Multimodal sensor fusion coal sample grinder and grinding control method

Abstract

The present application relates to a kind of coal sample grinder of multi-modal sensor fusion and grinding control method, coal sample grinder includes feeding mechanism, grinding mechanism, grinding control mechanism, vibration mechanism, sample discharge mechanism, sample container and weighing mechanism;Grinding mechanism is equipped with grinding cavity, sample outlet is connected sample container by sample discharge mechanism, and sample container is arranged on weighing mechanism;Grinding cavity is arranged on vibration mechanism, and the top of grinding cavity is equipped with compressed air inlet;Grinding control mechanism includes controller and sensor component, wherein sensor component includes one or several of gas pressure sensor, light transmittance sensor, image sensor and temperature sensor.The present application can realize the accurate control of coking coal and coke sample target particle size, meet the strict requirements of relevant standard specification to coal sample particle size, improve the degree of automation of coal sample grinding device on this basis, realize the upgrading iteration of coal sample preparation device.

Description

Technical Field

[0001] This invention relates to the field of coal and coke laboratory sample preparation technology, and in particular to a coal sample grinder and grinding control method with multimodal sensor fusion. Background Technology

[0002] Coal and coke experiments are typically used for property analysis, process research, and performance testing of related products, such as industrial analysis of coal and coke. Samples used in coal and coke experiments must be crushed, reduced in size, and dried strictly according to standards to avoid deviations in experimental results.

[0003] In coking experiments, the coal samples used for analyzing G-value, Y-value, b-value, and Gibbs freeness have particularly strict requirements for particle size. The target particle size in relevant standards and specifications is as follows:

[0004] The GB / T5447-2014 standard, "Method for Determination of Caking Index of Bituminous Coal", stipulates that the particle size requirement for the caking index sample is 0.1-0.2 mm, accounting for 20%-35%, while also being compatible with other samples such as the calorific value sample.

[0005] The GB / T25213-2010 standard, "Determination of plasticity of coal - Constant torque Gibbs plasticity tester method", stipulates that the particle size requirement for Gibbs flowability samples is 0.425 mm, of which the proportion of particles <0.2 mm is <50%.

[0006] The GB / T474-2008 standard, "Preparation Method of Coal Samples", stipulates that the particle size requirement for the plastic layer sample is 1.5 mm, of which the proportion of particles <0.2 mm is ≤30%.

[0007] The GB / T16773-2008 standard, "Preparation Methods for Coal Petrological Analysis Samples," stipulates that the particle size requirement for coal petrological analysis samples is 1.0 mm, of which the proportion of particles <0.1 mm is ≤10%.

[0008] There are various devices used for preparing coal samples for coking experiments, but most are simple in structure and lack a high degree of automation. Especially for the preparation of coal samples for coking coal analysis, most currently use sealed sample preparation machines and double-roll crushers, requiring manual operation throughout the entire process. This is not only labor-intensive, but the quality of the coal samples also heavily relies on the experience, cognitive ability, and sense of responsibility of the experimental personnel. The final particle size of the coal samples often deviates significantly from the standard requirements and is unstable, frequently causing substantial deviations in transaction settlement and production process guidance, resulting in high management costs.

[0009] This invention is designed to completely iterate upon traditional preparation methods. It enables rigorous, scientific sample preparation that meets relevant GB requirements, entirely through the equipment itself, bringing about a historic transformation in transaction settlement and guiding production process control. Summary of the Invention

[0010] To address the aforementioned issues, this invention provides a multimodal sensor fusion coal sample grinder and grinding control method, which enables precise control of the target particle size of coking coal and coke samples, meeting the stringent requirements of relevant standards and specifications for coal sample particle size. Furthermore, it improves the automation control level of the coal sample grinding device, achieving an upgrade and iteration of the coal sample preparation device.

[0011] To achieve the above objectives, the present invention employs the following technical solution:

[0012] A multimodal sensing fusion coal sample grinding mill includes a feeding mechanism, a grinding mechanism, a grinding control mechanism, a vibration mechanism, a sample discharging mechanism, a sample holding mechanism, and a weighing mechanism. The grinding mechanism has a grinding chamber, with a feed inlet at the top connected to the feeding mechanism, and a sample outlet at the bottom connected to the sample holding mechanism via the sample discharging mechanism. The sample holding mechanism is mounted on the weighing mechanism. The grinding chamber is mounted on the vibration mechanism and is a closed chamber with a compressed air inlet at the top. The grinding control mechanism includes a controller and a sensor assembly, wherein the sensor assembly includes one or more of a gas pressure sensor, a light transmittance sensor, an image sensor, and a temperature sensor. The gas pressure sensor, light transmittance sensor, image sensor, and temperature sensor are all located at the top of the grinding chamber and are respectively connected to the controller. The controller is also connected to the control terminal of the vibration mechanism.

[0013] The feeding mechanism includes a feeding hopper, a feeding valve A, a buffer pipe, and a feeding valve B; the top of the feeding hopper is equipped with a movable hopper cover, and the bottom of the feeding hopper is connected to the grinding chamber through the buffer pipe; the upper end of the buffer pipe is connected to the feeding hopper through the feeding valve A, and the lower end of the buffer pipe is connected to the grinding chamber through the material sample valve B.

[0014] The grinding chamber is equipped with a grinding ring and a grinding block; the sample dispensing mechanism includes a sample dispenser and a sample dispensing pipe, the sample dispenser is located at the bottom of the grinding chamber and is equipped with a sieve; the sample dispenser is connected to the sample holding mechanism through the sample dispensing pipe.

[0015] The vibration mechanism is an eccentric vibration mechanism, consisting of a vibration frame, a variable frequency motor, an eccentric wheel, and a support. The grinding chamber is located on the vibration frame, which is movably connected to the support. The vibration frame consists of a top plate, multiple springs, and a bottom plate, with the springs located between the top plate and the bottom plate. The variable frequency motor and the eccentric wheel are located inside the support, with the eccentric wheel on the motor shaft of the variable frequency motor.

[0016] The sample holding mechanism consists of a sample holding bucket and a telescopic moving mechanism. The sample holding bucket has two working positions: a sample receiving position and a sample taking position. Driven by the telescopic moving mechanism, the sample holding bucket moves back and forth between the two working positions.

[0017] A method for controlling coal sample grinding using multimodal sensor fusion includes the following steps:

[0018] 1) Start the variable frequency motor, the grinding chamber starts to operate, and the raw materials used to prepare coal samples enter the grinding chamber through the feeding mechanism. Then close the sample inlet valve B.

[0019] 2) Before grinding begins, the variable frequency motor runs at a low speed, and the grinding chamber is subjected to a weak vibration force. The raw material is stirred in the grinding chamber due to the low inertia of the grinding ring and grinding block. The vibration energy at this stage is insufficient to break the raw material. During the stirring process, some of the raw material that meets the particle size requirements is discharged through the sampler into the sample container.

[0020] 3) At the start of grinding, the variable frequency motor runs at high speed, using the grinding ring and grinding blocks to crush the raw material; compressed air is introduced into the grinding chamber at set time intervals to assist in sample output; during the grinding process, the particle size of the raw material is detected in real time by a light transmittance sensor and an image sensor; by comprehensively considering the flowability index, working condition characteristic parameters and coal sample hardness characteristics, the vibration frequency, compressed air flow rate and grinding time are dynamically adjusted using a finite state machine model and a fuzzy PID controller; thus achieving dynamic adjustment of grinding process parameters.

[0021] 4) Adjust the operating frequency of the variable frequency motor and the flow rate and frequency of the compressed air entering the grinding chamber according to the real-time detection of the coal sample output rate by the weighing mechanism.

[0022] 5) Using the coal sample mass measured by the weighing mechanism, the amount of remaining raw material in the grinding chamber is automatically identified until the coal sample mass matches the raw material mass. Once all the samples are discharged, the frequency converter motor is turned off, and the coal sample grinding process ends.

[0023] The formula for the comprehensive liquidity indicator is as follows;

[0024] ;

[0025] in, comb ∈[0,1]: Normalized liquidity indicator, 0 = complete blockage, 1 = optimal liquidity;

[0026] I(t): Light transmittance sensor reading;

[0027] I max : Maximum range of light transmittance sensor;

[0028] I min : Minimum range of light transmittance sensor;

[0029] G(t): Particle density extracted by the image sensor;

[0030] G ideal The limit of particle density that an image sensor can resolve;

[0031] ω1: Light transmittance weight;

[0032] ω2: Image density weight;

[0033] ω3: Sampling rate weight;

[0034] Q out (t): Real-time sampling rate;

[0035] Q max Maximum sample size;

[0036] Q min : Minimum sample volume;

[0037] ω1+ω2+ω3=1: Modal weighting coefficient.

[0038] The operating condition characteristic parameters include the following parameters:

[0039] (1) Stickiness coefficient:

[0040] ;

[0041] Among them, P cav (t): Measured air pressure in the grinding chamber;

[0042] P atm Atmospheric pressure;

[0043] P supply Gas supply pressure;

[0044] (2) Hardness characteristic quantity:

[0045] ;

[0046] Where T(t): real-time temperature;

[0047] T amb Ambient temperature;

[0048] Δt: Temperature rise monitoring time window;

[0049] m: Current coal sample quality being processed;

[0050] The hard coal compensation mode is triggered when H(t) > 0.15.

[0051] The formula for dynamically adjusting the vibration frequency is:

[0052] ;

[0053] Constraint: 0.5 ≤ f(t) ≤ 2000 Hz;

[0054] Where, f low Pre-screening at low frequency;

[0055] fbase Reference breaking frequency;

[0056] K1: Hardness gain;

[0057] K2: Pressure coupling coefficient;

[0058] K3: Liquidity compensation coefficient;

[0059] H(t): Hardness characteristic quantity;

[0060] C p (t): viscosity coefficient;

[0061] L comb Liquidity indicator, L comb ∈[0,1].

[0062] A virtual grinding machine model was established for digital twin verification, and the grinding control process was calibrated online. The specific process is as follows:

[0063] (1) The virtual grinding machine model is established as follows:

[0064] ;

[0065] Where x: the displacement between the grinding block and the grinding ring;

[0066] m: The equivalent mass of the vibration system, i.e., the total equivalent mass of the grinding block and the eccentric wheel;

[0067] C: Damping coefficient, including energy dissipation due to friction between the grinding chamber and the coal sample, as well as air resistance;

[0068] Inertial force, the product of mass and acceleration;

[0069] Damping force is proportional to velocity;

[0070] Kx: Elastic restoring force, which is proportional to displacement;

[0071] F0sin(2πft): Periodic external force, simulating the driving force of a motor;

[0072] μ(T): Temperature-dependent friction coefficient, μ(T) = 0.3 + 0.05(T − 20);

[0073] (2) The online calibration formula for grinding parameters is as follows:

[0074] ;

[0075] Where, θ old : Current model parameter values, i.e., parameters to be calibrated;

[0076] θ new : Updated model parameter values;

[0077] Q real Actual measured performance indicators, including sample yield and particle size qualification rate;

[0078] Q sim Performance metrics for digital twin model simulation;

[0079] Simulation output Q sim The gradient with respect to the parameter θ, i.e., the sensitivity;

[0080] It executes automatically according to the set cycle, ensuring parameter synchronization.

[0081] Compared with the prior art, the beneficial effects of the present invention are:

[0082] 1) It achieves precise control of the target particle size of coking coal and coke samples, which can meet the strict requirements of relevant standards and specifications for coal sample particle size, especially the particle size requirements of coal samples used for experiments analyzing G value, Y value, b value, Gibbs freeness, etc., ensuring the accuracy of the final experimental results and providing a reliable basis for coking production and transaction settlement.

[0083] 2) The coal sample grinding mill has a high degree of automation, enabling remote control from the terminal for unmanned operation on site;

[0084] 3) Precise particle size control: Through step-by-step crushing, grinding and screening, the prepared coal samples are guaranteed to meet the particle size requirements;

[0085] 4) Achieve 100% complete sample output to avoid sample segregation;

[0086] 5) The coal sample grinder adopts a fully enclosed structure, and there is no dust spillage during the entire operation process, which eliminates coal sample loss during sample preparation and improves the working environment.

[0087] 6) The coal sample grinder has an automatic cleaning capability to prevent cross-contamination of coal samples. Attached Figure Description

[0088] Figure 1 This is a three-dimensional structural diagram of the coal sample grinding machine described in this invention.

[0089] Figure 2 This is a partial structural schematic diagram of the coal sample grinding machine described in this invention.

[0090] Figure 3 This is the state transition logic diagram of the finite state machine (FSM) control logic described in this invention.

[0091] Figure 4This is an architecture diagram of the fuzzy PID control algorithm described in this invention.

[0092] In the diagram: 1-Compartment cover; 2-Air inlet; 3-Feeding bin; 4-Feeding valve A; 5-Buffer pipe; 6-Feeding valve B; 7-Sensor assembly; 7.1-Gas pressure sensor; 7.2-Light transmittance sensor; 7.3-Image sensor; 7.4-Temperature sensor; 8-Grinding chamber; 8.1-Grinding ring; 8.2-Grinding block; 8.3-Sampling device; 9-Vibration frame; 10-Spring; 11-Support; 12-Eccentric wheel; 13-Variable frequency motor; 14-Sample discharge mechanism; 15-Sample container; 16-Telescopic moving mechanism; 17-Weighing mechanism; 18-Compressed air inlet. Detailed Implementation

[0093] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings:

[0094] like Figure 1 , Figure 2 As shown, the coal sample grinding mill with multimodal sensor fusion according to the present invention includes a feeding mechanism, a grinding mechanism, a grinding control mechanism, a vibration mechanism, a sample discharging mechanism, a sample holding mechanism, and a weighing mechanism. The grinding mechanism has a grinding chamber, the feed inlet at the top of the grinding chamber is connected to the feeding mechanism, and the sample outlet at the bottom of the grinding chamber is connected to the sample holding mechanism through the sample discharging mechanism. The sample holding mechanism is located on the weighing mechanism. The grinding chamber is located on the vibration mechanism and is a closed chamber with a compressed air inlet at the top. The grinding control mechanism includes a controller and a sensor assembly, wherein the sensor assembly includes one or more of a gas pressure sensor, a light transmittance sensor, an image sensor, and a temperature sensor. The gas pressure sensor, light transmittance sensor, image sensor, and temperature sensor are all located at the top of the grinding chamber and are respectively connected to the controller. The controller is also connected to the control terminal of the vibration mechanism.

[0095] The feeding mechanism includes a feeding hopper, a feeding valve A, a buffer pipe, and a feeding valve B; the top of the feeding hopper is equipped with a movable hopper cover, and the bottom of the feeding hopper is connected to the grinding chamber through the buffer pipe; the upper end of the buffer pipe is connected to the feeding hopper through the feeding valve A, and the lower end of the buffer pipe is connected to the grinding chamber through the material sample valve B.

[0096] The grinding chamber is equipped with a grinding ring and a grinding block; the sample dispensing mechanism includes a sample dispenser and a sample dispensing pipe, the sample dispenser is located at the bottom of the grinding chamber and is equipped with a sieve; the sample dispenser is connected to the sample holding mechanism through the sample dispensing pipe.

[0097] The vibration mechanism is an eccentric vibration mechanism, consisting of a vibration frame, a variable frequency motor, an eccentric wheel, and a support. The grinding chamber is located on the vibration frame, which is movably connected to the support. The vibration frame consists of a top plate, multiple springs, and a bottom plate, with the springs located between the top plate and the bottom plate. The variable frequency motor and the eccentric wheel are located inside the support, with the eccentric wheel on the motor shaft of the variable frequency motor.

[0098] The sample holding mechanism consists of a sample holding bucket and a telescopic moving mechanism. The sample holding bucket has two working positions: a sample receiving position and a sample taking position. Driven by the telescopic moving mechanism, the sample holding bucket moves back and forth between the two working positions.

[0099] The coal sample grinding control method based on multimodal sensor fusion described in this invention includes the following steps:

[0100] 1) Start the variable frequency motor, the grinding chamber starts to operate, and the raw materials used to prepare coal samples enter the grinding chamber through the feeding mechanism. Then close the sample inlet valve B.

[0101] 2) Before grinding begins, the variable frequency motor runs at a low speed, and the grinding chamber is subjected to a weak vibration force. The raw material is stirred in the grinding chamber due to the low inertia of the grinding ring and grinding block. The vibration energy at this stage is insufficient to break the raw material. During the stirring process, some of the raw material that meets the particle size requirements is discharged through the sampler into the sample container.

[0102] 3) At the start of grinding, the variable frequency motor runs at high speed, using the grinding ring and grinding blocks to crush the raw material; compressed air is introduced into the grinding chamber at set time intervals to assist in sample output; during the grinding process, the particle size of the raw material is detected in real time by a light transmittance sensor and an image sensor; by comprehensively considering the flowability index, working condition characteristic parameters and coal sample hardness characteristics, the vibration frequency, compressed air flow rate and grinding time are dynamically adjusted using a finite state machine model and a fuzzy PID controller; thus achieving dynamic adjustment of the grinding process parameters.

[0103] 4) Adjust the operating frequency of the variable frequency motor and the flow rate and frequency of the compressed air entering the grinding chamber according to the real-time sampling rate of the coal sample detected by the weighing mechanism.

[0104] 5) Using the coal sample mass measured by the weighing mechanism, the amount of remaining raw material in the grinding chamber is automatically identified until the coal sample mass matches the raw material mass. Once all the samples are discharged, the frequency converter motor is turned off, and the coal sample grinding process ends.

[0105] The formula for the comprehensive liquidity indicator is as follows;

[0106] ;

[0107] in, comb ∈[0,1]: Normalized liquidity indicator, 0 = complete blockage, 1 = optimal liquidity;

[0108] I(t): Light transmittance sensor reading;

[0109] I max : Maximum range of light transmittance sensor;

[0110] Imin : Minimum range of light transmittance sensor;

[0111] G(t): Particle density extracted by the image sensor;

[0112] G ideal The limit of particle density that an image sensor can resolve;

[0113] ω1: Light transmittance weight;

[0114] ω2: Image density weight;

[0115] ω3: Sampling rate weight;

[0116] Q out (t): Real-time sampling rate;

[0117] Q max Maximum sample size;

[0118] Q min : Minimum sample volume;

[0119] ω1+ω2+ω3=1: Modal weighting coefficient.

[0120] The operating condition characteristic parameters include the following parameters:

[0121] (1) Stickiness coefficient:

[0122] ;

[0123] Among them, P cav (t): Measured air pressure in the grinding chamber;

[0124] P atm Atmospheric pressure;

[0125] P supply Gas supply pressure;

[0126] (2) Hardness characteristic quantity:

[0127] ;

[0128] Where T(t): real-time temperature;

[0129] T amb Ambient temperature;

[0130] Δt: Temperature rise monitoring time window;

[0131] m: Current coal sample quality being processed;

[0132] The hard coal compensation mode is triggered when H(t) > 0.15.

[0133] The formula for dynamically adjusting the vibration frequency is:

[0134] ;

[0135] Constraint: 0.5 ≤ f(t) ≤ 2000 Hz;

[0136] Among them, f low Pre-screening at low frequency;

[0137] f base Reference breaking frequency;

[0138] K1: Hardness gain;

[0139] K2: Pressure coupling coefficient;

[0140] K3: Liquidity compensation coefficient;

[0141] H(t): Hardness characteristic quantity;

[0142] C p (t): viscosity coefficient;

[0143] L comb Liquidity indicator, L comb ∈[0,1].

[0144] A virtual grinding machine model was established for digital twin verification, and the grinding control process was calibrated online. The specific process is as follows:

[0145] (1) The virtual grinding machine model is established as follows:

[0146] ;

[0147] Where x: the displacement between the grinding block and the grinding ring;

[0148] m: The equivalent mass of the vibration system, i.e., the total equivalent mass of the grinding block and the eccentric wheel;

[0149] C: Damping coefficient, including energy dissipation due to friction between the grinding chamber and the coal sample, as well as air resistance;

[0150] Inertial force, the product of mass and acceleration;

[0151] Damping force is proportional to velocity;

[0152] Kx: Elastic restoring force, which is proportional to displacement;

[0153] F0sin(2πft): Periodic external force, simulating the driving force of a motor;

[0154] μ(T): Temperature-dependent friction coefficient, μ(T) = 0.3 + 0.05(T − 20);

[0155] (2) The online calibration formula for grinding parameters is as follows:

[0156] ;

[0157] Where, θ old : Current model parameter values, i.e., parameters to be calibrated;

[0158] θ new : Updated model parameter values;

[0159] Q real Actual measured performance indicators, including sample yield and particle size qualification rate;

[0160] Q sim Performance metrics for digital twin model simulation;

[0161] Simulation output Q sim The gradient with respect to the parameter θ, i.e., the sensitivity;

[0162] It executes automatically according to the set cycle, ensuring parameter synchronization.

[0163] To more intuitively illustrate the present invention, the embodiments of the present invention will be further described in conjunction with the examples. The following examples are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any technical solutions that can be obviously obtained by those skilled in the art within the scope of the technology disclosed in the present invention, including simple variations or equivalent substitutions, are all within the scope of protection of the present invention.

[0164] Example:

[0165] In this embodiment, the overall structure of the coal sample grinding mill with multimodal sensor fusion is as follows: Figure 1 , Figure 2 As shown, the main components include a bin cover 1, an air inlet 2, a feed bin 3, a feed valve A4, a buffer pipe 5, a feed valve B6, a sensor assembly 7, a grinding chamber 8, a vibrating frame 9, a spring 10, a support 11, an eccentric wheel 12, a variable frequency motor 13, a sample dispensing mechanism 14, a sample container 15, a telescopic moving mechanism 16, a weighing mechanism 17, etc., and also include a grinding control mechanism.

[0166] The sensor assembly 7 mainly includes a gas pressure sensor 7-1, a light transmittance sensor 7-2, an image sensor 7-3, and a temperature sensor 7-4, etc., and is centrally installed at the top of the grinding chamber 8. The grinding chamber body contains a grinding ring 8.1, a grinding block 8.2, and a sampler 8.3, etc. The sampler 8.3 is arranged below the grinding ring 8.1 and the grinding block 8.2, and is equipped with a screen to allow coal samples that meet the standard particle size requirements to pass through. For different particle size requirements (generally according to the requirements of GB / T5447, GB / T25213, GB / T479, GB / T16773 and other specifications and standards), samplers 8.3 with different screen openings are set to control the sample particle size.

[0167] In this embodiment, the grinding ring 8.1 and grinding block 8.2 in the grinding chamber 8, the vibration mechanism composed of a variable frequency motor and an eccentric wheel, and the telescopic movement mechanism all adopt conventional structures. As long as they can complete the relevant grinding, vibration, and movement functions, they are acceptable and will not be limited or described in detail here.

[0168] In this embodiment, the multimodal sensor fusion coal sample grinder establishes an AI model by detecting various parameters from multimodal sensors, thereby achieving the purpose of automatically controlling the coal sample grinding process and accurately controlling the coal sample particle size.

[0169] Coal sample particle size control is the core of this invention, especially for coking coal, which has high viscosity. When crushed to a certain fineness, its flowability decreases, leading to difficulties in discharge. This causes coal particles meeting the particle size requirements to remain in the grinding chamber, resulting in over-grinding and excessively small particle sizes, ultimately leading to a finished coal sample particle size distribution that fails to meet standards. To solve these problems, this embodiment constructs a dynamic grinding control system through multi-dimensional process parameter monitoring. Utilizing multi-modal sensor fusion, finite state machine (FSM) logic control, and fuzzy PID algorithm, dynamic optimization of the coal sample grinding process is achieved. The core modules include:

[0170] (1) Multimodal sensing layer: including gas pressure sensor, light transmittance sensor, image sensor, temperature sensor, motor frequency converter, and weighing sensor;

[0171] (2) Decision control layer: Finite state machine (FSM) + fuzzy PID controller;

[0172] (3) Execution layer: including variable frequency motor, pneumatic valves (feed valve A, feed valve B), and auxiliary heating device (configured as needed);

[0173] (4) Verification layer: Digital twin simulation platform + reinforcement learning calibration module.

[0174] The deep learning algorithms used in the grinding process include, but are not limited to, multimodal data fusion and algorithm models constructed from variables such as vibration spectrum, cavity pressure fluctuation, light transmittance, real-time images, temperature gradient, real-time sample yield, and air pressure damping. Details are as follows:

[0175] I. Multimodal data fusion and algorithm model;

[0176] 1) The comprehensive liquidity index is calculated using the following formula:

[0177] ;

[0178] 2) The operating condition characteristic parameters are calculated using the following formula:

[0179] a. Viscosity coefficient:

[0180] ;

[0181] b. Hardness characteristic quantity:

[0182] ;

[0183] II. Finite State Machine (FSM) Control Logic;

[0184] 1) The state space is defined as shown in the table below:

[0185] S0 Initial stirring 5Hz low-frequency vibration + sinusoidal disturbance S1 Normal grinding Fuzzy PID dynamic frequency modulation S2 High hardness compensation Frequency enhancement + temperature measurement assistance S3 Blockage Treatment Airflow cleaning S4 Safe shutdown Emergency braking + alarm

[0186] 2) The state transition logic diagram is as follows: Figure 3 As shown in the figure:

[0187] comb ∈[0,1]: Normalized liquidity index;

[0188] H: Hardness characteristic quantity, the hardness characteristic quantity at a certain moment is denoted as H(t);

[0189] AnyState: Any state;

[0190] T > 120℃: When the temperature is > 120℃;

[0191] df / dt > 500Hz / s: When the rate of increase of the output vibration frequency is > 500Hz / s;

[0192] C_ P C represents the viscosity coefficient, which is denoted as C at a given moment. p (t);

[0193] Q_ OUT The sample output and the viscosity coefficient at a certain moment are denoted as Q. out (t);

[0194] When the coal sample grinder is working, initial agitation is performed in stage S0, based on the fluidity index. comb And correlate with hardness characteristic; after entering the grinding stage S1, when the viscosity coefficient C_ P A value >0.7 indicates blockage, and the process proceeds to S3 for viscosity treatment; while when the hardness characteristic value H >0.15 and is associated with the sample volume Q_ OUT When the temperature drops below the set value, a high-hardness compensation action is activated to enhance the output and maintain the output sample for a certain period of time before returning to state S1. This process is repeated. During operation, the temperature is continuously monitored, and a safety shutdown is initiated when the temperature exceeds the limit (120℃) or the vibration frequency reaches the set value (500Hz / s).

[0195] 3) Examples of state actions are as follows:

[0196] defS1_control():

[0197] # Calculate target frequency using fuzzy PID

[0198] f_target=fuzzy_pid(e=Q_set-Q_real,ec=dQ / dt)

[0199] #Air Pressure Compensation

[0200] valve_open = 80% - 20% * C_p(t)

[0201] #Frequency safety limit

[0202] returnclamp(f_target,200Hz,1800Hz)

[0203] III. Fuzzy PID control algorithm;

[0204] 1) Fuzzy PID control architecture diagram as follows Figure 4 As shown in the figure:

[0205] e(t): Control system error;

[0206] ec(t): Rate of change of error;

[0207] Fuzzify: blurs the precise value of the input;

[0208] Defuzzify: Converts fuzzy variables into precise values;

[0209] ΔKp, ΔKi, and ΔKd are output adjustment values: where ΔKp is the proportional coefficient adjustment value, ΔKi is the integral coefficient adjustment value, and ΔKd is the derivative coefficient adjustment value.

[0210] System working process:

[0211] Collect the accurate values of e(t) and e_c(t) in real time.

[0212] Convert the accurate values into membership degrees of fuzzy sets through the Fuzzify module.

[0213] The Inference module generates fuzzy outputs according to the rules in the Fuzzy Rule Base.

[0214] The Defuzzify module converts the fuzzy outputs into accurate PID parameter adjustment amounts (△Kp, ΔKi, ΔKd).

[0215] Optimize the PID controller in real time according to the adjustment amounts to improve the system response speed and stability.

[0216] 2) The parameter adaptation rules are as follows:

[0217] ;

[0218] In the formula, e(t): control error;

[0219] K P (t): proportional gain;

[0220] T i (t): integral gain;

[0221] T d (t): derivative gain;

[0222] Q out (t): sample quality per unit time;

[0223] e -0.1t : time decay term;

[0224] H(t): hardness characteristic quantity.

[0225] IV. Sample volume detection and closed-loop control;

[0226] 1) The sample rate calculation model is

[0227] ;

[0228] Among them, Q out (t): sample quality per unit time;

[0229] N: number of discrete time points used to calculate the average value;

[0230] w(tk): at time point t k At that time, the cumulative sample mass measured by the weighing sensor;

[0231] w(t k -1 ): At time point t k -1 At that time, the cumulative sample mass measured by the weighing sensor;

[0232] t k : point in time;

[0233] t k -1 :t k A point in time preceding a given point in time.

[0234] By using the sliding window averaging method, the instantaneous sampling rates (Δw / Δt) of N adjacent time windows are averaged to obtain a more stable real-time sampling rate Qout(t).

[0235] 2) The particle size-yield joint observation model is as follows:

[0236] ;

[0237] In the formula, Comprehensive efficiency indicators;

[0238] Q out (t): Sample mass per unit time;

[0239] f(t): Output vibration frequency;

[0240] d target Target granularity;

[0241] d actual (t): The actual average particle size of the coal sample.

[0242] Decision threshold:

[0243] Γ>1.2Γ>1.2 → Reduce frequency (e.g., reduce by 5%);

[0244] Γ<0.6Γ<0.6 → Start self-test program;

[0245] V. Equation for dynamic adjustment of vibration frequency;

[0246] ;

[0247] Constraint: 0.5 ≤ f(t) ≤ 2000 Hz;

[0248] The parameters are explained below:

[0249] flow Pre-screening at low frequencies, such as 5Hz;

[0250] f base : Reference breaking frequency, such as 200Hz;

[0251] K1: Hardness gain, such as 50Hz / kg℃;

[0252] K2: Pressure coupling coefficient, such as 0.8;

[0253] K3: Liquidity compensation coefficient, such as 1.2;

[0254] L comb Liquidity indicator, L comb ∈[0,1];

[0255] H(t): Hardness characteristic quantity;

[0256] C p (t): viscosity coefficient.

[0257] V. Digital Twin Verification Framework;

[0258] 1) The virtual grinding machine model is

[0259] ;

[0260] Where, x: displacement, the amount of displacement between the grinding block and the grinding ring;

[0261] m: Equivalent mass of the vibration system (total equivalent mass of moving parts such as grinding blocks and eccentric wheels);

[0262] C: Damping coefficient (energy dissipation due to friction between the grinding chamber and the coal sample, air resistance, etc.);

[0263] Inertial force (mass × acceleration);

[0264] Damping force (proportional to velocity);

[0265] Kx: Elastic restoring force (proportional to displacement);

[0266] F0sin(2πft): Periodic external force, simulating the driving force of a motor;

[0267] μ(T): Temperature-dependent friction coefficient, μ(T) = 0.3 + 0.05(T − 20).

[0268] 2) The online calibration formula for parameters is:

[0269] ;

[0270] θold : Current model parameter values ​​(parameters to be calibrated);

[0271] θ new : Updated model parameter values;

[0272] Q real Actual measured performance indicators (such as sample yield and particle size qualification rate);

[0273] Q sim Performance metrics for digital twin model simulation;

[0274] Simulation output Q sim The gradient (sensitivity) with respect to parameter θ;

[0275] Parameter synchronization is performed automatically at regular intervals (e.g., every 24 hours).

[0276] After the raw materials to be processed into coal samples enter the feed hopper 3 and the hopper cover is closed, the feed valve A is opened after the device is initialized and reaches the working conditions. The raw materials flow through the buffer pipe and feed valve B to reach the grinding chamber. In the grinding chamber, the raw materials are identified and ground under AI control. During the grinding process, coal samples that meet the particle size requirements enter the sample container through the sample discharge mechanism. After the variable frequency motor starts, it generates vibration through the eccentric wheel, causing the grinding chamber to generate crushing force using the grinding ring and grinding blocks during the vibration process, which is controlled by the output of the dynamic adjustment equation. After all the raw materials have been ground in the grinding chamber and entered the sample container (weighed by the weighing mechanism and calculated by the control system), the sample container is sent out through the telescopic mechanism, and the coal sample preparation process is completed.

[0277] If the amount of coal sample after grinding is less than the amount of raw material added, it indicates that there is still a small amount of residue in the grinding chamber. At this time, the control system calculates the amount of residue and starts the cleaning mode.

[0278] After grinding begins, the system uses the detection values ​​from light transmittance sensors and image sensors, as well as the sample yield and other variables calculated in real time by the control system based on the detection values ​​from the weighing mechanism. It employs models based on comprehensive flowability indices, operating condition characteristic parameters, and coal sample hardness characteristics, following a finite state machine (FSM) control logic, transitioning between S0, S1, S2, S3, and S4. In simpler terms, if the coal sample has poor flowability based on the agitation, the compressed air volume is increased; conversely, if it has good flowability, the compressed air volume is decreased. The variable frequency motor initially operates at a low speed, subjecting the grinding chamber to weak vibration. The raw material is agitated by the grinding rings and grinding blocks due to their low inertia. At this stage, the vibration energy is insufficient to crush the raw material. Under initial agitation, the portion of the raw material with the required particle size can be discharged through the sampler into the sample container. Then, based on the real-time rate of change in the feed rate detected by the weighing mechanism, the speed of the variable frequency motor is increased to crush the remaining larger particles. This method allows for the sorting of particles in the raw material that meet the coal sample size requirements before crushing, avoiding over-grinding and achieving step-by-step crushing.

[0279] To accelerate sample output, compressed air is used for auxiliary sampling. Positive-pressure compressed air is introduced into the grinding chamber at regular intervals through the compressed air inlet. By controlling the operating frequency of the variable frequency motor and the frequency of compressed air injection, the particle size of the raw material is crushed to the target requirements. During the grinding process, the image sensor and the measured total mass of the coal sample are used to automatically determine whether there is any residual material in the grinding chamber and the amount of residual material. The machine stops only after all samples have been discharged, with the goal of stopping only after all coal samples have been discharged.

[0280] It should be noted that different coal samples have different hardness, viscosity, and moisture content. Therefore, the grinding time, i.e., the grinding cycle, is entirely determined by the properties of the coal sample itself. Thus, the determination of whether the grinding is finally complete can only be made after the corresponding model conditions are met and the sample has been completely extracted.

[0281] During the grinding process, from the moment the feed hopper is closed, the entire grinding process is in a completely enclosed state. This is not only to prevent dust from spilling out during the grinding process and to improve the working environment, but more importantly, to prevent the loss of coal samples after grinding.

[0282] The coal sample grinder of this invention is equipped with a self-cleaning mode. After grinding is completed, the system will automatically calculate the amount of coal sample loss and the amount of raw material residue in the grinding chamber, and autonomously decide whether to introduce a cleaning mode. When the cleaning mode is activated, the vacuum dust collector connected to the discharge mechanism generates negative pressure for suction and discharge, allowing external air to enter through the air inlet and pass through the raw material flow channel to clean the residual dust on the inner surface of the equipment that comes into contact with the raw material.

[0283] The multimodal sensing fusion parameters of the coal sample grinding mill described in this invention have high real-time requirements and a large computational load, and are not guided by a single algorithm. Therefore, employing a self-learning optimization algorithm is particularly important, and high-performance edge computing is preferred for computation. With the development of artificial intelligence technology, the equipment can be miniaturized and individualized, and after training, it can achieve targeted specialization. Furthermore, it can be connected to AI cloud computing to achieve learning-based data computation. This makes the equipment more intelligent and allows for continuous optimization.

[0284] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A coal sample grinding control method based on multimodal sensor fusion, implemented using a coal sample grinder based on multimodal sensor fusion, characterized in that, The multimodal sensing fusion coal sample grinding machine includes a feeding mechanism, a grinding mechanism, a grinding control mechanism, a vibration mechanism, a sample discharge mechanism, a sample holding mechanism, and a weighing mechanism; the grinding mechanism is provided with a grinding chamber, the feed inlet at the top of the grinding chamber is connected to the feeding mechanism, the sample outlet at the bottom of the grinding chamber is connected to the sample holding mechanism through the sample discharge mechanism, and the sample holding mechanism is located on the weighing mechanism; The grinding chamber is located on the vibration mechanism. The grinding chamber is a closed chamber with a compressed air inlet at the top. The grinding control mechanism includes a controller and a sensor assembly, wherein the sensor assembly includes one or more of a gas pressure sensor, a light transmittance sensor, an image sensor, and a temperature sensor; the gas pressure sensor, light transmittance sensor, image sensor, and temperature sensor are all located at the top of the grinding chamber and are respectively connected to the controller; the controller is also connected to the control end of the vibration mechanism. The multimodal sensor fusion-based coal sample grinding control method includes the following steps: 1) Start the variable frequency motor, the grinding chamber starts to operate, and the raw materials used to prepare coal samples enter the grinding chamber through the feeding mechanism. Then close the sample inlet valve B. 2) Before grinding begins, the variable frequency motor runs at a low speed, and the grinding chamber is subjected to a weak vibration force. The raw material is stirred in the grinding chamber due to the low inertia of the grinding ring and grinding block. The vibration energy at this stage is insufficient to break the raw material. During the stirring process, a portion of the raw material that meets the particle size requirements is discharged through the sampler into the sample container; 3) At the start of grinding, the variable frequency motor runs at high speed, using the grinding ring and grinding blocks to crush the raw materials; Compressed air is introduced into the grinding chamber at set time intervals to assist in sample output; During the grinding process, the particle size of the raw material is detected in real time by light transmittance sensor and image sensor; the vibration frequency, compressed air flow rate and grinding time are dynamically adjusted by comprehensively considering flowability index, working condition characteristic parameters and coal sample hardness characteristics, using finite state machine model and fuzzy PID controller. To achieve dynamic adjustment of grinding process parameters; The formula for the comprehensive liquidity indicator is as follows; in, comb ∈[0,1]: Normalized liquidity indicator, 0 = complete blockage, 1 = optimal liquidity; I(t): Light transmittance sensor reading; I max : Maximum range of light transmittance sensor; I min : Minimum range of light transmittance sensor; G(t): Particle density extracted by the image sensor; G ideal The limit of particle density that an image sensor can resolve; ω1: Light transmittance weight; ω2: Image density weight; ω3: Sampling rate weight; Q out (t): Real-time sampling rate; Q max Maximum sample size; Q min : Minimum output quantity; ω1+ω2+ω3=1: Modal weighting coefficient; 4) Adjust the operating frequency of the variable frequency motor and the flow rate and frequency of the compressed air entering the grinding chamber according to the real-time detection of the coal sample output rate by the weighing mechanism. 5) Using the coal sample mass measured by the weighing mechanism, the amount of remaining raw material in the grinding chamber is automatically identified until the coal sample mass matches the raw material mass. Once all the samples are discharged, the frequency converter motor is turned off, and the coal sample grinding process ends.

2. The coal sample grinding control method based on multimodal sensor fusion according to claim 1, characterized in that, The feeding mechanism includes a feeding hopper, a feeding valve A, a buffer pipe, and a feeding valve B; the top of the feeding hopper is equipped with a movable hopper cover, and the bottom of the feeding hopper is connected to the grinding chamber through the buffer pipe; the upper end of the buffer pipe is connected to the feeding hopper through the feeding valve A, and the lower end of the buffer pipe is connected to the grinding chamber through the material sample valve B.

3. The coal sample grinding control method based on multimodal sensor fusion according to claim 1, characterized in that, The grinding chamber is equipped with a grinding ring and a grinding block; the sample dispensing mechanism includes a sample dispenser and a sample dispensing tube, the sample dispenser is located at the bottom of the grinding chamber, and the sample dispenser is equipped with a sieve. The sample dispenser is connected to the sample holding mechanism via a sample dispensing tube.

4. The coal sample grinding control method based on multimodal sensor fusion according to claim 1, characterized in that, The vibration mechanism is an eccentric vibration mechanism, consisting of a vibration frame, a variable frequency motor, an eccentric wheel, and a support. The grinding chamber is located on the vibration frame, which is movably connected to the support. The vibration frame consists of a top plate, multiple springs, and a bottom plate, with the springs located between the top plate and the bottom plate. The variable frequency motor and the eccentric wheel are located inside the support, with the eccentric wheel on the motor shaft of the variable frequency motor.

5. The coal sample grinding control method based on multimodal sensor fusion according to claim 1, characterized in that, The sample holding mechanism consists of a sample holding bucket and a telescopic moving mechanism. The sample holding bucket has two working positions: a sample receiving position and a sample taking position. Driven by the telescopic moving mechanism, the sample holding bucket moves back and forth between the two working positions.

6. The coal sample grinding control method based on multimodal sensor fusion according to claim 1, characterized in that, The operating condition characteristic parameters include the following parameters: (1) Stickiness coefficient: Among them, P cav (t): Measured air pressure in the grinding chamber; P atm Atmospheric pressure; P supply Gas supply pressure; (2) Hardness characteristic quantity: Where T(t): real-time temperature; T amb Ambient temperature; Δt: Temperature rise monitoring time window; m: Current coal sample quality being processed; The hard coal compensation mode is triggered when H(t) > 0.

15.

7. The coal sample grinding control method based on multimodal sensor fusion according to claim 1, characterized in that, The formula for dynamically adjusting the vibration frequency is: Constraint: 0.5 ≤ f(t) ≤ 2000 Hz; Among them, f low Pre-screening at low frequency; f base Reference breaking frequency; K1: Hardness gain; K2: Pressure coupling coefficient; K3: Liquidity compensation coefficient; H(t): Hardness characteristic quantity; C p (t): viscosity coefficient; L comb Liquidity indicator, L comb ∈[0,1].

8. The coal sample grinding control method based on multimodal sensor fusion according to claim 1, characterized in that, A virtual grinding machine model was established for digital twin verification, and the grinding control process was calibrated online. The specific process is as follows: (1) The virtual grinding machine model is established as follows: Where x: the displacement between the grinding block and the grinding ring; m: The equivalent mass of the vibration system, i.e., the total equivalent mass of the grinding block and the eccentric wheel; C: Damping coefficient, including energy dissipation due to friction between the grinding chamber and the coal sample, as well as air resistance; Inertial force, the product of mass and acceleration; Damping force is proportional to velocity; Kx: Elastic restoring force, which is proportional to displacement; F0sin(2πft): Periodic external force, simulating the driving force of a motor; μ(T): Temperature-dependent friction coefficient, μ(T) = 0.3 + 0.05(T − 20); (2) The online calibration formula for grinding parameters is as follows: Where, θ old : Current model parameter values, i.e., parameters to be calibrated; θ new : Updated model parameter values; Q real Actual measured performance indicators, including sample yield and particle size qualification rate; Q sim Performance metrics for digital twin model simulation; Simulation output Q sim The gradient with respect to the parameter θ, i.e., the sensitivity; It executes automatically according to the set cycle, ensuring parameter synchronization.

Images