Ball mill for aluminum silicate production

Abstract

The invention relates to the technical field of ball mills, and particularly discloses a ball mill for aluminum silicate production, which comprises a rack, an outer cylinder, a driving assembly, a discharge pipe, a feed pipe and a control interaction unit, and the two sides of the top of the rack are fixedly connected with a first bearing seat and a second bearing seat respectively; the outer cylinder is rotatably connected to the rack through a first bearing seat and a second bearing seat, an inner cylinder is fixedly connected in the outer cylinder, a plurality of steel balls are arranged in the inner cylinder, and counterweight adjusting assemblies are arranged in the steel balls; according to the invention, the inertia and impact force of the steel ball can be changed through the arrangement of the counterweight adjusting assembly, so that the ball mill can flexibly adjust the working state of the steel ball according to the properties and grinding requirements of different materials.

Description

Technical Field

[0001] The invention relates to the technical field of ball mills, in particular to a ball mill for producing aluminum silicate. Background Art

[0002] Aluminum silicate, an important inorganic non-metallic material widely used in refractories and thermal insulation, is produced from raw materials such as bauxite, kaolin, and quartz sand, all of which have high hardness. Product performance is closely related to the raw material particle size distribution. For example, aluminum silicate powder used in refractory fibers typically requires an average particle size of 5-10μm or even finer. Industrial production also requires a continuous process of "crushing-grinding-mixing-molding-sintering," placing high demands on the grinding equipment's production capacity and particle size control capabilities. Overflow ball mills, based on the principle of "cylinder rotation + steel ball media grinding," use the rotating cylinder to drive the steel balls to produce a holistic crushing effect, including throwing, grinding, and impact. Their unique discharge structure, lacking a grate and relying on the slurry's own weight to overflow, allows the fine powder to remain in the cylinder longer, enabling more thorough grinding and a concentrated product particle size distribution. This effectively meets the needs of aluminum silicate fine grinding and stringent particle size requirements, making them suitable for industrial continuous production processes. In the prior art, the steel balls in the ball mill are all solid, and the working state of the steel balls cannot be adjusted according to the different medium materials. Summary of the Invention

[0003] Aiming at the technical problem in the prior art that the steel balls in the ball mill are all solid and the working state of the steel balls cannot be adjusted according to the different medium materials, the present invention provides a ball mill for aluminum silicate production.

[0004] The technical solution adopted by the present invention is: a ball mill for aluminum silicate production, the ball mill comprising: a frame, an outer cylinder, a driving assembly, a discharge pipe, a feed pipe, and a control interaction unit, the top two sides of the frame are respectively fixedly connected with a first bearing seat and a second bearing seat, the outer cylinder is rotatably connected to the frame through the first bearing seat and the second bearing seat, the outer cylinder is fixedly connected with an inner cylinder, a plurality of steel balls are provided in the inner cylinder, a counterweight adjustment assembly is provided in the steel balls, the driving assembly is located on the frame, and is used to drive the outer cylinder to rotate, the discharge pipe is fixedly connected to the outside of the first bearing seat, and a discharge port is provided at the side end of the outer cylinder corresponding to the position of the discharge pipe, the feed pipe is fixedly connected to the outside of the second bearing seat, and a feed port is provided at the position of the feed pipe corresponding to the side end of the outer cylinder, and the control interaction unit monitors and adjusts the operating status of the ball mill and performs fault diagnosis.

[0005] The present invention is further configured as follows: the drive assembly includes a first motor, a gear box, a first gear and a ring gear, the ring gear is fixedly connected to the outer cylinder, the first gear is rotatably connected to the frame, the gear box is fixedly connected to the frame, the second gear and the third gear are rotatably connected inside the gear box, the first motor is fixedly connected to the frame, the output end of the first motor is fixedly connected to the second gear, the third gear is coaxially fixedly connected to a synchronous shaft, the synchronous shaft is coaxially fixedly connected to the first gear, and the first gear is meshed with the ring gear.

[0006] The present invention is further configured as follows: the outside of the discharge pipe is fixedly connected to a first fixing frame, the outside of the feed pipe is fixedly connected to a second fixing frame, the outside of the first fixing frame is fixedly connected to a second motor, the outside of the second fixing frame is fixedly connected to a third motor, the output end of the second motor is fixedly connected to a first spiral conveying rod, the output end of the third motor is fixedly connected to a second spiral conveying rod, a feed hopper is provided above the feed pipe, the output end of the feed hopper is fixedly connected to the feed pipe, inlets and outlets are provided in the outer cylinder and the inner cylinder, sealing covers are provided in the inlets and outlets, the sealing covers are buckled in the inlets and outlets, and long bolts are passed through the inlets and outlets and the sealing covers.

[0007] The present invention is further configured such that a spiral blade is provided on the outside of the steel ball, an inner ball is provided inside the steel ball, the outside of the inner ball is fixedly connected to a connecting rod, the connecting rod is fixedly connected to the steel ball, and the inner ball is provided with multiple sets of sliding holes, and the counterweight adjustment assembly includes a first threaded rod, a second threaded rod and a third threaded rod rotatably connected to the steel ball, the bottom of the first threaded rod is fixedly connected to a third bevel gear, the outside of the second threaded rod is fixedly connected to the second bevel gear, one end of the third threaded rod is fixedly connected to the first bevel gear, the third bevel gear and the second bevel gear are meshed with each other, and the second bevel gear and the first bevel gear are meshed with each other, the first threaded rod is a bidirectional threaded rod with opposite thread groove directions, the outsides of the first threaded rod, the second threaded rod and the third threaded rod are all threadedly connected to a counterweight ball, and the counterweight balls on the outside of the first threaded rod are provided with two, and multiple sets of guide rods are also fixedly connected to the steel ball, and the guide rods pass through the counterweight ball.

[0008] The present invention is further configured such that the outsides of the first threaded rod, the second threaded rod and the third threaded rod corresponding to the outside of the steel ball are all fixedly connected to a fixed block, wherein a positioning groove is provided in one of the fixed blocks, one end of the first threaded rod extends into the positioning groove and is fixedly connected to a rotating block, a positioning plate is provided on the outside of the positioning groove, a positioning block is fixedly connected to the bottom of the positioning plate, a positioning bolt is threadedly connected in the positioning plate, the positioning bolt is threadedly connected to the fixed block, and a hexagonal groove is provided at the position of the positioning block corresponding to the outside of the rotating block.

[0009] The present invention is further configured such that the control interaction unit includes a ball mill operation status monitoring module, a ball mill grinding process mathematical model establishment module, an objective function optimization module, a parameter optimization module based on an intelligent algorithm, a real-time control and adjustment module, a fault diagnosis and early warning module, a ball mill operation data recording module, a human-computer interaction interface module, and a system integration module.

[0010] The present invention is further configured such that the ball mill operation status monitoring module includes a vibration sensor, a temperature sensor, and a rotation speed sensor;

[0011] The vibration sensor is mounted on the first bearing seat or the second bearing seat, the temperature sensor is mounted on the outside of the first motor, the first bearing seat or the second bearing seat, and the speed sensor is located outside the frame to measure the speed of the outer cylinder. The vibration signal is represented in the time domain as V(t), the temperature signal collected by the temperature sensor is T(t), and the speed collected by the speed sensor is n(t), where t represents time. The raw signals collected by the sensors are preprocessed. In addition to filtering, noise reduction, and normalization, multi-scale data fusion technology is introduced. The vibration signal V(t) and the temperature signal T(t) are decomposed at multiple scales, and wavelet transform is used to obtain signal characteristics at different scales.

[0012] For the vibration signal V(t), the low-frequency component A is obtained after j-layer wavelet decomposition j (t) and high frequency component D j,k (t)(k=1,2,…,j), extract features of components at different scales and construct a multi-scale feature vector V ms :V ms =[mean(A j ),std(A j ),mean(D j,k ),std(D j,k )]; where mean represents the mean calculation function, and std represents the standard deviation calculation function;

[0013] The temperature signal T(t) is normalized and the formula is: Among them, T norm (t) is the normalized temperature value, T min and T max are the minimum and maximum values ​​of the temperature signal during the monitoring period; a dynamic noise compensation algorithm is introduced to perform adaptive compensation for noise characteristics under different working conditions, and the noise energy ratio ρ is calculated based on the spectrum analysis of the vibration signal. noise ; Where V(f) is the frequency domain representation of the vibration signal, F noise is the frequency range where the noise is mainly distributed, F totalFor the entire spectrum range; develop a spatiotemporal feature fusion network to simultaneously process spatially distributed multi-sensor data and time series features, and construct a spatiotemporal feature tensor X st ;X st =[X s1 (t),X s2 (t),…,X sn (t)]; where X si (t) represents the feature vector of the i-th sensor at time t, and the spatiotemporal features are extracted through the spatiotemporal convolutional network;

[0014]

[0015] The present invention is further configured such that the mathematical model building module of the ball mill grinding process includes introducing a nonlinear dynamic model; assuming that the pulverization process of the material in the ball mill conforms to the improved secondary pulverization dynamic equation: Among them, X i is the content of particles with a particle size of i in the material, k i is the first-order crushing rate constant of particles with a particle size of i, k' i is the secondary crushing rate constant; a grinding efficiency prediction model based on deep learning is established, and a convolutional neural network is used to learn a large amount of operating data. The input is the sensor data and operating parameters after feature extraction, and the output is the grinding efficiency prediction value. Build a dynamic energy consumption prediction model, the actual output power P of the first motor out The calculation is as follows:

[0016] P out =P in ·η motor ·η trans Among them, P in is the input power, η motor is the first motor efficiency, η trans is the transmission efficiency; the first motor efficiency η motor and transmission efficiency η trans Calculated by the following formulas:

[0017]

[0018] Among them, P rated is the rated power of the first motor, k1, k2, k3 are empirical coefficients, and n0 is the optimal speed point;

[0019] Combining quantum computing with neural networks, the output of the quantum layer is calculated as: q =σ(W q ·U θ |0>+β); where Wq is the quantum weight matrix, U θ To parameterize the quantum circuit, |0> is the initial quantum state, β is the bias, σ is the activation function, and the quantum layer output is combined with the classical layer: Among them, W c is the classic weight matrix, b is the bias, and f is the output function. A multi-scale grinding dynamics model is constructed. Considering the differences in the crushing behaviors of materials with different particle sizes, the materials are divided into K intervals according to the particle size. The crushing dynamics equation for each interval is: Among them, X k is the material content in the kth particle size interval, k k and k′ k is the crushing rate constant in this interval, k jk is the conversion rate constant from the jth interval to the kth interval; the objective function optimization module introduces equipment wear W as the optimization target based on the original maximization of grinding efficiency η and minimization of energy consumption E, and establishes a three-objective optimization function: Among them, α, β and γ are weight coefficients, and α+β+γ=1, η max is the maximum grinding efficiency, E min is the minimum energy consumption, W max is the maximum allowable equipment wear;

[0020] Equipment wear W is based on the wear of the steel ball m w And the inner cylinder wear thickness δ is calculated:

[0021] Among them, λ1 and λ2 are weight coefficients, m w,max is the maximum allowable wear of the steel ball, δ max is the maximum allowable wear thickness of the inner cylinder; the energy consumption E is related to the power P and running time t of the ball mill, and the calculation formula is:

[0022] The power P of the ball mill is calculated by the following empirical formula:

[0023] P=k p (M+ρV m )ω 2 R; where k p is the power coefficient; M is the loading capacity; ρ is the density of the material; V m is the volume of the material in the inner cylinder; ω is the angular velocity of the ball mill, R is the radius of the inner cylinder; the quality stability index Q is introduced s , measure the degree of fluctuation of product quality, calculate the standard deviation of particle size distribution σ through real-time monitoring of product particle size distribution: Q s = Where σ0 is the target standard deviation; the quality stability index is added to the optimization objective function: Among them, δ is the mass stability weight coefficient;

[0024] A dynamic weight adaptive optimization algorithm is proposed to automatically adjust the objective function weight according to the operating status. The weight adjustment formula is:

[0025]

[0026]

[0027] Among them, η α ,η β ,η γ ,η δ is the learning rate, is the partial derivative of the objective function with respect to each weight; the carbon emission index V is introduced to establish the green optimization objective function: Among them, C min is the minimum carbon emission target, ∈ is the carbon emission weight coefficient, and the carbon emission C is calculated as: C = Σ i E i ·f i ; Among them, E i is the consumption of the i-th energy, f i is the carbon emission factor of the i-th energy source; wherein, the parameter optimization module based on the intelligent algorithm uses the improved quantum particle swarm optimization algorithm to optimize the operating parameters of the ball mill;

[0028] In the quantum particle swarm optimization algorithm, quantum behavior and chaotic search strategy are introduced, and the position update formula of each particle in the quantum space is:

[0029] in, is the quantum rotating gate parameter, which is dynamically adjusted through chaotic mapping; δ is the step size factor; L(α) is the quantum Levy flight path; p id (t) is the individual optimal position of the i-th particle; g d (t) is the global optimal position of the entire particle swarm;

[0030] An adaptive learning rate optimization algorithm is proposed to dynamically adjust the learning parameters of the quantum particle swarm. The learning rate η(t) changes dynamically with the number of iterations t:

[0031] Among them, η0 is the initial learning rate, η ∞ is the final learning rate, T is the decay period;

[0032] A population diversity maintenance mechanism is introduced to calculate the average Euclidean distance D between particles: Where N is the number of particles, x i and x j is the particle position vector;

[0033] Develop a multi-agent collaborative optimization framework that uses communication and collaboration between agents to: Among them, J i is the objective function of the ith agent, u i Its control input, u -i is the control input for other agents, and each agent uses a deep deterministic policy gradient algorithm to make decisions:

[0034] μ * (s)=argmax μ Q π (s,μ(s)); where μ * is the optimal strategy, Q π is the action value function, s is the state, and a hybrid optimization method of quantum genetic algorithm and particle swarm algorithm is proposed. The update formula of quantum chromosome is: θ i (t+1)=θ i (t)+Δθ·sin(2πft+φ); where, θ i is the phase angle of the i-th quantum bit, Δθ is the phase adjustment amount, f is the frequency, φ is the initial phase, and chromosome evolution is achieved through quantum rotation gate operation:

[0035] Among them, the real-time control and adjustment module controls the feed speed, rotation speed and inner drum loading of the ball mill in real time according to the obtained optimal operating parameter combination;

[0036] Adaptive fuzzy PID control algorithm is introduced to adjust the outer cylinder speed;

[0037] Assume that the actual speed of the outer cylinder is n actual , the target speed is n target , the adjustment control quantity Δu of the frequency converter is calculated as follows: First, according to the speed deviation e=n target -n actual and the rate of change of deviation Adjusting PID control parameter k using fuzzy inference rules p 、k i and k d :

[0038] k p =k p0 +Δk p (e,ec);

[0039] k i =k i0 +Δki (e,ec);

[0040] k d =k d0 +Δk d (e,ec);

[0041] Among them, k p0 、k i0 and k d0 is the initial PID control parameter, Δk p (e,ec),Δk i (e,ec) and Δk d (e,ec) is the parameter adjustment amount obtained by fuzzy reasoning; then, the adjustment control amount Δu is calculated; For the control of feeding speed, the adaptive fuzzy PID control algorithm is also used. According to the real-time feedback of grinding efficiency, the feeding speed of the second screw conveyor rod is adjusted to keep the feeding speed at the optimal value.

[0042] Design a multivariable collaborative control strategy, consider the coupling effect between each control parameter, and calculate the coupling degree between parameters by establishing the coupling coefficient matrix C: Among them, y i is the i-th output variable, u j is the jth control variable; dynamically adjust the control parameters according to the coupling coefficient matrix to improve the system response speed and stability; develop a predictive control compensation algorithm to predict the system state at the future moment; use the long short-term memory network to predict the system output in the next k steps based on the current state and historical data Among them, y is the system output, u is the control input, n and m are the lengths of historical data, and the control parameters are adjusted in advance according to the prediction results;

[0043] A dynamic adaptive control architecture is proposed to automatically switch the control strategy according to the system operation status, and the system operation status index S is defined as:

[0044] Among them, s i is the i-th state feature, w i is the corresponding weight, and switches between different control strategies according to the state index S:

[0045]

[0046] Among them, u i is the i-th control strategy, R i For the corresponding state area; develop a distributed cooperative control algorithm to assign control tasks to multiple controllers, each controller is responsible for local control and exchanges information through a communication network: Among them, K i is the control law of the i-th controller, e i is the local error, T ij is the communication weight, is the predicted output of the j-th controller.

[0047] The present invention is further configured such that the fault diagnosis and early warning module establishes a fault diagnosis model based on transfer learning;

[0048] First, a deep neural network model is pre-trained on a large amount of general mechanical equipment fault data, and then fine-tuned for the specific fault data of a ball mill. The collected sensor data and extracted feature parameters are used as input, and the trained fault diagnosis model is used to determine whether the ball mill has a fault and the type of fault. An improved convolutional long short-term memory network is used as the main structure of the fault diagnosis model, which can simultaneously process the spatial and temporal features of time series data.

[0049] The output of the model is the failure probability vector P = [p1, p2, ..., p m ], where p i It represents the probability of the i-th fault of the ball mill; when the maximum probability p max = When max(P) exceeds the set threshold τ, a fault warning signal is issued and the fault type and cause are provided;

[0050] Develop a fault evolution prediction model to predict fault development trends based on the Hidden Markov Model. Train the HMM using historical fault data to obtain the state transition probability matrix A and observation probability matrix B:

[0051] A ij =P(q t+1 =j|q t =i);

[0052] B jk =P(o t =k|q t =j);

[0053] Among them, q t is the hidden state at time t, o t is the observation value at time t; the Viterbi algorithm is used to calculate the state sequence and predict the fault evolution path and remaining life; a multi-dimensional fault feature fusion model is constructed to fuse the time domain, frequency domain and time-frequency domain features, and the essential representation of each domain feature is extracted through the autoencoder, and then feature splicing is performed: F fusion =[AE time (F time ),AE freq (F freq ),AEtimefreq (F timefreq )]; where F time 、F freq 、F timefreq They are time domain, frequency domain and time-frequency domain features, AE time AE freq AE timefreq is the corresponding autoencoder;

[0054] Design a quantum entanglement fault diagnosis algorithm and use the principle of quantum entanglement to improve the distinguishability of fault characteristics; define the quantum feature vector φ: Among them, |f i > is the i-th fault characteristic state, c i is the corresponding probability amplitude; the fault diagnosis result is obtained through quantum measurement: P(k)=|<φ k ∣ψ>| 2 ; Among them, |φ k > is the eigenstate of the jth fault; develop a federated learning fault diagnosis system to achieve knowledge sharing among multiple ball mills; each ball mill locally trains a fault diagnosis model and uploads the model parameters: Among them, θ i is the model parameter of the i-th ball mill, n i For the sample size, Improve fault diagnosis accuracy by aggregating global models through federated learning.

[0055] The present invention is further configured such that a ball mill operation data recording module records various operating parameters, sensor data, control instructions, and fault information of the ball mill in real time; a reinforcement learning algorithm is introduced to analyze the recorded data and learn an optimal control strategy through continuous interaction with the environment; a state space S is defined as a set of operating parameters and sensor data of the ball mill, an action space S is defined as a set of adjustable operating parameters, and a reward function R is designed based on indicators such as grinding efficiency, energy consumption, and equipment wear;

[0056] Develop a knowledge graph construction system to convert operating data into structured knowledge. Through entity recognition, relationship extraction, and knowledge fusion, construct a ball mill knowledge graph G = (V, E), where V is the entity set and E is the relationship set. Use the knowledge graph for fault diagnosis and decision support: Where D is the set of possible decisions, O is the observed anomaly, and P(d|O,G) is the probability that decision d explains anomaly O under the knowledge graph G.

[0057] Design an online learning mechanism to enable the system to adapt to changes in working conditions in real time; use a concept drift detection algorithm to trigger model updates when changes in data distribution are detected: Where f(x) is the characteristic function, N and M are the window sizes of new and old data respectively, and ∈ is the drift threshold. A digital twin-driven reinforcement learning framework is developed, and the digital twin model is used to generate simulation data for policy training. The state transition function of the digital twin model is:

[0058] s t+1 =f(s t ,a t ,ξ t ); where s t is the state, a t For action, t For random disturbances, the control strategy is optimized through the Actor-Critic architecture: Among them, π θ is the policy function, Q π is the action value function;

[0059] A causal inference analysis method is proposed to identify the causal relationship between operating parameters; by constructing a causal graph G = (V, E), the causal effect is calculated using Do-calculus: P(Y = y | do(X = x)) = Σ z P(Y=y|X=x,Z=z)P(Z=z);

[0060] Here, X is the intervention variable, Y is the outcome variable, and Z is the confounding variable. The system integration module introduces digital twin technology during system operation to establish a digital twin model of the ball mill. The digital twin model is updated by synchronizing the ball mill's operating data in real time. Virtual simulation experiments are conducted using the digital twin model to test the impact of different control strategies and parameter adjustments on the ball mill's operating performance. Furthermore, based on new material properties and production process requirements, combined with the simulation results of the digital twin model, the fault diagnosis model is retrained and operating parameters are updated.

[0061] The system calculates the expected return E(R) under different maintenance strategies; E(R) = Σ s∈S P(s|a)·R(s,a); where s is the system state, a is the maintenance action, P(s|a) is the probability that the system is in state s after taking action a, and R(s,a) is the benefit of taking action a in state s.

[0062] Develop a multi-agent collaborative optimization framework: Among them, J i is the objective function of the ith agent, u i Its control input, u -iIt provides control input for other intelligent agents; designs a quantum secure communication protocol to ensure the data transmission security of the ball mill control system; uses quantum key distribution technology to generate the encryption key: K = QKF(A, B), where A and B are the communicating parties and K is the shared key; uses the characteristics of quantum entanglement to detect potential eavesdropping: Among them, P(a i |b i ) is the conditional probability, P(a i ) is the unconditional probability, N is the number of measurements, and ∈ is the threshold; develop a blockchain-driven device management system to achieve full life cycle management of devices; upload device information and operation records to the chain: Block = (Header, Data);

[0063] The Header contains the hash value and timestamp of the previous block, and the Data contains the device status, maintenance records, etc. The maintenance plan is automatically executed using smart contracts: Execute(M) = ifCthenAelseB; where M is the maintenance contract, C is the trigger condition, and A and B are the execution actions.

[0064] The beneficial effects of the present invention are:

[0065] First, compared to the existing technology, the present invention utilizes a counterweight adjustment assembly to vary the inertia and impact force of the steel balls, enabling the ball mill to flexibly adjust its operating conditions based on the properties of different materials and grinding requirements. For materials with higher hardness, increasing the steel balls' moment of inertia and impact force results in more effective crushing. For materials with lower hardness, however, the grinding process is more stable, improving grinding uniformity and fineness. Compared to traditional fixed-structure steel balls, this adjustable design significantly enhances the versatility and adaptability of the ball mill, enabling it to meet diverse production needs.

[0066] 2. Compared with the existing technology, the present invention significantly improves the control accuracy of the ball mill by controlling the setting of the interactive unit, making the operation parameter control more precise and the grinding process more stable and efficient; effectively reduces energy consumption and balances equipment wear, extending the service life of the equipment; greatly enhances the fault diagnosis capability and provides early warning of potential faults; realizes deep intelligent application, and provides strong support for the application of ball mills in complex scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a structural schematic diagram of the present invention;

[0068] Figure 2 It is a structural schematic diagram of the frame and outer cylinder in the present invention;

[0069] Figure 3 Schematic diagram of the internal structure of the gearbox in the present invention;

[0070] Figure 4 It is a structural schematic diagram of the outer cylinder in the present invention;

[0071] Figure 5 It is a structural schematic diagram of the inner cylinder in the present invention;

[0072] Figure 6 It is a structural diagram of the steel ball in the present invention;

[0073] Figure 7 This is a schematic diagram of the main structure of the steel ball in the present invention;

[0074] Figure 8 Schematic diagram of the cross-sectional structure of the steel ball in the present invention;

[0075] Figure 9 It is a schematic structural diagram of the rotating block in the steel ball of the present invention;

[0076] Figure 10 It is a structural diagram of the positioning block in the present invention.

[0077] The following are marked in the figure:

[0078] 1. Frame; 2. Outer cylinder; 3. First bearing seat; 4. First gear; 5. Ring gear; 6. Sealing cover; 7. Long bolt; 8. Inlet and outlet; 9. First motor; 10. Gearbox; 11. Second gear; 12. Third gear; 13. Synchronous shaft; 14. Discharge pipe; 15. Discharge port; 16. First fixed frame; 17. Second motor; 18. First screw conveyor rod; 19. Second fixed frame; 20. Third motor; 21. Feed pipe; 22. Second screw conveyor rod; 2 3. Feed hopper; 24. Second bearing seat; 25. Inner cylinder; 26. Steel ball; 27. Spiral blade; 28. Fixed block; 29. ​​Positioning plate; 30. Positioning bolt; 31. Inner ball; 32. Connecting rod; 33. Sliding hole; 34. First threaded rod; 35. Positioning groove; 36. Rotating block; 37. Second threaded rod; 38. Third threaded rod; 39. Guide rod; 40. Counterweight ball; 41. First bevel gear; 42. Second bevel gear; 43. Third bevel gear; 44. Positioning block. DETAILED DESCRIPTION

[0079] In the description of the present invention, it should be noted that the terms "front", "up", "down", "left", "right", "vertical", "horizontal", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, they cannot be understood as limiting the present invention.

[0080] The following is combined with Figure 1-10The present invention is further described.

[0081] Example 1:

[0082] In order to solve the problems existing in the background technology, the present application proposes the following technical solutions: A ball mill for aluminum silicate production, comprising: a frame 1, an outer cylinder 2, a driving assembly, a discharge pipe 14, a feed pipe 21 and a control interaction unit, the top two sides of the frame 1 are fixedly connected with a first bearing seat 3 and a second bearing seat 24, the outer cylinder 2 is rotatably connected to the frame 1 through the first bearing seat 3 and the second bearing seat 24, the outer cylinder 2 is fixedly connected with an inner cylinder 25, and a plurality of steel balls 26 are arranged in the inner cylinder 25, the driving assembly is located on the frame 1, and is used to drive the outer cylinder 2 to rotate, the discharge pipe 14 is fixedly connected to the outside of the first bearing seat 3, and a discharge port 15 is provided at the side end of the outer cylinder 2 corresponding to the position of the discharge pipe 14, the feed pipe 21 is fixedly connected to the outside of the second bearing seat 24, and a feed port is provided at the position of the feed pipe 21 corresponding to the side end of the outer cylinder 2, and the control interaction unit monitors and adjusts the operating status of the ball mill and performs fault diagnosis.

[0083] In this embodiment, the drive assembly includes a first motor 9, a gearbox 10, a first gear 4, and a ring gear 5. The ring gear 5 is fixedly connected to the outer cylinder 2. The first gear 4 is rotatably connected to the frame 1. The gearbox 10 is fixedly connected to the frame 1. A second gear 11 and a third gear 12 are rotatably connected within the gearbox 10. The first motor 9 is fixedly connected to the frame 1. The output end of the first motor 9 is fixedly connected to the second gear 11. The third gear is coaxially fixedly connected to a synchronizing shaft 13. The synchronizing shaft 13 is coaxially fixedly connected to the first gear 4, which meshes with the ring gear 5. The first motor 9, as the power source of the drive assembly, features high power and high efficiency, capable of stable and powerful power output, providing a solid power foundation for the operation of the ball mill. It is fixedly connected to the frame 1 to ensure stability during operation and reduce power loss and equipment damage caused by vibration. The configuration of the gearbox 10 optimizes and transmits the power output of the motor. The second gear 11 and the third gear 12 within the gearbox 10 achieve multi-stage power transmission and speed change through precise meshing design. The second gear 11 is directly connected to the output end of the first motor 9, receives the rotational power of the motor and transmits it to the third gear 12, and the third gear 12 then transmits the power to the first gear 4 through a coaxially fixed synchronous shaft 13.

[0084] The first gear 4 is rotatably connected to the frame 1 and meshes with the ring gear 5 fixed to the outer cylinder 2. This gear transmission method offers the advantages of a precise transmission ratio, high power transmission, and high efficiency. During power transmission, the rotation of the first gear 4 precisely drives the ring gear 5, thereby causing the outer cylinder 2 and inner cylinder 25 to rotate synchronously. By combining and designing the various gears within the gearbox 10, the speed and torque of the first gear 4 can be adjusted to meet the speed and power requirements of the ball mill at different grinding stages. For example, in the initial grinding stage, a higher torque is required to crush the material. In this case, the transmission ratio of the gearbox 10 can be adjusted to increase the torque transmitted to the ring gear 5 by the first gear 4. In the later stages of grinding, the speed of the outer cylinder 2 can be appropriately increased to achieve a finer grinding of the material.

[0085] Furthermore, the enclosed design of the gearbox 10 effectively prevents the ingress of dust and impurities, protecting the normal operation of the internal gears and extending their service life. The entire drive assembly features a compact and rational design and a scientific layout. While ensuring efficient power transmission, it also reduces operating noise, creating a relatively quiet working environment for operators. Its stable operating performance also helps improve the grinding quality and production efficiency of the ball mill, ensuring the continuity and stability of the aluminum silicate production process.

[0086] In this embodiment, the outside of the discharge pipe 14 is fixedly connected to the first fixing frame 16, the outside of the feed pipe 21 is fixedly connected to the second fixing frame 19, the outside of the first fixing frame 16 is fixedly connected to the second motor 17, the outside of the second fixing frame 19 is fixedly connected to the third motor 20, the output end of the second motor 17 is fixedly connected to the first spiral conveying rod 18, the output end of the third motor 20 is fixedly connected to the second spiral conveying rod 22, a feed hopper 23 is provided above the feed pipe 21, the output end of the feed hopper 23 is fixedly connected to the feed pipe 21, inlet and outlet 8 are provided in the outer cylinder 2 and the inner cylinder 25, and a sealing cover 6 is provided in the inlet and outlet 8, which is buckled in the inlet and outlet 8, and a long bolt 7 is passed through the inlet and outlet 8 and the sealing cover 6; the first fixing frame 16 and the second fixing frame 19 are respectively fixed to the outside of the discharge pipe 14 and the feed pipe 21, providing a stable installation foundation for the second motor 17, the third motor 20 and the spiral conveying rod. They are made of high-strength materials and can withstand the vibrations generated by the motor operation and the forces exerted by the screw conveyor rod when conveying materials, ensuring that the entire conveying system remains stable during operation without shaking or displacement. The second motor 17 and the third motor 20 serve as the power devices for material conveying, and have good speed regulation performance and stable torque output.

[0087] The second screw conveyor 22, driven by a third motor 20 and connected to the feed pipe 21, is responsible for conveying material from the feed hopper 23 to the inner drum 25. The design of the feed hopper 23 increases the material storage space, facilitating centralized loading of materials by the operator, reducing the need for frequent additions, and improving work efficiency. Driven by the motor, the second screw conveyor 22 rotates its spiral blades 27, pushing the material evenly and continuously into the inner drum 25, ensuring a timely and stable supply of material to meet the grinding requirements of the ball mill.

[0088] The first screw conveyor 18, driven by the second motor 17, cooperates with the discharge pipe 14 to perform screening and return conveyance. When the ground slurry overflows from the discharge port 15 into the discharge pipe 14, it may contain some incompletely ground bulk material or steel balls 26. Driven by the second motor 17, the first screw conveyor 18 rotates in the opposite direction, pushing these materials and steel balls 26 from the discharge pipe 14 back into the inner drum 25 for further grinding and processing. This design avoids material waste, improves grinding efficiency and the quality of the finished product, and also ensures that the steel balls 26 are not lost during discharge, reducing production costs.

[0089] Among them, the inlet and outlet 8 provided on the outer cylinder 2 and the inner cylinder 25, as well as the matching blocking cover 6 and long bolts 7, provide convenience for the maintenance of the equipment and the addition and replacement of the steel balls 26. When the equipment needs to be repaired or the steel balls 26 need to be added, the operator only needs to unscrew the long bolts 7, remove the blocking cover 6, and then perform the corresponding operations through the inlet and outlet 8. The design of the long bolts 7 ensures a tight connection between the blocking cover 6 and the inlet and outlet 8, which can effectively prevent material leakage during the operation of the ball mill and ensure the sealing and safety of the equipment. This design not only meets the needs of daily maintenance and operation of the equipment, but also ensures the stability and reliability of the ball mill during normal operation, making the entire production process more efficient and convenient.

[0090] In this embodiment, a spiral leaf 27 is provided on the outside of the steel ball 26, an inner ball 31 is provided in the steel ball 26, the outside of the inner ball 31 is fixedly connected to a connecting rod 32, the connecting rod 32 is fixedly connected to the steel ball 26, and a plurality of sliding holes 33 are provided in the inner ball 31. The counterweight adjustment assembly includes a first threaded rod 34, a second threaded rod 37 and a third threaded rod 38 that are rotatably connected to the steel ball 26. The bottom of the first threaded rod 34 is fixedly connected to a third bevel gear 43, the outside of the second threaded rod 37 is fixedly connected to a second bevel gear 42, and one end of the third threaded rod 38 is fixedly connected to the first bevel gear 43. Gear 41 and the third bevel gear 43 mesh with the second bevel gear 42, which in turn meshes with the first bevel gear 41. The first threaded rod 34 is a bidirectional threaded rod with thread grooves in opposite directions. Counterweight balls 40 are threadedly connected to the exterior of each of the first, second, and third threaded rods 34, 37, and 38. Two counterweight balls 40 are provided on the exterior of the first threaded rod 34. Multiple sets of guide rods 39 are also fixedly connected to the interior of the steel ball 26, extending through the counterweight balls 40. The spiral blades 27 on the exterior of the steel ball 26 play a unique role in the operation of the ball mill. As the outer and inner cylinders 25 rotate, the steel ball 26 moves with it. The spiral blades 27 alter the contact pattern and trajectory between the steel ball 26 and the material. They guide the material flow between the steel balls 26, distributing it more evenly within the grinding area, thus avoiding localized accumulation or inadequate grinding. At the same time, the spiral blades 27 can also assist the steel balls 26 in performing preliminary crushing and mixing of the materials during the friction and stirring process with the materials, thereby improving the grinding efficiency of the ball mill.

[0091] In this embodiment, the inner ball 31 is fixedly connected to the steel ball 26 via a connecting rod 32, providing stable support for the internal adjustment mechanism. Multiple sets of sliding holes 33 provided in the inner ball 31 provide a track for the movement of the counterweight ball 40, ensuring that the counterweight ball 40 can slide smoothly inside the steel ball 26. The first threaded rod 34, the second threaded rod 37, and the third threaded rod 38 are meshed with each other through bevel gears to form an interlocking adjustment system. When one of the threaded rods is rotated, the other two threaded rods will rotate synchronously under the transmission of the bevel gears, realizing the simultaneous adjustment of the positions of multiple counterweight balls 40.

[0092] In this embodiment, the first threaded rod 34 serves as a bidirectional threaded rod, and the two counterweight balls 40 disposed on its exterior can be moved to the sides or the center, respectively, and cooperate with the counterweight balls 40 on the second threaded rod 37 and the third threaded rod 38 to more flexibly adjust the center of gravity and mass distribution of the steel ball 26. The counterweight balls 40 are made of a high-density metal material and have a large mass. By changing their position within the steel ball 26, the inertia and impact force of the steel ball 26 can be significantly changed. The provision of the guide rod 39 provides guidance and limitation for the movement of the counterweight balls 40, preventing the counterweight balls 40 from deflecting or getting stuck during movement, thereby ensuring a smooth and stable adjustment process.

[0093] The design of the steel balls 26 with adjustable counterweights allows the ball mill to flexibly adjust the working state of the steel balls 26 according to the properties of different materials and grinding requirements. For materials with higher hardness, moving the counterweight balls 40 toward the outside of the steel balls 26 increases the moment of inertia and impact force of the steel balls 26, enabling more effective material crushing. For materials with lower hardness, moving the counterweight balls 40 toward the center of the steel balls 26 allows the steel balls 26 to grind more smoothly, improving the uniformity and fineness of the grinding. Compared to traditional fixed-structure steel balls 26, this adjustable design greatly enhances the versatility and adaptability of the ball mill, meeting diverse production needs, reducing the time and cost associated with replacing different types of steel balls 26, and improving production efficiency.

[0094] In this embodiment, the outsides of the first threaded rod 34, the second threaded rod 37 and the third threaded rod 38 corresponding to the outside of the steel ball 26 are all fixedly connected to a fixed block 28, one of which is provided with a positioning groove 35, one end of the first threaded rod 34 extends into the positioning groove 35 and is fixedly connected to a rotating block 36, and a positioning plate 29 is provided on the outside of the positioning groove 35, and a positioning block 44 is fixedly connected to the bottom of the positioning plate 29, and a positioning bolt 30 is threadedly connected in the positioning plate 29, and the positioning bolt 30 is threadedly connected to the fixed block 28, and a hexagonal groove is provided at the position of the positioning block 44 corresponding to the outside of the rotating block 36; the fixed block 28, the positioning groove 35, the positioning plate 29, the positioning bolt 30 and the like arranged on the outside of the steel ball 26 together constitute a reliable positioning and locking system, ensuring that the position of the adjusted counterweight ball 40 will not change during the operation of the ball mill, thereby ensuring the stable working state of the steel ball 26. The fixing blocks 28 are respectively fixed to the outside of the first threaded rod 34, the second threaded rod 37 and the third threaded rod 38, providing additional support for the threaded rods, enhancing the stability of the threaded rods during rotation, and preventing the adjustment accuracy and reliability of the counterweight ball 40 from being affected by the shaking of the threaded rods.

[0095] The positioning groove 35 provided in one of the fixed blocks 28 cooperates with the rotating block 36 at the end of the first threaded rod 34. The rotating block 36 can rotate freely in the positioning groove 35, making it convenient for the operator to insert a tool into the hexagonal groove outside the rotating block 36 to rotate the first threaded rod 34, thereby adjusting the position of the counterweight ball 40. When the counterweight ball 40 is adjusted to the appropriate position, the positioning plate 29 comes into play. The positioning block 44 at the bottom of the positioning plate 29 matches the shape of the hexagonal groove outside the rotating block 36. The positioning plate 29 is installed on the fixed block 28 so that the positioning block 44 is embedded in the hexagonal groove. Then, by tightening the positioning bolt 30, the positioning plate 29 is firmly fixed to the fixed block 28. In this way, the positioning plate 29 and the positioning block 44 form a limit for the rotating block 36, preventing the first threaded rod 34 from continuing to rotate, thereby locking the position of the counterweight ball 40.

[0096] This positioning and locking design has multiple advantages. On the one hand, it ensures that during the high-speed operation of the ball mill, even if it is subjected to strong vibrations and impacts, the counterweight ball 40 will not be displaced by the rotation of the threaded rod, ensuring that the steel ball 26 always maintains the set center of gravity and mass distribution, maintaining a stable grinding effect. On the other hand, the operation is simple and quick. When the operator needs to adjust the position of the counterweight ball 40, he only needs to loosen the positioning bolt 30, remove the positioning plate 29, and easily rotate the threaded rod to adjust it; after the adjustment is completed, the positioning plate 29 is reinstalled and the positioning bolt 30 is tightened to quickly complete the positioning and locking. In addition, the design has a compact structure, takes up little space, and will not affect the normal movement of the steel ball 26 in the ball mill. This reliable positioning and locking system, combined with the adjustable counterweight structure inside the steel ball 26, not only ensures the flexibility of the steel ball 26 adjustment, but also ensures the stability of the working state, providing a solid guarantee for the ball mill to efficiently and accurately grind materials of different properties, further improving the performance and reliability of the ball mill, and meeting the demand for high-quality grinding equipment in the production process of aluminum silicate.

[0097] The method of using this embodiment is as follows: add the material into the feed hopper 23, the third motor 20 drives the second threaded rod 37 to rotate, so that the material can be transported to the inner cylinder 25, and water or grinding liquid is added for synchronous grinding; after the transportation is completed, the steel ball 26 is transported to the inner cylinder 25 through the inlet and outlet 8, and then the inlet and outlet 8 are closed by the blocking cover 6, and then the blocking cover 6 is fixed by the long bolt 7; start the first motor 9 to work, the first motor 9 drives the second gear 11 to rotate, the second gear 11 drives the third gear 12 to rotate, and the third gear 12 drives The first gear 4 rotates, driving the ring gear 5, which in turn rotates the outer cylinder 2 and the inner cylinder 25. The impact of the moving steel balls 26 on the material achieves grinding. The ground slurry overflows into the discharge pipe 14 and is discharged through the discharge pipe 14. The second motor 17 drives the first screw conveyor rod 18 to rotate in the opposite direction, pushing the overflowing steel balls 26 and bulk material back into the inner cylinder 25 for further processing. It should be noted that the counterweight ball 40 used for the steel balls 26 is made of a high-density metal material, such as lead or tungsten alloy. By rotating the first threaded rod 34 with a tool, the position of the counterweight ball 40 within the steel balls 26 can be changed, thereby adjusting the center of gravity and mass distribution of the steel balls 26. When grinding harder materials, the counterweight ball 40 can be moved toward the outside of the steel balls 26 to increase the impact force of the steel balls 26 during collision. For softer materials, the counterweight ball 40 can be moved toward the center of the steel balls 26 to improve grinding uniformity. This design of adjustable counterweight enables the steel balls 26 to adapt to the grinding requirements of materials of different properties, greatly improving the versatility and working efficiency of the ball mill.

[0098] Example 2:

[0099] The control interaction unit includes a ball mill operation status monitoring module, a ball mill grinding process mathematical model building module, an objective function optimization module, a parameter optimization module based on an intelligent algorithm, a real-time control and adjustment module, a fault diagnosis and early warning module, a ball mill operation data recording module, a human-computer interaction interface module, and a system integration module. The ball mill operation status monitoring module includes a vibration sensor, a temperature sensor, and a speed sensor. The speed sensor is a non-contact speed sensor, which is an existing product and is used to measure the speed of the outer cylinder 2. The vibration sensor is installed on the first bearing seat 3 or the second bearing seat 24, and the temperature sensor is installed on the outside of the first motor 9, the first bearing seat 3, or the second bearing seat 24. The speed sensor is located on the outside of the frame 1 and is used to measure the speed of the outer cylinder 2. Assume that the vibration signal is represented as V(t) in the time domain, the temperature signal collected by the temperature sensor is T(t), and the speed collected by the speed sensor is n(t), where t represents time.

[0100] In this embodiment, a comprehensive operation status monitoring system is constructed by installing multiple sensors such as vibration, temperature, and speed at key positions of the ball mill. Compared with traditional monitoring methods, this system can collect key data during the operation of the ball mill in real time and accurately. The collected data covers multiple dimensions of equipment operation, providing rich and accurate basic information for subsequent analysis and control. On this basis, a dynamic noise compensation algorithm and a spatiotemporal feature fusion network are introduced to further improve data quality and feature extraction capabilities. The dynamic noise compensation algorithm can adaptively suppress noise according to different working conditions to ensure that the collected signals truly reflect the operating status of the equipment; the spatiotemporal feature fusion network can simultaneously process the spatial distribution and time series characteristics of multi-sensor data, effectively explore the potential correlation between data, thereby more comprehensively and deeply understanding the operating status of the ball mill, laying a solid foundation for subsequent precise control and fault diagnosis.

[0101] The original signal collected by the sensor is preprocessed. In addition to filtering, noise reduction and normalization, multi-scale data fusion technology is introduced. The vibration signal V(t) and temperature signal T(t) are decomposed into multiple scales, and the wavelet transform is used to obtain the signal characteristics at different scales. For the vibration signal V(t), the low-frequency component A is obtained after j-layer wavelet decomposition. j (t) and high frequency component D j,k (t)(k=1,2,…,j), extract features of components at different scales and construct a multi-scale feature vector V ms ; V ms =[mean(A j ),std(A j ),mean(D j,k ),std(D j,k )]; where mean represents the mean calculation function, and std represents the standard deviation calculation function. Through this multi-scale feature extraction, the detailed information of the ball mill operating status can be captured more comprehensively;

[0102] The temperature signal T(t) is normalized and the formula is: Among them, T norm (t) is the normalized temperature value, T min and T max are the minimum and maximum values ​​of the temperature signal during the monitoring period;

[0103] Normalization processing helps to unify the scale of different sensor data, which is convenient for subsequent analysis and processing. Dynamic noise compensation algorithm is introduced to perform adaptive compensation for noise characteristics under different working conditions. According to the spectrum analysis of vibration signal, the noise energy ratio ρ is calculated. noise ; Where V(f) is the frequency domain representation of the vibration signal, F noiseis the frequency range where the noise is mainly distributed, F total For the entire spectrum range; dynamically adjust the filtering parameters according to the noise energy ratio to improve signal quality.

[0104] Develop a spatiotemporal feature fusion network to simultaneously process spatially distributed multi-sensor data and time series features, and construct a spatiotemporal feature tensor X st ;X st =[X s1 (t),X s2 (t),…,X sn (t)]; where X si (t) represents the feature vector of the i-th sensor at time t, and the spatiotemporal features are extracted through the spatiotemporal convolutional network;

[0105] Combining spatial convolution and temporal convolution, it can capture the spatial correlation and temporal evolution characteristics between sensors;

[0106] Among them, the mathematical model building module of the ball mill grinding process includes the introduction of nonlinear dynamics model;

[0107] Assume that the pulverization process of the material in the ball mill conforms to the improved two-stage pulverization kinetic equation:

[0108] Among them, X i is the content of particles with a particle size of i in the material, k i is the first-order crushing rate constant of particles with a particle size of i, k' i is the secondary crushing rate constant. These two constants are related to factors such as the ball mill speed n, loading capacity M, and material properties. They can be obtained by fitting experimental data combined with nonlinear regression algorithm.

[0109] In the above technical solution, conventional processing methods such as filtering, noise reduction, and normalization are adopted, and multi-scale data fusion technology is introduced to perform multi-scale decomposition and feature extraction on vibration and temperature signals, which can more comprehensively capture the detailed information of the ball mill's operating status. The dynamic noise compensation algorithm dynamically adjusts the filtering parameters based on the vibration signal spectrum analysis to effectively improve the signal quality. The application of the spatiotemporal feature fusion network combines the spatial and temporal features of multi-sensor data, and extracts key features through the spatiotemporal convolutional network. Compared with a single data processing method, it can more accurately extract effective features reflecting the operating status of the ball mill, providing high-quality data support for subsequent model establishment and parameter optimization, and significantly improving the accuracy and effectiveness of data processing.

[0110] The grinding efficiency η of the ball mill is related to parameters such as feed rate v and rotation speed n. A grinding efficiency prediction model based on deep learning is established. A convolutional neural network is used to learn a large amount of operating data. The input is sensor data and operating parameters after feature extraction, and the output is the grinding efficiency prediction value.

[0111] Construct a dynamic energy consumption prediction model, comprehensively consider factors such as the first motor efficiency, transmission loss and material characteristics, and the actual output power P of the first motor out The calculation is as follows: out =P in ·η motor ·η trans Among them, P in is the input power, η motor is the first motor efficiency, η trans is the transmission efficiency; the first motor efficiency η motor and transmission efficiency η trans Calculated by the following formulas:

[0112]

[0113] Among them, P rated is the rated power of the first motor, k1, k2, k3 are empirical coefficients, and n0 is the optimal speed point;

[0114] Combining quantum computing with neural networks improves the model prediction accuracy. The output of the quantum layer is calculated as: q =σ(W q ·U θ |0>+β); where W q is the quantum weight matrix, U θ To parameterize the quantum circuit, |0> is the initial quantum state, β is the bias, σ is the activation function, and the quantum layer output is combined with the classical layer:

[0115] Among them, W c is the classic weight matrix, b is the bias, and f is the output function. A multi-scale grinding dynamics model is constructed. Considering the differences in the crushing behaviors of materials with different particle sizes, the materials are divided into K intervals according to the particle size. The crushing dynamics equation for each interval is:

[0116] Among them, X k is the material content in the kth particle size interval, k k and k′ k is the crushing rate constant in this interval, k jk is the conversion rate constant from the jth interval to the kth interval.

[0117] In the above technical solution, the nonlinear characteristics of material crushing are fully taken into consideration, and an improved secondary crushing kinetic equation is adopted, which is more in line with the actual grinding process than the traditional model. The grinding efficiency prediction model based on deep learning uses convolutional neural networks to learn a large amount of operating data, which can accurately predict the grinding efficiency. The innovative quantum hybrid neural network model integrates quantum computing and classical neural networks to greatly improve the model prediction accuracy; the multi-scale grinding kinetic model takes into account the differences in the crushing behavior of materials with different particle sizes, making the model's description of the grinding process more detailed and accurate. The establishment of these models provides a strong theoretical basis and technical support for in-depth understanding of the ball mill grinding process and optimization of operating parameters, effectively improving the controllability of the grinding process and the accuracy of efficiency prediction.

[0118] Among them, the objective function optimization module introduces equipment wear W as the optimization target on the basis of maximizing the original grinding efficiency η and minimizing the energy consumption E, and establishes a three-objective optimization function: Among them, α, β and γ are weight coefficients, and α + β + γ = 1. Their values ​​are adjusted according to actual production needs to balance the relationship between grinding efficiency, energy consumption and equipment wear; η max is the maximum grinding efficiency obtained under experimental conditions, E min is the minimum energy consumption, W max is the maximum allowable equipment wear;

[0119] Equipment wear W is based on the wear of the steel ball m w And the inner cylinder wear thickness δ is calculated: Among them, λ1 and λ2 are weight coefficients, m w,max is the maximum allowable wear of the steel ball, δ max The maximum allowable wear thickness of the inner cylinder. The energy consumption E is related to the power P and running time t of the ball mill. The calculation formula is:

[0120] The power P of the ball mill can be calculated by the following empirical formula: P = k p (M)+ρV m )ω 2 R; where k p is the power coefficient, determined by experiment; M is the load (unit: kg); ρ is the density of the material (unit: kg / m 3 );V m is the volume of the material in the inner cylinder (unit: m 3 ); ω is the angular velocity of the ball mill (unit: rad / s), R is the radius of the inner cylinder (unit: m);

[0121] Introducing quality stability index Q s, measure the degree of fluctuation of product quality, calculate the standard deviation σ of particle size distribution through real-time monitoring of product particle size distribution: Among them, σ0 is the target standard deviation, and the quality stability index is added to the optimization objective function: Among them, δ is the mass stability weight coefficient.

[0122] A dynamic weight adaptive optimization algorithm is proposed to automatically adjust the objective function weight according to the operating status. The weight adjustment formula is:

[0123]

[0124] Among them, η α ,η β ,η γ ,η δ is the learning rate, is the partial derivative of the objective function with respect to each weight; introduce the carbon emission index C and establish the green optimization objective function:

[0125] Among them, C min is the minimum carbon emission target, ∈ is the carbon emission weight coefficient, and the carbon emission C is calculated as: C = Σ i E i ·f i ;

[0126] Among them, E i is the consumption of the i-th energy, f i is the carbon emission factor of the i-th energy source.

[0127] In the above technical solution, in the process of determining the optimization objective function, not only the maximization of grinding efficiency and the minimization of energy consumption are considered, but also the equipment wear and quality stability indicators are introduced to establish a multi-objective optimization function. Compared with the traditional single-objective optimization, it is more in line with actual production needs. The dynamic weight adaptive optimization algorithm can automatically adjust the objective function weight according to the operating status, making the optimization process more flexible and intelligent; the carbon emission index is introduced to construct a green optimization objective function, respond to the requirements of energy conservation and emission reduction, and promote the green production of ball mills. These innovations make the optimization of ball mill operating parameters more comprehensive and scientific, and can achieve a comprehensive balance of grinding efficiency, energy consumption, equipment wear, quality stability and carbon emissions under different production needs, thereby improving the economic and social benefits of the enterprise. Among them, the parameter optimization module based on the intelligent algorithm uses an improved quantum particle swarm optimization algorithm to optimize the operating parameters of the ball mill;

[0128] In the quantum particle swarm optimization algorithm, quantum behavior and chaotic search strategy are introduced to improve the global search capability and convergence speed of the algorithm. The position update formula of each particle in the quantum space is: δL(α); where is the quantum rotating gate parameter, which is dynamically adjusted through chaotic mapping; δ is the step size factor; L(α) is the quantum Levy flight path, which is used to enhance the local search capability of the algorithm; p id (t) is the individual optimal position of the i-th particle; g d (t) is the global optimal position of the entire particle swarm. Using feed rate v, rotational speed n, and loading M as optimization variables, an improved quantum particle swarm optimization algorithm was used to search for the optimal parameter combination that achieves the objective function F, while satisfying the constraints of ball mill operation safety and process requirements.

[0129] An adaptive learning rate optimization algorithm is proposed to dynamically adjust the learning parameters of the quantum particle swarm. The learning rate η(t) changes dynamically with the number of iterations t: Among them, η0 is the initial learning rate, η ∞ is the final learning rate, and T is the decay period. Through this adaptive adjustment, the global search and local search capabilities are balanced.

[0130] A population diversity maintenance mechanism is introduced to calculate the average Euclidean distance D between particles:

[0131] Where N is the number of particles, x i and x j is the particle position vector. When the diversity is lower than the threshold, the population perturbation operation is triggered to prevent the algorithm from falling into the local optimum.

[0132] Develop a multi-agent collaborative optimization framework to collaboratively optimize the ball mill control system and the entire production line, achieving global optimization through communication and collaboration between agents: Among them, J i is the objective function of the ith agent, u i Its control input, u -i is the control input of other agents, and each agent uses the Deep Deterministic Policy Gradient (DDPG) algorithm to make decisions: μ * (s)=argmax μ Q π (s,μ(s)); where μ * is the optimal strategy, Q π is the action value function, s is the state, and a hybrid optimization method of quantum genetic algorithm (QGA) and particle swarm optimization is proposed. The update formula of quantum chromosome is: θ i (t+1)=θ i (t)+Δθ·sin(2πft+φ);

[0133] Among them, θ iis the phase angle of the i-th quantum bit, Δθ is the phase adjustment amount, f is the frequency, φ is the initial phase, and chromosome evolution is achieved through quantum rotation gate operation:

[0134]

[0135] In the above technical solution, an improved quantum particle swarm optimization algorithm is used, combined with quantum behavior and chaotic search strategies. Compared with traditional optimization algorithms, it has stronger global search and convergence capabilities. The newly developed multi-agent collaborative optimization framework realizes the collaborative optimization of the ball mill and the production line, improving production efficiency from an overall level; the hybrid optimization method of the quantum genetic algorithm and the particle swarm algorithm combines the advantages of the two algorithms to further improve the parameter optimization effect. The application of these intelligent algorithms can quickly and accurately find the optimal combination of ball mill operating parameters, which not only improves the operating efficiency of the ball mill itself, but also realizes collaborative operation with the production line, optimizes the entire production process, and improves the production efficiency of the enterprise. Among them, the real-time control and adjustment module controls the feed speed, rotation speed and inner drum loading of the ball mill in real time according to the obtained optimal operating parameter combination; the adaptive fuzzy PID control algorithm is introduced to adjust the outer drum rotation speed;

[0136] Assume that the actual speed of the outer cylinder is n actual , the target speed is n target , the adjustment control quantity Δu of the inverter is calculated as follows:

[0137] First, according to the speed deviation e=n target -n actual and the rate of change of deviation Adjusting PID control parameter k using fuzzy inference rules p 、k i and k d ;k p =k p0 +Δk p (e,ec);k i =k i0 +Δk i (e,ec);k d =k d0 +Δk d (e,ec);

[0138] Among them, k p0 、k i0 and k d0 is the initial PID control parameter, Δk p (e,ec),Δk i (e,ec) and Δk d (e,ec) is the parameter adjustment amount obtained based on fuzzy reasoning.

[0139] Then, the adjustment control amount Δu is calculated:

[0140] For the control of feeding speed, the adaptive fuzzy PID control algorithm is also used. According to the real-time feedback of grinding efficiency, the feeding speed of the second screw conveyor is adjusted to keep the feeding speed at the optimal value. A multivariable collaborative control strategy is designed, considering the coupling effect between each control parameter. By establishing the coupling coefficient matrix C, the coupling degree between the parameters is calculated:

[0141] Among them, y i is the i-th output variable, u j For the jth control variable, dynamically adjust the control parameters according to the coupling coefficient matrix to improve the system response speed and stability, develop a predictive control compensation algorithm, predict the system state at the future moment, and use the long short-term memory network (LSTM) to predict the system output in the next k steps based on the current state and historical data.

[0142] Where y is the system output, u is the control input, n and m are the lengths of historical data, and the control parameters are adjusted in advance according to the prediction results to reduce system lag;

[0143] A dynamic adaptive control architecture is proposed to automatically switch the control strategy according to the system operation status, and the system operation status index S is defined as:

[0144] Among them, s i is the i-th state feature, w i is the corresponding weight, and switches between different control strategies according to the state index S:

[0145]

[0146] Among them, u i is the i-th control strategy, R i For the corresponding state areas, a distributed cooperative control algorithm is developed to assign control tasks to multiple controllers. Each controller is responsible for local control and exchanges information through a communication network:

[0147] Among them, K i is the control law of the i-th controller, e i is the local error, T ij is the communication weight, is the predicted output of the j-th controller;

[0148] In the above technical solution, an adaptive fuzzy PID control algorithm is introduced in the real-time control and regulation link, which can dynamically adjust the control parameters according to the speed deviation and change rate. Compared with traditional PID control, it has higher control accuracy and faster response. The designed multivariable collaborative control strategy takes into account the coupling effect between parameters to improve system stability; the predictive control compensation algorithm uses LSTM to predict the system state, adjust parameters in advance, and reduce system lag. The dynamic adaptive control architecture can automatically switch the control strategy according to the system operating state, making the control more flexible; the distributed collaborative control algorithm assigns control tasks to multiple controllers to achieve efficient collaborative control. The comprehensive application of these control technologies ensures that the ball mill can maintain stable and efficient operation under various working conditions, significantly improving the dynamic performance and robustness of the system.

[0149] Among them, the fault diagnosis and early warning module establishes a fault diagnosis model based on transfer learning. First, the deep neural network model is pre-trained on a large amount of general mechanical equipment fault data, and then fine-tuned for the specific fault data of the ball mill. The sensor data collected in step one and the feature parameters extracted in step two are used as input. The trained fault diagnosis model is used to determine whether the ball mill has a fault and the type of fault.

[0150] The improved convolutional long short-term memory network (ConvLSTM) is used as the main structure of the fault diagnosis model, which can simultaneously process the spatial and temporal features of time series data. The output of the model is the fault probability vector P = [p1, p2, ..., p m ], where p i It represents the probability of the i-th fault of the ball mill. When the maximum probability p max =max(P) exceeds the set threshold τ, a fault warning signal is issued, and the fault type and possible fault cause are provided.

[0151] Develop a fault evolution prediction model based on the Hidden Markov Model (HMM) to predict the fault development trend. Train the HMM with historical fault data to obtain the state transition probability matrix A and observation probability matrix B: ij =P(q t+1 =j|q t =i); B jk =P(o t =k|q t =j); where q t is the hidden state at time t, o t is the observed value at time t, and the Viterbi algorithm is used to calculate the most likely state sequence to predict the fault evolution path and remaining life.

[0152] A multi-dimensional fault feature fusion model is constructed to fuse the time domain, frequency domain, and time-frequency domain features. The autoencoder (AE) is used to extract the essential representation of each domain feature, and then feature splicing is performed:

[0153] F fusion =[AE time (F time ),AE freq (F freq ),AE timefreq (F timefreq )]; where F time 、F freq 、F timefteq They are time domain, frequency domain and time-frequency domain features, AE time AE freq AE timefreq is the corresponding autoencoder.

[0154] A quantum entanglement fault diagnosis algorithm is designed. The principle of quantum entanglement is used to improve the discrimination of fault characteristics and the quantum characteristic vector ψ is defined as: Among them, |f i > is the i-th fault characteristic state, c i is the corresponding probability amplitude, and the fault diagnosis result is obtained through quantum measurement: P(k)=|<φ k ∣ψ>| 2 ; Among them, |φ k > is the eigenstate of the kth fault.

[0155] Develop a federated learning fault diagnosis system to achieve knowledge sharing among multiple ball mills. Each ball mill trains a fault diagnosis model locally and uploads the model parameters:

[0156] Among them, θ i is the model parameter of the i-th ball mill, n i For the sample size, Improve fault diagnosis accuracy by aggregating global models through federated learning.

[0157] In the above technical solution, the fault diagnosis model based on transfer learning is pre-trained with general equipment fault data and fine-tuned for the ball mill, which improves the efficiency and accuracy of fault diagnosis. The improved ConvLSTM model effectively processes the characteristics of time series data. The quantum entanglement fault diagnosis algorithm uses quantum characteristics to improve the discrimination of fault features. The federated learning fault diagnosis system realizes multi-machine knowledge sharing, further improving the diagnosis accuracy. The fault evolution prediction model can predict the development trend of faults. The multi-dimensional fault feature fusion model integrates the characteristics of each domain.

[0158] Among them, the ball mill operation data recording module records the various operating parameters, sensor data, control instructions and fault information of the ball mill in real time. The reinforcement learning algorithm is introduced to analyze the recorded data. Through continuous interaction with the environment, the optimal control strategy is learned. The state space S is defined as the operating parameters and sensor data set of the ball mill, and the action space A is the adjustable operating parameter set. The reward function R is designed based on indicators such as grinding efficiency, energy consumption and equipment wear. The control strategy is continuously optimized through the reinforcement learning algorithm, so that the ball mill can achieve efficient and stable operation under different working conditions. A knowledge graph construction system is developed to convert the operating data into structured knowledge. Through entity recognition, relationship extraction and knowledge fusion, a ball mill knowledge graph G = (V, E) is constructed, where V is the entity set and E is the relationship set. The knowledge graph is used for fault diagnosis and decision support:

[0159] Where D is the set of possible decisions, O is the observed anomaly, and P(d|O,G) is the probability that decision d explains anomaly O under the knowledge graph G. An online learning mechanism is designed to enable the system to adapt to changes in working conditions in real time. A concept drift detection algorithm is used to trigger model updates when changes in data distribution are detected: Where f(x) is the characteristic function, N and M are the window sizes of new and old data respectively, and ∈ is the drift threshold. Develop a digital twin-driven reinforcement learning framework and use the digital twin model to generate simulation data for strategy training. The state transition function of the digital twin model is: s t+1 =f(s t ,a t ,ξ t ); where s t is the state, a t For action, t For random disturbances, the control strategy is optimized through the Actor-Critic architecture:

[0160] Among them, π θ is the policy function, Q π For the action value function, a causal inference analysis method is proposed to identify the causal relationship between operating parameters. By constructing a causal graph G = (V, E), the causal effect is calculated using Do-calculus: P(Y = y | do(X = x)) = ∑ z P(Y=y|X=x,Z=z)P(Z=z); where X is the intervention variable, Y is the outcome variable, and Z is the confounding variable. The control strategy is optimized based on the causal relationship to improve system performance.

[0161] In the above technical solution, the operation data recording system established by the operation data recording and analysis steps records various types of ball mill data in real time, providing rich materials for data analysis, introducing reinforcement learning algorithms to learn optimal control strategies and improve the operation efficiency of the ball mill. The knowledge graph construction system converts data into structured knowledge to support fault diagnosis and decision-making; the online learning mechanism enables the system to adapt to changes in working conditions, and the digital twin-driven reinforcement learning framework uses simulation data training strategies to improve learning efficiency; the causal inference analysis method identifies parameter causal relationships and optimizes control strategies. These realize the in-depth mining and utilization of ball mill operation data, and provide strong support for continuously optimizing the operation performance of the ball mill and improving the level of production management.

[0162] Among them, the human-computer interaction interface module can be used by the operator to view the operating status, various parameters, fault warning information, etc. of the ball mill in real time. At the same time, as another embodiment, virtual reality (VR) technology can also be introduced, and the operator can use VR equipment to immersively view the internal structure and operating status of the ball mill, which is convenient for equipment inspection and maintenance. In addition, the human-computer interaction interface integrates an intelligent voice assistant function, supports voice command input and operation prompts, and improves the convenience of operation. The interface also has functions such as data query, report generation and printing, which is convenient for production management and data analysis. An augmented reality (AR) assisted maintenance system is developed to display equipment maintenance information in real time through AR glasses. The system superimposes virtual information with actual equipment to guide operators to perform maintenance operations: I AR =I real +α·Proj(I virtual ,P,C); among them, I AR For augmented reality images, I real is a real scene image, I virtual is virtual information, α is transparency, Proj is projection function, P is camera parameter, C is coordinate transformation matrix, and the emotional computing interaction module is designed to adjust the interface display and interaction mode by analyzing the operator's voice tone, facial expression, etc.: M adjust =f(E(S),H); where E(S) is the emotional state extracted from speech and expression, H is the historical interaction record, f is the emotional response function, and M adjust Adjust the strategy for the interface.

[0163] At the same time, as another embodiment, a holographic projection interactive interface can also be designed to display the operating status of the ball mill in three-dimensional space through holographic projection technology, and the operator can interact with the holographic interface through gestures and voice: holo =HoloProj(V 3D ,P view ); where V 3D For the three-dimensional data model, Pview To observe point parameters, the interface supports multi-user collaborative operation to improve team collaboration efficiency.

[0164] In the above technical solutions, in the human-computer interaction interface design link, traditional VR, intelligent voice assistant and other technologies have improved the convenience of operation.

[0165] Among them, the system integration module forms a complete ball mill control and management system. During the operation of the system, digital twin technology is introduced to establish a digital twin model of the ball mill. By synchronizing the operating data of the ball mill in real time, the digital twin model is updated and optimized. The digital twin model is used to conduct virtual simulation experiments to test the impact of different control strategies and parameter adjustments on the operating performance of the ball mill, providing a reference for the optimization of the actual system. At the same time, according to the new material properties and production process requirements, combined with the simulation results of the digital twin model, the fault diagnosis model is retrained and the operating parameters are optimized; the control algorithm is adjusted and optimized to improve the control accuracy and response speed of the system.

[0166] Build an intelligent maintenance decision-making system to recommend the best maintenance strategy based on the digital twin model and reinforcement learning. The system calculates the expected benefits E(R) under different maintenance strategies: E(R) = ∑ s∈S P(s|a)·R(s,a); where s is the system state, a is the maintenance action, P(s|a) is the probability that the system is in state s after taking action a, and R(s,a) is the benefit of taking action a in state s. A multi-agent collaborative optimization framework is developed to collaboratively optimize the ball mill control system and the entire production line. Through communication and collaboration between agents, the global optimal solution is achieved:

[0167] Among them, J i is the objective function of the ith agent, u i Its control input, u -i It is the control input for other intelligent agents. A quantum secure communication protocol is designed to ensure the data transmission security of the ball mill control system, and the encryption key is generated using quantum key distribution (QKD) technology: K = QKD(A, B);

[0168] Among them, A and B are the communicating parties, and K is the shared key. Using the characteristics of quantum entanglement to detect potential eavesdropping: Among them, P(a i |b i ) is the conditional probability, P(a i ) is the unconditional probability, N is the number of measurements, and ∈ is the threshold.

[0169] Develop a blockchain-driven equipment management system to manage the entire equipment lifecycle, and upload equipment information and operation records to the chain: Block = (Header, Data); where the Header contains the hash value and timestamp of the previous block, and the Data contains the equipment status, maintenance records, etc. Smart contracts are used to automatically execute maintenance plans: Execute(M) = ifCthenAelseB;

[0170] Here, M represents the maintenance contract, C represents the trigger condition, and A and B represent the execution actions. System integration modules form a complete system. Digital twin technology enables virtual simulation and system optimization. The newly designed quantum secure communication protocol utilizes quantum key distribution technology to ensure data transmission security and prevent information leakage and tampering. The blockchain-driven equipment management system manages the entire equipment lifecycle, utilizing smart contracts to automatically execute maintenance plans and improve the transparency and automation of equipment management. The application of these technologies not only enhances the overall performance and reliability of the ball mill control system but also provides innovative solutions for equipment management and data security, making the ball mill control and management system more intelligent, secure, and efficient.

[0171] In summary, in this embodiment, in terms of operating status monitoring, a variety of sensors are installed to build a comprehensive monitoring system. Combined with the dynamic noise compensation algorithm and the spatiotemporal feature fusion network, it can not only accurately collect key data in real time, but also effectively suppress noise and explore potential data correlations, providing high-quality basic information for subsequent analysis and control. In the data preprocessing and feature extraction links, multi-scale data fusion and other technologies deeply explore signal details, and the spatiotemporal feature fusion network realizes the integration of spatiotemporal features of multi-sensor data, improves data processing accuracy, and lays a solid foundation for model establishment and parameter optimization. In the construction of the mathematical model of the grinding process, the improved secondary pulverization kinetics equation and the grinding efficiency prediction model based on deep learning are more in line with the actual working conditions. The introduction of the quantum hybrid neural network model and the multi-scale grinding kinetics model, when determining the optimization objective function, comprehensively considers multiple factors such as grinding efficiency, energy consumption, equipment wear, quality stability and carbon emissions, establishes a multi-objective optimization function, and flexibly adjusts the weights through the dynamic weight adaptive optimization algorithm to achieve multi-objective balance. Intelligent algorithm-based parameter optimization utilizes an improved quantum particle swarm optimization algorithm, combined with a multi-agent collaborative optimization framework and hybrid optimization methods, significantly enhancing global search and convergence capabilities and enabling coordinated operation of the ball mill and production line. Real-time control and adjustment utilizes an adaptive fuzzy PID control algorithm, a multivariable collaborative control strategy, and a predictive control compensation algorithm to improve control accuracy and response speed, reduce system lag, and utilize a dynamic adaptive control architecture and distributed collaborative control algorithms to enable the system to flexibly adapt to different operating conditions and maintain stable and efficient operation. Regarding fault diagnosis and early warning, a transfer learning-based fault diagnosis model, combined with a quantum entanglement fault diagnosis algorithm and a federated learning fault diagnosis system, significantly improves fault diagnosis efficiency, accuracy, and feature discrimination. A fault evolution prediction model and a multi-dimensional fault feature fusion model can predict fault trends in advance, building a comprehensive and accurate fault diagnosis and early warning system and reducing equipment failure risks and maintenance costs. Operational data recording and analysis utilizes a recording system and algorithms such as reinforcement learning to enable in-depth data mining and utilization. A digital twin-driven reinforcement learning framework and causal inference analysis methods further optimize control strategies and continuously improve ball mill performance.

[0172] While the embodiments of the present invention have been shown and described, it will be apparent to those skilled in the art that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A ball mill for aluminum silicate production, characterized in that: The ball mill comprises: A frame (1), wherein a first bearing seat (3) and a second bearing seat (24) are fixedly connected to both sides of the top of the frame (1); An outer cylinder (2), the outer cylinder (2) being rotatably connected to the frame (1) via a first bearing seat (3) and a second bearing seat (24), an inner cylinder (25) being fixedly connected to the outer cylinder (2), a plurality of steel balls (26) being provided in the inner cylinder (25), and a counterweight adjustment assembly being provided in the steel balls (26); A driving assembly, the driving assembly being located on the frame (1) and being used to drive the outer cylinder (2) to rotate; A discharge pipe (14), the discharge pipe (14) is fixedly connected to the outside of the first bearing seat (3), and a discharge port (15) is provided at a side end of the outer cylinder (2) corresponding to the position of the discharge pipe (14); A feed pipe (21), the feed pipe (21) is fixedly connected to the outside of the second bearing seat (24), and a feed port is provided at a position of the feed pipe (21) corresponding to the side end of the outer cylinder (2); A control interaction unit is used to monitor and adjust the operating status of the ball mill and perform fault diagnosis.

2. A ball mill for aluminum silicate production according to claim 1, characterized in that: The driving assembly comprises a first motor (9), a gear box (10), a first gear (4) and a ring gear (5), wherein the ring gear (5) is fixedly connected to the outer cylinder (2), the first gear (4) is rotatably connected to the frame (1), the gear box (10) is fixedly connected to the frame (1), a second gear (11) and a third gear (12) are rotatably connected in the gear box (10), the first motor (9) is fixedly connected to the frame (1), an output end of the first motor (9) is fixedly connected to the second gear (11), the third gear is coaxially fixedly connected to a synchronous shaft (13), the synchronous shaft (13) is coaxially fixedly connected to the first gear (4), and the first gear (4) is meshed with the ring gear (5).

3. A ball mill for aluminum silicate production according to claim 2, characterized in that: The outside of the discharge pipe (14) is fixedly connected to a first fixing frame (16), the outside of the feed pipe (21) is fixedly connected to a second fixing frame (19), the outside of the first fixing frame (16) is fixedly connected to a second motor (17), the outside of the second fixing frame (19) is fixedly connected to a third motor (20), the output end of the second motor (17) is fixedly connected to a first spiral conveying rod (18), the output end of the third motor (20) is fixedly connected to a second spiral conveying rod (22), a feed hopper (23) is provided above the feed pipe (21), the output end of the feed hopper (23) is fixedly connected to the feed pipe (21), an inlet and outlet (8) are provided in the outer cylinder (2) and the inner cylinder (25), a blocking cover (6) is provided in the inlet and outlet (8), the blocking cover (6) is buckled in the inlet and outlet (8), and a long bolt (7) is passed through the inlet and outlet (8) and the blocking cover (6).

4. The ball mill for producing aluminum silicate according to claim 1, characterized in that: The steel ball (26) is provided with a spiral leaf (27) on the outside, an inner ball (31) is provided in the steel ball (26), the outer part of the inner ball (31) is fixedly connected with a connecting rod (32), the connecting rod (32) is fixedly connected to the steel ball (26), a plurality of sliding holes (33) are provided in the inner ball (31), the counterweight adjustment assembly comprises a first threaded rod (34), a second threaded rod (37) and a third threaded rod (38) which are rotatably connected to the steel ball (26), the bottom of the first threaded rod (34) is fixedly connected with a third bevel gear (43), the outer part of the second threaded rod (37) is fixedly connected with a second bevel gear (42), the third threaded rod (38) is fixedly connected with the second bevel gear (43), the third threaded rod (38) is fixedly connected with the second bevel gear (42), the third threaded rod (34) is fixedly connected with the second bevel gear (42 ... One end of the threaded rod (38) is fixedly connected to the first bevel gear (41), the third bevel gear (43) and the second bevel gear (42) are meshed with each other, and the second bevel gear (42) and the first bevel gear (41) are meshed with each other. The first threaded rod (34) is a bidirectional threaded rod with opposite thread groove directions. The outsides of the first threaded rod (34), the second threaded rod (37) and the third threaded rod (38) are all threadedly connected to a counterweight ball (40). There are two counterweight balls (40) on the outside of the first threaded rod (34). A plurality of guide rods (39) are fixedly connected to the steel ball (26), and the guide rods (39) pass through the counterweight ball (40).

5. A ball mill for aluminum silicate production according to claim 4, characterized in that: The outsides of the first threaded rod (34), the second threaded rod (37) and the third threaded rod (38) corresponding to the outside of the steel ball (26) are all fixedly connected with a fixed block (28), wherein a positioning groove (35) is provided in one of the fixed blocks (28), one end of the first threaded rod (34) extends into the positioning groove (35) and is fixedly connected with a rotating block (36), a positioning plate (29) is provided outside the positioning groove (35), a positioning block (44) is fixedly connected to the bottom of the positioning plate (29), a positioning bolt (30) is threadedly connected in the positioning plate (29), and the positioning bolt (30) is threadedly connected to the fixed block (28), and a hexagonal groove is provided at the position of the positioning block (44) corresponding to the outside of the rotating block (36).

6. A ball mill for aluminum silicate production according to claim 5, characterized in that: The control interaction unit includes a ball mill operation status monitoring module, a ball mill grinding process mathematical model establishment module, an objective function optimization module, a parameter optimization module based on an intelligent algorithm, a real-time control and adjustment module, a fault diagnosis and early warning module, a ball mill operation data recording module, a human-computer interaction interface module, and a system integration module.

7. A ball mill for aluminum silicate production according to claim 6, characterized in that: The ball mill operation status monitoring module includes a vibration sensor, a temperature sensor, and a speed sensor; The vibration sensor is mounted on the first bearing seat (3) or the second bearing seat (24), the temperature sensor is mounted on the outside of the first motor (9), the first bearing seat (3) or the second bearing seat (24), and the speed sensor is located outside the frame (1) and is used to measure the speed of the outer cylinder (2); Assume that the vibration signal is expressed as V(t) in the time domain, the temperature signal collected by the temperature sensor is T(t), and the speed collected by the speed sensor is n(t), where t represents time; The raw signals collected by the sensor are pre-processed. In addition to filtering, noise reduction and normalization, multi-scale data fusion technology is introduced; The vibration signal V(t) and temperature signal T(t) are decomposed into multiple scales, and the signal features at different scales are obtained by wavelet transform. For the vibration signal V(t), the low-frequency component A is obtained after j-layer wavelet decomposition j (t) and high frequency component D j,k (t)(k=1,2,…,j), extract features of components at different scales and construct a multi-scale feature vector V ms : V ms =[mean(A j ),std(A j ),mean(D j,k ),std(D j,k )]; Among them, mean represents the mean calculation function, and std represents the standard deviation calculation function; The temperature signal T(t) is normalized and the formula is: Among them, T norm (t) is the normalized temperature value, T min and T max are the minimum and maximum values ​​of the temperature signal during the monitoring period; A dynamic noise compensation algorithm is introduced to perform adaptive compensation for noise characteristics under different working conditions. The noise energy ratio ρ is calculated based on the spectrum analysis of the vibration signal. noise : Among them, V(f) is the frequency domain representation of the vibration signal, F noise is the frequency range where the noise is mainly distributed, F total For the entire spectrum range; Develop a spatiotemporal feature fusion network to simultaneously process spatially distributed multi-sensor data and time series features, and construct a spatiotemporal feature tensor X st : X st =[X s1 (t),X s2 (t),…,X sn (t)]; Among them, X si (t) represents the feature vector of the i-th sensor at time t, and the spatiotemporal features are extracted through the spatiotemporal convolutional network; 8. A ball mill for aluminum silicate production according to claim 6, It is characterized by: in, The mathematical model building module of ball mill grinding process includes the introduction of nonlinear dynamic model; Assume that the pulverization process of the material in the ball mill conforms to the improved two-stage pulverization kinetic equation: Among them, X i is the content of particles with a particle size of i in the material, k i is the first-order crushing rate constant of particles with a particle size of i, k' i is the secondary crushing rate constant; Establish a grinding efficiency prediction model based on deep learning, use convolutional neural network to learn a large amount of operating data, input is sensor data and operating parameters after feature extraction, and output is the grinding efficiency prediction value Build a dynamic energy consumption prediction model, the actual output power P of the first motor out The calculation is as follows: P out =P in ·or motor ·or trans ; Among them, P in is the input power, η motor is the first motor efficiency, η trans is the transmission efficiency; the first motor efficiency η motor and transmission efficiency η trans Calculated by the following formulas: Among them, P rated is the rated power of the first motor, k1, k2, k3 are empirical coefficients, and n0 is the optimal speed point; Combining quantum computing with neural networks, the output of the quantum layer is calculated as: y q =σ(W q ·U θ |0>+b); Among them, W q is the quantum weight matrix, U θ To parameterize the quantum circuit, |0> is the initial quantum state, β is the bias, σ is the activation function, and the quantum layer output is combined with the classical layer: Among them, W c is the classic weight matrix, b is the bias, and f is the output function. A multi-scale grinding dynamics model is constructed. Considering the differences in the crushing behaviors of materials with different particle sizes, the materials are divided into K intervals according to the particle size. The crushing dynamics equation for each interval is: Among them, X k is the material content in the kth particle size interval, k k and k′ k is the crushing rate constant in this interval, k jk is the conversion rate constant from the jth interval to the kth interval; Among them, the objective function optimization module introduces equipment wear W as the optimization target on the basis of maximizing the original grinding efficiency η and minimizing the energy consumption E, and establishes a three-objective optimization function: Among them, α, β and γ are weight coefficients, and α+β+γ=1, η max is the maximum grinding efficiency, E min is the minimum energy consumption, W max is the maximum allowable equipment wear; The wear degree of the equipment W is based on the wear amount m of the steel ball (26) w And the wear thickness δ of the inner cylinder (25) is calculated as follows: Among them, λ1 and λ2 are weight coefficients, m w,max is the maximum allowable wear of the steel ball (26), δ max is the maximum allowable wear thickness of the inner cylinder (25); Energy consumption E is related to the power P and operating time t of the ball mill. The calculation formula is: The power P of the ball mill is calculated by the following empirical formula: P=k p (M+ρV m )oh 2 R; Among them, k p is the power coefficient; M is the loading capacity; ρ is the density of the material; V m is the volume of the material in the inner cylinder (25); ω is the angular velocity of the ball mill, R is the radius of the inner cylinder (25); Introducing quality stability index Q s , measure the degree of fluctuation of product quality, calculate the standard deviation σ of particle size distribution through real-time monitoring of product particle size distribution: Where σ0 is the target standard deviation; the quality stability index is added to the optimization objective function: Among them, δ is the mass stability weight coefficient; A dynamic weight adaptive optimization algorithm is proposed to automatically adjust the objective function weight according to the operating status. The weight adjustment formula is: Among them, η α ,η β ,η γ ,η δ is the learning rate, is the partial derivative of the objective function with respect to each weight; the carbon emission index C is introduced to establish the green optimization objective function: Among them, C min is the minimum carbon emission target, ∈ is the carbon emission weight coefficient, and the carbon emission C is calculated as: Among them, E i is the consumption of the i-th energy, f i is the carbon emission factor of the i-th energy source; Among them, the parameter optimization module based on intelligent algorithm uses improved quantum particle swarm optimization algorithm to optimize the operating parameters of the ball mill; In the quantum particle swarm optimization algorithm, quantum behavior and chaotic search strategy are introduced, and the position update formula of each particle in the quantum space is: in, is the quantum rotating gate parameter, which is dynamically adjusted through chaotic mapping; δ is the step size factor; L(α) is the quantum Levy flight path; p id (t) is the individual optimal position of the i-th particle; g d (t) is the global optimal position of the entire particle swarm; According to the adaptive learning rate optimization algorithm, the learning parameters of the quantum particle swarm are dynamically adjusted, and the learning rate η(t) changes dynamically with the number of iterations t: Among them, η0 is the initial learning rate, η ∞ is the final learning rate, T is the decay period; A population diversity maintenance mechanism is introduced to calculate the average Euclidean distance D between particles: Where N is the number of particles, x i and x j is the particle position vector; Develop a multi-agent collaborative optimization framework that uses communication and collaboration between agents to: Among them, J i is the objective function of the ith agent, u i Its control input, u -i is the control input for other agents, and each agent uses a deep deterministic policy gradient algorithm to make decisions: μ * (s)=argmax μ Q π (s,μ(s)); Among them, μ * is the optimal strategy, Q π is the action value function, s is the state, and a hybrid optimization method of quantum genetic algorithm and particle swarm algorithm is proposed. The update formula of quantum chromosome is: i i (t+1)=θ i (t)+Δθ·sin(2πft+φ); Among them, θ i is the phase angle of the i-th quantum bit, Δθ is the phase adjustment amount, f is the frequency, φ is the initial phase, and chromosome evolution is achieved through quantum rotation gate operation: The real-time control and regulation module controls the feed rate, rotation speed and inner drum (25) loading capacity of the ball mill in real time according to the obtained optimal operating parameter combination; Adaptive fuzzy PID control algorithm is introduced to adjust the speed of the outer cylinder (2); Assume that the actual speed of the outer cylinder (2) is n actual , the target speed is n target , the adjustment control quantity Δu of the inverter is calculated as follows: First, according to the speed deviation e=n target -n actual and the rate of change of deviation Adjusting PID control parameter k using fuzzy inference rules p 、k i and k d : k p =k p0 +Δk p (e,ec); k i =k i0 +Δk i (e,ec); k d =k d0 +Δk d (e,ec); Among them, k p0 、k i0 and k d0 is the initial PID control parameter, Δk p (e,ec),Δk i (e,ec) and Δk d (e,ec) is the parameter adjustment amount obtained based on fuzzy reasoning; Then, the adjustment control amount Δu is calculated: For the control of the feeding speed, an adaptive fuzzy PID control algorithm is also used to adjust the feeding speed of the second screw conveying rod (22) according to the real-time feedback of the grinding efficiency; Design a multivariable collaborative control strategy, consider the coupling effect between each control parameter, and calculate the coupling degree between parameters by establishing the coupling coefficient matrix C: Among them, y i is the i-th output variable, u j is the jth control variable; Dynamically adjust control parameters based on the coupling coefficient matrix, develop a predictive control compensation algorithm, and predict the system state at future moments; Based on the current state and historical data, a long short-term memory network is used to predict the system output for the next k steps. Among them, y is the system output, u is the control input, n and m are the lengths of historical data, and the control parameters are adjusted in advance according to the prediction results; A dynamic adaptive control architecture is proposed to automatically switch the control strategy according to the system operation status, and the system operation status index S is defined as: Among them, s i is the i-th state feature, w i is the corresponding weight, and switches between different control strategies according to the state index S: Among them, u i is the i-th control strategy, R i For the corresponding state area; develop a distributed cooperative control algorithm to assign control tasks to multiple controllers, each controller is responsible for local control and exchanges information through a communication network: Among them, K i is the control law of the i-th controller, e i is the local error, T ij is the communication weight, is the predicted output of the j-th controller.

9. A ball mill for aluminum silicate production according to claim 6, It is characterized by: in, The fault diagnosis and early warning module establishes a fault diagnosis model based on transfer learning. The method is as follows: First, a deep neural network model is pre-trained on a large amount of general mechanical equipment failure data, and then fine-tuned for the specific failure data of ball mills; The collected sensor data and extracted feature parameters are used as input to determine whether the ball mill has a fault and the type of fault using the trained fault diagnosis model; The improved convolutional long short-term memory network is used as the main structure of the fault diagnosis model, which can simultaneously process the spatial and temporal features of time series data; The output of the model is the failure probability vector P = [p1, p2, ..., p m ], where p i represents the probability of the ball mill experiencing the i-th fault; When the maximum probability p max = When max(P) exceeds the set threshold τ, a fault warning signal is issued and the fault type and cause are provided; Develop a fault evolution prediction model to predict fault development trends based on the Hidden Markov Model; By training HMM with historical fault data, we can obtain the state transition probability matrix A and observation probability matrix B: A ij =P(q t+1 =j|q t =i); B jk =P(o t =k|q t =j); Among them, q t is the hidden state at time t, o t is the observed value at time t; the state sequence is calculated using the Viterbi algorithm to predict the fault evolution path and remaining life; Construct a multi-dimensional fault feature fusion model to fuse the time domain, frequency domain, and time-frequency domain features. Use the autoencoder to extract the essential representation of each domain feature, and then perform feature splicing: F fusion =[AE time (F time ),AE freq (F freq ),AE timefreq (F timefreq )]; Among them, F time 、F freq 、F timefreq They are time domain, frequency domain and time-frequency domain features, AE time AE freq AE timefreq is the corresponding autoencoder; Design a quantum entanglement fault diagnosis algorithm and use the principle of quantum entanglement to improve the distinguishability of fault characteristics; Define the quantum eigenvector ψ: Among them, |f i > is the i-th fault characteristic state, c i is the corresponding probability amplitude; the fault diagnosis result is obtained through quantum measurement: P(k)=|<φ k ∣ψ>| 2 ; Among them, |φ k > is the eigenstate of the kth fault; Develop a federated learning fault diagnosis system to achieve knowledge sharing among multiple ball mills; each ball mill locally trains a fault diagnosis model and uploads the model parameters: Among them, θ i is the model parameter of the i-th ball mill, n i For the sample size, Aggregating global models through federated learning.

10. A ball mill for producing aluminum silicate according to claim 6, It is characterized by: in, The ball mill operation data recording module records the ball mill's operating parameters, sensor data, control instructions and fault information in real time; A reinforcement learning algorithm is introduced to analyze the recorded data and learn the optimal control strategy through continuous interaction with the environment. The state space S is defined as the ball mill's operating parameters and sensor data set, the action space A is the set of adjustable operating parameters, and the reward function R is designed based on grinding efficiency, energy consumption, and equipment wear indicators. Develop a knowledge graph construction system to transform operational data into structured knowledge; Through entity recognition, relationship extraction and knowledge fusion, a ball mill knowledge graph G = (V, E) is constructed, where V is the entity set and E is the relationship set; Leveraging knowledge graphs for fault diagnosis and decision support: Where D is the decision set, O is the observed anomaly, and P(d|O,G) is the probability that decision d explains anomaly O under the knowledge graph G; Design an online learning mechanism to enable the system to adapt to changes in working conditions in real time; use a concept drift detection algorithm to trigger model updates when changes in data distribution are detected: Where f(x) is the characteristic function, N and M are the window sizes of new and old data respectively, and ∈ is the drift threshold; Develop a digital twin-driven reinforcement learning framework and use the digital twin model to generate simulation data for policy training; the state transition function of the digital twin model is: s t+1 =f(s t ,a t ,x t ); Among them, s t is the state, a t For action, t For random disturbances, the control strategy is optimized through the Actor-Critic architecture: Among them, π θ is the policy function, Q π is the action value function; Propose a causal inference analysis method to identify the causal relationship between operating parameters; By constructing a causal graph G = (V, E), the causal effect is calculated using Do-calculus: Among them, X is the intervention variable, Y is the outcome variable, and Z is the confounding variable; Among them, the system integration module introduces digital twin technology during system operation to establish a digital twin model of the ball mill; By synchronizing the ball mill's operating data in real time, the digital twin model is updated and virtual simulation experiments are conducted using the digital twin model to test the impact of different control strategies and parameter adjustments on the ball mill's operating performance. At the same time, based on the new material characteristics and production process requirements, combined with the simulation results of the digital twin model, the fault diagnosis model is retrained and the operating parameters are updated; The system calculates the expected benefits E(R) under different maintenance strategies: Where s is the system state, a is the maintenance action, P(s|a) is the probability that the system is in state s after taking action a, and R(s,a) is the benefit of taking action a in state s. Develop a multi-agent collaborative optimization framework: Among them, J i is the objective function of the ith agent, u i Its control input, u -i Provide control input for other agents; Design a quantum secure communication protocol to ensure the data transmission security of the ball mill control system; use quantum key distribution technology to generate encryption keys: K = QKD(A,B); Among them, A and B are the communicating parties, K is the shared key; the characteristics of quantum entanglement are used to detect potential eavesdropping: Among them, P(a i |b i ) is the conditional probability, P(a i ) is the unconditional probability, N is the number of measurements, and ∈ is the threshold; Develop a blockchain-driven equipment management system to achieve full life cycle management of equipment; upload equipment information and operation records to the chain: Block = (Header, Data); The Header contains the hash value and timestamp information of the previous block, and the Data contains the device status, maintenance records, etc. Smart contracts are used to automatically execute maintenance plans: Execute(M)=ifCthenAelseB; Among them, M is the maintenance contract, C is the trigger condition, and A and B are the execution actions.

Images