer boundary;

[0067] The path optimization process is completed in a variable weight graph search mode. The path optimization process is based on the comprehensive evaluation of the stress distribution performance and spatial obstacle avoidance ability of each path in the candidate path set and the core structure model. The path risk scoring rule is adaptively adjusted according to the lithology input parameters, so that the final selected path has the minimum disturbance risk and the highest shape fitting degree in the overall structure.

[0068] After the construction of the candidate path set (corresponding to the construction of the candidate path set in the foregoing), further optimization and verification in structural space and physical logic are needed to screen the final execution path.

[0069] In the final confirmation process of the cutting path, considering that the actual core may have spatial attitude changes such as deformation, deviation, or rotation during transportation, collection, or fixation, a coordinate fitting mechanism needs to be introduced in the candidate path set. This mechanism includes two key operations:

[0070] Core end face matching: Obtain the direction of the normal vector of the two end faces through scanning, and fit the attitude of the core in three-dimensional space; long axis direction modeling: form the main axis based on the connection of the end face center line, and construct a local coordinate system with the axis as the reference.

[0071] This process forms a local reference coordinate frame for “restoring” the core to an ideal geometric state, facilitating subsequent path comparison (corresponding to the use of the “core three-dimensional structure model” in the foregoing).

[0072] In this coordinate space, the original candidate path set is mapped to a subset of trajectories in a high-dimensional structure, where each trajectory retains its orientation in the actual space of the core, but eliminates the noise caused by attitude deviation. This step ensures the comparability between paths. Next, in the optimization process, the shortest path stability and the ability to avoid interlayer boundaries are used as dual target parameters. This means that when selecting the final path, not only the shortest geometric distance is considered, but also:

[0073] The path should avoid high-risk areas of interlayers as much as possible (corresponding to the construction of the “interlayer structure potential distribution map and the interlayer probability field”, and the construction of the interlayer probability field in the core overall model), the overall stress distribution of the path should be uniform, and there should be no obvious concentration points to reduce structural disturbance during cutting (corresponding to the screening logic of the “local stress concentration coefficient”).

[0074] Path optimization adopts a discrete point array construction method for trajectory iterative optimization, that is, the entire feasible path space is divided into a discrete coordinate grid, different path point combinations are simulated by enumeration, and points are arranged at a higher density near the interlayer boundary (i.e., “non-uniform encryption”) to improve local path precision and obstacle avoidance flexibility. This encryption mechanism can be automatically triggered by setting an interlayer probability threshold, for example, in areas where the interlayer probability value is greater than 0.6, the point array density is 2-3 times that of other areas.

[0075] After the point array is constructed, path traversal and screening are completed through variable weight graph search. The variable weight graph refers to assigning a weight value to each path (between two discrete points) when constructing the graph model. The weight value is related to the delamination risk value, stress concentration coefficient, and geometric curvature of the area where the path is located. The specific value can be generated by a weighting function. The graph search algorithm can use variants of the Dijkstra or A* path algorithm in existing technologies, and the weighted optimal value is obtained among multiple weight factors.

[0076] Path optimization ultimately relies on a path risk scoring rule, in which the proportion of each scoring factor (such as delamination avoidance priority, path smoothness, and geometric distance) can be automatically adjusted according to the input lithology parameters. For example, for high brittle cores (such as shale), the delamination avoidance weight is higher; and for sandstone with strong structural integrity, the path smoothness and processing efficiency weight is higher.

[0077] Through this series of multi-objective, multi-dimensional, and multi-parameter constraint optimization, a cutting trajectory with the smallest disturbance risk, optimal path stability, and the most suitable geometric shape in the overall structure is finally selected from the candidate path set, providing a high reliability basis for the subsequent cutting execution stage (corresponding to position tracking and point-by-point fine tuning in the subsequent cutting execution).

[0078] Suppose a core sample is a weakly cemented sand-shale interbed, and the delamination dense area in the delamination probability field appears in the middle section of the core. There are 12 feasible paths in the candidate path set, of which path 6 and path 9 pass through the delamination dense area and are assigned high weight values. In the path point array, the delamination area is set to 0.25 mm, and the remaining area is set to 0.5 mm. In the path risk scoring, the lithology parameter is set to "high vulnerability", so the delamination avoidance weight in the scoring function is 0.6, the path smoothness is 0.3, and the stress balance is 0.1. After performing variable weight graph search, path 3 is selected as the final trajectory, with an overall disturbance score of 0.21, which is significantly lower than the average value of 0.48 of other paths, and has significant structural superiority and execution stability.

[0079] The trajectory deviation judgment process is realized by comparing the spatial position data collected in the continuous cutting process with the preset trajectory. The trajectory comparison process uses the minimum curvature change calculation method between continuous time discrete sampling points. The deviation judgment process introduces comparison mechanisms for three different stages before, during, and after cutting. It automatically marks the error burst point and expected error turning point positions. Through three error clustering maps combined with multi-threshold judgment criteria, it determines whether to execute the trajectory correction instruction.

[0080] In order to ensure the execution accuracy and stability of the trajectory in the core cutting process, the present application proposes a trajectory deviation judgment process, which aims to identify the trajectory deviation trend in real time during the cutting process, and accordingly judge whether to start the subsequent trajectory fine tuning or correction operation (corresponding to the path execution and fine tuning mechanism).

[0081] The input data of this process comes from the spatial position data collected in real time during the continuous cutting process, which is usually obtained through high-precision encoders, displacement measuring devices or attitude sensors, and records the actual position sequence of the cutter head during the execution process. The position sequence is discretized at a certain sampling frequency to form a "cutting trajectory point set" arranged in time sequence.

[0082] The actual trajectory point set is compared with the preset trajectory planned before cutting, and the preset trajectory is derived from the finally optimized and confirmed path, so it has the minimum stress disturbance and the best morphological fitting degree. The trajectory comparison method is not simply comparing the coordinate difference of the two paths, but using the minimum curvature change calculation method between the discrete sampling points in continuous time, that is, calculating the curvature of the micro-arc segment formed by two adjacent sampling points, and then comparing the curvature changes of the actual trajectory and the preset trajectory at the corresponding positions. If the curvature difference of the two trajectories at the same position point exceeds the preset threshold, it is considered that there is deviation.

[0083] In order to more accurately identify the space-time characteristics of the trajectory deviation, the judgment process introduces a three-stage comparison mechanism for the whole cutting process, which is:

[0084] Pre-cutting stage: evaluate the positioning error of the equipment start-up and the early stage of cutter head movement, and exclude the initial deviation factor;

[0085] Mid-cutting stage: compare the sampling data in real time to identify the time point when the error starts to grow;

[0086] Post-cutting stage: comprehensive analysis of the overall difference between the trajectory after cutting and the target path to confirm the final deviation accumulation result.

[0087] During the execution process, if the mutation trend of the trajectory curvature change is detected, the "error burst point" will be automatically marked, that is, the first key point where the trajectory obviously starts to deviate; at the same time, combined with the derivative of the error with respect to time, the "expected error turning point" is further determined, that is, the time point when the error is predicted to enter the growth acceleration zone.

[0088] All error data of the sampling points are used to construct a three-section error clustering map, that is, the error points are clustered into three subsets according to the time interval (before, during and after), and the mean, variance, change trend and other statistical analysis are performed in each subset. This map helps to reveal the time evolution pattern of the error behavior and provides supporting evidence for subsequent judgment.

[0089] Finally, based on the error clustering map, combined with multi-threshold judgment criteria, determine whether to execute the trajectory correction instruction. For example, when the error growth rate continuously exceeds the threshold β, and the frequency of burst points exceeds the threshold ρ per unit time, the trajectory correction is triggered. Further, a more preferred embodiment can be obtained: that is, the judgment criteria can not only consider the absolute value of the current error, but also introduce composite indicators such as error growth rate, fluctuation amplitude, and burst point density, and use a weighted logic function for judgment.

[0090] Taking a core of a sandwich complex section as an example, the trajectory sampling frequency is set to 100 Hz, the total cutting time is 30 seconds, and 3000 actual trajectory points are collected. Comparing these points with the optimized path, it is found that from the 1120th point, the curvature difference continuously increases, and at the 1170th point, it reaches 0.028 / m, which exceeds the set threshold 0.025 / m, and is automatically marked as an “error burst point”. Subsequently, at the 1210th point, the error change trend changes from slow increase to rapid growth, and is recorded as a “turning point”. Based on this, in the error clustering map, the error density of the middle section exceeds the corresponding threshold, and the fluctuation frequency increases, meeting the multi-threshold logic judgment criteria, and issuing a trajectory fine-tuning instruction to perform rolling position adjustment on the path between the 1100th and 1250th points (corresponding to the fine-tuning execution mechanism part).

[0091] The trajectory deviation judgment process can realize early identification of the small deviation trend in the execution path, avoiding structural damage or cutting failure caused by error accumulation. The three-stage comparison and multi-index fusion judgment method has strong robustness, and is particularly suitable for processing core structures with interlayers and large stress fluctuations, and cooperates with the path fine-tuning mechanism to build a complete dynamic closed-loop control system.

[0092] The dynamic adjustment process of the contact angle and the feed speed in the cutting action is combined with the core surface response characteristics and the current position of the cutter to make a joint decision, and the calculation of the real-time adjustment parameters is completed in each time slice, and the possible contact form of the next step is judged through a limited prediction window;

[0093] This joint adjustment strategy does not use a fixed value switching method, but continuously adjusts based on a local target optimal control model, where the control quantity can be determined jointly based on three indicators:

[0094] The current position of the tangent direction of the interlayer prediction boundary, the instantaneous change of the cutting force increment of the cutter, and the core surface friction feedback signal form a weighted function form of the local adjustment cost function, realizing the continuous surface of the parameters.

[0095] To adapt to the complexity of the core structure and the uncertainty of the local response, the execution parameters of the cutter head need to be dynamically adjusted during the cutting process, especially the contact angle and the feed speed, which are two key control variables. This dynamic adjustment process is not a static preset value call, but a real-time decision based on the current position of the cutter head and the response characteristics of the core surface (corresponding to the integration analysis process of the "position data collected in real time in the path execution phase" and the "potential distribution map of the interlayer structure").

[0096] During the cutting process, the "time slice" is the smallest control unit, and each time slice corresponds to a control parameter refresh cycle. In each time slice, the optimal feed speed and contact angle at the current time are calculated based on the spatial position of the current cutter head, the local properties of the core structure, and the feedback data in the previous cycle. This strategy avoids the traditional fixed rate or angle method and can flexibly respond to complex structures in areas with structure mutations or interlayer adjacent areas.

[0097] In addition, a limited prediction window is set to determine the "contact pattern" that the current cutting behavior may encounter in the next time slice, i.e., whether the cutter head will enter the interlayer, the hard area, or cross the risk area of stiffness mutation. The prediction window is generally set to 3-5 time slices in the future, and the data comes from the interlayer prediction map (corresponding to the content of the "potential distribution map of the interlayer structure and the interlayer probability field") and the historical moving direction deduction model of the cutter head. This joint adjustment strategy clearly proposes not to use traditional fixed threshold switching logic (such as setting the rule of "interlayer area speed reduction by half"), but to use a local target optimal control model, i.e., to find the optimal parameter combination for the current position structure condition at each time, forming a "continuous and derivable adjustment path".

[0098] In this model, the three core control variables constitute the basis for joint decision-making:

[0099] The tangent direction of the current position to the interlayer prediction boundary: that is, to judge the relative position relationship and directional angle between the current cutter head position and the nearest interlayer boundary. If the cutter head is about to cut into the interlayer area, the angle should be adjusted to be parallel to the interlayer rather than orthogonal to reduce the risk of structure damage (corresponding to the interlayer boundary modeling part);

[0100] Instantaneous incremental change of cutting force of the cutter head: that is, to calculate the rising trend of the cutting resistance in the current time slice. If there is a sharp rise, it means that it may enter a high-hardness area, and the feed speed or angle should be reduced to avoid tool jamming or rebound;

[0101] Core surface friction feedback signal: this signal comes from the friction force measurement in the contact area. When entering the area with sudden changes in surface roughness, the friction signal will also change, which is an important reference to characterize the structure state change.

[0102] These three control variables are combined into a "local adjustment cost function" in the form of a weighted function. Each control variable corresponds to a weight factor, and the specific weight is set according to the lithology type and task target (for example, the friction response weight is higher in shale, and the cutting force change weight is higher in sandstone). The smaller the output value of the cost function, the better the parameter combination.

[0103] Finally, the cost function is constructed as a parameter continuous surface, that is, the adjustment of the feed speed and the contact angle is no longer a jump value, but a continuous and smooth adjustment along the surface, achieving intelligent progressive control of the cutting behavior.

[0104] The control model takes the three structural feedbacks at the current time as input variables, and outputs a set of optimal contact angle-feed speed combinations at the minimum value position of the "local adjustment cost function". This means that the contact angle and feed speed are not calculated or controlled separately, but are determined by the same multi-input multi-output control model (MIMO), and are interrelated and optimized together.

[0105] The three control variables are denoted as: the current cutter position and the tangent angle θ of the predicted boundary of the interlayer, the instantaneous change ΔF of the cutting force, and the surface friction feedback signal, i.e. the friction coefficient M of the cutter and the core contact surface. These parameters are combined by a weighted function to form a cost function C, that is:

[0106] C = w1·f1(θ) + w2·f2(ΔF) + w3·f3(M);

[0107] w1, w2, w3 are the preset non-zero control coefficients corresponding to the three control variables, f1, f2, f3 are normalization functions (mapping each input to a unified interval such as [0, 1]), and C represents the cost of the parameter combination at this time, and the smaller the value, the better the final goal is to find the "contact angle-feed speed" combination corresponding to the minimum value point of C in this model surface. Assuming that the data at the current time slice is calculated as follows:

[0108] C = 0.5*0.67 + 0.3*0.6 + 0.2*0.67 = 0.335 + 0.18 + 0.134 = 0.649; compared with the results of other time slices in history, if the value is higher, it means that there is a greater disturbance risk in the current feed speed and angle parameter combination; the control logic will slide and adjust to a combination with a smaller cost, achieving dynamic optimization of the cutting behavior.

[0109] Take a certain section of dense shale sample with interlayer as an example, the current position of the cutter head is located at the center of the core, and the prediction window shows that there is a high probability of interlayer boundary 3 mm ahead. In the previous time slice, the tangent angle remains unchanged, but the cutting force rises from 22 N to 29 N, with an instantaneous incremental change of 7 N; the friction signal rises from 0.8 to 1.5. In the cost function, the cutting force weight, i.e. the cutting force control coefficient, is set to 0.5, the friction feedback weight, i.e. the friction feedback control coefficient, is set to 0.3, and the interlayer direction weight, i.e. the tangent angle control coefficient, is set to 0.2. After combining the three indicators, a cost value of 0.42 is obtained, which is lower than the set adjustment threshold of 0.5, so a small angle reduction operation is performed, and the feed speed is reduced from 1.0 mm / s to 0.75 mm / s. In the subsequent time slice, the cutting force tends to be stable, the friction signal decreases, and the normal feed rate is gradually restored, realizing intelligent response to the structural change section.

[0110] If it is judged to be a path reconstruction trigger state, the historical cutting trajectory, error record, and interlayer exposure position information will be automatically archived, and a time sequence cutting behavior trajectory atlas will be formed to generate the next round of path correction scheme; the atlas is constructed by using event chain representation method, all key turning nodes are defined by structural state change points, which constitute the decision tree input data set, and are used as prior conditions in the next path planning.

[0111] During the cutting process, if the trajectory deviation judgment process identifies that the trajectory execution error exceeds the tolerance, or the cutting result quality score is lower than the threshold (corresponding to the aforementioned "after cutting is completed, the structural integrity of the target cutting surface is analyzed, and the quality score is calculated combined with the trajectory fitting rate, if it is lower than the preset threshold, the path reconstruction and re-cutting instruction is executed"), the path reconstruction operation will be triggered. At this time, historical execution data needs to be called to provide data support and behavior guidance for the next round of path reconstruction.

[0112] Therefore, all dynamic behavior data closely related to the cutting process will be archived, including: historical cutting trajectory: records the complete path point sequence of the cutter head position in the three-dimensional coordinate system during each actual execution process; error record: including the path deviation value, curvature change, error burst point and turning point position in each time slice (corresponding to the aforementioned "error burst point and expected error turning point marking"); interlayer exposure position information: the start and end area of the interlayer exposed in the actual cutting process is collected through structural imaging or cutting section feedback, and the position is calibrated combined with the previously constructed interlayer probability field (corresponding to the "generation of interlayer structure potential distribution map" part).

[0113] The above three types of data will be organized into a time-ordered structure, forming a time-sequential cutting behavior trajectory atlas. This atlas is an abstraction of the entire cutting process, with the following characteristics: the recording granularity can reach 10-50 milliseconds per behavior sample, ensuring the details of behavior evolution; the trajectory atlas is not just a path set, but a data chain with behavior and result labels. The atlas construction uses an event chain representation, which divides the cutting process into a series of continuous "event segments", each triggered by a state change, such as: "interlayer entry", "friction increase", "trajectory burst error", "error correction execution", "angle rapid adjustment", "score below threshold", etc. Each event node is a "structural state change point", which is extracted as a key behavior inflection point and defined as a core node in the trajectory atlas.

[0114] All key nodes are organized into a data set that can be used for behavior prediction, forming a decision tree input data set with a branching structure. In this data set, each behavior path (such as "error -> correction -> interlayer exposure") can be used as a historical sample for path optimization, for training models or directly for conditional screening. In the next path planning or trajectory fine-tuning (corresponding to the aforementioned "cutting path construction" and "point-by-point fine-tuning"), the data in this atlas will be imported as prior conditions, including:

[0115] In the path generation phase, avoid the historical interlayer exposure area; in the parameter adjustment phase, dynamically load certain error-correction relationship patterns; in the trajectory screening phase, give high cost marks to paths that have failed before. Taking a core sample as an example, in the initial cutting, path P3 has an error surge at the 25th second, with trajectory point number 32003400, error value increasing from 0.8mm to 1.9mm, and score result 68 (lower than threshold 70). This segment of path is recorded as a failed segment, with the interlayer boundary exposed between 7884mm on the z-axis. The event chain representation is as follows: node 1: start -> normal cutting; node 2: error rise -> Δerror > 1mm; node 3: score low -> score = 68; node 4: trigger re-cut -> enter P6 path; In the second round of path planning, the trajectory optimization phase automatically marks z=78-84mm as a high-risk area, increases the avoidance probability, and removes the P3 path segment from the candidate set.

[0116] Through the construction and structured representation of the cutting behavior trajectory atlas, not only the review and archiving of the execution history are realized, but also the behavior-level knowledge guidance is provided for future path planning and adjustment. The event chain representation enhances the structuralization of trajectory data, which can support intelligent behaviors such as model training, decision tree construction, and parameter prediction, and form a path scheduling logic with learning ability.

[0117] The structural integrity analysis includes quantitative evaluation of the flatness of the cutting surface, the number of structural mutation edges, and the extent of the exposed interlayer. The flatness evaluation uses the average residual of the fitted surface, the number of mutation edges is counted by detecting regions where the first derivative changes by more than a set threshold, and the extent of the exposed interlayer is calculated by integrating the predicted interlayer density through the cutting surface. The quality score is output by a pre-set machine learning model based on the above calculation results and the trajectory fitting rate.

[0118] After the cutting behavior is completed, in order to evaluate the effect of the cutting operation on the structure protection, a structural integrity analysis of the final cutting surface is required. This analysis process is not only a result detection means, but also provides a basis for judging whether the subsequent path needs to be reconstructed (corresponding to the aforementioned "if below the preset threshold, execute path reconstruction and re-cut instruction" content). The analysis includes three main quantitative indicators:

[0119] Cutting surface flatness definition: the degree of deviation between the overall shape of the cutting surface and the ideal plane.

[0120] Calculation method: Perform plane fitting on the actual three-dimensional scanning data of the cutting surface to construct a theoretical reference plane (such as a least squares fitting plane), project the actual point cloud onto the plane, and calculate the vertical residual of each point to obtain a set of residual vectors.

[0121] Final evaluation value: residual mean (i.e., the average distance of all points from the fitted plane), usually in millimeters or microns.

[0122] Indicator meaning: The smaller the value, the more flat the cutting surface; if the residual mean is higher than the set flatness tolerance threshold, it may indicate trajectory fluctuations or structural fractures.

[0123] Number of structural mutation edges definition: the number of edges on the cutting surface that exhibit abrupt changes in structure, which is an important evaluation factor for microstructure continuity.

[0124] Calculation method: Perform curvature derivative analysis along the tangent direction of the cutting surface, extract the first derivative of the tangent, and detect the number of regions where the change is greater than a set threshold in the spatial sequence.

[0125] First derivative mutation determination method: For any three consecutive measurement points, let their heights be h i-1 , h i , h i+1 , and the height data of measurement point i and its adjacent two measurement points, then the first derivative D i of measurement point i is: Δx is the horizontal distance between the two adjacent sampling points, also known as the horizontal sampling interval. If |D i | is greater than the derivative mutation judgment threshold, then the point is counted as a mutation point.

[0126] Final statistical value: the number of point area segments (adjacent points can be combined into a segment) that meet the mutation conditions.

[0127] Laminate exposure degree definition: whether the final cutting surface exposes the original predicted laminate structure, and the exposure degree reflects the quality of laminate protection.

[0128] Calculation method: Based on the previously generated laminate structure potential distribution map (corresponding to the "Generation of Laminate Structure Potential Distribution Map" section), the space region crossed by the cutting surface is obtained, the laminate probability field is called, and the laminate probability integral value of the region corresponding to the cutting surface is calculated.

[0129] Result significance: The larger the integral value, the more exposed laminate regions; if it is higher than the set laminate exposure risk threshold, it will be judged as a laminate damage behavior.

[0130] Trajectory fitting rate is an index used to measure the similarity between the actual cutting trajectory and the preset target path, reflecting the path accuracy and deviation control level in the execution process. The calculation method of this index is as follows:

[0131] First, a theoretical path is planned before cutting begins, which consists of a series of ordered spatial position points, called target trajectory point list;

[0132] During the actual cutting process, the spatial coordinates of the tool head at consecutive time points are obtained through the position tracking device, forming an actual trajectory point list with the same number and interval as the theoretical path;

[0133] Next, for each pair of points with the same number, i.e. a target trajectory point and its corresponding actual cutting point, the straight-line distance between the two points in three-dimensional space is calculated, and the spatial deviation value between all point pairs is recorded;

[0134] The spatial deviation values are accumulated and then divided by the total length of the target trajectory to represent the "total deviation ratio";

[0135] Finally, the trajectory fitting rate is obtained by "1 minus the total deviation ratio", which is between 0 and 1. The closer the value is to 1, the better the fit, indicating that the actual trajectory is highly consistent with the theoretical path. If the value is significantly lower than 1, it indicates that there is a significant deviation or path execution error.

[0136] Suppose the target path consists of 100 points, and the actual cutting path also collects 100 corresponding position points. In the calculation, it is found that the average deviation between each pair of points is 0.2 mm, and the total length of the target path is 100 mm, so the total deviation ratio is 0.002, and the final trajectory fitting rate is 0.998, indicating that the executed trajectory is very accurate.

[0137] After the four indicators are quantitatively calculated, a preset machine learning model is input, which is trained based on historical data and has the ability to score the structural integrity. Input features: flatness residual mean; number of abrupt edges; probability integral of interlayer exposure; trajectory fitting rate; output results: quality score Q ∈ [0, 100], the higher the value, the better the cutting effect, set the score acceptance threshold, if it is lower than the preset threshold, execute the path reconstruction and re-cut instruction. Model type: support vector regression (SVR); gradient boosting regression tree (GBRT); multilayer perceptron (MLP) neural network, etc. The model is trained by manually labeled data set, and the score output result is known and controllable.

[0138] For example: in a shale interlayer sample, the cutting surface analysis results after cutting are as follows: flatness residual mean: 0.35 mm; number of abrupt edges: 5 sections; interlayer exposure integral value: 0.46; trajectory fitting rate: 0.81; input the above values into the trained support vector machine model, output the integrity score as 64, which is lower than the preset threshold (set as 70), and it is determined that the structural damage risk is high, so the trajectory is immediately archived, and the candidate path optimization link is retriggered to execute the path reconstruction and re-cut instruction.

[0139] The point-by-point fine-tuning process of the cutting path adopts a dynamic optimization strategy based on a rolling window. In the execution interval continuously sliding along the cutting direction, the stress change rate and cutting error trend of the current path point are used to determine whether to make a slight adjustment to the position of the path point and its adjacent points at each time;

[0140] To ensure that the cutting trajectory achieves high-precision local correction while maintaining the overall structural stability, the adjustment range of each adjustment should not exceed the standard deviation range of the path curvature change in the historical cutting record, which is generated based on the trajectory variation statistical characteristics of samples of the same lithology type in the historical path set.

[0141] During the cutting process, due to the complex structure and uneven material of the core, the cutter head may deviate slightly while moving along the initial path due to uneven stress, friction changes or execution errors. In order to ensure the high consistency of the actual path and the theoretical path, and at the same time avoid structural disturbance, a point-by-point fine-tuning mechanism is introduced in the path execution to correct the execution trajectory in real time (corresponding to the "trajectory deviation judgment process" part).

[0142] The fine-tuning mechanism controls based on a "rolling window" as the basic unit. During the execution along the cutting direction, a fixed length cutting interval is continuously defined as the current window, which slides forward as a whole every time it advances one position unit, and updates the data state in this interval in real time. Therefore, this strategy is a continuous dynamic optimization mode, rather than a phased or manual intervention mode.

[0143] At each moment, that is, in each rolling window, the current path point and its adjacent points before and after are taken as candidate adjustment points, and two factors are taken as the basis for judgment:

[0144] The stress change rate is an index for measuring whether the structure stress at a certain position is significantly fluctuating in a continuous time slice, indicating whether the stress of the current region is rapidly increasing and thus may enter the core interlayer boundary or high stiffness mutation region.

[0145] In actual application, the calculation method of the stress change rate is as follows: record the stress feedback value generated by the cutting process of the tool head at a certain position at the current moment, which is usually collected by a force sensor or a torque sensor; consult the stress feedback value at the same position or adjacent position in the last time slice; divide the stress difference between the two time slices by the time interval between the two time slices to obtain the stress change amplitude per unit time; compare it with a pre-set judgment standard. If the stress change amplitude per unit time is higher than the set standard, it can be determined that the structure state of the position has mutated, and there may be a risk of local extrusion, structure weakening or cutting into a high impedance region. When the stress change rate is high, it indicates that the cutting region is rapidly changing the structure state, such as crossing the edge of the interlayer in the core, the stiffness interface or the fracture, and trajectory correction should be performed at this point or its adjacent points to prevent further deviation of the path and cause damage

[0146] The cutting error trend is used to evaluate whether the deviation between the actual motion trajectory of the tool head and the pre-set cutting path is continuously expanding, which is an important basis for judging the risk of deviation. The calculation steps are as follows: record the spatial distance between the actual position of the tool head and the planned path in the current time slice as the "instantaneous error" of the current position; consult the error records of the same position in the previous time slice or multiple consecutive time slices; compare the change trend of these error values. If the error values show a gradual increase, especially if the error values in consecutive multiple time slices are higher than those in the previous time slice, it can be judged that the error is in a "growth trend"; if the growth amplitude exceeds the set safe error increase standard, it can be considered that the trajectory may have deviated from the trend line of the original path, and timely correction is needed. When the error trend is continuously expanding, it indicates that the tool head has gradually deviated from the ideal path, and if it is not adjusted, it may further deviate from the safe trajectory, even cross the interlayer or damage the boundary, so the fine tuning mechanism should be immediately enabled.

[0147] The stress change rate and the cutting error trend can be judged separately or simultaneously. If any of the indicators exceeds the respective allowable range, it can be used as a basis for judging that the current path point has a high risk signal, triggering the fine-tuning operation. If both are abnormal, the priority is higher, and the fine-tuning amplitude or area range can be appropriately relaxed, but the spatial adjustment amount must still be controlled within the historical statistical range (corresponding to the "not exceeding the standard deviation range of the path curvature change in the historical cutting record" content)

[0148] The setting of this limit boundary comes from the statistical analysis of the trajectory behavior of the same type of core samples in the historical trajectory set (corresponding to the "cutting behavior trajectory map construction" part described above). The specific method is as follows: for multiple core samples that have been executed and have good results, extract the geometric curvature change between their path points; calculate the standard deviation of these change values, and set it as the curvature stability threshold of this type of rock (hereinafter referred to as CR threshold); any fine-tuning of the current execution trajectory is not allowed to break through this CR threshold, to ensure that the overall structural stability is not destroyed.

[0149] Taking a section of sand-shale interbedded sample as an example, the path rolling window is set to slide every 10 mm, covering 5 path points. At 42 seconds, the 3rd point in the window detects a stress change rate of 0.26 (unit: MPa / s), and the error of this point in the previous time window increases from 0.6 mm to 1.1 mm, showing a growing trend.

[0150] It is determined that the point meets the adjustment condition, and the position correction operation is performed. Historical data shows that in this type of rock, the average curvature change is 0.08, the standard deviation is 0.02, and the CR threshold is set to 0.02. The pre-adjustment angle offset of the current point is 0.017, which is lower than the CR threshold, so the adjustment is effective. The error of this path segment tends to converge subsequently, and the score is normal.

[0151] This fine-tuning mechanism controls the trajectory correction accuracy within a feasible and safe spatial range, realizes continuous monitoring through rolling window, risk identification through double-index judgment, and "continuous feedback + continuous adjustment + avoidance of over-adjustment" through adjustment amplitude constraint based on historical statistical boundaries. Taking "stress mutation" and "error expansion" as the core trigger basis, combined with rolling window strategy and dynamic judgment standard, it constitutes a self-stabilization control means at the level of the smallest controllable unit in cutting behavior, which not only enhances the path execution precision, but also improves the response capability to structural mutation areas.

[0152] The above formulas are dimensionless numerical calculations. The formula is obtained by software simulation of a large amount of data to obtain the most recent real situation. The preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0153] It should be understood that the magnitude of the sequence of the above processes does not mean the order of execution, the execution order of the processes should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0154] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0155] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described devices and units can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0156] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited to this. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for accurate core cutting based on intelligent control, characterized in that, The method comprises the following steps: Collecting the surface morphology, texture characteristics and internal acoustic response signals of the core sample, reconstructing a three-dimensional space model through multi-source information fusion, and generating a potential distribution map of the interlayer structure; Based on the interlayer distribution map and the core shape model, a cutting path that fits the real shape of the core is constructed using spatial curvature analysis. The cutting path is evaluated for shortest structural stability, and a continuous path outside the interlayer region is selected; During the cutting process, the spatial offset data between the cutting tool and the core are obtained in real time through position tracking, which are input into the trajectory deviation judgment process for point-by-point fine tuning of the cutting path; When performing the cutting action, the contact angle and feed speed are dynamically adjusted to adapt to the local structural stiffness changes of the core, thereby realizing a stable, continuous and non-destructive cutting process of the interlayer; After the cutting is completed, the structural integrity of the target section is analyzed, and the quality score is calculated based on the trajectory fitting rate. If the score is below the preset threshold, the path reconstruction and re-cutting instructions are executed; The construction process of the cutting path includes generating a plurality of candidate path sets. Each path is initially fitted based on the spatial curvature features extracted from the three-dimensional structure model of the core, forming a trajectory sample with cutting feasibility; For each candidate path, the local stress concentration coefficient corresponding to the path is calculated. The coefficient is defined as the ratio of the average stress gradient of the region crossed by the path under simulated cutting load to the core stiffness response. The top several paths sorted by stress concentration coefficient from small to large are selected as the final candidate set to reduce the risk of structural disturbance during cutting and improve overall stability; The generation of the potential distribution map of the interlayer structure is based on multi-source data joint modeling. The acoustic response signals, image texture changes and surface microstructure scanning data are combined after boundary judgment to form an interlayer judgment function. The function integrates the confidence levels of different source signals through logical integration and constructs an interlayer probability field in the overall core model; The integral expression in the interlayer judgment function adopts one of the three typical combination methods: based on the logical product weighting rule, the maximum confidence selection rule or the Bayesian confidence updating rule; The final confirmation of the cutting path is based on the constructed candidate path set. The coordinate fitting mechanism is introduced into the candidate set. The spatial attitude matching is performed for the core end face and the long axis direction to establish a local reference coordinate frame to eliminate the influence of core deformation. In this coordinate space, the candidate path set is mapped to a feasible trajectory subset in the high-dimensional structure. The shortest stability and interlayer boundary obstacle avoidance ability of the path are used as target parameters. The discrete point array construction method is used for trajectory iterative optimization, and the point array is non-uniformly encrypted near the interlayer boundary; The path optimization process is completed in a variable weight graph search manner. The path optimization process is based on the comprehensive evaluation of the stress distribution performance and spatial obstacle avoidance ability of each path in the core structure model and the candidate path set. The path risk score rule is adaptively adjusted according to the lithology input parameters, so that the final selected path has the minimum disturbance risk and the highest shape fitting degree in the overall structure.

2. The intelligent control based accurate core cutting method as claimed in claim 1, wherein, The trajectory deviation judgment process is realized by comparing the spatial position data collected in the continuous cutting process with the preset trajectory. The trajectory comparison process adopts the minimum curvature change calculation method between continuous time discrete sampling points. The deviation judgment process introduces the comparison mechanism of three different stages before, during and after cutting. The error burst point and the expected error turning point position are automatically marked during execution. Whether to execute the trajectory correction instruction is determined by combining the three-stage error clustering atlas and the multi-threshold judgment standard.

3. The intelligent control based accurate core cutting method as claimed in claim 2, wherein, The dynamic adjustment process of the contact angle and the feed speed in the cutting action combines the core surface response characteristics and the current position of the cutter head to make a joint decision. The calculation of the real-time adjustment parameters is completed in each time slice, and the possible contact form of the next step is judged through a limited prediction window. The joint adjustment strategy does not use a fixed value switching method, but continuously adjusts based on a local target optimal control model. The control amount can be determined jointly based on three indicators: The current position of the tangent direction of the interlayer predicted boundary, the instantaneous incremental change of the cutting force of the cutter head, and the core surface friction feedback signal. These three indicators form a local adjustment cost function in the form of a weighted function, realizing the continuous surface of the parameters.

4. The intelligent control based accurate core cutting method as claimed in claim 3, wherein, If it is judged that the path reconstruction trigger state is triggered, the historical cutting trajectory, error record, and interlayer exposure position information will be automatically archived, and a time sequence cutting behavior trajectory atlas will be formed to generate the next round of path correction scheme. The atlas construction uses event chain representation, all key turning nodes are defined as structural state change points, which constitute the decision tree input data set, and are used as prior conditions in the next path planning.

5. The intelligent control based accurate core cutting method as claimed in claim 4, wherein, The structural integrity analysis includes quantitative evaluation of the flatness of the cutting surface, the number of structural mutation edges and the exposure degree of the interlayer. The flatness evaluation adopts the average value of the fitting surface residual, the number of mutation edges is counted by the number of regions with a first derivative change greater than a certain threshold in the cutting surface, and the exposure degree of the interlayer is calculated by the integral of the predicted interlayer density through the cutting surface. The quality score is output by the preset machine learning model based on the above calculation results and the trajectory fitting rate.

6. The intelligent control based accurate core cutting method as claimed in claim 5, wherein, The point-by-point fine tuning process of the cutting path uses a dynamic optimization strategy based on a rolling window during execution. In the execution interval continuously sliding along the cutting direction, whether to make a slight adjustment to the path point and its adjacent points is judged according to the stress change rate of the current path point and the cutting error trend at each time. To ensure that the cutting trajectory realizes high-precision local correction while maintaining the overall structural stability, the adjustment amplitude of each adjustment must not exceed the standard deviation range of the path curvature change in the historical cutting record. This range is generated based on the trajectory variation statistical characteristics of samples of the same lithology type in the historical path set.

Images