A stable commutating control system for a DC motor with a multi-point rolling brush structure

Abstract

This invention relates to the field of DC motor control technology, specifically to a smooth commutation control system for a DC motor with a matched multi-point rolling brush structure. It includes a data acquisition module, an observer module, a prediction module, a control decision module, and a motor drive module. The system acquires current, speed, and mechanical vibration feedback data, extracts high-frequency jumping characteristics matching the rolling frequency, calculates the contact degradation entropy, and calculates its approximation rate to the failure boundary. When the contact degradation entropy is below a dangerous threshold, a first control command is output to maintain the current torque output; when it is above or equal to the dangerous threshold, a second control command is output, combined with the approximation rate, to inject damping current, thereby reducing the speed and rebuilding mechanical contact stability, reducing the risk of continuous sparking, torque loss, and drill bit jamming.

Description

Technical Field

[0001] This invention relates to the field of DC motor control technology, specifically to a smooth commutation control system for a DC motor with a matched multi-point rolling brush structure. Background Technology

[0002] Smooth commutation control of DC motors refers to the monitoring and adjustment of the commutation state during operation of DC motors equipped with multi-point rolling brushes to maintain stable motor output. Current DC motor commutation control methods mainly include current feedback-based control, speed feedback-based control, and closed-loop regulation based on conventional electromechanical parameters. However, when using existing technologies for commutation control, on the one hand, the identification of micro-bounce, contact degradation, and mechanical vibration disturbances of multi-point rolling brushes in the commutation interval is not accurate enough. On the other hand, under strong vibration conditions such as drilling equipment in deep-earth environments, there is a lack of early judgment and targeted intervention on the trend of the contact state approaching the failure boundary, which can easily lead to increased commutation sparks, unstable torque output, and reduced operational reliability of DC motors. Summary of the Invention

[0003] The purpose of this invention is to provide a smooth commutation control system for a DC motor with a matched multi-point rolling brush structure, addressing the following technical problems: Existing DC motor commutation control methods are not accurate enough in identifying micro-jumping, contact degradation, and mechanical vibration disturbances of multi-point rolling brushes in the commutation interval, and lack early judgment and targeted intervention for the trend of contact state approaching the failure boundary under strong vibration conditions such as deep-earth drilling equipment. To address these technical problems, this invention provides a smooth commutation control system for a DC motor with a matched multi-point rolling brush structure that can accurately integrate multimodal feedback data to extract high-frequency jumping characteristics to calculate contact degradation entropy, combine the contact degradation entropy to provide early warning of the approach rate to the preset failure boundary, and dynamically inject damping current to rebuild mechanical contact stability. The objective of this invention can be achieved through the following technical solutions: A smooth commutation control system for a DC motor with a multi-point rolling brush structure, the DC motor including the multi-point rolling brush structure, the system comprising: The data acquisition module is used to acquire multimodal feedback data of the DC motor; the multimodal feedback data includes current feedback data, speed feedback data, and mechanical vibration feedback data containing the microscopic collision characteristics generated by the multi-point rolling brush structure in the commutation range of the DC motor. The observer module is used to extract high-frequency jumping features that match the rolling frequency of the brush structure based on the multimodal feedback data, and calculate the contact degradation entropy that characterizes the degree of brush contact state degradation accordingly. The prediction module is used to calculate the rate at which the contact degradation entropy approaches a preset failure boundary within a preset historical time window, wherein the preset failure boundary is greater than a preset danger threshold. The control decision module is used to output a first control command to maintain the current torque when the contact degradation entropy is lower than the preset danger threshold; and to output a second control command in combination with the approximation rate when the contact degradation entropy is higher than or equal to the preset danger threshold, thereby injecting damping current into the DC motor to reduce the speed and rebuild mechanical contact stability. The motor drive module is used to receive the first control command or the second control command to adjust the drive voltage or inject the damping current.

[0004] Optionally, the observer module includes: The feature extraction unit is used to filter the multimodal feedback data to fuse and separate the high-frequency fluctuation features from the current feedback data, the rotational speed feedback data and the mechanical vibration feedback data. The entropy calculation unit is used to input the high-frequency fluctuating features into a preset deep learning regression model and output the contact degradation entropy. The preset deep learning regression model includes a one-dimensional convolutional layer, a long short-term memory network layer, and a fully connected regression layer connected in sequence. The high-frequency fluctuating features are truncated by a preset time window to construct a two-dimensional tensor with temporal and feature channel dimensions, and then input into the one-dimensional convolutional layer. The fully connected regression layer maps the tensor to a dimensionless scalar in the range of 0 to 1 as the contact degradation entropy.

[0005] Optionally, the prediction module is specifically used for: Obtain the contact degradation entropy within a preset historical time window; Calculate the gradient of the change in the contact degradation entropy within the historical time window; The difference between the preset failure boundary and the current contact degradation entropy is calculated as the residual margin. When the contact degradation entropy shows an upward trend and the remaining margin is greater than zero, the ratio of the change gradient to the remaining margin is calculated as the approximation rate.

[0006] Optionally, the first control command is a torque sustaining command generated based on a sliding mode control algorithm, using the deviation between the speed feedback data and the preset target speed as input. Specifically, the sliding mode control algorithm constructs a linear sliding surface using the deviation between the speed feedback data and the preset target speed and its derivative with respect to time as state variables, and calculates the sliding mode control law using an exponential reaching law combined with a sign function. Finally, the corresponding drive voltage adjustment is calculated through the output of the sliding mode control law as the torque sustaining command.

[0007] Optionally, when the control decision module outputs the second control command, it specifically executes: The target damping frequency is determined by multiplying the approximation rate by a preset frequency mapping coefficient. The waveform parameters of the damping current are generated based on the target damping frequency; The waveform parameters are encapsulated into the second control command and output; wherein the damping current is a high-frequency pulsating ripple current superimposed on the DC main drive current; the waveform parameters specifically include pulsation frequency, pulsation amplitude and pulse duty cycle, wherein the pulsation frequency is the target damping frequency, the pulsation amplitude has a preset positive correlation with the approximation rate, and the pulse duty cycle is set to a fixed constant.

[0008] Optionally, it also includes a closed-loop recovery module, the closed-loop recovery module being used for: After outputting the second control command, the contact degradation entropy updated in real time by the observer module is continuously acquired; Determine the relationship between the updated contact degradation entropy and the preset danger threshold; If the updated contact degradation entropy is lower than the preset danger threshold, a recovery command is generated; wherein, the recovery command is used to trigger the control decision module to switch to outputting the first control command; If the updated contact degradation entropy is higher than or equal to the preset danger threshold, the second control command is maintained.

[0009] Optionally, the preset failure boundary is the contact degradation entropy critical value corresponding to the continuous commutation failure state of the DC motor; wherein, the continuous commutation failure state is defined as the duration of torque loss of the DC motor being greater than or equal to a preset time threshold.

[0010] Optionally, the DC motor is used in deep-earth drilling equipment; wherein the mechanical vibration feedback data includes non-periodic mechanical resonance noise caused by geological faults.

[0011] Compared with the prior art, the present invention has the following beneficial effects: 1. This system acquires multimodal feedback data covering current, rotational speed, and mechanical vibration through a data acquisition module, and extracts high-frequency jumping characteristics matching the brush rolling frequency using an observer module to calculate the contact degradation entropy. This mechanism overcomes the technical defects of traditional control that relies on only a single electrical parameter, effectively filters out complex noise interference such as non-periodic mechanical resonance waves caused by geological faults in deep-earth environments, and achieves accurate and disturbance-resistant identification of micro-collisions and contact degradation levels of multi-point rolling brushes in the commutation interval. It directly solves the problem that existing technologies are not accurate enough in identifying micro-jumps and disturbances. 2. This system calculates the ratio of the change gradient of contact degradation entropy within a historical time window to the remaining margin through a prediction module, and then calculates the approach rate to the preset failure boundary. This design improves the traditional passive over-limit alarm into a dynamic assessment of the instability evolution trend, and can provide an accurate advance measure of the deterioration trend before the duration of torque loss of the motor exceeds the preset threshold and triggers a continuous failure state of commutation. This makes up for the deficiency of existing technology in not being able to judge the trend of approaching the failure boundary in advance, and provides the system with a time margin for early intervention. 3. This system achieves smooth mode switching through the control decision module: within the safe range, the first control command generated based on the sliding mode control algorithm maintains the torque, ensuring that the motor accurately follows the target speed under load fluctuations; when the contact degradation entropy reaches the dangerous threshold, a high-frequency pulsating ripple current containing a specific pulsation frequency and amplitude is generated by combining the approximation rate as a damping current injected into the motor; this targeted second control command can accurately match the brush bounce rhythm, actively reduce the speed and rebuild the mechanical contact stability, effectively solving the problems of increased commutation sparks and unstable torque output caused by the lack of targeted intervention in existing technologies; 4. This system introduces a closed-loop recovery module. After the damping current injection is performed, it continuously tracks and judges the contact degradation entropy updated in real time by the observer module. When it is determined that the contact state has improved and the updated contact degradation entropy has fallen back below the danger threshold, the system generates a recovery command in a timely manner to trigger the control decision module to switch back to the first control command. This mechanism establishes a scientific dynamic balance between risk concession and performance recovery, avoiding the efficiency decay of the motor caused by over-protection or long-term restricted speed reduction, and improving the overall operating efficiency and reliability of the DC motor under strong vibration extreme conditions. Attached Figure Description

[0012] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0014] like Figure 1 As shown, a smooth commutation control system for a DC motor with a multi-point rolling brush structure is disclosed. The DC motor includes a multi-point rolling brush structure, and the system comprises: The data acquisition module is used to acquire multimodal feedback data of the DC motor. The multimodal feedback data includes current feedback data, speed feedback data, and mechanical vibration feedback data containing the microscopic collision characteristics generated by the multi-point rolling brush structure in the commutation range of the DC motor. The observer module is used to extract high-frequency jumping features matching the rolling frequency of the brush structure based on multimodal feedback data, and calculate the contact degradation entropy that characterizes the degree of brush contact state degradation. The prediction module is used to calculate the rate at which the contact degradation entropy approaches a preset failure boundary within a preset historical time window, where the preset failure boundary is greater than a preset danger threshold. The control decision module is used to output a first control command to maintain the current torque when the contact degradation entropy is lower than a preset danger threshold; when the contact degradation entropy is higher than or equal to the preset danger threshold, it outputs a second control command in combination with the approximation rate to inject damping current into the DC motor to reduce the speed and rebuild mechanical contact stability. The motor drive module is used to receive a first control command or a second control command to adjust the drive voltage or inject damping current.

[0015] This embodiment provides a smooth commutation control mechanism for a DC motor with a multi-point rolling brush structure. Specifically, this embodiment places the system in the main drive link of a deep-earth drilling equipment, with the DC motor responsible for driving the drill bit to continuously break rock. Due to the step changes in rock hardness, the existence of temperature gradients downhole, and the mechanical vibration amplitude exceeding the preset normal threshold during drilling, although the multi-point rolling brush can reduce traditional sliding friction, it is prone to micro-bouncing, sudden changes in contact resistance, and instantaneous sparks in the commutation interval. Therefore, the system no longer uses the deviation of rotational speed or current from the target value as the sole control basis, but introduces continuous observation of the mechanical contact state of the brush, and incorporates whether to continue maintaining the current high torque output and whether to actively yield to rebuild contact stability into the control process. To avoid ambiguity in names, the drive module mentioned below refers to the motor drive module in the aforementioned embodiments. The dangerous threshold and failure boundary also have fixed meanings: the former is used to trigger the control boundary of whether to switch from the first control command to the second control command, and the latter is used for the prediction module to evaluate the remaining safety margin when the contact degradation entropy approaches the upper limit of the disaster. The two are not the same threshold and always satisfy the condition that the failure boundary is greater than the dangerous threshold. The data acquisition module may include a Hall current sensor, an encoder or tachogenerator, and a miniature vibration sensor arranged near the brush holder or end cap; current feedback data reflects the continuity of the armature circuit during commutation; speed feedback data reflects whether the motor output can be maintained after the drill bit load changes; mechanical vibration feedback data directly carries the high-frequency impact information of the contact surface between the brush body and the commutator; for multi-point rolling brushes, the core factor inducing commutation failure is not only the increase in average current, but also the short-term discrete collisions that occur when the rolling contact point is locally unstable. These collisions occur before the loss of torque and are manifested by both vibration and current ripple. Therefore, the observer module extracts high-frequency jumping features that match the rolling frequency from the above multimodal feedback and classifies them into contact degradation entropy; this contact degradation entropy is used to characterize whether the current brush contact state is in stable rolling, slightly unstable, or has entered a high-risk stage that is about to induce a continuous electric arc. To facilitate understanding, a simplified example can be used to illustrate the data flow. Assume a continuous sampling period is divided into three segments, denoted as S1, S2, and S3. S1 corresponds to the normal rock cutting stage, S2 corresponds to the drill bit entering a hard interlayer stage, and S3 corresponds to the downhole resonance enhancement stage. In S1, the current waveform has no abrupt spikes, the rotational speed fluctuation amplitude is below the preset stable lower limit, and the vibration contains only low-amplitude rolling components. In S2, the current exhibits intermittent spikes, and high-frequency components adjacent to the rolling frequency begin to appear in the vibration. In S3, current spikes, slight rotational speed fluctuations, and high-frequency impact vibrations simultaneously intensify. The observer module does not observe a single signal in isolation, but rather aligns these three segments in time and performs correlation analysis. When high-frequency fluctuations simultaneously exhibit corresponding characteristic components in the current, rotational speed, and vibration channels, it can be determined that they are more likely to originate from brush contact deterioration, rather than simply from casing vibration or measurement noise. After receiving continuously updated contact degradation entropy, the prediction module assesses its tendency to approach the failure boundary. The focus of this prediction module is to assess the accumulation rate of contact deterioration: that is, to determine whether it is slow accumulation or rapid deterioration in a short period of time. If the contact degradation entropy has increased but the change is gradual, it indicates that the brush body still has a certain mechanical buffering capacity. If the contact degradation entropy surges rapidly within a preset reference time window, it indicates that the brush body may lose stable conduction conditions due to micro-arc ablation, bounce amplification, or resonant coupling. Based on this, the system outputs two types of control strategies: one is the first control command to continue maintaining the current torque output; the other is to switch to the second control command when necessary, actively reducing the speed by injecting damping current, suppressing repeated bouncing of the brush body, and prompting the rolling contact to re-fit. In abnormal situations, when the current sensor experiences short-term sampling loss due to strong electromagnetic interference, or when the vibration sensor's data reliability decreases due to dust accumulation or impact saturation at the installation location, the observer module can reduce the weight of the abnormal channel and maintain a conservative estimate based only on the remaining two channels. If two or more channels are abnormal at the same time, the system does not directly judge it as an escalation of contact deterioration, but instead enters a steady-state protection mode, freezing the rate of change of risk level for a short time to avoid unnecessary damping injection due to sensor distortion. Furthermore, if the contact degradation entropy is close to the danger threshold but the approach rate is low, the system can prioritize maintaining the current drive voltage and increasing the sampling refresh rate; if the contact degradation entropy has exceeded the danger threshold and the approach rate continues to increase, the system directly enters damping control to prevent the continuous sparking caused by maintaining torque with a large current. Furthermore, the above-mentioned process of maintaining the current drive voltage and increasing the sampling refresh rate is still a conservative execution method under the framework of the first control command. Its purpose is to improve the reliability of state identification when the triggering condition of the second control command is not met, rather than replacing the damping injection logic in the second control command. When deep-earth drilling operations enter a section below 1500 meters, the drill bit cuts into a region interspersed with fault fragments from relatively uniform rock strata. Although the main drive motor can still maintain the predetermined torque, the vibration sensor near the end cap has detected high-frequency collision components with amplitudes greater than the preset disturbance threshold. At this time, a spike corresponding to the reversal interval appears in the current waveform, and the speed feedback shows slight but continuous jitter. After the system integrates the above three types of information, it determines that the multi-point rolling brush has further developed from the critical micro-jump to the contact deterioration acceleration zone. If maintaining the maximum torque is still the only goal, the local ablation of the brush body will be further aggravated. Therefore, the control decision module outputs a second control command, and the motor drive module injects damping current without completely cutting off the power, so that the motor speed is reduced for a short time and the brush contact surface regains stable clamping conditions. The purpose of this step is to expand the traditional commutation control centered on output performance into a risk hedging control that takes into account the stability of mechanical contact, so as to realize the early identification and timely intervention of the micro instability state of multi-point rolling brushes and reduce the risk of drill bit jamming caused by continuous sparks and torque loss. In this embodiment, the observer module includes: The feature extraction unit is used to filter the multimodal feedback data to extract high-frequency fluctuation features from the current feedback data, speed feedback data and mechanical vibration feedback data. The entropy calculation unit is used to input high-frequency fluctuating features into a preset deep learning regression model and output contact degradation entropy. The preset deep learning regression model includes a one-dimensional convolutional layer, a long short-term memory network layer, and a fully connected regression layer connected in sequence. The high-frequency fluctuating features are truncated by a preset time window to construct a two-dimensional tensor with temporal and feature channel dimensions, and then input into the one-dimensional convolutional layer. The fully connected regression layer maps the tensor to a dimensionless scalar in the range of 0 to 1 as the contact degradation entropy.

[0016] This embodiment provides an observer refinement mechanism. Specifically, in the aforementioned drilling mainline scenario, directly judging whether the brush is unstable based solely on raw current, rotation speed, and vibration data is easily affected by complex mechanical noise downhole. For example, overall torsional vibration of the drill string, meshing impact of the deceleration mechanism, and random oscillations caused by formation fracturing may all manifest as high-frequency disturbances on the sensor. If feature extraction and separation are not performed first, the system may mistakenly identify impacts unrelated to the brush as contact deterioration, thereby prematurely reducing power output and affecting drilling continuity. The following is a detailed description: The feature extraction unit performs synchronization and filtering on the data from each channel. The current channel mainly retains the commutation ripple and transient spike features, the speed channel mainly focuses on the minor fluctuations that do not conform to the load inertia in a short period of time, and the vibration channel extracts the impact components near the rolling contact frequency band of the brush body. The so-called high-frequency jumping feature that matches the rolling frequency has the following physical meaning: when the brush body rolls stably, the effect of the contact point on the commutator is continuous and predictable; when bouncing occurs, it will form discrete impacts superimposed on the periodic basis. These impacts are manifested as fine pulses in vibration, as spikes or discontinuous ripples in current, and as small but dense instantaneous disturbances in speed. After the feature extraction unit aligns the three in time, it only retains the components that have mutual corroboration relationships, thereby separating out the high-frequency jumping feature that truly reflects the deterioration of the contact. Furthermore, in order to make the rolling frequency have a stable and unique engineering meaning, it can be understood as the reference frequency corresponding to the commutation-related contact rhythm when the rolling contact point of the multi-point rolling brush structure sweeps across the current speed. This reference frequency is preferably determined by real-time speed feedback data combined with the number of brush rolling contact points, commutator segment pitch or pre-calibrated structural rhythm parameters. Specifically, the feature extraction unit does not retain all high-frequency components across the entire frequency band. Instead, it first establishes candidate frequency bands around the reference rolling frequency and its adjacent tolerance frequency bands, and then filters out the impact components that appear synchronously with current spikes and speed fluctuations from the candidate frequency bands. This avoids mistaking irrelevant broadband noise as high-frequency fluctuation features. A simplified example can be used to illustrate this. Suppose that within a certain time period, the current characteristic sequence is denoted as A = [smooth, spike, spike], the rotational speed characteristic sequence is denoted as B = [stable, slight jitter, slight jitter], and the vibration characteristic sequence is denoted as C = [low impact, high impact, high impact]. If the three sequences synchronously exhibit a combination of spike-slight jitter-high impact within the same sampling segment, then that segment is more likely to correspond to brush micro-jumping. If only the vibration channel exhibits high impact, while the current and rotational speed show no corresponding changes, then that segment is more likely to be external structural vibration rather than contact deterioration. Based on this, the feature extraction unit outputs the fused high-frequency jitter features for use by the subsequent entropy value calculation unit. Furthermore, if the real-time rotational speed changes, the aforementioned candidate frequency bands are updated synchronously to ensure that the rolling frequency always matches the rolling contact rhythm under the current operating conditions, rather than a fixed single frequency band. The entropy calculation unit can use a preset deep learning regression model, which is trained through historical test samples. The samples include normal rolling contact, slight bouncing, obvious ablation, and edge states close to continuous arcing. The labeling rules for historical test samples are as follows: the labels of normal rolling contact samples are calibrated close to 0, the labels of commutation failure samples close to continuous arcing are calibrated close to 1, and samples in the intermediate transition state are calibrated by linear or nonlinear interpolation between 0 and 1 based on their rotational speed fluctuation amplitude and current peak density. The contact degradation entropy output by the model is not a simple single-point signal amplitude, but a comprehensive characterization of contact uncertainty, instability propagation trend, and local ablation risk. In essence, the higher the contact degradation entropy, the more difficult it is for the system to ensure safe commutation by maintaining torque in a conventional way, because each further increase in driving force may amplify contact instability in the opposite direction. In abnormal situations, when the deep learning regression model encounters uncovered extreme conditions, such as changes in sensor installation location, batch changes in brush material, or abnormally low downhole temperature causing sudden changes in vibration propagation characteristics, it can simultaneously output a model confidence assessment flag. If the confidence is higher than the predetermined lower limit, the model result is used. If the confidence is lower than the predetermined lower limit, the entropy calculation unit switches to conservative estimation mode, generating an alternative risk level based on the envelope strength and continuity of high-frequency fluctuation features, avoiding the model's uncertain output from directly interfering with control decisions. If there are obvious saturation, discontinuity, or clock asynchrony issues in the original data, data correction is performed first. If correction fails, the update of contact degradation entropy is paused, only the stable value of the previous cycle is retained, and a data maintenance alarm is triggered. Furthermore, the aforementioned model confidence assessment label serves only as an auxiliary criterion for whether to use the deep learning regression model results, and does not change the module division relationship where high-frequency fluctuation features are output by the feature extraction unit and contact degradation entropy is output by the entropy value calculation unit. After the drill bit cuts into the fault fracture zone, the overall vibration of the downhole equipment is significantly enhanced. If fusion and separation are not performed, the system may misidentify the broadband vibration caused by drill string collision as a brush failure. At this stage, the feature extraction unit finds that although the impact of the vibration channel increases, only some of the impacts occur synchronously with current spikes and speed fluctuations, so only these synchronous segments are retained. The entropy calculation unit identifies, based on historical training samples, that these features are closer to the state of contact surface bouncing accompanied by mild ablation than the shell resonance itself, and outputs the corresponding contact degradation entropy. The purpose of this step is to improve the physical relevance of the observation results by separating first and then calculating, so as to achieve reliable identification of the real contact state of multi-point rolling brushes and reduce misjudgments caused by external mechanical noise. In this embodiment, the prediction module is specifically used for: Obtain the contact degradation entropy within a preset historical time window; Calculate the gradient of the change in contact degradation entropy within a historical time window; The difference between the preset failure boundary and the current contact degradation entropy is calculated as the residual margin. When the contact degradation entropy shows an upward trend and the remaining margin is greater than zero, the ratio of the change gradient to the remaining margin is calculated as the approximation rate.

[0017] This embodiment provides a failure approach prediction mechanism. Specifically, in the aforementioned system, there are limitations to obtaining only the current instantaneous value of the contact degradation entropy. In drilling scenarios, the key to inducing system commutation instability often lies not in the absolute value of the current degradation state, but in the rate at which the degradation state approaches the preset failure boundary. If the control system only makes on / off judgments based on the current instantaneous value, it may miss the best intervention time before the contact deterioration truly goes out of control. The following is a detailed description: The prediction module continuously acquires the contact degradation entropy within the historical time window to observe its evolution trend; the changing gradient, which represents the physical meaning of the rate of contact state deterioration: if the gradient is lower than the preset gradient threshold, it indicates that although there is local bounce in the brush body, the contact surface can still maintain basic stability; if the gradient exceeds the preset gradient threshold, it indicates that microscopic ablation, thermal accumulation and mechanical rebound are forming a mutually reinforcing relationship, and the subsequent degradation rate may further increase. The so-called residual margin corresponds to how much available buffer is left between the current contact state and the actual danger boundary; the approximation rate obtained by combining the two can be understood as how much reaction time the system has left before the critical state of commutation instability; to avoid misunderstandings of this ratio in implementation, the processing order of the prediction module can be further clarified as follows: first, read the contact degradation entropy sequence in chronological order within the historical time window, then form a change gradient based on the overall upward, flat, or downward trend of the sequence, and obtain the residual margin by subtracting the current contact degradation entropy from the failure boundary; only when the residual margin is still a positive safety interval, the ratio of the change gradient to the residual margin is used as the approximation rate; Therefore, a high approach rate indicates both a rapid increase in contact degradation entropy and proximity to the failure boundary; a low approach rate indicates slower deterioration or sufficient remaining safety margin. A simplified example can illustrate this: assuming four consecutive contact degradation entropy states are obtained within the historical window, corresponding to E1, E2, E3, and E4 respectively, where the rate of change from E1 to E2 is below the first preset threshold, the rate of change from E2 to E3 exceeds the second preset threshold, and the rate of change from E3 to E4 increases rapidly; if the current state E4 is only a safety interval below the preset lower limit from the failure boundary, then even if E4 has not yet crossed the boundary, its approach rate should be considered high; conversely, if E4 is greater than the preset safety margin from the failure boundary, and the changes in the previous few cycles are below the preset rate of change threshold, then the system can be considered to temporarily have room to maintain power output; the control timing is determined by the changing trend and the remaining margin, rather than relying on a single numerical result. Furthermore, to avoid ambiguity in engineering implementation, the gradient of change can be understood as a comprehensive measure of the direction and intensity of change in contact degradation entropy within a historical time window. When the overall contact degradation entropy increases, the gradient of change takes a positive deterioration trend; when the overall contact degradation entropy remains flat or decreases, it is considered as a low approximation or de-approximation state. The remaining margin is preferably understood as a positive safety interval between the current contact degradation entropy and the failure boundary, and it only participates in the ratio calculation when the current state has not yet reached the failure boundary. Therefore, the approach rate not only reflects the rate of deterioration, but also inherently includes the constraint of how much buffer is left before failure. Furthermore, if multiple local fluctuation peaks exist simultaneously within a historical time window, it is preferable to determine the change gradient based primarily on the overall trend and secondarily on individual peaks, thereby avoiding the direct amplification of a single incidental shock into a continuous approach. In engineering implementation, the historical time window can cover multiple reversal cycles to avoid misjudging a single incidental shock as a precursor to disaster. If the window time is less than the first preset lower time limit, a single external shock may lead to an overly aggressive trend judgment. If the window time is greater than the second preset upper time limit, it will mask the accelerating contact deterioration. Therefore, the time window is generally matched with several complete commutation sequences under the rated speed of the motor, so that the prediction module can both perceive the real evolution of the brush body state and retain sufficient real-time performance. In abnormal situations, if there are a large number of missing values, misaligned data stamps or insufficient confidence of the model output within the historical time window, the prediction module will not directly form an approximation rate, but will mark the current state as having a high uncertainty metric. In this case, a more conservative control strategy can be adopted: if the current contact degradation entropy is close to the danger threshold, then enter the finite damping state first; if the current is still far below the danger threshold, then maintain the original control and prioritize the restoration of sampling integrity; the finite damping state is a restricted mode for executing the second control command. In this state, the amplitude of the injected damping current is forcibly limited to a preset percentage of the rated operating current of the DC motor, and its target damping frequency is fixed at the lower limit of the preset initial target damping frequency to avoid excessive torque impact on the motor due to data uncertainty. Furthermore, if a one-time spike occurs within the historical window followed by a rapid decline, the prediction module can identify it as an impact event rather than a continuous deterioration, thus avoiding repeated switching of control modes due to short-term mechanical impacts. Furthermore, if the current contact degradation entropy has reached or exceeded the failure boundary, causing the remaining margin to approach zero or no longer have the significance of a normal safety interval, the prediction module will no longer amplify the ratio. Instead, it will directly process the approximation rate according to the highest risk level and output an out-of-bounds flag, so that the control decision module can directly execute the most conservative suppression strategy and avoid abnormal amplification or distorted judgment caused by an excessively small denominator. If the current contact degradation entropy has not exceeded the boundary but the remaining margin is less than the preset minimum safety margin, the ratio calculation result can also be constrained by an upper limit before outputting it to prevent abnormal spike results that are not of engineering feasibility due to an excessively small remaining margin. During the continuous advancement of the drilling equipment, the contact state of the brushes initially deteriorated slowly. However, as the formation entered the vicinity of the fractured fault, the contact deterioration entropy rapidly increased within several consecutive commutation cycles, and the safety margin between it and the failure boundary continuously decreased. Although the drive motor could still maintain rotation and the drill bit had not completely stalled, the prediction module had already determined that the system was approaching the dangerous range before the critical commutation instability. Therefore, the high approximation rate result was sent to the control decision module to gain advance time for subsequent active deceleration and damping injection. The purpose of this step is to expand the judgment from whether it is already dangerous to whether it is rapidly becoming dangerous, so as to achieve early warning of commutation disaster, rather than post-disaster remediation. In this embodiment, the first control command is a torque sustaining command generated based on the sliding mode control algorithm, with the deviation between the speed feedback data and the preset target speed as input. Specifically, the sliding mode control algorithm constructs a linear sliding surface using the deviation between the speed feedback data and the preset target speed and its derivative with respect to time as state variables, and calculates the sliding mode control law using an exponential reaching law combined with a sign function. Finally, the corresponding drive voltage adjustment amount is calculated as the torque sustaining command through the output of the control law.

[0018] This embodiment provides a first control command generation mechanism; specifically, when drilling is in normal condition or when the contact degradation entropy is within a safe range, the system does not need to actively degrade its operation, but should maintain the stable torque and speed required by the drilling equipment as much as possible; since the load fluctuation in deep drilling is significant, ordinary linear control is prone to slow response or overshoot when encountering sudden changes in rock hardness, so the first control command preferably adopts the sliding mode control method to generate torque maintenance command. The engineering significance of sliding mode control lies in enhancing the adaptability to parameter uncertainties and external load shocks. For drilling motors, the goal is not to keep the speed absolutely constant at every moment, but to enable the motor to quickly follow the set drilling conditions without causing commutation instability. When the speed feedback is lower than the target speed, the controller appropriately increases the drive action to compensate for the drill bit load. When the speed feedback is higher than the target speed, the controller converges the output accordingly to avoid a sudden increase in no-load speed. Because sliding mode control has a strong tolerance for system parameter drift, it can maintain a strong torque tracking capability even if changes in downhole temperature cause deviations in resistance, inductance, or mechanical resistance. A simplified example can be used to illustrate this; assuming the target speed corresponds to state R0, real-time feedback shows R1, R2, and R3 in sequence, where R1 is lower than R0 and the difference is within the first preset range, R2 is lower than R0 and the difference exceeds the first preset range, and R3 rises to close to R0; the first control command performs small compensation in stage R1, strong compensation in stage R2, and gradually converges in stage R3, so that the torque output returns to stability; The control logic aims to limit the system to prioritize drilling continuity and maintain stable output through disturbance rejection control when the contact state is still safe. However, this mechanism alone has its limitations. If the brush contact is close to instability, simply pursuing rapid compensation for speed deviation may increase contact degradation through higher driving stress. Therefore, the first control command is only applicable when the contact degradation entropy is below the danger threshold. Once the observation results indicate that the brush contact stability is insufficient, the system will no longer insist on rapidly reducing the speed error, but should switch to the second control strategy that prioritizes maintaining commutation stability. As a backup protection mechanism or under extreme conditions, if there is a significant jump in speed feedback but the current and vibration channels do not show a corresponding load change, it may be due to instantaneous pulse loss of the speed sensor or encoder interference. In this case, the first control command does not directly amplify the compensation amount, but performs a slow-release processing on the speed deviation. If the speed sensor is continuously abnormal, the system can form an alternative speed based on the armature current and back electromotive force estimation values ​​to maintain torque control for a short time until the speed measurement link is restored. If the contact degradation entropy crosses the danger threshold during the control process, the strengthening trend of the first control command is immediately stopped to avoid continuing to push torque under erroneous conditions. When the drill bit is advancing in a uniform and dense rock layer, the motor load is large but changes relatively smoothly, and the contact degradation entropy is continuously kept within a safe range. At this time, due to the occasional increase in local cutting resistance of the drill bit, the real-time speed is briefly lower than the target value. The first control command increases the drive voltage through sliding mode control, so that the torque is quickly made up and the drill bit continues to feed stably. Throughout the process, the system did not inject damping current or actively reduce speed, because the brush contact state still had the mechanical stability to withstand conventional load adjustments. The purpose of this step was to provide high disturbance-resistant torque maintenance capability when the contact state was controllable, thereby achieving continuous output and rapid following under drilling conditions, and providing a basic control layer for normal system operation. In this embodiment, when the control decision module outputs the second control command, it specifically executes: The target damping frequency is determined by multiplying the approximation rate by a preset frequency mapping coefficient. The waveform parameters of the damping current are generated based on the target damping frequency; The waveform parameters are encapsulated into a second control command and output; the damping current is represented as a high-frequency pulsating ripple current superimposed on the DC main drive current; the waveform parameters specifically include the pulsation frequency, pulsation amplitude and pulse duty cycle, where the pulsation frequency is the target damping frequency, the pulsation amplitude and the approximation rate have a preset positive correlation proportional mapping relationship, and the pulse duty cycle is set to a fixed constant.

[0019] This embodiment provides a second control command refinement mechanism; specifically, based on the above, simply deciding to inject damping current is not enough to stabilize the multi-point rolling brush, because the brush body bounce rhythm corresponding to different deterioration speeds is not the same; if the damping current frequency is not properly selected, it will not be able to achieve the suppression effect, but will easily couple with mechanical jump, aggravating commutation sparks. Therefore, this embodiment further introduces a mapping relationship between the approximation rate and the target damping frequency, so that the damping effect can be matched to the current contact instability rhythm. As described below, the higher the approximation rate, the faster the contact deterioration evolves, the more dense the micro-jumps of the brush body are, and the easier it is for local heat accumulation and discrete collisions to superimpose in a short period of time. Based on this, the control decision module determines the target damping frequency through a preset mapping coefficient. Its physical essence is to make the injected damping current align with the bounce interval of the brush body that needs to be suppressed in terms of time rhythm. Based on the target damping frequency, the waveform parameters of the damping current are generated, such as waveform period, duty cycle, rise and fall slope, so that the current output by the drive module no longer only provides torque, but also has mechanical stabilization functions such as adsorption contact, reduction of rebound, and smooth commutation. Furthermore, to make this mapping relationship more executable, the preset frequency mapping coefficient can be understood as a scale obtained from bench calibration. Its function is to convert the approximation rate into the target damping frequency range that the drive module can output. Specifically, the control decision module first forms an initial target damping frequency based on the approximation rate, and then, in combination with the driver switching capability, the motor electromagnetic time constant, and the known structural resonance forbidden zone, limits and avoids this initial result, finally obtaining the target damping frequency that can be actually executed. This maintains the main logic determined by multiplication in the embodiment, and avoids the target frequency falling into the invalid region, the too low region, or the dangerous resonance region due to extreme approximation rates. A simplified example can be used to illustrate this; assume that the system identifies two different risk states: state T1 is a medium approach rate and state T2 is a high approach rate; for T1, the control decision module can generate a damping waveform with a corresponding first preset amplitude and frequency, which mainly reduces mild bounce; for T2, a damping waveform with a higher frequency and larger amplitude is generated to suppress continuous discrete impacts on the brush body in a shorter time. For example, waveform parameters can be divided into a first parameter P1 representing frequency, a second parameter P2 representing amplitude envelope, and a third parameter P3 representing pulse duty cycle. Under different approximation rates, these three sets of parameters are combined according to a preset mapping rule to form different second control commands. The preferred generation order here is: first, the first parameter P1 is determined by the approximation rate and frequency mapping coefficient, and then the second parameter P2 and the third parameter P3 are selected to match the suppression rhythm corresponding to the first parameter P1, so that the amplitude and duration serve the frequency target, rather than changing arbitrarily away from the frequency target. In engineering, the frequency mapping coefficient can be obtained through bench tests or downhole operation data calibration. The calibration is not based on simple theoretical derivation, but on the actual mechanical response of the brush body, commutator, spring preload and installation structure under different instability stages. For example, a certain frequency band is most effective in suppressing mild bounce, while another frequency band is more suitable for suppressing bounce before continuous sparking. The control system selects the damping current that matches the dynamics of contact degradation accordingly. Under extreme operating conditions, if the approach rate increases abnormally but the current motor temperature rise is already close to the heat dissipation limit, the second control command should also include thermal protection constraints when generating the command to limit the damping current amplitude or shorten the continuous injection time to prevent overheating of the windings due to suppression of contact deterioration. If the mapped target damping frequency falls into the known structural resonance forbidden zone of the equipment, the control decision module should automatically bias to the adjacent safe frequency band. If the sensor signal indicates that the collision vibration has not weakened but rather strengthened after damping injection, it is determined that the current waveform parameters are mismatched, and a more conservative waveform combination should be selected in a timely manner. Furthermore, if the approach rate is very low but the contact degradation entropy has exceeded the danger threshold, the second control command can prioritize the use of a low-frequency, low-amplitude exploratory damping waveform to observe the subsequent decline of the contact degradation entropy while ensuring that the opportunity for re-fitting is not lost. If the contact degradation entropy continues not to decrease, the damping strength can be gradually increased to avoid using an excessively strong waveform in the initial stage, which could cause additional thermal shock or mechanical disturbance. When the drill bit encounters a fault-crossing zone, the observer and prediction module jointly determine that the contact degradation entropy has exceeded the danger threshold and the approach rate is greater than the preset historical average threshold. Based on this, the control decision module no longer simply lowers the voltage, but instead generates a damping current waveform that matches the current brush bounce rhythm by combining the mapping relationship. After the drive module executes this, the motor speed decreases in a controlled manner, but the high-frequency collisions in the commutation zone begin to weaken, and the current spikes change from dense to sparse, indicating that the brush contact surface is returning to a controllable state from the edge of instability. The purpose of this step is to improve the damping control from whether to apply it to how to apply it, so as to achieve targeted suppression of the brush bounce rhythm at different degradation rates and improve the effectiveness of rebuilding mechanical contact stability. In this embodiment, a closed-loop recovery module is also included, which is used for: After outputting the second control command, the contact degradation entropy updated in real time by the observer module is continuously acquired; Determine the relationship between the updated contact degradation entropy and the preset danger threshold; If the updated contact degradation entropy is lower than the preset danger threshold, a recovery command is generated; the recovery command is used to trigger the control decision module to switch to outputting the first control command. If the updated contact degradation entropy is higher than or equal to the preset danger threshold, continue to output the second control command.

[0020] This embodiment provides a closed-loop recovery mechanism. Specifically, in the aforementioned scheme, the second control command is an active degradation control strategy, the purpose of which is to rebuild the stability of the system's mechanical contact by temporarily limiting the power output. However, if the system continues to maintain damped deceleration for a long time after the contact state has improved, it will cause a decrease in drilling efficiency and unnecessary energy consumption. Conversely, if the recovery is too early, it may cause the brush body, which has just stabilized, to destroy the contact state. Therefore, a closed-loop recovery module is needed to continuously track the contact state after damping control. Specifically, after the second control command takes effect, the closed-loop recovery module continues to obtain real-time updated contact degradation entropy from the observer. When the updated value drops below the danger threshold, it indicates that the mechanical contact between the brush and the commutator has been initially restored, and the risks of discrete collisions and micro-arcs have been significantly reduced. At this point, a recovery command can be generated, causing the control decision module to switch back to the first control command and gradually return to the normal torque maintenance mode. If the updated value is still higher than or equal to the danger threshold, it indicates that the brush has not completely escaped the instability zone, and damping control still needs to be maintained. A simplified example can be used to illustrate this; assuming that after the damping control is started, three consecutive observation segments yield states Q1, Q2, and Q3, where Q1 is still above the danger threshold, Q2 is close to the threshold, and Q3 is stably below the threshold; the closed-loop recovery module continues to maintain the second control command during the Q1 and Q2 stages, without rushing to switch; only when Q3 shows that it has fallen back to the safe zone and remains continuously stable, does it output the recovery command; the technical principle is that brush contact recovery usually has inertia, and a brief fall in a single sampling segment does not represent true stability, while continuous fall is more reliable in engineering; Furthermore, the recovery process is preferably a gradual switching rather than an instantaneous jump. Specifically, after the recovery command is triggered, the system can first reduce the damping current, and then slowly increase the drive voltage and torque command, so that the brush body can smoothly bear the load on the recovered contact surface, instead of immediately returning to the extreme operating condition. This can avoid inducing secondary contact failure. Furthermore, in order to ensure consistency with the switching logic of the embodiment, the triggering condition of the recovery command is still based on the premise that the updated contact degradation entropy is lower than the danger threshold. The continuous stability criterion, hysteresis interval, and gradual damping removal are only used as engineering execution refinements to improve switching reliability, and do not change the main judgment basis for switching from the second control command to the first control command. Under extreme operating conditions, if the contact degradation entropy fluctuates around the danger threshold, the closed-loop recovery module can set a hysteresis range or a continuous stability time criterion to prevent the control mode from frequently switching between the two commands. If the contact degradation entropy does not decrease significantly after damping control has been in place for a period of time, it indicates that the current damping strategy may not be sufficient to restore mechanical stability. The system can further reduce the upper speed limit or link the host computer to issue a maintenance prompt. If a rapid increase in the approach rate is detected again during the recovery process, the recovery process is immediately canceled and the second control command is re-executed. If the drive side detects that the speed rises too quickly, the current spikes become dense again, or the high-frequency components of vibration rise a second time after the recovery is triggered, the closed-loop recovery module can determine that the recovery is incomplete, immediately cancel the recovery command, and return to the damping hold state. After the fault disturbance ended, the downhole mechanical resonance gradually weakened. The closed-loop recovery module detected a decrease in the number of current spikes, a reduction in high-frequency vibration impacts, and a continuous decline in contact degradation entropy. The system did not immediately restore high torque upon the first decline, but instead maintained short-term damping until several consecutive reversing cycles showed safety before issuing a recovery command to allow the motor to re-enter torque maintenance mode. After recovery, the drill bit feed rate increased, but the reversing state remained stable. The purpose of this step is to establish a closed-loop balance between risk concession and performance recovery, thereby achieving adaptive operation of mode switching and avoiding secondary risks caused by over-protection or premature recovery. In this embodiment, the preset failure boundary is the contact degradation entropy critical value corresponding to the continuous failure state of the DC motor commutation; wherein, the continuous failure state of the commutation is defined as the duration of torque loss of the DC motor being greater than or equal to a preset time threshold.

[0021] This embodiment provides a failure boundary definition mechanism. Specifically, in high-risk drilling scenarios, if the danger threshold and failure boundary do not have clear physical meaning, the control system will find it difficult to be neither oversensitive nor slow to react. Therefore, this embodiment directly anchors the failure boundary to the continuous failure state of reversing and uses the duration of torque loss exceeding a preset time threshold as the catastrophic criterion, so that the upper limit reference of the contact degradation entropy has clear engineering consequences. The so-called continuous commutation failure state is not a single spark, short-term fluctuation, or instantaneous stall, but rather a continuous loss of effective commutation capability over a period of time due to continuous bouncing of the contact surface and arc propagation. At this time, although the DC motor still has armature current, the torque cannot be stably transmitted to the drill bit, and the drilling link enters an effective torque loss state. If this torque loss continues for more than a preset time threshold, the drill bit is prone to lose its self-recovery cutting capability in deep rock formations, causing the drill to get stuck or even the entire drill string to remain downhole. Therefore, setting the contact degradation entropy critical value corresponding to this catastrophic state as the failure boundary enables the prediction module and the control decision module to work together around the actual failure consequences. A simplified example can be used to illustrate this; assuming there are three states in bench tests or historical downhole data: state F1 is short-lived sparking but torque can be recovered, state F2 is continuous jitter and significant torque decay, and state F3 is torque loss that continues to exceed a preset time threshold; then the failure boundary is not taken from F1, nor simply from the average level of F2, but from the critical range of contact degradation entropy just before entering F3; thus, the danger threshold can be set before this boundary as an early intervention point, while the failure boundary serves as the truly insurmountable control limit; when specifically determining the time threshold, the fault tolerance of the drilling process should be considered; For example, in continuous rock breaking operations, instantaneous torque drops below this threshold may still be absorbed by the drill bit inertia and downhole energy storage, preventing immediate drill bit jamming; however, once the threshold is exceeded, frictional lock between the drill bit and the rock formation may be quickly established, and even if the motor is restarted, it will be difficult for the drill bit to break free on its own. Therefore, this time threshold is an engineering parameter jointly determined by the mechanical link and process tolerance, rather than an arbitrarily set mathematical constant. Under extreme conditions, if a certain type of drilling equipment changes its reduction mechanism, drill string length, or drill bit type, the original time threshold and failure boundary should be recalibrated and cannot be mechanically copied. If the field data is insufficient to accurately determine the critical value, a lower failure boundary can be taken first based on the conservative principle to prioritize the survivability of the equipment. After accumulating sufficient operating data, the boundary can be updated through calibration. If a short-term torque loss occurs but the duration does not exceed the threshold, the system should classify it as a high-risk event record rather than directly treating it as a failure, for use in subsequent model correction and maintenance assessment. During a deep well operation, the motor experienced continuous sparking and a significant drop in output torque. Although the current remained high, the drill bit failed to resume effective cutting within a preset timeframe, ultimately being classified as a reversing failure event. The system subsequently used the contact degradation entropy trajectory before and after this event to calibrate the failure boundary, enabling subsequent similar equipment to initiate risk retreat measures before approaching the critical zone, rather than passively shutting down after the drill bit was locked by the rock formation. The purpose of this step is to establish a physical boundary for the contact degradation entropy that corresponds to the actual catastrophic consequences, thereby achieving consistency between risk assessment and drilling failure mechanisms and improving the engineering reliability of control threshold settings. In this embodiment, a DC motor is used in a deep-earth drilling equipment; wherein, the mechanical vibration feedback data includes non-periodic mechanical resonance wave noise caused by geological faults.

[0022] This embodiment provides an adaptation mechanism for deep drilling conditions. Specifically, after the aforementioned system enters the downhole installation stage, the most representative complex interference is not a constant load, but non-periodic mechanical resonance noise caused by geological faults, fracture zones, and the transmission of the long drill string. This type of noise is sudden, wide-bandwidth, and time-varying, and does not present a fixed frequency like regular rotational imbalance. Therefore, it is easier to mask or distort the real contact signal of the multi-point rolling brush. The non-periodic mechanical resonance waves are transmitted step by step through the drill bit, drill rod, reduction mechanism, and motor housing, eventually affecting the brush holder structure and commutator area. The mechanism that causes commutation abnormalities is as follows: initially, it may not cause macroscopic stall, nor may it immediately lead to a significant increase in average current, but it will form high-frequency, intermittent, and variable-direction micro-impacts on the brush contact surface. Under this excitation, the multi-point rolling brush may rapidly develop from the originally acceptable critical micro-jump to continuous bouncing and local arcing. Under this condition, although the macroscopic torque output of the system is still maintainable, the actual contact reference has been destroyed. If the control system continues to treat the load fluctuations as normal and continues to enhance the drive, it may actually accelerate the contact deterioration. In response to this characteristic, the processing of mechanical vibration feedback data not only focuses on the stable frequency components, but also on the temporal density, persistence and coupling degree with the commutation interval of non-periodic impacts; specifically, the system does not simply filter out all irregular vibrations as noise, but identifies which impacts, although originating from the external geological environment, have already begun to substantially affect the brush contact. In this way, the control strategy can be truly adapted to deep drilling rather than ordinary laboratory environments. A simplified example can be used to illustrate this. Suppose that the vibration channel has three sets of irregular impact segments K1, K2 and K3 within a certain period of time. K1 only manifests as shell shaking, and the current and rotation speed do not change synchronously. K2 is irregular, but it appears synchronously with the current spike. K3 not only appears synchronously with the current spike, but is also accompanied by fine vibrations in the rotation speed. The system can regard K1 as an external disturbance background, K2 as a warning segment that may begin to affect commutation, and K3 as a dangerous segment that has interfered with the mechanical contact of the brush. It can be seen that even if the noise itself comes from the formation fault, whether it enters the control closed loop depends on whether it has left verifiable evidence of contact instability at the electromechanical coupling level. Under extreme operating conditions, if the downhole fault activity is severe, causing the vibration channel to be in a high noise background for a long time, the system can appropriately raise the confirmation threshold for a single vibration segment. The contact degradation entropy can only be increased after the corresponding anomalies of the current and speed channels are met simultaneously. If the vibration channel becomes inaccurate due to aging caused by long-term high impact, the system can automatically activate redundant sensors or construct alternative observations through current and speed characteristics according to the maintenance strategy. If the external resonance has exceeded the allowable limit of the equipment, even if the contact degradation entropy has not yet reached the failure boundary, the upper drilling rig control system can be linked to perform load reduction or drilling stoppage to prevent further expansion of structural damage. During a deep drilling operation, the drill bit entered a complex fault intersection area, and obvious non-periodic resonance waves began to appear in the mechanical vibration feedback. Initially, this type of resonance did not cause the motor to stall, and drilling continued in a macroscopic kinematic state. However, the system found that some of the impacts had formed a synchronous relationship with the current spikes and speed fluctuations in the commutation interval, and thus identified it as a dangerous factor that was eroding the mechanical contact reference of the brush. The control decision no longer simply pursues maintaining the maximum drilling speed, but actively engages damping control based on the contact degradation entropy and approach rate, so that the equipment can maintain recoverable commutation stability under fault disturbances. The purpose of this step is to make the system truly applicable to non-ideal working conditions in deep-earth environments, thereby enabling the identification, isolation and utilization of fault-induced aperiodic resonance noise, and preventing external mechanical disturbances from evolving into a continuous commutation failure state through brush contact instability.

[0023] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A smooth commutation control system for a DC motor with a matched multi-point rolling brush structure, characterized in that, The DC motor includes a multi-point rolling brush structure, and the system includes: The data acquisition module is used to acquire multimodal feedback data of the DC motor; the multimodal feedback data includes current feedback data, speed feedback data, and mechanical vibration feedback data containing the microscopic collision characteristics generated by the multi-point rolling brush structure in the commutation range of the DC motor. The observer module is used to extract high-frequency jumping features that match the rolling frequency of the brush structure based on the multimodal feedback data, and calculate the contact degradation entropy that characterizes the degree of brush contact state degradation accordingly. The prediction module is used to calculate the rate at which the contact degradation entropy approaches a preset failure boundary within a preset historical time window, wherein the preset failure boundary is greater than a preset danger threshold. The control decision module is used to output a first control command to maintain the current torque when the contact degradation entropy is lower than the preset danger threshold; and to output a second control command in combination with the approximation rate when the contact degradation entropy is higher than or equal to the preset danger threshold, thereby injecting damping current into the DC motor to reduce the speed and rebuild mechanical contact stability. The motor drive module is used to receive the first control command or the second control command to adjust the drive voltage or inject the damping current.

2. The DC motor smooth commutation control system with a matched multi-point rolling brush structure according to claim 1, characterized in that, The observer module includes: The feature extraction unit is used to filter the multimodal feedback data to fuse and separate the high-frequency fluctuation features from the current feedback data, the rotational speed feedback data and the mechanical vibration feedback data. The entropy calculation unit is used to input the high-frequency fluctuating features into a preset deep learning regression model and output the contact degradation entropy. The preset deep learning regression model includes a one-dimensional convolutional layer, a long short-term memory network layer, and a fully connected regression layer connected in sequence. The high-frequency fluctuating features are truncated by a preset time window to construct a two-dimensional tensor with temporal and feature channel dimensions, and then input into the one-dimensional convolutional layer. The fully connected regression layer maps the tensor to a dimensionless scalar in the range of 0 to 1 as the contact degradation entropy.

3. The DC motor smooth commutation control system with a matched multi-point rolling brush structure according to claim 1, characterized in that, The prediction module is specifically used for: Obtain the contact degradation entropy within a preset historical time window; Calculate the gradient of the change in the contact degradation entropy within the historical time window; The difference between the preset failure boundary and the current contact degradation entropy is calculated as the residual margin. When the contact degradation entropy shows an upward trend and the remaining margin is greater than zero, the ratio of the change gradient to the remaining margin is calculated as the approximation rate.

4. The DC motor smooth commutation control system with a matched multi-point rolling brush structure according to claim 1, characterized in that, The first control command is a torque sustaining command generated based on a sliding mode control algorithm, using the deviation between the speed feedback data and the preset target speed as input. Specifically, the sliding mode control algorithm constructs a linear sliding surface using the deviation between the speed feedback data and the preset target speed and its derivative with respect to time as state variables, and calculates the sliding mode control law using an exponential reaching law combined with a sign function. Finally, the corresponding drive voltage adjustment is calculated through the output of the sliding mode control law as the torque sustaining command.

5. A DC motor smooth commutation control system with a matched multi-point rolling brush structure according to claim 1, characterized in that, When the control decision module outputs the second control command, it specifically executes the following: The target damping frequency is determined by multiplying the approximation rate by a preset frequency mapping coefficient. The waveform parameters of the damping current are generated based on the target damping frequency; The waveform parameters are encapsulated into the second control command and output; wherein the damping current is a high-frequency pulsating ripple current superimposed on the DC main drive current; the waveform parameters specifically include pulsation frequency, pulsation amplitude and pulse duty cycle, wherein the pulsation frequency is the target damping frequency, the pulsation amplitude has a preset positive correlation with the approximation rate, and the pulse duty cycle is set to a fixed constant.

6. The DC motor smooth commutation control system with a matched multi-point rolling brush structure according to claim 1, characterized in that, It also includes a closed-loop recovery module, which is used for: After outputting the second control command, the contact degradation entropy updated in real time by the observer module is continuously acquired; Determine the relationship between the updated contact degradation entropy and the preset danger threshold; If the updated contact degradation entropy is lower than the preset danger threshold, a recovery command is generated; wherein, the recovery command is used to trigger the control decision module to switch to outputting the first control command; If the updated contact degradation entropy is higher than or equal to the preset danger threshold, the second control command is maintained.

7. A DC motor smooth commutation control system with a matched multi-point rolling brush structure according to claim 1, characterized in that, The preset failure boundary is the contact degradation entropy critical value corresponding to the continuous failure state of the DC motor commutation; wherein, the continuous failure state of the commutation is defined as the duration of torque loss of the DC motor being greater than or equal to a preset time threshold.

8. A DC motor smooth commutation control system with a matched multi-point rolling brush structure according to claim 1, characterized in that, The DC motor is used in deep-earth drilling equipment; wherein the mechanical vibration feedback data includes non-periodic mechanical resonance noise caused by geological faults.

Images