A coal mine crushing device for coal mining and a method of using the same

Abstract

The application discloses a coal mine crushing device for coal mining and a use method thereof, and relates to the technical field of coal mine crushing devices, which comprises a sensing layer, a control layer, an execution layer and a predictive maintenance layer; the sensing layer collects multi-source working condition data through a vibration monitoring unit, a current monitoring unit, a visual recognition unit and a position detection unit; a fuzzy logic reasoner is built in the control layer to fuse and identify the multi-source working condition data to output a working condition type, and a strategy generation module generates corresponding control instructions; the execution layer adjusts the rotating speed and torque of a crushing motor, adjusts the real-time gap value of a crushing mechanism and executes a foreign matter removal action according to the control instructions; and the predictive maintenance layer constructs a digital twin and predicts the remaining service life based on vibration spectrum data; through multi-source data fusion and working condition adaptive control, the application realizes accurate identification, cooperative adjustment and predictive maintenance of the crushing operation under all working conditions, and effectively improves the crushing efficiency, equipment safety and operation continuity.

Description

Technical Field

[0001] This invention relates to the field of coal mining technology, and in particular to a coal mining crushing device and its usage method. Background Technology

[0002] Traditional coal mine crushing equipment mainly includes toothed roll crushers, hammer crushers, and impact crushers. Their working principle relies on the mechanical impact and squeezing action between the crushing mechanism and the material. To adapt to the complex working conditions underground, existing equipment is usually equipped with simple overload protection devices (such as shear pins and hydraulic couplers) and alarm systems based on a single parameter (such as motor current). Some upgraded equipment also integrates variable frequency speed control function, which can adjust the operating parameters to a certain extent through manual or simple logic control.

[0003] First, coal seams have complex occurrence conditions, with highly heterogeneous coal and rock hardness, gangue content and distribution. Furthermore, during mining, foreign metal objects such as anchor bolts and rails inevitably get mixed in. Existing equipment relies heavily on manual inspections or single current thresholds to determine operating conditions. When material properties change abruptly or foreign objects enter, the system cannot accurately identify different operating conditions such as "heavy load," "overload," or "jamming" within milliseconds. This often results in the equipment experiencing severe vibration, stalling, or even toothed roller breakage before the protection device is passively triggered, causing unplanned shutdowns and equipment damage.

[0004] Second, the adjustment capability of the existing equipment is not matched with the changes in working conditions. For example, when the particle size distribution of the feed becomes finer and the hardness of the material decreases (light load condition), the equipment still runs at a constant speed and cannot automatically increase the crushing frequency to increase the throughput. When the gangue content increases sharply (heavy load condition), it is difficult to fundamentally reduce the torque load of the crushing mechanism by simply reducing the feed rate. This can easily lead to the motor overheating for a long time and shorten the equipment life.

[0005] Third, the existing system lacks the ability to adjust the key parameters of the crushing mechanism in a closed loop. The gap value usually needs to be adjusted manually by stopping the machine, and it cannot be dynamically changed according to the real-time material characteristics, resulting in poor stability of crushed particle size. Moreover, when metal foreign objects enter, the crushing mechanism with a fixed gap is very prone to jamming, and the handling process is cumbersome and has high safety risks.

[0006] Fourth, existing technologies for assessing equipment health status are mostly based on cumulative operating time or periodic manual inspections, lacking predictive maintenance methods based on real-time vibration, current and other multi-source data. This makes it difficult to detect abnormalities in the early stages of toothed roller wear, and sudden failures often affect the continuous production of the mining face.

[0007] Therefore, in response to the problems mentioned above, this invention proposes a coal mine crushing device and its usage method for coal mining. Summary of the Invention

[0008] To overcome the problems of existing coal mine crushing devices, such as lagging condition perception, insufficient adaptive adjustment capability, lack of active foreign object removal mechanism, and lack of health management, this invention proposes a coal mine crushing device and its usage method for coal mining. The system integrates multi-dimensional perception data such as vibration, current, vision, and position to construct a real-time condition identification model based on fuzzy logic. Based on the identification results, it dynamically and collaboratively adjusts the speed, torque, and crushing gap of the crushing mechanism. At the same time, it integrates active foreign object removal and predictive maintenance functions, thereby realizing adaptive and efficient operation and autonomous health management of crushing operations under all working conditions.

[0009] The technical solution of this invention is: a coal mine crushing device for coal mining, comprising a sensing layer, a control layer, an execution layer, and a predictive maintenance layer.

[0010] The sensing layer includes: The vibration monitoring unit is used to collect vibration spectrum data of the crushing mechanism in real time. The sampling frequency of the vibration spectrum data is not less than 20kHz, and the monitoring frequency band covers 10-10kHz. The current monitoring unit is used to collect stator current data of the crushing motor in real time, with a sampling frequency of not less than 1kHz, and to extract the root mean square value of the current in the time domain and the characteristic components in the frequency domain. A visual recognition unit is used to collect material image data from the feed inlet and identify the material particle size distribution, material hardness characteristics, and probability of foreign matter presence in the material image data based on a deep learning model. The visual recognition unit includes a line laser projector, a binocular camera, and a 3D point cloud reconstruction module for reconstructing the 3D point cloud data of the material. The deep learning model simultaneously outputs the bounding box of the material, the estimated material volume calculated based on the point cloud data, and the hardness label of high manganese steel or sandstone based on texture analysis. A position detection unit is used to detect the real-time gap value of the toothed roller or hammer of the crushing mechanism. The position detection unit is a magnetostrictive displacement sensor or a laser rangefinder sensor with a resolution of not less than 0.01 mm. The control layer is communicatively connected to the perception layer, and the control layer includes: The data fusion module is used to receive vibration spectrum data, stator current data, material particle size distribution, material hardness characteristics, probability of foreign object presence and real-time gap value, and synchronize and align the above data on timestamps to construct a feature vector reflecting the current crushing condition. The state determination module has a built-in fuzzy logic inferencer. The fuzzy logic inferencer takes the feature vector as input and outputs the current working condition type. The working condition type includes at least light load working condition, heavy load working condition, overload working condition and jamming working condition. The fuzzy logic inference engine includes: The membership function library stores Gaussian membership functions corresponding to each parameter in the feature vector. The rule base stores multiple "if-then" fuzzy rules. These fuzzy rules are used to define the correspondence between the amplitude of the low-frequency component (5-50Hz) in the vibration spectrum data exceeding a preset first threshold and the heavy-load condition. They also define the correspondence between the stator current data having a mutation rate exceeding 30% of the rated current within 50ms and the overload condition. The defuzzification interface is used to convert the fuzzy inference result into a definite working condition type using the centroid method; The strategy generation module is used to generate corresponding control commands based on the operating condition type. When the working condition is light load, a first speed increase command is generated to control the frequency converter drive unit to increase the speed of the crushing motor to a preset first speed threshold (110% to 120% of the rated speed). When the working condition is heavy load, a speed reduction command and a gap increase command are generated. These commands control the variable frequency drive unit to reduce the speed of the crushing motor to a preset second speed threshold (70% to 85% of the rated speed). The hydraulic adjustment unit is also controlled to increase the real-time gap value by 15-30% of the current gap value. When the working condition is the overload condition, an emergency stop and reverse command is generated to control the frequency converter drive unit to stop forward rotation and perform reverse rotation within 50ms. The reverse rotation duration is not less than 2 seconds. At the same time, a bypass start command is generated. The execution layer is communicatively connected to the control layer and is used to execute control instructions. The execution layer includes: The variable frequency drive unit is used to adjust the output speed and output torque of the crushing motor according to the control command, thereby changing the impact energy of the crushing mechanism; A hydraulic adjustment unit is used to drive the hydraulic system according to control commands to adjust the real-time gap value of the toothed roller or hammer of the crushing mechanism. The hydraulic adjustment unit includes a proportional servo valve, which is used to control the real-time gap value in a closed loop according to the gap adjustment amount in the control command, so that the real-time gap value responds to the change of the control command within 0.1 seconds. The foreign object removal unit is used to activate the reverse rotation logic or the bypass channel to discharge uncrushable objects when a control command corresponding to the over-iron condition is received. The foreign object removal unit includes a controllable flap installed on the side wall of the crushing chamber and a metal storage bin connected to the controllable flap. When a control command corresponding to the over-iron condition is received, the controllable flap is controlled to open within 0.3 seconds. The predictive maintenance layer is communicatively connected to the control layer and is used to record the time distribution of the operating condition type, and to construct a wear prediction model based on the changing trend of the vibration spectrum data, outputting the remaining service life prediction value and maintenance suggestions. The predictive maintenance layer includes: The digital twin and simulation module are synchronized with the physical entity of the crushing mechanism and are used to receive real-time data from the sensing layer. The simulation module is used to simulate the wear effect of changing control commands on the physical entity of the crushing mechanism in the digital twin. The predicted remaining service life is calculated based on the simulation results of the simulation module and the amplitude attenuation rate of the sideband components in the vibration spectrum data.

[0011] This invention proposes a method for using a coal mine crushing device in coal mining, comprising the following steps: S1, execute the self-test procedure to confirm that the communication between the perception layer, control layer, execution layer and predictive maintenance layer is normal, and load the fuzzy rule base in the state determination module; S2, real-time acquisition of vibration spectrum data from vibration monitoring unit, stator current data from current monitoring unit, material image data from visual recognition unit and real-time gap value from position detection unit, wherein vibration spectrum data is continuously acquired at a sampling rate of not less than 20kHz, stator current data is continuously acquired at a sampling rate of not less than 1kHz, and visual recognition unit acquires material images at a rate of not less than 5 frames / second. S3, the data obtained in step S2 is timestamped in the control layer, a feature vector is constructed, input to the fuzzy logic inference engine, and the current working condition type is output. The complete cycle of the working condition identification does not exceed 100ms. S4. Based on the working condition type output in step S3, call the strategy generation module to generate control commands and send them to the execution layer. In the case of light load conditions, the speed increase control logic is executed to increase the speed of the crushing motor to 110% to 120% of the rated speed, thereby increasing the crushing frequency; In heavy-load conditions, the speed reduction and gap expansion control logic is executed to reduce the speed of the crushing motor to 70% to 85% of the rated speed, while increasing the real-time gap value by 15% to 30% of the current gap value to reduce the torque of the crushing mechanism. If it is a case of passing iron, execute the emergency stop reverse and bypass discharge control logic, stop the forward rotation within 50ms and execute the reverse rotation. The reverse rotation lasts for no less than 2 seconds. At the same time, control the controllable flap to open within 0.3 seconds to discharge the metal foreign object into the metal temporary storage bin. If the work is stuck, execute the high-frequency forward and reverse jitter control logic, and alternate between forward and reverse rotation with a cycle of 0.5 seconds for 3 to 5 cycles to remove material accumulation; S5, the control execution results from step S4 are fed back to the predictive maintenance layer to update the wear prediction model, and the control parameters in the strategy generation module are optimized based on the control effect, specifically including: S51, construct a digital twin of the crushing mechanism, and correct the stiffness matrix and damping matrix of the digital twin based on real-time vibration spectrum data; S52 uses a digital twin simulation module to predict the impact coefficient of heavy-load operation on the life of the toothed roller under the current wear condition. S53, when the predicted remaining service life is lower than the preset maintenance threshold, a predictive maintenance work order containing specific maintenance components, maintenance time windows and spare parts list is generated and sent to the remote monitoring center. The calculation of the predicted remaining service life is based on the amplitude attenuation rate of the sideband components in the vibration spectrum data. When the amplitude of the sideband components attenuates by more than 15% compared with the initial value, it is determined to be in the middle of wear. When the attenuation exceeds 30%, it is determined to be in the maintenance state.

[0012] The beneficial effects of this invention are: 1. This invention introduces a state determination module with a built-in fuzzy logic inferencer into the control layer. Through the collaborative operation of the membership function library, rule library and defuzzification interface, it achieves millisecond-level accurate identification of light load, heavy load, overload, and jamming conditions. This solves the problem that existing technologies are unable to respond quickly when the operating conditions change abruptly. In particular, the joint fuzzy inference of stator current change rate and low-frequency vibration components significantly improves the identification accuracy of overload conditions.

[0013] 2. The strategy generation module and the execution layer of this invention work together to realize adaptive collaborative control based on working condition type. Under light load conditions, it automatically increases speed to increase throughput. Under heavy load conditions, it synchronously executes speed reduction and gap widening to reduce load. Under iron crossing conditions, it completes emergency stop reversal and bypass discharge in milliseconds. Under jammed conditions, it removes accumulation through high-frequency forward and reverse shaking, forming a differentiated intelligent control strategy covering all working conditions.

[0014] 3. The foreign object removal unit of this invention includes a controllable flap and a metal storage bin. Under the condition of passing iron, it is executed in conjunction with the reverse rotation logic of the frequency conversion drive unit, realizing the active and rapid discharge of metal foreign objects. This avoids the problem of equipment damage or long-term downtime caused by metal foreign objects stuck in traditional devices, and greatly improves the continuity and safety of crushing operations.

[0015] 4. This invention uses a digital twin and simulation module in the predictive maintenance layer to correct model parameters using real-time vibration spectrum data and calculate the remaining service life by combining the attenuation rate of sideband components. This enables online assessment and early warning of equipment health status, transforming traditional post-maintenance or periodic maintenance into predictive maintenance based on the actual condition of the equipment, effectively reducing the frequency of unplanned downtime. Attached Figure Description

[0016] Figure 1 The diagram shown is a schematic representation of the sensing layer structure of the present invention. Figure 2 The diagram shown is a schematic representation of the control layer structure of the present invention. Figure 3 The diagram shown is a schematic representation of the execution layer structure of the present invention. Figure 4 The diagram shown is a schematic representation of the predictive maintenance layer structure of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] This invention provides an embodiment: a coal mine crushing device for coal mining, comprising: Please see Figure 1 In this embodiment, the perception layer will be described in detail: The vibration monitoring unit employs a triaxial accelerometer with a range of ±50g and a frequency response range of 0.1-15kHz. This unit continuously acquires vibration acceleration signals during the operation of the crushing mechanism at a sampling frequency of 20kHz. During operation, the vibration signal is processed by an anti-aliasing filter and then converted into a 24-bit digital signal by an analog-to-digital converter. The control layer performs a Fast Fourier Transform on the raw vibration signal every 100 milliseconds, extracting spectral data within the 10-10kHz frequency band. The low-frequency band (5-50Hz) primarily reflects the overall load state of the crushing mechanism; vibration energy in this band is positively correlated with the degree of material accumulation and the meshing resistance of the toothed rollers. The mid-frequency band (200-2000Hz) primarily reflects the local impact state of the toothed rollers or hammers. The high-frequency band above 2000Hz primarily reflects abnormal states such as tooth surface wear and early bearing failure. The spectral data output by the vibration monitoring unit is stored in the data fusion module in the form of feature vectors.

[0019] The current monitoring unit employs the closed-loop Hall effect principle, achieving a measurement accuracy of 0.5%. This unit acquires the three-phase stator current data of the crusher motor in real time at a sampling frequency of 1kHz. In the control layer, the current data undergoes digital filtering, and the root mean square value, peak factor, and spectral characteristics of the three-phase current are calculated in real time. The current monitoring unit focuses on extracting the current mutation rate index, which is the ratio of the current value change amplitude every 50 milliseconds to the rated current. When the current mutation rate exceeds 30%, this index is marked as one of the key criteria for overload conditions. Simultaneously, the current monitoring unit also acquires the fundamental frequency component and harmonic components (2nd and 3rd harmonics) of the current to assist in determining the motor's operating state and load characteristics.

[0020] The visual recognition unit first projects a line of structured light onto the material surface at the feed inlet using a line laser projector with a wavelength of 660nm. A binocular camera simultaneously acquires images of the material with the structured light stripes; the camera's resolution is no less than 2 megapixels, and the frame rate is 5 frames per second. After processing the acquired images using a stereo matching algorithm, a 3D point cloud reconstruction module generates dense 3D point cloud data of the material surface, achieving a point cloud density of over 5000 points per square meter. Based on the reconstructed 3D point cloud, the system can calculate the material's bulk volume, maximum particle size, and particle size distribution curve.

[0021] The visual recognition unit incorporates a deep learning model, which uses an improved YOLO object detection network as its infrastructure. Compared to the traditional YOLO model, this invention improves the network structure, specifically: A spatial attention module was added to the backbone network to enhance the extraction of surface texture features of materials. Three parallel branches were added to the detection head to output the bounding box of the material, the estimated volume of the material, and the hardness label, respectively. During the model training phase, a dataset of coal and mineral material images containing different coal types, different gangue contents, different block sizes, and foreign objects such as anchor bolts, road spikes, and metal mesh was pre-collected. The dataset contained over 100,000 images, each manually labeled. During inference, the trained model can simultaneously identify each material block in the image, outputting its bounding box coordinates, the estimated volume calculated based on point cloud projection, and the hardness level derived from texture feature analysis. For metallic foreign objects, the model identifies their unique rectangular outline, metallic luster reflection characteristics, and significant differences from the coal and rock background, outputting a probability value for the presence of the foreign object. When this probability value exceeds 0.8, the system determines that an unbreakable metallic foreign object is present.

[0022] The position detection unit employs a magnetostrictive displacement sensor with a resolution of 0.01 mm. The measurement stroke is set to 50-200 mm according to the crusher specifications. The sensor's non-contact measurement method avoids mechanical wear, ensuring long-term operational reliability. The position detection unit outputs the real-time gap value of the crusher's toothed rollers or hammers at a 100 Hz update rate and transmits this value to the data fusion module in the control layer. When the hydraulic adjustment unit executes a gap adjustment command, the position detection unit feeds back the deviation between the actual gap value and the target gap value to the proportional servo valve controller in real time, achieving high-precision gap control.

[0023] Please see Figure 2 In this embodiment, the control layer will be described in detail: After receiving vibration spectrum data, stator current data, material particle size distribution, material hardness characteristics, foreign object presence probability, and real-time gap value from the sensing layer, the data fusion module first performs timestamp alignment. Since the sampling frequencies of each sensing unit are different, the data fusion module uses an interpolation alignment algorithm to unify all data into a 10-millisecond control cycle. The aligned data is then spliced ​​together according to preset feature dimensions to form a high-dimensional feature vector containing 32 feature components. This feature vector covers information on four dimensions: the dynamic state, electrical state, material state, and mechanical state of the equipment, and can comprehensively reflect the crushing conditions at the current moment.

[0024] The state determination module's built-in fuzzy logic inference engine consists of three parts: a membership function library, a rule base, and a defuzzification interface. The membership function library stores Gaussian membership functions corresponding to the parameters in the eigenvector: Taking the low-frequency component (energy in the 5-50Hz band) in vibration spectrum data as an example, its membership function is defined as three fuzzy sets: low energy, medium energy, and high energy. The center value of the low-energy set is 20mm. 2 / s 4 The standard deviation is 8mm. 2 / s 4 The central value of the energy set is 50 mm. 2 / s 4 The standard deviation is 12 mm. 2 / s 4 The center value of the high-energy set is 90 mm. 2 / s 4 The standard deviation is 15 mm. 2 / s 4 .

[0025] Taking the mutation rate of stator current data as an example, its fuzzy set is defined as low mutation rate, medium mutation rate and high mutation rate, among which the membership function of high mutation rate reaches its peak when the mutation rate exceeds 30%.

[0026] Taking the probability of foreign object presence as an example, its fuzzy set is defined as no foreign object, possible foreign object presence, and certain foreign object presence. The membership function of certain foreign object presence approaches 1 when the probability value exceeds 0.8.

[0027] The rule base stores over 50 "if-then" fuzzy rules, which are derived from expert knowledge of coal mine crushing processes and a large amount of historical operational data. Typical rules include: (1) If the energy of the low-frequency component of vibration is high, the root mean square value of the stator current is high, and the material hardness is hard, then the working condition is a heavy load condition with a confidence level of 0.9.

[0028] (2) If the stator current mutation rate is high, the probability of the presence of foreign matter is certain, and non-periodic impact characteristics appear in the vibration spectrum, then the working condition type is over-iron working condition with a confidence level of 0.95.

[0029] (3) If the low-frequency component energy of the vibration is high, the root mean square value of the stator current is high, and the real-time gap value continues to decrease to the minimum value, then the working condition is a jamming working condition with a confidence level of 0.85.

[0030] (4) If the energy of the low-frequency component of vibration is low, the root mean square value of the stator current is low, and the particle size distribution of the material is fine, then the working condition is a light load condition with a confidence level of 0.88.

[0031] After receiving the feature vector, the fuzzy inference engine activates all applicable fuzzy rules in parallel. Each rule outputs a fuzzy conclusion based on the degree to which its preconditions are met. The defuzzification interface uses the centroid method to calculate a weighted average of the fuzzy conclusions of all activated rules, resulting in a precise output of the operating condition type. The entire inference process is completed within 50 milliseconds, ensuring the system can respond quickly to changes in operating conditions.

[0032] The strategy generation module generates corresponding control commands based on the operating condition type output by the status determination module and according to the preset control strategy. This module maintains a control strategy table internally, the details of which are as follows: When the operating condition is light load, the strategy generation module generates a first speed increase command. This command includes a target speed value, which is set to 110-120% of the rated speed of the crusher motor. The specific value is dynamically determined based on the particle size distribution and hardness characteristics of the material. When the particle size is smaller than the design value and the hardness is low, the upper limit of 120% is used; when the particle size is close to the design value, the lower limit of 110% is used. At the same time, the strategy generation module also generates a feeding rate increase suggestion signal, which is sent to the upstream feeding system through the collaborative control interface, suggesting that the feeding rate be increased to 120-150% of the current feeding rate.

[0033] When the operating condition is heavy load, the strategy generation module generates a speed reduction command and a gap increase command. The speed reduction command sets the target speed of the crusher motor to 70-85% of the rated speed, with the specific value dynamically determined based on the specific value of the low-frequency vibration energy component; the higher the energy, the lower the target speed. The gap increase command sets the increase in the real-time gap value to 15-30% of the current gap value, with the specific value determined based on the material hardness characteristics; the higher the hardness, the greater the increase. Simultaneously, the strategy generation module sends a material rate reduction signal to the upstream feeding system through the collaborative control interface, requiring the feed rate to be reduced to below 50% of the current feed rate within 1 second to prevent further material accumulation.

[0034] When the operating condition is an overload condition, the strategy generation module generates an emergency stop / reverse command and a bypass start command. The emergency stop / reverse command requires the frequency converter drive unit to stop forward rotation within 50ms and immediately start reverse rotation. The reverse rotation speed is set to 30% of the rated speed, and the reverse rotation duration is not less than 2 seconds, thereby disengaging the metal foreign object entering the crushing chamber from the meshing position and creating conditions for subsequent discharge. The bypass start command requires the foreign object removal unit to open the controllable flap within 0.3 seconds, allowing the metal foreign object to fall into the metal temporary storage bin under gravity.

[0035] When the operating condition is a jammed condition, the strategy generation module generates high-frequency forward and reverse jitter control logic. This instruction requires the frequency converter drive unit to alternate between forward and reverse rotation at a cycle of 0.5 seconds, with each direction's running time being 0.5 seconds. A dead time of 0.1 seconds is set during the forward and reverse switching process, lasting for 3 to 5 cycles. This high-frequency jitter control method can effectively loosen the material accumulated in the crushing chamber, thus relieving the jamming state. If the jamming state is not relieved after 3 cycles, the cycle time is increased to 5 cycles. If the jamming still cannot be relieved, the system issues a serious fault alarm, prompting manual intervention.

[0036] Please see Figure 3 In this embodiment, the execution layer will be described in detail: The variable frequency drive unit adopts a high-performance vector control inverter with an output frequency range of 0-120Hz and a control accuracy of 0.01Hz. This inverter has a built-in dual closed-loop control structure with speed and current loops, enabling it to adjust the motor speed to the target value within 100 milliseconds of receiving a speed command. When an emergency stop / reverse command is received, the inverter first reduces the motor speed to zero within 20 milliseconds using energy-consumption braking, and then completes the direction switching and accelerates in the opposite direction to the target speed within 30 milliseconds. The variable frequency drive unit also monitors the motor's output torque in real time. When the torque exceeds 120% of the motor's rated torque, it automatically triggers limiting protection to ensure the motor is not damaged.

[0037] The hydraulic adjustment unit includes a hydraulic pump station, a proportional servo valve, and a hydraulic cylinder. The proportional servo valve has a response frequency of no less than 50Hz and can control the displacement of the hydraulic cylinder in a closed loop according to the gap adjustment amount in the control command, thereby achieving precise adjustment of the gap value in real time. After receiving the gap increase command, the valve core of the proportional servo valve reaches the set opening degree within 30 milliseconds, the hydraulic cylinder pushes the movable bearing seat to move, and the position detection unit provides real-time feedback on the actual position, forming a position closed-loop control. The entire gap adjustment process is completed within 0.1 seconds, with an adjustment accuracy of ±0.02mm. The hydraulic system is also equipped with an accumulator for rapid power supply in emergencies, ensuring rapid response of gap adjustment during iron-crossing conditions.

[0038] The foreign object removal unit comprises a controllable flap, a metal storage bin, and a flap drive mechanism. The controllable flap is located on the side wall of the crushing chamber, forming a bypass channel with the feed chute. Under normal circumstances, the controllable flap is closed, and the material enters the downstream after normal crushing in the crushing chamber. Upon receiving a bypass opening command, the rapid-acting cylinder in the flap drive mechanism pushes the controllable flap to rotate 90 degrees within 0.3 seconds, opening the bypass channel. Simultaneously, the frequency converter drive unit executes a reverse rotation command, pushing out the metal foreign object stuck between the crushing teeth through reverse rotation, causing it to fall into the metal storage bin along the bypass channel. The metal storage bin is equipped with a level sensor; when the metal accumulates to the set level, the system issues a clearing prompt.

[0039] Please see Figure 4 In this embodiment, the predictive maintenance layer is described in detail: The digital twin of this layer is constructed based on the physical structure, material properties, and dynamic parameters of the crushing mechanism. During system initialization, the digital twin loads the geometric model, finite element model, and initial stiffness and damping matrices of the crushing mechanism. During system operation, the digital twin receives vibration spectrum data, current data, and position data from the sensing layer in real time via the OPC UA protocol, and uses this real-time data to correct its model parameters. Specifically, based on the modal frequency information in the real-time vibration spectrum data, the digital twin uses a model update algorithm to reverse-calculate the equivalent stiffness and equivalent damping of the crushing mechanism under the current state, and corrects the initial stiffness and damping matrices online. The corrected digital twin can more accurately reflect the true state of the physical entity.

[0040] The simulation module, based on a modified digital twin, simulates the impact of different control strategies on the wear of the crushing mechanism. For example, when the system plans to execute heavy-load control logic, the simulation module pre-simulates the impact of this heavy-load process on the contact stress at the tooth tips of the toothed rollers in the digital twin, and calculates the amount of wear on the toothed rollers due to this operation based on fatigue cumulative damage theory. In this way, the system can assess the long-term impact before actually executing control commands, and adjust control parameters as needed while ensuring production efficiency, thereby extending equipment life.

[0041] The calculation of the remaining service life prediction integrates the simulation results from the simulation module with the sideband components from the measured vibration spectrum data. Sideband components refer to the modulation sidebands around the gear meshing frequency or bearing fault characteristic frequency in the vibration spectrum; their amplitude reflects the degree of tooth surface wear or bearing damage. The system uses the amplitude of the sideband components in the initial state as a benchmark and monitors their attenuation rate in real time. When the amplitude of the sideband components attenuates by more than 15% compared to the initial value, it is determined to be in the middle stage of wear, and the system prompts for increased attention; when the attenuation exceeds 30%, it is determined to be in a maintenance-required state, and the system automatically generates maintenance recommendations. Simultaneously, based on the current wear state and combined with future production plans, the simulation module predicts the specific value of the remaining service life, outputting it in the form of operating hours or estimated tonnage that can be processed.

[0042] When the predicted remaining service life is lower than the preset maintenance threshold, the predictive maintenance layer automatically generates a predictive maintenance work order. This work order includes the name of the specific component to be maintained (left toothed roller, right toothed roller, bearing, etc.), the suggested maintenance time window (based on idle periods in the production plan), and a spare parts list (including the model, quantity, and procurement recommendations of the spare parts to be replaced). This work order is sent to the remote monitoring center via the industrial internet platform and executed after review by maintenance management personnel.

[0043] At runtime, specifically: S1. After the system powers on, the main controller of the control layer executes a self-test program, sequentially sending handshake signals to each unit in the sensing layer and each unit in the execution layer to confirm normal communication. During the self-test, the system also checks whether the initial readings of each sensor are within a reasonable range; for example, the initial vibration value of the vibration monitoring unit should be less than 0.5 mm / s. 2 The initial gap value of the position detection unit should deviate from the preset value by no more than 0.1 mm. After the self-test passes, the system loads the fuzzy rule base in the state determination module, reads the rule base file from non-volatile memory into memory, and establishes a rule index table to improve inference efficiency. At the same time, the predictive maintenance layer server starts the digital twin service and loads the geometric model and initial stiffness matrix of the crushing mechanism.

[0044] S2, the system executes data acquisition tasks cyclically with a 10-millisecond control cycle. Within each cycle, the vibration monitoring unit acquires a 0.5-second vibration signal at a 20kHz sampling rate and calculates its spectral data; the current monitoring unit acquires continuous current waveforms at a 1kHz sampling rate and calculates the root mean square value and mutation rate of the current; the visual recognition unit acquires material images at a rate of 5 frames per second, and each frame is processed by a deep learning model to output material feature data. Since deep learning inference takes approximately 50 milliseconds, the actual update frequency of the visual data is approximately 10 frames per second; the position detection unit outputs real-time gap values ​​at an update rate of 100Hz. All acquired data is timestamped and sent to the buffer of the data fusion module. The buffer uses a circular queue structure, retaining the most recent 2 seconds of historical data.

[0045] In step S3, the data fusion module extracts the currently aligned data from the cache and constructs a feature vector containing 32 feature components. This feature vector is input into the fuzzy logic inference engine. The inference engine first fuzzifies each feature component, calculates its membership degree to each fuzzy set, and then activates all rules in the rule base that satisfy the preconditions. The activation strength of each rule is the minimum membership degree of each precondition. Finally, the fuzzification is defuzzified using the centroid method to calculate the precise output value of the current operating condition. The entire operating condition identification process takes no more than 100 milliseconds, and the system latency from data acquisition to operating condition output is controlled within 120 milliseconds, meeting real-time control requirements.

[0046] S4. Based on the operating condition type output in step S3, the strategy generation module generates corresponding control commands and sends them to the execution layer. The specific execution process under each operating condition is described below: Under light load conditions, the system executes speed-up control logic. After receiving the first speed-up command, the frequency converter drive unit gradually increases the crusher motor speed from the rated speed to 110-120% of the rated speed, with the speed-up ramp time set to 5 seconds to avoid impact. Simultaneously, the coordination control interface sends a material rate increase signal to the upstream feeding system. Upon receiving the signal, the upstream feeding system increases the vibration frequency of the feeder or the belt speed by 20-30%, increasing the feed rate. After executing speed-up control, the impact frequency of the crushing mechanism increases, the material throughput speed accelerates, and the overall system processing capacity increases by 15-25%. The control layer continuously monitors changes in operating conditions. If the light load condition persists for more than 30 minutes, the system further optimizes control parameters to maximize output while ensuring equipment safety.

[0047] Under heavy-load conditions, the system executes speed reduction and gap widening control logic. First, the frequency converter drive unit smoothly reduces the crushing motor speed to 70-85% of the rated speed within 3 seconds. Simultaneously, the hydraulic adjustment unit starts, and the proportional servo valve controls the hydraulic cylinder movement according to the command value, increasing the real-time gap value by 15-30% of the current gap value. The gap widening process adopts an S-shaped acceleration and deceleration curve to ensure smooth operation and avoid mechanical impact. Under the combined effect of speed reduction and gap widening, the throughput per unit time of the crushing mechanism decreases, the meshing torque of a single toothed roller decreases, and the motor current recovers to below 90% of the rated current. The collaborative control interface simultaneously sends a material rate reduction signal to the upstream feeding system, which reduces the feed rate to below 50% of the current feed rate within 1 second to prevent further material accumulation in the crushing chamber.

[0048] Under the condition of overload, the system executes emergency stop, reverse rotation, and bypass discharge control logic. When the status determination module outputs the overload condition, the control layer immediately interrupts all current control tasks and executes the overload processing flow with the highest priority. The variable frequency drive unit first completes the forward rotation stop within 50 milliseconds. This process is achieved through the DC braking function of the frequency converter, with the braking current set to 150% of the motor's rated current. Then, the variable frequency drive unit completes the direction switching within 30 milliseconds and accelerates in the reverse direction to 30% of the rated speed. The reverse rotation lasts for 2 to 3 seconds. Simultaneously with the start of the reverse rotation, the controllable flap of the foreign object removal unit opens within 0.3 seconds, forming a bypass channel. During the reverse rotation, the metal foreign objects stuck between the crushing teeth are pushed out in the reverse direction and fall into the metal temporary storage bin along the bypass channel. After completing the reverse rotation, the variable frequency drive unit stops running, and the system pauses for 5 seconds to observe the operating condition. If the vibration monitoring unit and current monitoring unit show that the working condition has returned to normal, the system will restart forward rotation and resume crushing operations; if the abnormality still exists, the over-iron processing procedure will be executed again, up to three times. If it is still ineffective, a serious alarm will be issued.

[0049] Under jammed conditions, the system executes high-frequency forward and reverse jitter control logic. Unlike the emergency stop and reverse rotation under overload conditions, the jitter control under jammed conditions adopts a periodic forward and reverse switching method. The frequency converter drive unit alternates between forward and reverse rotation with a period of 0.5 seconds, running for 0.5 seconds in each direction. A dead time of 0.1 seconds is set during the forward and reverse switching, lasting for a total of 3 cycles. This high-frequency jitter can loosen the accumulated material and disrupt its stable stacking structure. If the condition returns to normal after 3 cycles, the system switches to heavy-load processing logic; if the jammed state is still not resolved, it continues to execute for 5 cycles; if it still cannot be resolved after 5 cycles, the system stops running and issues a manual intervention alarm.

[0050] S5: During system operation, all control execution results, operating condition change records, and sensing data are synchronized to the predictive maintenance layer.

[0051] S51, the predictive maintenance layer server receives vibration spectrum data and location data from the sensing layer every 10 minutes. It uses this data to correct the stiffness matrix and damping matrix of the digital twin online. The correction algorithm uses Kalman filtering and model update technology to ensure that the corrected model can accurately reflect the dynamic characteristics of the physical entity.

[0052] S52, using a modified digital twin, simulates the heavy-load conditions the system may face in the future. Based on the fatigue cumulative damage theory, it calculates the contribution of each heavy-load condition to the fatigue damage of the toothed roller tip. Damage accumulation adopts a linear accumulation model. When the accumulated damage reaches a critical value, it is determined to be the end of the service life. The system also monitors the sideband components in the vibration spectrum and uses their amplitude attenuation rate as a supplementary indicator of wear status.

[0053] S53: When the predicted remaining service life is lower than a preset maintenance threshold, such as a predicted remaining service life of less than 30 days or less than 50,000 tons of throughput, the system automatically generates a predictive maintenance work order. The work order includes: the specific name of the component requiring maintenance, a suggested maintenance time window (automatically selected based on the production plan to determine the most suitable downtime), and a spare parts list (including the name, quantity, inventory status, and procurement recommendations of the spare parts to be replaced). This work order is sent to the remote monitoring center via the industrial internet platform and pushed to the mobile terminals of relevant maintenance personnel.

[0054] This invention provides an embodiment: This example uses an existing toothed roller crusher as the experimental object, with a rated processing capacity of 500 tons / hour and an initial toothed roller gap of 50mm. This example adds a sensing layer, a control layer, an execution layer, and a predictive maintenance layer to the crusher. During the test, the system automatically collects and records all operating data. Three operating conditions were set up for verification: light load, heavy load, and overload. Verification of each condition used a comparative method, recording the system's performance indicators under intelligent control mode and traditional control mode. Traditional control mode refers to operation relying solely on manual inspection and fixed parameters, as detailed below: Under light load conditions, on the 8th day of continuous operation, the feed particle size distribution showed a significant refinement, with materials smaller than 30mm accounting for 75%. The coal was relatively soft. After the system's visual recognition unit detected the change in particle size distribution and the decrease in hardness, the status judgment module output "light load condition," and the strategy generation module issued an acceleration command, increasing the crusher motor speed from the rated speed of 1480 r / min to 1628 r / min. Simultaneously, the coordination control interface sent a material lifting signal to the upstream feeding system, increasing the feed rate from 450 tons / hour to 540 tons / hour. This system operated continuously for 2 hours under these conditions, recording indicators such as throughput, unit energy consumption, and product particle size. Compared to the traditional control mode (maintaining rated speed and fixed feed rate), the intelligent control mode increased the crushing system's throughput by 18.5%, reduced unit energy consumption by 12.3%, and maintained a product particle size qualification rate above 95%, without any particle size exceeding the standard due to increased throughput.

[0055] Under heavy load conditions, on the 25th day of continuous operation, the gangue content in the feed suddenly increased to 40%. The gangue had high hardness, and the system vibration monitoring unit detected low-frequency vibration energy of 5Hz to 50Hz, starting from the initial 45mm. 2 / s 4 Rise to 98mm 2 / s 4 The current monitoring unit detected that the root mean square value of the motor current increased from 65% to 112% of the rated current. The status judgment module output a heavy load condition. The strategy generation module issued a speed reduction and gap widening command. The motor speed decreased to 1184 r / min within 3 seconds, and the toothed roller gap increased from 50 mm to 65 mm. At the same time, the upstream feeding system reduced the feed rate from 500 tons / hour to 250 tons / hour within 1 second. The system operated under this condition until the vibration and current indicators returned to normal. Compared with the traditional control mode (which relies solely on manual shutdown after discovering the current overload), the average processing time for heavy load conditions under the intelligent control mode was shortened from 15 minutes (including manual judgment and processing time) to 2 minutes. The duration of motor overload was reduced from more than 3 minutes in the traditional mode to no more than 30 seconds, effectively avoiding motor overheating and equipment damage.

[0056] On the 42nd day of continuous operation under the condition of passing through iron, a metal anchor bolt weighing approximately 15 kg entered the crushing chamber along with the material. The visual recognition unit detected the metal foreign object with a probability value of 0.92. The status determination module output that the condition of passing through iron was met, and the system completed the forward rotation stop and started reverse rotation within 50 milliseconds. The reverse rotation lasted for 2.5 seconds, and the controllable flap opened within 0.28 seconds. Under the action of reverse rotation, the metal foreign object was dislodged from between the crushing teeth and fell into the metal temporary storage bin along the bypass channel. The entire process from the foreign object entering the crushing chamber to its discharge took approximately 3.8 seconds. The crushing equipment did not suffer any damage, and the system automatically resumed normal crushing operation after the foreign object was discharged. In the traditional control mode, the entry of such metal foreign objects often caused the crushing teeth to jam, requiring manual shutdown for handling. The average handling time exceeded 30 minutes, and it was often accompanied by equipment damage such as broken tooth rollers or chipped tooth tips.

[0057] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A coal mine crushing device for coal mining, characterized in that, It includes the perception layer, control layer, execution layer, and predictive maintenance layer: The sensing layer includes: The vibration monitoring unit is used to collect vibration spectrum data of the crushing mechanism in real time during operation; The current monitoring unit is used to collect the stator current data of the crushing motor in real time; The visual recognition unit is used to collect material image data from the feed inlet and identify the particle size distribution, hardness characteristics and probability of foreign matter presence in the material image data based on a deep learning model. The position detection unit is used to detect the real-time gap value of the toothed roller or hammer of the crushing mechanism. The control layer is communicatively connected to the perception layer, and the control layer includes: The data fusion module is used to receive vibration spectrum data, stator current data, material particle size distribution, material hardness characteristics, probability of foreign object presence and real-time gap value, and synchronize and align the above data on timestamps to construct a feature vector reflecting the current crushing condition. The state determination module has a built-in fuzzy logic inferencer. The fuzzy logic inferencer takes the feature vector as input and outputs the current working condition type. The working condition type includes at least light load working condition, heavy load working condition, overload working condition and jamming working condition. The strategy generation module is used to generate corresponding control commands based on the operating condition type. The execution layer is communicatively connected to the control layer and is used to execute control instructions. The execution layer includes: The variable frequency drive unit is used to adjust the output speed and output torque of the crushing motor according to the control command, thereby changing the impact energy of the crushing mechanism; The hydraulic adjustment unit is used to drive the hydraulic system according to control commands and adjust the real-time gap value of the toothed roller or hammer of the crushing mechanism. The foreign object removal unit is used to activate the reverse rotation logic or the bypass channel to remove uncrushable objects when it receives the control command corresponding to the over-iron working condition. The predictive maintenance layer is communicatively connected to the control layer and is used to record the time distribution of operating conditions. Based on the changing trend of vibration spectrum data, it constructs a wear prediction model and outputs the predicted value of remaining service life and maintenance suggestions.

2. A coal mine crushing device for coal mining according to claim 1, characterized in that: The fuzzy logic inference engine in the state determination module includes a membership function library, a rule library, and a defuzzification interface. The membership function library stores Gaussian membership functions corresponding to each parameter in the feature vector. The rule library stores multiple "if-then" fuzzy rules, which are used to define the correspondence between low-frequency components in the vibration spectrum data and heavy-load conditions, and also define the correspondence between the mutation rate of stator current data and overload conditions. The defuzzification interface is used to convert the fuzzy inference results into a definite operating condition type using the centroid method.

3. A coal mine crushing device for coal mining according to claim 1, characterized in that, The strategy generation module includes: When the working condition is light load, a first speed increase command is generated to control the frequency converter drive unit to increase the speed of the crushing motor to the preset first speed threshold, thereby increasing the throughput. When the working condition is heavy load, a speed reduction command and a gap increase command are generated to control the frequency conversion drive unit to reduce the speed of the crushing motor to the preset second speed threshold, and at the same time control the hydraulic adjustment unit to increase the real-time gap value, thereby reducing the crushing load. When the working condition is an overload condition, an emergency stop and reverse command is generated to control the frequency converter drive unit to stop rotating in the forward direction and perform reverse rotation within a preset time. At the same time, a bypass opening command is generated to control the foreign object removal unit to start.

4. A coal mine crushing device for coal mining according to claim 1, characterized in that: The visual recognition unit includes a line laser projector for projecting structured light onto the surface of the material, a binocular camera for acquiring material images with structured light stripes, and a three-dimensional point cloud reconstruction module for reconstructing the material based on the material images. The deep learning model is an improved YOLO model, which simultaneously outputs the bounding box of the material, the estimated volume of the material calculated based on point cloud data, and the hardness label of high manganese steel or sandstone based on texture analysis.

5. A coal mine crushing device for coal mining according to claim 1, characterized in that: The predictive maintenance layer also includes a digital twin and a simulation module, wherein the physical entity of the crushing mechanism in the digital twin is synchronized to receive real-time data from the perception layer; the simulation module is used to simulate the wear effect of changing control commands on the physical entity of the crushing mechanism in the digital twin.

6. A coal mine crushing device for coal mining according to claim 1, characterized in that: The predicted remaining service life is calculated based on the simulation results from the simulation module and the sideband components in the vibration spectrum data.

7. A coal mine crushing device for coal mining according to claim 1, characterized in that: The foreign object removal unit includes a controllable flap installed on the side wall of the crushing chamber and a metal storage bin connected to the controllable flap. When the foreign object removal unit receives a control command corresponding to the over-iron working condition, it controls the controllable flap to open, so that the uncrushable object falls into the metal storage bin under the action of gravity or mechanical thrust.

8. A coal mine crushing device for coal mining according to claim 1, characterized in that: The position detection unit is a magnetostrictive displacement sensor or a laser rangefinder with a resolution of not less than 0.01 mm. The hydraulic adjustment unit includes a proportional servo valve, which is used to control the real-time gap value in a closed loop according to the gap adjustment amount in the control command, so that the real-time gap value responds to the change of the control command within 0.1 seconds.

9. A method of using a coal mine crushing device for coal mining, based on a coal mine crushing device according to any one of claims 1-8, characterized in that, Includes the following steps: S1, execute the self-test procedure to confirm that the communication between the perception layer, control layer, execution layer and predictive maintenance layer is normal, and load the fuzzy rule base in the state determination module; S2, real-time acquisition of vibration spectrum data from vibration monitoring unit, stator current data from current monitoring unit, material image data from visual recognition unit, and real-time gap value from position detection unit; S3, align the data obtained in step S2 with timestamps in the control layer, construct a feature vector, input it into the fuzzy logic inference engine, and output the current working condition type; S4. Based on the working condition type output in step S3, call the strategy generation module to generate control commands and send them to the execution layer; if it is a light-load working condition, execute the speed-up control logic to increase the crushing frequency. In heavy-load conditions, the speed reduction and gap widening control logic is executed to reduce the torque of the crushing mechanism; in overload conditions, the emergency stop reverse and bypass discharge control logic is executed to protect the crushing mechanism; in jammed conditions, the high-frequency forward and reverse shaking control logic is executed to release material accumulation. S5. Feed back the control execution results from step S4 to the predictive maintenance layer, update the wear prediction model, and optimize the control parameters in the strategy generation module based on the control effect.

10. The method of using a coal mine crushing device according to claim 9, characterized in that, Step S5 specifically also includes: S51, construct a digital twin of the crushing mechanism, and correct the stiffness matrix and damping matrix of the digital twin based on real-time vibration spectrum data; S52 uses a digital twin simulation module to predict the impact coefficient of heavy-load operation on the life of the toothed roller under the current wear condition. S53 When the predicted remaining service life is lower than the preset maintenance threshold, a predictive maintenance work order containing specific maintenance components, maintenance time windows and spare parts list is generated and sent to the remote monitoring center.

Images