CN-122014255-A - Coal mining machine navigation cutting planning method and system based on working face ladder model

CN122014255ACN 122014255 ACN122014255 ACN 122014255ACN-122014255-A

Abstract

The invention discloses a coal mining machine navigation cutting planning method and system based on a working face ladder model, which are used for directly representing step mutation of a coal seam floor by constructing a discretization working face ladder model, realizing space quantization of uncertainty of geological information based on three-level geological cognitive states of completed cutting times and prediction line-of-sight division, establishing a trusted geological model dynamic evolution mechanism driven from priori driving to data driving, modeling the floor height as a first-order Markov hidden state variable on the basis, generating probability distribution of a future floor section by adopting Kalman filtering and multi-step Bayesian prediction, carrying out on-line compensation by combining exponential decay weighted historical deviation, and finally, generating cylinder target height by self-adaptive weighted fusion by taking a prediction section as a main track and using an adjustment compensation quantity output by a cutting current layering system.

Inventors

WANG XUEWEN
MENG TONGTONG
HUANG YIJIE
SONG YUKE
XIE JIACHENG
LI XIAOPENG
LI YANG
ZHANG PEILIN
WANG XIAOTING
Song Zhengshuo

Assignees

太原理工大学

Dates

Publication Date: 20260512
Application Date: 20260403

Claims (10)

1. The coal mining machine navigation cutting planning method based on the working face ladder model is characterized by comprising the following steps of: step one, constructing a working face ladder model and a three-level geological cognitive state model, wherein the step one comprises the following steps: Discretizing the coal bed bottom plate into a transverse section sequence corresponding to the cutting times along the advancing direction, wherein each transverse section consists of a discrete bed bottom point set distributed along the cutting direction to form a working face ladder model, Based on the number of times of the current completed cutters and a preset prediction vision distance, a three-level geological cognitive state model is constructed, and a working face is divided into three cognitive areas along the advancing direction: The high-confidence actual measurement area is an area which is actually cut and completes data acquisition; the central trust deduction area is a range which is not cut but can be predicted through extrapolation of a historical profile; The low-confidence blind detection zone is a far-end zone exceeding the cognitive ability of the current system; with the propulsion of the coal mining machine, the boundaries of the three cognitive areas synchronously move forward, so that the dynamic migration of the cognitive state is realized; step two, a dynamic evolution mechanism of the trusted geological model is established, which comprises the following steps: constructing an initial discretized three-dimensional bottom plate height field as a priori model based on multi-source static geological data; mapping the measured data to the grid cells of the prior model through interpolation after each cutting is completed, executing parameter coverage operation on the grid cells covered by the measured data, and updating the confidence level state; Step three, performing multi-step probability prediction and immediate updating based on Bayesian state estimation, including: modeling the bottom plate height as a first-order Markov hidden state variable based on the historical actual track and the priori geologic model; Adopting Kalman filtering recursion to solve posterior distribution, and outputting the optimal estimation and variance of the current bottom plate height; Performing multi-step Bayesian prediction based on a state transition model, generating probability distribution of future H-tool bottom plate heights, and performing boundary correction by using roadway real heights; Introducing an outer ring instant updating mechanism of exponential decay weighted historical deviation, performing online compensation on prediction, and outputting a future bottom plate prediction section with probability information; step four, executing dynamic cutting bottom plate generation and automatic height adjustment control, including: The method comprises the steps of merging multi-sensor data to calculate the pose of a roller in real time, generating a six-degree-of-freedom dynamic cutting block by adopting incremental modeling, and forming a continuously evolving dynamic bottom plate model; And taking the Bayesian prediction section as a main track, taking the heightening compensation quantity output based on the cutting current layering system as feedback correction, generating the roller target height through self-adaptive weighting fusion, and issuing a heightening instruction after kinematic constraint and time sequence verification.
2. The coal mining machine navigation cutting planning method based on the working face ladder model of claim 1 is characterized by comprising the steps of carrying out sensor error simulation and system verification, constructing an error coal mining machine model, superposing preset errors comprising inertial navigation drift, random noise and inclination sensor errors on theoretical real poses, generating simulated sensing data driving digital twin scenes, and evaluating the suppression capability of sensor noise by comparing track deviation and heightening response of the theoretical coal mining machine and the error coal mining machine.
3. The method for planning navigation cutting of coal mining machine based on working face ladder model according to claim 1 or 2, wherein in the first step, the working face ladder model is defined as an ordered set: ; Wherein x k is the pushing position corresponding to the kth knife, and s k is the section corresponding to the kth knife, which can be expressed as a function defined on [ z s ,z e ]; the working face step model takes actual cutting action as an index, no continuity constraint exists between adjacent sections, local abrupt change is directly reflected, and a structural basis is provided for subsequent cognitive state division.
4. The method for planning navigation cutting of coal mining machine based on working face ladder model of claim 3, wherein in the second step, the parameter coverage operation is performed by replacing the original prior height value of the grid unit with the actual measured height value, replacing the original prior variance with the sensor measurement variance, and retaining the original prior geological data for the uncovered blind area grid.
5. The method of claim 4 wherein in step two, the updated confidence level state is a region in which the cutting is completed and the measured data is filled, the state is shifted from the confidence deduction region to the high confidence determination region, the newly exposed distal region is marked as "trusted", and the newly exposed distal region is marked as "to be verified" by being brought into the confidence deduction region from the low confidence blind detection region.
6. The method for planning navigation cutting of coal mining machine based on working face ladder model according to claim 1 or 5, wherein in the third step, the multi-step probability prediction based on Bayesian state estimation is specifically: At the fixed cutting position z, modeling the bottom plate height as a hidden state variable x k (z) evolving along with the tool number k, describing the state evolution rule by adopting a first-order Markov process in consideration of certain random fluctuation characteristics of the bottom plate height between adjacent tools, and establishing a state evolution equation: ; Wherein w k (z) is process noise and is used for representing random change of the heights of the bottom plates between adjacent poles caused by local non-stationarity of a geological structure; establishing an observation equation based on the state evolution equation: ; Wherein A k (z) is the actual cutting track observation value of the kth knife, v k (z) is observation noise and is used for representing the measurement error of the sensor; The posterior distribution of the state is solved by adopting Kalman filtering recursion to obtain posterior mean value Sum of the post-test square difference ; Performing multi-step Bayesian prediction based on a state transition model, wherein the probability prediction of the k+m-th knife bottom plate height is as follows: ; Its mean value constitutes the preliminary predicted section ; ; Introducing an uncertainty threshold And a high mutation threshold If the post-test variance is poor Or adjacent order difference amplitude The predicted gradient at the corresponding position z is zeroed out.
7. The method for planning navigation cutting of coal mining machine based on working face ladder model of claim 6, wherein in step three, the outer ring instant update mechanism specifically comprises: After each complete knife cut, the deviation between the actual bottom plate trajectory a k (z) of the knife and the final predicted profile P k (z) taken before cutting is calculated: the stored history buffer forms a bias sequence ; When the prediction of the subsequent nth tool is generated, weighting and fusing the historical deviation by adopting an exponential decay weight to obtain a compensation term c n (z); ; Wherein the weight is And lambda is an attenuation coefficient, and the compensation term is superimposed on the basic prediction P n (z) to obtain a final prediction track after instant updating.
8. The mining machine navigation cutting planning method based on the working face ladder model of claim 1 or 7, wherein in the fourth step, the cutting current layering system comprises the following steps: establishing a multi-level threshold mapping table of a cutting current range and a coal rock firmness coefficient, subdividing the current range into a plurality of levels, wherein different levels correspond to different heightening strategies; Introducing a time duration criterion, and judging the working condition according to the current value and the duration time; and correcting the current threshold interval by recording the actual cutting effect after current alarm.
9. The mining machine navigation cutting planning method based on the working face ladder model of claim 8, wherein in the fourth step, the self-adaptive weighting fusion is performed to generate a roller target height h target (z); ; wherein P k+1 (z) is a predicted section, Δh is an up-regulation compensation quantity output by a current layering system, and a (z) is an adaptive weight coefficient and is determined by the current base plate post-test variance and a current value.
10. A coal mining machine navigation cutting planning system based on a working face ladder model is characterized in that the method of any one of claims 1-9 is executed.

Description

Coal mining machine navigation cutting planning method and system based on working face ladder model Technical Field The invention belongs to the technical field of coal mining, and particularly relates to a coal mining machine navigation cutting planning method and system based on a working face ladder model. Background The coal mine intellectualization is the core direction of industry development, and unmanned operation of a fully mechanized working face requires that a coal mining machine has autonomous sensing and self-adaptive cutting capability on geological conditions of a coal bed. Autonomous and stable cutting of the coal mining machine under complex geological conditions is realized, and the method is a core target of intelligent mining. Early coal cutter cutting mainly relies on a memory cutting mode, a standard track of a cutter is recorded through manual teaching, and the track is repeated by subsequent cutters. The method has certain applicability under the condition of stable coal seam, but is difficult to adapt to fluctuation and change of the coal seam, and under-cutting or over-cutting is often caused under the condition of complex coal seam, so that resource waste or equipment damage is caused. The existing coal cutter self-adaptive cutting track planning method still has the following defects in practical application: (1) Most of the existing methods rely on preset geological models or historical data for track planning, but under the condition of complex coal seams, geological conditions have obvious concealment and sudden changes. The real-time performance and the accuracy of various sensing means are difficult to ensure due to the fact that the sensing system is limited by underground severe environments, and the system is difficult to respond to unpredictable geological changes in real time. (2) Most of the existing methods focus on optimization of geologic models and track planning layers, and have insufficient attention to physical response in the cutting process. When the coal mining machine encounters gangue inclusion, hard coal or rock stratum mutation in actual cutting, an autonomous elevation decision mechanism based on real-time physical performance data such as cutting current, vibration and the like is lacking. The roller height adjustment depends on a preset track or an off-line strategy, and cannot respond to the sudden change of the load in real time, so that the risk of overstock, understock or overload of equipment is caused, and the autonomous adaptability of the cutting process is restricted. (3) Most methods adopt a mode of planning a next tool track based on the previous tool data, and essentially belong to post-correction rather than real-time self-adaption. When the trend or trend of the coal bed is changed drastically, the hysteresis planning cannot realize real follow-up adjustment, and the cutting process still has the risk of undercutting or overdriving. (4) The prior art attempts to fuse multidimensional data such as virtual pre-planning, real-time perception, equipment pose and the like, but lacks a dynamic weighing mechanism based on confidence when information conflicts. The system has difficulty in judging whether the macro model is based on the microscopic detection or not, so that uncertainty exists in control decision. (5) The existing method is usually used for modeling or training aiming at a specific working surface, and the model has strong scene dependence. When the coal mining machine is transferred to coal beds with different geological conditions, the adaptability of the original model is greatly reduced, data are required to be collected again for calibration or training, and the knowledge migration capability across the working face is lacked. Disclosure of Invention The invention aims to provide a coal mining machine navigation cutting planning method and system based on a working face ladder model, which at least solve the technical problems that the step mutation of a coal seam floor is difficult to effectively characterize under the condition of a complex coal seam and the planned path lacks dynamic adaptability. In order to achieve the above object, according to one aspect of the present invention, there is provided a shearer navigation cutting planning method based on a face ladder model, comprising: step one, constructing a working face ladder model and a three-level geological cognitive state model, wherein the step one comprises the following steps: Discretizing the coal bed bottom plate into a transverse section sequence corresponding to the cutting times along the advancing direction, wherein each transverse section consists of a discrete bed bottom point set distributed along the cutting direction to form a working face ladder model, Based on the number of times of the current completed cutters and a preset prediction vision distance, a three-level geological cognitive state model is constructed, and a working face is divided in