A smart anti-jamming method and system for a fully computerized three-arm rock drilling rig

Abstract

This invention discloses an intelligent anti-jamming method and system for a fully computerized three-arm drilling rig, relating to the field of three-arm drilling rig control technology. The method includes: collecting multi-dimensional sensor parameters; performing real-time posterior probability estimation of geological types based on the multi-dimensional sensor parameters to generate posterior probabilities for geological types; adaptively weighting and fusing the multi-dimensional sensor parameters using the posterior probabilities of geological types to form a composite state feature vector; predicting jamming precursor probabilities based on the time series of the composite state feature vector; constructing a time-varying multi-dimensional reward signal based on the jamming precursor probabilities and the posterior probabilities of geological types; generating a collaborative optimal anti-jamming strategy command for the three drill arms under the guidance of the time-varying multi-dimensional reward signal; precisely adjusting the advance speed, rotation speed, and impact frequency of the three drill arms according to the collaborative optimal anti-jamming strategy command; and performing differentiated proactive prevention actions for jamming risks corresponding to different geological types.

Description

Technical Field

[0001] This invention belongs to the field of three-arm rock drilling rig control technology, and more specifically, relates to an intelligent anti-jamming method and system for a fully computerized three-arm rock drilling rig. Background Technology

[0002] The fully computerized three-arm drilling rig is a core piece of equipment in mining, tunneling, and geotechnical engineering. With its high degree of automation, high drilling efficiency, and wide operating range, it is widely used in various tunneling operations, including railways, highways, national defense, and water conservancy projects. It enables the drill arm to automatically drill according to a pre-designed hole layout, precisely controlling hole position, advance beam angle, and hole depth, effectively reducing over-drilling and under-drilling, and improving blasting efficiency. However, in actual operation, "drill rod jamming" is a key bottleneck restricting the operational efficiency and safety of the fully computerized three-arm drilling rig. Jamming not only interrupts drilling operations but can also cause the drill rod to bend and break, damaging the drilling mechanism, increasing equipment maintenance costs, and in severe cases, leading to safety accidents.

[0003] Existing anti-jamming systems for fully computerized three-arm rock drilling rigs mostly employ traditional control methods, relying on fixed pressure thresholds or simple logic judgments to achieve anti-jamming control, which has the following three limitations: First, insufficient geological adaptability. Geological conditions are complex and varied, with significant differences in the hardness, fissure development, and cave distribution of different rock strata. Fixed thresholds cannot adapt to diverse geological environments, which can easily lead to misjudgments or delayed responses, or remedial measures may be taken only after the drill bit gets stuck, thus missing the opportunity for prevention; or anti-jamming actions may be frequently triggered, resulting in ineffective drilling and reduced work efficiency.

[0004] Second, there is a lack of early warning sign recognition capabilities. Existing systems rely solely on a single pressure threshold to extract early warning signs of splintering, failing to comprehensively mine early patterns before splintering occurs from multi-dimensional sensor signals. The early warning sign recognition window is short, the recognition rate is low, and true proactive prevention cannot be achieved.

[0005] To address the aforementioned three technical issues, while some existing technologies have introduced single reinforcement learning algorithms into rock drilling control, they generally suffer from the following shortcomings: First, the state representation dimension is singular, failing to integrate multimodal sensing information such as acoustic emission signals, thus limiting the timeliness and accuracy of identifying early signs of drill jamming; second, the reward function uses static weights, making it impossible to dynamically adjust the priority of each target according to real-time geological conditions; and third, the three arms are controlled independently without establishing an inter-arm coordination mechanism, resulting in significant hydraulic interference issues when multiple arms are operating simultaneously.

[0006] Therefore, developing a fully closed-loop intelligent anti-jamming drill system that integrates multimodal perception, online geological classification, temporal precursor prediction, geological adaptive reward, and three-arm collaborative decision-making has become a key technological direction for improving the overall operational performance of a fully computerized three-arm rock drilling rig. Summary of the Invention

[0007] To address the above technical problems, this invention proposes an intelligent anti-jamming system for a fully computerized three-arm rock drilling rig, comprising a status perception module, an intelligent decision-making module, and an execution control module, including: The state perception module includes a multimodal sensor component, an online Bayesian geological classification unit, and a geological adaptive state coding unit. The multimodal sensor component collects multidimensional sensor parameters, the online Bayesian geological classification unit performs real-time posterior probability estimation of geological types based on the multidimensional sensor parameters, and generates posterior probabilities of geological types. The geological adaptive state coding unit performs adaptive weighted fusion of multidimensional sensor parameters based on the posterior probabilities of geological types to form a composite state feature vector. The intelligent decision-making module includes a temporal convolutional network precursor prediction unit, a geological adaptive dynamic reward function unit, and a three-arm collaborative Nash equilibrium strategy unit. The temporal convolutional network precursor prediction unit predicts the probability of drill string jamming precursors based on the time series of composite state feature vectors. The geological adaptive dynamic reward function unit constructs a time-varying multi-dimensional reward signal based on the probability of drill string jamming precursors and the posterior probability of geological type. The three-arm collaborative Nash equilibrium strategy unit generates a collaborative optimal anti-jamming strategy instruction for the three drill arms under the guidance of the time-varying multi-dimensional reward signal. The execution control module precisely adjusts the advance speed, rotation speed and impact frequency of the three drill arms according to the collaborative optimal anti-jamming strategy command, and performs differentiated proactive prevention actions for the risk of jamming corresponding to different geological types.

[0008] Furthermore, the intelligent decision-making module also includes: a digital twin synchronization unit; During actual operation, the actual state of the trolley is synchronized in real time. The digital twin synchronization unit calculates the adaptive synchronization gain based on the state error between the actual state of the trolley and the digital twin state vector of the digital twin environment, and corrects the digital twin state vector.

[0009] Furthermore, the posterior probability of generating geological types includes: in, For the first Geological types The conditional likelihood function, For at any time The observed feature vector, The total number of dimensions, The number of geological types, For the first Geological types The conditional likelihood function, For the first Each geological type at time The posterior probability, For the first Geological types The next The standard deviation of the observed eigenvectors For the first Geological types Next moment The 3D observation feature vector, For the first Geological types The next The mean of the observed eigenvectors, For the first Each geological type at time The posterior probability, For the first Each geological type at time The posterior probability.

[0010] Furthermore, the formation of the composite state feature vector includes: in, For the first The drilling arm was at a constant time The composite state feature vector, For the first Each geological type at time Adaptive coding weights, As the first weight, For the first The drilling arm was at a constant time The pressure to advance For the first The drilling arm was at a constant time The sliding average of propulsion pressure, For the first The slip standard deviation of the propulsion pressure of the No. 1 drill arm As the second weight, For the first The drilling arm was at a constant time The rate of change of propulsion pressure, To push forward the design upper limit of the pressure change rate, As the third weight, For the first The drilling arm was at a constant time The speed of advancement, To push the speed design upper limit, As the fourth weight, For the first The drilling arm was at a constant time rotational speed, An upper limit is designed for the rotational speed. As the fifth weight, For the first The drilling arm was at a constant time The longitudinal acceleration of the drill rod, This is the acceleration due to gravity.

[0011] Furthermore, calculate the first... Each geological type at time Adaptive coding weights include: in, For the first Sensing coding sensitivity coefficients corresponding to each geological condition.

[0012] Furthermore, time-varying multi-dimensional reward signals include: at time... Drilling efficiency bonus At any moment Energy efficiency bonus At any moment Parameter smoothness reward At any moment Smooth slag discharge reward and at the moment Total reward ; Calculate at time Total reward : in, For at any time The first reward weight, which changes dynamically with geological conditions. For at any time The second reward weight, which changes dynamically with geological conditions. For the moment The third reward weight, which changes dynamically with geological conditions. For the moment The fourth reward weight, which changes dynamically with geological conditions. This is the first benchmark scaling factor. This is the second benchmark scaling factor. This is the third benchmark scaling factor. For at any time Early warning signs of a stuck pin, For at any time Penalty for chip malfunction, For at any time Equipment overload penalty.

[0013] Furthermore, calculations are performed at time... Reward weights that change dynamically with geological conditions include: in, For the first The baseline reward weight for each reward component. For the first Each geological type at time The posterior probability, For the first The geological modulation sensitivity coefficient of the reward component, For the first The reward amount is in the first Adjustment factor matrix elements under various geological conditions The number of geological types.

[0014] Furthermore, the instructions for generating the cooperative optimal anti-jamming strategy for the three drill arms include: in, For the first The drilling arm was at a constant time The optimal action vector, For the first A centralized evaluation network for the No. 1 drill arm. For at any time The global state, For the first The motion vector of the drill arm For the remaining two drill arms at time The action vector, The penalty coefficient for inter-arm coordination conflict. For the first The cost function of the multi-arm conflict caused by the drill arm, after iterative convergence, yields the combination of coordinated actions of the three arms. As a collaborative optimal anti-blocking strategy instruction.

[0015] This invention also proposes an intelligent anti-jamming method for a fully computerized three-arm rock drilling rig, comprising: Collect multi-dimensional sensor parameters, perform real-time posterior probability estimation of geological types based on multi-dimensional sensor parameters, generate posterior probabilities of geological types, and adaptively weight and fuse multi-dimensional sensor parameters using the posterior probabilities of geological types to form a composite state feature vector. Based on the time series of composite state feature vectors, the probability of drill bit jamming precursors is predicted. A time-varying multi-dimensional reward signal is constructed based on the probability of drill bit jamming precursors and the posterior probability of geological type. Under the guidance of the time-varying multi-dimensional reward signal, a collaborative optimal anti-jamming strategy instruction for the three drill arms is generated. Based on the collaborative optimal anti-jamming strategy command, the advance speed, rotation speed and impact frequency of the three drill arms are precisely adjusted, and differentiated proactive prevention actions are performed for the jamming risk corresponding to different geological types.

[0016] Furthermore, during actual operation, the actual state of the trolley is synchronized in real time. Based on the state error between the actual state of the trolley and the state vector of the digital twin environment, the adaptive synchronization gain is calculated, and the state vector of the digital twin is corrected.

[0017] In summary, the technical solutions conceived by this invention have the following beneficial effects compared with the prior art: 1. Multimodal sensing fusion significantly improves the accuracy of precursor identification. This is achieved by introducing an acoustic emission sensor and constructing a multimodal observation feature vector. By combining online Bayesian geological classification units, real-time dynamic identification of geological types is achieved. The feature dimensions for identifying pinhole jamming precursors have been expanded from 5 dimensions (single hydraulic pressure + motion parameters) in the original scheme to 8 dimensions. The average lead time for identifying pinhole jamming precursors can reach 2.5 seconds, providing ample response time for preventive intervention.

[0018] 2. Geological adaptive state coding enables dynamic focusing of sensor information. The geological adaptive state coding unit combines Bayesian posterior probability with sensor sensitivity coefficients through Softmax weighting, enabling the composite state features to be dynamically focused. The system automatically highlights the most critical sensing components under different geological conditions, enhances the ability of state representation to distinguish the geological environment, and avoids irrelevant signals from interfering with the agent's decision-making.

[0019] 3. Temporal convolutional anomalous prediction, balancing long-term temporal dependence and real-time performance. The 6-layer dilated causal convolutional structure of TCN maintains parallel computing efficiency while covering approximately 12.7 seconds of historical information in the receptive field, with an inference latency of less than 5 ms, meeting the real-time requirement of a 10 Hz control frequency. Compared with LSTM anomalous prediction methods, the training speed is improved by about 3 times.

[0020] 4. Geologically adaptive dynamic rewards eliminate the geological adaptation limitations of static weights. Reward weights. With real-time Bayesian posterior probability dynamic adjustment, the agent automatically raises safety priority in karst areas and automatically strengthens efficiency targets in intact rock masses, eliminating the adaptability defects of traditional static weighted reward functions under complex and variable geological conditions, and improving overall operational efficiency by about 18%.

[0021] 5. Three-arm cooperative Nash equilibrium strategy to eliminate multi-arm hydraulic interference. The three-arm cooperative strategy unit based on the CTDE framework establishes a multi-arm conflict cost function. By incorporating the hydraulic coupling interference of the three arms into the strategy optimization objective, the mutual interference problem caused by the independent control of the three arms is fundamentally solved, and the measured standard deviation of the hydraulic pressure fluctuation of the trolley is reduced by about 30%.

[0022] 6. Digital twin synchronous pre-training eliminates the risk of cold start on physical machines. The adaptive synchronous gain mechanism ensures high-fidelity tracking of the physical system state by the digital twin model. Virtual pre-training completes policy initialization in a safe environment, reducing the failure rate of the chip during the cold start phase on physical machines by more than 60%, while also avoiding damage to equipment due to insufficient algorithm exploration. Attached Figure Description

[0023] Figure 1 This is a system structure diagram of Embodiment 1 of the present invention; Figure 2 This is a flowchart of the method in Embodiment 2 of the present invention; Figure 3 This is a system structure diagram of Embodiment 4 of the present invention; Figure 4 This is a schematic diagram of a three-arm rock drilling rig. Detailed Implementation

[0024] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0025] The method provided by this invention can be implemented in a terminal environment that may include one or more of the following components: a processor, a storage medium, and a display screen. The storage medium stores at least one instruction, which is loaded and executed by the processor to implement the method described in the following embodiments.

[0026] A processor may include one or more processing cores. The processor uses various interfaces and lines to connect various parts of the terminal, and performs various functions and processes data by running or executing instructions, programs, code sets or instruction sets stored in the storage medium, and by calling data stored in the storage medium.

[0027] Storage media can include random access memory (RAM) or read-only memory (ROM). Storage media can be used to store instructions, programs, code, code sets, or instructions.

[0028] The display screen is used to show the user interface of each application.

[0029] In addition, those skilled in the art will understand that the structure of the terminal described above does not constitute a limitation on the terminal. The terminal may include more or fewer components, or combine certain components, or have different component arrangements. For example, the terminal may also include radio frequency circuits, input units, sensors, audio circuits, power supplies, and other components, which will not be described in detail here.

[0030] Example 1 like Figure 1 As shown, this embodiment proposes a fully computerized three-arm rock drilling rig (such as...). Figure 4 The intelligent anti-jamming system (shown) includes a status perception module, an intelligent decision-making module, and an execution control module, comprising: The state perception module includes a multimodal sensor component, an online Bayesian geological classification unit, and a geological adaptive state coding unit. The multimodal sensor component collects multidimensional sensor parameters, the online Bayesian geological classification unit performs real-time posterior probability estimation of geological types based on the multidimensional sensor parameters, and generates posterior probabilities of geological types. The geological adaptive state coding unit performs adaptive weighted fusion of multidimensional sensor parameters based on the posterior probabilities of geological types to form a composite state feature vector. Preferably, in addition to the existing pressure sensor, velocity sensor, and displacement sensor, an additional set of acoustic emission sensors is installed at the fixed end of the rock drilling mechanism and the shank connection of each drill arm to collect stress wave signals generated by rock fracturing during drilling. The acoustic emission sensor sampling frequency is not less than 500 kHz, and effective acoustic emission features are extracted in conjunction with a bandpass filter (passband range 50 kHz~400 kHz).

[0031] Therefore, the multidimensional sensor parameters include hydraulic parameters, motion parameters, and acoustic emission parameters.

[0032] Preferably, to achieve real-time identification of geological types, this embodiment introduces an online Bayesian geological classification unit. The geological type set is defined as follows: ,in These correspond to intact rock masses, fractured rock masses, karst cave areas, hard rock areas, and soft rock areas, respectively, and are denoted as... .

[0033] At each control moment Constructing observation feature vectors from multidimensional operating parameters ,in These are, respectively, the average propulsion pressure, the rate of change of propulsion pressure, the average slewing pressure, the average impact pressure, the normalized value of propulsion speed, the normalized value of slewing speed, the root mean square of drill bit acceleration, and the cumulative ringing count of acoustic emissions. Under different geological types, The components of follow a Gaussian distribution with a mean of . Standard deviation is ( , The parameters are obtained through offline pre-training using historical job data.

[0034] Specifically, based on the posterior probability of the previous time step. As the current prior, the posterior probability of each geological type at the current moment is calculated using the Bayesian recursive formula. That is, the posterior probability of generating a geological type includes: in, For the first Geological types The conditional likelihood function, For at any time The observed feature vector, The total number of dimensions, The number of geological types, For the first Geological types The conditional likelihood function, For the first Each geological type at time The posterior probability, For the first Geological types The next The standard deviation of the observed eigenvectors For the first Geological types Next moment The 3D observation feature vector, For the first Geological types The next The mean of the observed eigenvectors, For the first Each geological type at time The posterior probability, For the first Each geological type at time The posterior probability, During system initialization, the prior is set to a uniform distribution, i.e. , .

[0035] Geological category probability vector The data is synchronously output to the geological adaptive state coding unit and the geological adaptive dynamic reward function unit, serving as a unified information basis for subsequent modules to perceive geological conditions.

[0036] Specifically, the formation of composite state feature vectors includes: in, For the first The drilling arm was at a constant time The composite state feature vector, For the first Each geological type at time Adaptive coding weights, As the first weight, For the first The drilling arm was at a constant time The pressure to advance For the first The drilling arm was at a constant time The sliding average of propulsion pressure, For the first The slip standard deviation of the propulsion pressure of the No. 1 drill arm As the second weight, For the first The drilling arm was at a constant time The rate of change of propulsion pressure, To push forward the design upper limit of the pressure change rate, a value of 2.5 MPa / s is adopted. As the third weight, For the first The drilling arm was at a constant time The speed of advancement, To set the upper limit for propulsion speed, a value of 0.08 m / s is adopted. As the fourth weight, For the first The drilling arm was at a constant time rotational speed, The upper limit for the rotational speed is set at 300 r / min. As the fifth weight, For the first The drilling arm was at a constant time The longitudinal acceleration of the drill rod, For gravitational acceleration, the above weight values ​​are given as examples. , , , , At any moment global state Defined as the concatenation of three-arm composite features and geological posterior probabilities: ,in .

[0037] Specifically, the definition of the first Sensing coding sensitivity coefficients corresponding to geological conditions (Dimensionless), its numerical examples are as follows: (Integrity of rock mass) (Rock mass with well-developed fissures) (Cave area) (Hard rock area) (Soft rock area), calculate the first Each geological type at time Adaptive coding weights include: in, For the first Sensing coding sensitivity coefficients corresponding to each geological condition The larger the value, the more sensitive the perceived posterior probability of the geological type is to the influence of the encoding weight, thus assigning a larger state space attention weight under the corresponding geological conditions.

[0038] The intelligent decision-making module includes a temporal convolutional network precursor prediction unit, a geological adaptive dynamic reward function unit, and a three-arm collaborative Nash equilibrium strategy unit. The temporal convolutional network precursor prediction unit predicts the probability of drill string jamming precursors based on the time series of composite state feature vectors. The geological adaptive dynamic reward function unit constructs a time-varying multi-dimensional reward signal based on the probability of drill string jamming precursors and the posterior probability of geological type. The three-arm collaborative Nash equilibrium strategy unit generates a collaborative optimal anti-jamming strategy instruction for the three drill arms under the guidance of the time-varying multi-dimensional reward signal. Preferably, based on the geological adaptive state coding output, this invention introduces a Temporal Convolutional Network (TCN) precursor prediction unit to accurately predict the probability of stuck probe precursors. The TCN adopts a causal dilated convolutional structure and has a total of Layer, number Layer expansion ratio ,Right now The length of the receptive field coverage is At any moment ( For the kernel size, take The receptive field is 127 time points, which corresponds to approximately 12.7 seconds of historical information at a sampling frequency of 10 Hz.

[0039] TCN with length as Historical composite state feature sequence of steps For input, the first Hidden state vectors Computed by dilated causal convolution with residual connections: in, Expansion ratio The causal dilated convolution operator guarantees that future information is not used; : No. Layer convolution kernel tensor, The number of hidden layer channels is taken as... ; : No. Layer bias vector; Linear modified activation function; residual connection Ensure stable gradient propagation; the input of layer 0 is... (extended by linear projection to) ); go through After layer convolution, the probability of a pin tether malfunction is... Calculated by the output layer: in, The Sigmoid activation function maps the output to... ; This is the output layer weight vector; This is the output layer bias scalar. Indicates the first The drilling arm was at a constant time The predicted probability of a probe jamming precursor is input as a penalty term into the geological adaptive dynamic reward function and simultaneously displayed on the risk level interface of the human-computer interaction module.

[0040] Specifically, time-varying multi-dimensional reward signals include: at time... Drilling efficiency bonus At any moment Energy efficiency bonus At any moment Parameter smoothness reward At any moment Smooth slag discharge reward and at the moment Total reward ; Calculate at time Total reward : in, For at any time The first reward weight, which changes dynamically with geological conditions. For at any time The second reward weight, which changes dynamically with geological conditions. For the moment The third reward weight, which changes dynamically with geological conditions. For the moment The fourth reward weight, which changes dynamically with geological conditions. This is the first benchmark scaling factor. This is the second benchmark scaling factor. This is the third benchmark scaling factor. For at any time Early warning signs of a stuck pin, For at any time Penalty for chip malfunction, For at any time The equipment overload penalty, the example value of the above benchmark proportional coefficient is [value missing]. , , (All are dimensionless).

[0041] Preferably, calculate drilling efficiency bonus : in, Single step time The incremental drilling depth, in meters, is calculated differentially from the displacement sensor. Control the time step, in seconds. s; : Based on the propulsion speed, take m / s, then m is the reference single-step drilling depth; Efficiency nonlinear index (dimensionless), with a value greater than 1 to increase the incentive for high-speed drilling; Indicator function, when The value is 1 if the drilling stops, and 0 otherwise, to avoid giving a positive reward when the drilling stops. Calculate energy efficiency reward : in, Current control step (time) Total power consumption of the three-arm hydraulic system of the entire vehicle, in kW; Energy consumption penalty coefficient, unit m / kW; m: Prevents division by zero and normal quantities; The lower the energy consumption per unit drilling depth, the closer this value is to 1; Calculate parameter smoothness reward : in, The difference between the current action vector and the previous action vector. The normalized joint motion vector (dimensionless) of the three-arm propulsion speed, rotation speed, and impact frequency. It is the Euclidean norm (dimensionless); Smoothness sensitivity coefficient (dimensionless); The smoother the change in motion, the closer this value is to 1, effectively suppressing frequent and drastic parameter switching; Calculate the reward for smooth slag discharge : in, The slag discharge smoothness index (dimensionless) is estimated by fusing the flushing water flow sensor and the return air dust concentration sensor. A higher value indicates smoother slag discharge. The calculation formula is as follows: in, The current flushing water volumetric flow rate (L / min) is collected by an electromagnetic flow meter installed on the flushing water pipeline; L / min is the upper limit of the designed flushing water flow rate; Let be the water flow excitation coefficient (dimensionless), and let its value be such that... achieve At 80% of the time, the Sigmoid output approaches 0.99; This is a sigmoid function, with both the independent variable and the output being dimensionless; The current dust concentration (mg / m³) in the return air tunnel is collected by a laser dust sensor installed at the return air inlet. mg / m³ is the reference dust concentration under normal slag discharge conditions, determined by statistical analysis of historical operation data; mg / m³ represents the standard deviation of dust concentration tolerance; the water flow rate term uses a sigmoid function to characterize the monotonic relationship that "the more abundant the flow, the smoother the slag discharge," while the dust term uses a sigmoid function. The Gaussian kernel centered on the dust concentration gives a low score when the dust concentration is too low (dust is not fully carried) or too high (orifice blockage). Reference threshold for slag discharge smoothness (dimensionless). Logistic kurtosis coefficient (dimensionless); subtracting 0.5 makes... ,when Rewards reset to zero; calculation card trigger penalty. : in, Take the highest probability of a stuck pin in the three arms (dimensionless). , These are linear and quadratic penalty coefficients (dimensionless), respectively. The quadratic term causes the penalty to increase sharply when the probability of an aura is high, thus strengthening the avoidance of high-risk states.

[0042] Calculate the penalty for chip failure : in, The baseline penalty amplitude (dimensionless) for chip failure is the highest-order negative incentive in the reward function, ensuring safety as the priority. : Cumulative fault aggravation coefficient (dimensionless), which means that for each time a pin jamming fault occurs, the baseline penalty increases by 50%, and the cumulative penalty mechanism strengthens the constraint on repeated pin jamming. Deadline The cumulative number of times the drill bit gets stuck in the current working hole (dimensionless integer); Overload penalty for computing devices : in, The upper limit of the propulsion pressure safety is set at 25 MPa. The overload penalty exponent (dimensionless), with a power greater than 1, makes the penalty more severe the more the upper limit is exceeded, thus acting as a soft boundary constraint. Specifically, the calculation is performed at time... Reward weights that change dynamically with geological conditions include: in, For the first The baseline reward weight for each reward component, with values ​​for each component being [value]. , , , ,satisfy , For the first Each geological type at time The posterior probability, For the first The geological modulation sensitivity coefficient of the reward component is set to a value of , , , , For the first The reward amount is in the first Adjustment factor matrix elements under various geological conditions The number of geological types.

[0043] Preferred, Numerical examples (with drilling efficiency reward weights) For example): in intact rock mass Down (Increased weighting of efficiency encourages rapid drilling); fractured rock mass Down Cave area Down (Efficiency weight reduced, safety crossing prioritized); Hard rock area Down ; soft rock area Down . The effective range is determined by This constraint ensures that the weight is always positive.

[0044] Preferred total reward As training signals, used to fit a centralized evaluation network. Evaluate network parameters By minimizing the following based Update the Bellman residual loss: in, Mathematical expectation operator Priority experience replay buffer, storing interaction samples accumulated during system operation. The sampling probability is determined by the priority experience replay mechanism in Section 5; Discount factor (dimensionless); The target evaluation network parameters (dimensionless) are determined by the evaluation network parameters. via soft update Slow tracking, These are soft update coefficients (dimensionless), used to stabilize the training objective; : Target network parameters The calculated target Q value is used to construct the Bellman target and avoid excessive bootstrap bias during training. : The mean squared expected value of the Bellman residuals (dimensionless). Specifically, when three drilling arms operate simultaneously, their hydraulic systems are coupled through the trolley's central hydraulic station. The movement of one arm can interfere with the hydraulic parameters of the other two. To address the multi-arm collaborative decision-making problem, this embodiment, based on the Centralized Training with Decentralized Execution (CTDE) framework, innovatively introduces a multi-arm conflict cost function with spatial decay, considering the hydraulic spatial coupling characteristics of the three-arm rock drilling mechanism. This constructs a three-arm collaborative Nash equilibrium strategy unit, achieving multi-arm collaborative anti-jamming control that surpasses existing CTDE frameworks. Specifically, the instructions for generating the optimal collaborative anti-jamming strategy for the three drilling arms include: in, For the first The drilling arm was at a constant time The optimal action vector, For the first A centralized evaluation network for the No. 1 drill arm. For at any time The global state, For the first The motion vector of the drill arm For the remaining two drill arms at time The action vector, The penalty coefficient for inter-arm coordination conflict. For the first The cost function of the multi-arm conflict caused by the drill arm, after iterative convergence, yields the combination of coordinated actions of the three arms. As a collaborative optimal anti-blocking strategy instruction.

[0045] Preferred, the first Cost function of multi-arm conflict caused by drill arm No. 1 for: in, : No. Drill Arm Motion Vector The resulting change in propulsion pressure, in MPa. MPa: Upper limit of propulsion pressure safety, which has the same meaning as overload penalty; : No. Number and The physical distance between the fixed ends of the drill arm, in meters; m: Spatial coupling attenuation length scale (unit: m); The above formula shows that the closer the arm distance ( The larger the term, the more significant the conflict cost caused by the difference in pressure changes between the two arms, prompting the agent to automatically avoid action combinations where adjacent arms simultaneously adjust pressure significantly. The three-arm actions are solved for Nash equilibrium using an iterative optimal response method. In each iteration, the actions of the other two arms are fixed, and the action of the current arm is optimized. After convergence, the coordinated action combination of the three arms is obtained. And then send it to the execution control module.

[0046] The execution control module precisely adjusts the advance speed, rotation speed and impact frequency of the three drill arms according to the collaborative optimal anti-jamming strategy command, and performs differentiated proactive prevention actions for the risk of jamming corresponding to different geological types.

[0047] Specifically, the intelligent decision-making module further includes: a digital twin synchronization unit; During actual operation, the actual state of the trolley is synchronized in real time. The digital twin synchronization unit calculates the adaptive synchronization gain based on the state error between the actual state of the trolley and the digital twin state vector of the digital twin environment, and corrects the digital twin state vector.

[0048] Preferably, the present invention constructs a digital twin model of a fully computerized three-arm rock drilling rig, which is used to perform large-scale virtual pre-training of reinforcement learning algorithms before deployment and to synchronize the rig status in real time during actual operation, thereby assisting in the online verification of strategies and parameter updates.

[0049] Define the normalized synchronization error between the actual state of the trolley and the state vector of its digital twin: In the formula, For at any time The actual condition of the trolley For digital twin state vectors, The dimension is the state vector.

[0050] To accelerate state correction in digital twin models when synchronization errors are large, an adaptive synchronization gain is introduced: in, Nominal synchronous gain (dimensionless); Gain modulation amplitude (dimensionless). Synchronization error threshold (dimensionless), when Time gain close ,when Time gain trend This accelerates the convergence of the twin model's state. The digital twin state is updated once per control step: in, This is a dynamic model for a digital twin. During the virtual pre-training phase, the agent completes at least [a certain task] in the digital twin environment. By using interactive steps to obtain initialization strategy parameters before migrating to the physical system, the failure rate of the chip jammer during the cold start phase of the physical machine can be reduced by more than 60%.

[0051] Preferably, the execution control module receives the optimal combination of coordinated actions output by the three-arm coordinated Nash equilibrium strategy unit. The normalized motion vector is mapped to hydraulic valve opening command (%), motor speed command (r / min), and impact frequency command (Hz), which are then precisely executed by the electro-proportional valve control system. The execution control module receives the predicted probability of the preceding events. Upon receiving an alarm signal exceeding 0.70, the system automatically enters the pin jamming warning execution mode and responds accordingly based on the following classification strategies: Prevention of cave-related drilling blockages: Online Bayesian geological classification unit identification When the probability of encountering a cave exceeds 0.60, the control module will reduce the propulsion force to 50% of its current value, the impact pressure to 40%, and the rotation speed to 30 r / min. This "light impact + slow rotation" mode will smoothly traverse the cave area, preventing rapid accumulation of debris that could jam the drill rod. Real-time monitoring will be conducted during the traversal. When the probability of the precursor drops below 0.20, the parameters return to normal.

[0052] Gradual jamming prevention: When the pressure rise slope (provided by the least-squares fitting slope of the feature extraction unit) exceeds 1.2 MPa / s, the execution control module intervenes in the early stages of precursor development, reducing the propulsion speed by 15% (in... Adjust within the range of 10% to 20%, while increasing the rotational speed by 8% (within...). (Adjust within the range of 5% to 10%), find the drilling path with the least resistance, and resolve the risk before the drill bit gets stuck.

[0053] Prevention of crack / hard rock jamming: Geological classification and identification (Crack development) or (In hard rock areas) the execution control module adjusts the impact power (adjustment range is 70%~130% of the rated value) and rotation speed (adjustment range is...) in real time according to the current rotation pressure. (40%~100%), avoiding "dead ends" in hard rock and "air blasts" in fractures, while reducing equipment wear. The above adjustment values ​​are determined by the output action vector of the three-arm collaborative Nash equilibrium strategy unit. The execution control module is only responsible for mapping and execution, and does not perform rule judgment independently.

[0054] Example 2 like Figure 2 As shown, this embodiment proposes an intelligent anti-jamming method for a fully computerized three-arm rock drilling rig, including: Step S1: Collect multi-dimensional sensor parameters, perform real-time posterior probability estimation of geological type based on multi-dimensional sensor parameters, generate posterior probability of geological type, and adaptively weight and fuse multi-dimensional sensor parameters through posterior probability of geological type to form composite state feature vector. Step S2: Based on the time series of the composite state feature vector, predict the probability of drill bit jamming precursors; construct a time-varying multi-dimensional reward signal based on the probability of drill bit jamming precursors and the posterior probability of geological type; and generate a collaborative optimal anti-jamming strategy instruction for the three drill arms under the guidance of the time-varying multi-dimensional reward signal. Step S3: Based on the collaborative optimal anti-jamming strategy command, the advance speed, rotation speed and impact frequency of the three drill arms are precisely adjusted, and differentiated proactive prevention actions are performed for the jamming risk corresponding to different geological types.

[0055] Since Example 2 is based on Example 1, it will not be described again.

[0056] Example 3 This invention also proposes a storage medium storing multiple instructions for implementing the intelligent anti-jamming method for a fully computerized three-arm rock drilling rig.

[0057] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0058] Optionally, in this embodiment, the storage medium is configured to store program code for performing the method steps of Embodiment 1.

[0059] Example 4 This invention also proposes an electronic device, such as... Figure 3 As shown, it includes a processor and a storage medium connected to the processor. The storage medium stores multiple instructions, which can be loaded and executed by the processor to enable the processor to execute the intelligent anti-jamming method for a fully computerized three-arm rock drilling rig.

[0060] Specifically, the electronic device in this embodiment can be a computer terminal, which may include one or more processors and a storage medium.

[0061] The storage medium can be used to store software programs and modules, such as the intelligent anti-jamming method for a fully computerized three-arm rock drilling rig in this embodiment of the invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the storage medium, thus realizing the aforementioned intelligent anti-jamming method for a fully computerized three-arm rock drilling rig. The storage medium may include high-speed random access storage media, and may also include non-volatile storage media, such as one or more magnetic storage systems, flash memory, or other non-volatile solid-state storage media. In some instances, the storage medium may further include storage media remotely configured relative to the processor, which can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0062] The processor can execute the method steps of Embodiment 1 by calling the information and application stored in the storage medium through the transmission system.

[0063] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0064] In the several embodiments provided by this invention, it should be understood that the disclosed technical content can be implemented in other ways. The system embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between units or modules, and may be electrical or other forms.

[0065] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0066] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0067] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, optical disks, and other media capable of storing program code.

[0068] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. An intelligent anti-jamming system for a fully computerized three-arm rock drilling rig, comprising a status perception module, an intelligent decision-making module, and an execution control module, characterized in that, include: The state perception module includes a multimodal sensor component, an online Bayesian geological classification unit, and a geological adaptive state coding unit. The multimodal sensor component collects multidimensional sensor parameters, the online Bayesian geological classification unit performs real-time posterior probability estimation of geological types based on the multidimensional sensor parameters, and generates posterior probabilities of geological types. The geological adaptive state coding unit performs adaptive weighted fusion of multidimensional sensor parameters based on the posterior probabilities of geological types to form a composite state feature vector. The intelligent decision-making module includes a temporal convolutional network precursor prediction unit, a geological adaptive dynamic reward function unit, and a three-arm collaborative Nash equilibrium strategy unit. The temporal convolutional network precursor prediction unit predicts the probability of drill string jamming precursors based on the time series of composite state feature vectors. The geological adaptive dynamic reward function unit constructs a time-varying multi-dimensional reward signal based on the probability of drill string jamming precursors and the posterior probability of geological type. The three-arm collaborative Nash equilibrium strategy unit generates a collaborative optimal anti-jamming strategy instruction for the three drill arms under the guidance of the time-varying multi-dimensional reward signal. The execution control module precisely adjusts the advance speed, rotation speed and impact frequency of the three drill arms according to the collaborative optimal anti-jamming strategy command, and performs differentiated proactive prevention actions for the risk of jamming corresponding to different geological types.

2. The intelligent anti-jamming system for a fully computerized three-arm rock drilling rig as described in claim 1, characterized in that, The intelligent decision-making module also includes: a digital twin synchronization unit; During actual operation, the actual state of the trolley is synchronized in real time. The digital twin synchronization unit calculates the adaptive synchronization gain based on the state error between the actual state of the trolley and the digital twin state vector of the digital twin environment, and corrects the digital twin state vector.

3. The intelligent anti-jamming system for a fully computerized three-arm rock drilling rig as described in claim 1, characterized in that, The posterior probability of generating a geological type includes: in, For the first Geological types The conditional likelihood function, For at any time The observed feature vector, The total number of dimensions, The number of geological types, For the first Geological types The conditional likelihood function, For the first Each geological type at time The posterior probability, For the first Geological types The next The standard deviation of the observed eigenvectors For the first Geological types Next moment The 3D observation feature vector, For the first Geological types The next The mean of the observed eigenvectors, For the first Each geological type at time The posterior probability, For the first Each geological type at time The posterior probability.

4. The intelligent anti-jamming system for a fully computerized three-arm rock drilling rig as described in claim 3, characterized in that, The formation of a composite state feature vector includes: in, For the first The drilling arm was at a constant time The composite state feature vector, For the first Each geological type at time Adaptive coding weights, As the first weight, For the first The drilling arm was at a constant time The pressure to advance For the first The drilling arm was at a constant time The sliding average of propulsion pressure, For the first The slip standard deviation of the propulsion pressure of the No. 1 drill arm As the second weight, For the first The drilling arm was at a constant time The rate of change of propulsion pressure, To push forward the design upper limit of the pressure change rate, As the third weight, For the first The drilling arm was at a constant time The speed of advancement To push the speed design upper limit, As the fourth weight, For the first The drilling arm was at a constant time rotational speed, An upper limit is designed for the rotational speed. As the fifth weight, For the first The drilling arm was at a constant time The longitudinal acceleration of the drill rod, This is the acceleration due to gravity.

5. The intelligent anti-jamming system for a fully computerized three-arm rock drilling rig as described in claim 4, characterized in that, Calculate the first Each geological type at time Adaptive coding weights include: in, For the first Sensing coding sensitivity coefficients corresponding to each geological condition.

6. The intelligent anti-jamming system for a fully computerized three-arm rock drilling rig as described in claim 1, characterized in that, Time-varying multidimensional reward signals include: at time... Drilling efficiency bonus At any moment Energy efficiency bonus At any moment Parameter smoothness reward At any moment Smooth slag discharge reward and at the moment Total reward ; Calculate at time Total reward : in, For at any time The first reward weight, which changes dynamically with geological conditions. For at any time The second reward weight, which changes dynamically with geological conditions. For the moment The third reward weight, which changes dynamically with geological conditions. For the moment The fourth reward weight, which changes dynamically with geological conditions. This is the first benchmark scaling factor. This is the second benchmark scaling factor. This is the third benchmark scaling factor. For at any time Early warning signs of a stuck pin, For at any time Penalty for chip malfunction. For at any time Equipment overload penalty.

7. The intelligent anti-jamming system for a fully computerized three-arm rock drilling rig as described in claim 6, characterized in that, Calculate at time Reward weights that change dynamically with geological conditions include: in, For the first The baseline reward weight for each reward component. For the first Each geological type at time The posterior probability, For the first The geological modulation sensitivity coefficient of the reward component, For the first The reward amount is in the first Adjustment factor matrix elements under various geological conditions The number of geological types.

8. The intelligent anti-jamming system for a fully computerized three-arm rock drilling rig as described in claim 6, characterized in that, The instructions for generating the cooperative optimal anti-jamming strategy for the three drill arms include: in, For the first The drilling arm was at a constant time The optimal action vector, For the first A centralized evaluation network for the No. 1 drill arm. For at any time The global state, For the first The motion vector of the drill arm For the remaining two drill arms at time The action vector, The penalty coefficient for inter-arm coordination conflict. For the first The cost function of the multi-arm conflict caused by the drill arm, after iterative convergence, yields the combination of coordinated actions of the three arms. As a collaborative optimal anti-blocking strategy instruction.

9. A method for preventing the drill bit from jamming in a fully computerized three-arm rock drilling rig, characterized in that, include: Collect multi-dimensional sensor parameters, perform real-time posterior probability estimation of geological types based on multi-dimensional sensor parameters, generate posterior probabilities of geological types, and adaptively weight and fuse multi-dimensional sensor parameters using the posterior probabilities of geological types to form a composite state feature vector. Based on the time series of composite state feature vectors, the probability of drill bit jamming precursors is predicted. A time-varying multi-dimensional reward signal is constructed based on the probability of drill bit jamming precursors and the posterior probability of geological type. Under the guidance of the time-varying multi-dimensional reward signal, a collaborative optimal anti-jamming strategy instruction for the three drill arms is generated. Based on the collaborative optimal anti-jamming strategy command, the advance speed, rotation speed and impact frequency of the three drill arms are precisely adjusted, and differentiated proactive prevention actions are performed for the jamming risk corresponding to different geological types.

10. The intelligent anti-jamming method for a fully computerized three-arm rock drilling rig as described in claim 9, characterized in that, During actual operation, the actual state of the trolley is synchronized in real time. Based on the state error between the actual state of the trolley and the state vector of the digital twin environment, the adaptive synchronization gain is calculated, and the state vector of the digital twin is corrected.

Images