Powder mixing processing control method, device and equipment and storage medium

Abstract

The invention belongs to the technical field of industrial control, and discloses a powder mixing processing control method, device and equipment and a storage medium, and the method comprises the steps: collecting target data in a powder mixing processing process in real time; inputting the target data into a prediction model to obtain a probability distribution prediction result of the state of the mixed production line in a plurality of time steps in the future, the prediction model being constructed by adopting a dynamic Bayesian network in advance based on the mixed production line; based on the probability distribution prediction result and the target data, a model prediction control algorithm is adopted to optimize a real-time control strategy, and a control instruction is obtained; and controlling the mixed production line according to the control instruction. The powder is uniformly mixed, and the dynamic adaptability is high.

Description

Technical Field

[0001] The invention belongs to the technical field of industrial control, and in particular relates to a powder mixing processing control method, device, equipment and storage medium. Background Art

[0002] As a key process in industrial production, powder mixing is widely used in chemical, pharmaceutical, food, building materials and other industries. Its main goal is to mix powders with different characteristics in a certain proportion so that the mixture meets the uniformity requirements, thereby ensuring the quality and performance of downstream processes (such as tableting, sintering, packaging, etc.). However, the existing powder mixing processing technology faces many challenges and shortcomings in practice, and it is difficult to meet the needs of modern industry for high-precision, automated and flexible production.

[0003] Currently, many mixing equipment adopts fixed parameter control strategies, such as preset stirring speed and mixing time. However, this method ignores the impact of dynamic changes of materials on the mixing process, often leading to control rigidity, unable to dynamically adjust stirring parameters according to real-time working conditions, and difficult to adapt to the characteristics of different batches of raw materials; lack of feedback mechanism for key process parameters (such as particle distribution and material fluidity), the final mixing effect is more volatile, resulting in insufficient mixing uniformity.

[0004] Therefore, the current powder mixing has problems of insufficient mixing uniformity and poor dynamic adaptability. Summary of the invention

[0005] The object of the present invention is to provide a powder mixing processing control method, device, equipment and storage medium, which can solve the problems of insufficient mixing uniformity and poor dynamic adaptability when mixing powders.

[0006] The first aspect of the present invention discloses a powder mixing processing control method, comprising:

[0007] Real-time collection of target data during powder mixing process;

[0008] Inputting the target data into a prediction model to obtain a probability distribution prediction result of the state of the hybrid production line at a certain time step in the future, wherein the prediction model is pre-constructed based on the hybrid production line using a dynamic Bayesian network;

[0009] Based on the probability distribution prediction result and the target data, a model predictive control algorithm is used to optimize the real-time control strategy to obtain a control instruction;

[0010] The mixed production line is controlled according to the control instructions.

[0011] In some embodiments, in a dynamic Bayesian network, a state transition model is used to describe the relationship between the state and control variables of the hybrid production line at the current time step and the state at the next time step to obtain the prediction model; the probability distribution prediction result includes the predicted value of the state of the hybrid production line and the confidence interval corresponding to the predicted value.

[0012] In some embodiments, obtaining the probability distribution prediction results of the state of the hybrid production line at several future time steps includes:

[0013] Using the Bayesian Information Criterion to balance model complexity and fitting ability to optimize the network topology of the prediction model;

[0014] Calculating the weights of all edges in the network topology of the prediction model and removing edges below a preset threshold;

[0015] The maximum likelihood estimation and expectation maximization algorithms are used to optimize the parameters, and the parameters are iterated repeatedly until convergence, so as to obtain the probability distribution prediction result.

[0016] In some embodiments, based on the probability distribution prediction result and the target data, a model predictive control algorithm is used to optimize the real-time control strategy to obtain a control instruction, including

[0017] According to the probability distribution prediction result and the target data, a system state model is used to predict the behavior of the hybrid production line in several future time steps, and the system state model is used to describe the state of the hybrid production line in the current time step and the influence of the control variables on the state in the next time step;

[0018] Optimize the objective function of the model predictive control algorithm based on the prediction results of the behavior of the hybrid production line in several future time steps to obtain the optimal control variables in several future time steps;

[0019] The optimal control variable of the current time step is extracted from the optimal control variables of several future time steps to obtain the control instruction.

[0020] In some embodiments, the objective function of the model predictive control algorithm is optimized according to the prediction results of the behavior of the hybrid production line in the future several time steps to obtain the optimal control variables in the future several time steps, including:

[0021] Under the constraints, the objective function of the model predictive control algorithm is optimized using a quadratic programming algorithm according to the predicted results of the behavior of the hybrid production line at several future time steps, and the optimal control variables are output.

[0022] In some embodiments, the constraints include: control variable constraints, state variable constraints, process time constraints, equipment safety constraints, and powder property constraints.

[0023] The second aspect of the present invention discloses a powder mixing processing control device, comprising:

[0024] The acquisition module is used to collect target data in the powder mixing process in real time;

[0025] A prediction module, used for inputting the target data into a prediction model to obtain a probability distribution prediction result of the state of the hybrid production line at a certain time step in the future, wherein the prediction model is pre-constructed based on the hybrid production line using a dynamic Bayesian network;

[0026] A strategy optimization module, used to optimize the real-time control strategy using a model predictive control algorithm based on the probability distribution prediction result and the target data to obtain a control instruction;

[0027] A control module is used to control the hybrid production line according to the control instructions.

[0028] In some embodiments, the prediction module includes a network topology unit, an edge unit and a parameter optimization unit. The network topology unit is used to use the Bayesian information criterion to balance the model complexity and fitting ability to optimize the network topology of the prediction model. The edge unit is used to calculate the weights of all edges in the network topology of the prediction model and remove edges below a preset threshold. The parameter optimization unit is used to use maximum likelihood estimation and expectation maximization algorithms to perform parameter optimization, repeatedly iterate until the parameters converge, and obtain the probability distribution prediction result.

[0029] The third aspect of the present invention discloses an electronic device, comprising a memory storing executable program code and a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the powder mixing processing control method disclosed in the first aspect.

[0030] A fourth aspect of the present invention discloses a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the powder mixing processing control method disclosed in the first aspect.

[0031] The beneficial effects of the present invention are that the dynamic correlation between variables in the mixing process is modeled by DBN, the state of the mixing production line in the future time step is predicted, the mixing equipment is dynamically adjusted by MPC using the prediction results and real-time feedback, real-time control is combined with long-term optimization, real-time monitoring, dynamic prediction and closed-loop optimization control can be achieved, and the powder mixing is uniform and the dynamic adaptability is strong. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] The accompanying drawings herein show specific examples of the technical solutions described in the present invention, and together with the specific implementation methods, constitute a part of the specification, and are used to explain the technical solutions, principles and effects of the present invention.

[0033] Unless otherwise specified or defined, the same reference numerals in different drawings represent the same or similar technical features, and the same or similar technical features may also be represented by different reference numerals.

[0034] Figure 1 is a structural diagram of a powder mixing processing control system according to an embodiment of the present invention;

[0035] Figure 2 It is a flow chart of a powder mixing processing control method disclosed in an embodiment of the present invention;

[0036] Figure 3 is a flowchart of DBN optimization according to an embodiment of the present invention;

[0037] Figure 4 is a flow chart of optimization of a model predictive control algorithm according to an embodiment of the present invention;

[0038] Figure 5 is a schematic structural diagram of a powder mixing processing control device according to an embodiment of the present invention;

[0039] Figure 6 It is a structural schematic diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] Unless otherwise specified or defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art. In the case of combining the technical solution of the present invention with realistic scenarios, all technical and scientific terms used herein may also have meanings corresponding to the purpose of implementing the technical solution of the present invention. The "first, second..." used herein is only used to distinguish the names and does not represent a specific quantity or order. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0041] It should be noted that when a component is considered to be "fixed to" another component, it can be directly fixed to the other component or there can be a central component; when an component is considered to be "connected to" another component, it can be directly connected to the other component or there can be a central component at the same time; when an component is considered to be "installed on" another component, it can be directly installed on the other component or there can be a central component at the same time. When an component is considered to be "set on" another component, it can be directly set on the other component or there can be a central component at the same time.

[0042] Unless otherwise specified or defined, the "said" and "the" used in this document refer to the technical features or technical contents mentioned or described before the corresponding position, and the technical features or technical contents may be the same as or similar to the technical features or technical contents mentioned therein. In addition, the terms "including" and "having" and any variations thereof used in this document are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or devices.

[0043] Differences in powder size, density, humidity, etc. can lead to stratification and segregation during the mixing process, making it difficult to achieve ideal uniformity. Traditional mixing processes rely on fixed stirring parameters (such as preset stirring speed and mixing time) to operate, and fail to monitor and dynamically adjust the mixing process in real time, making it difficult to ensure uniformity; during the powder mixing process, the physical interaction characteristics between particles will change with changes in environmental conditions (such as temperature and humidity). Existing technologies lack a response mechanism to these dynamic changes, resulting in insufficient mixing uniformity.

[0044] Moreover, the operating parameters of traditional powder mixing equipment are mostly set by manual experience. The adjustment of key parameters such as mixing time, speed, and direction usually requires repeated experiments, which is time-consuming and has unstable effects. For different powders (such as light powder, heavy powder or high viscosity powder) or different batches of production, the equipment parameters need to be readjusted, which increases switching costs and uncertainties, resulting in parameter optimization relying on experience.

[0045] In order to solve the above problems, the present invention combines model predictive control (MPC) with dynamic Bayesian network (DBN) to establish a powder mixing process control method that combines real-time regulation with trend prediction in parallel, fundamentally improving the powder mixing uniformity, adaptability and efficiency. Among them, DBN is used to model the dynamic correlation between variables in the mixing process and predict the state of the mixing production line in future time steps; MPC uses the prediction results and real-time feedback to dynamically adjust the mixing equipment to form a powder mixing process control system that combines real-time control with long-term optimization. Through real-time monitoring, dynamic prediction and closed-loop optimization control of the system, the problems of insufficient powder mixing uniformity and poor dynamic adaptability are solved.

[0046] like Figure 1As shown, the powder mixing processing control system of this embodiment includes a perception layer, a decision layer and an execution layer. Among them, the perception layer uses sensors or equipment such as flow meters, temperature and humidity sensors, bin level sensors, uniformity detectors, etc. to collect key data in the powder mixing process in real time, such as silo status, feeding rate, mixing uniformity, environmental parameters, etc., to provide accurate status information for the decision layer; the decision layer includes DBN and MPC, and DBN uses historical data and real-time data to establish a dynamic Bayesian network to model the dynamic behavior of the system. By modeling the uncertainty and dynamic changes of the system, a probability distribution prediction of the state of the system (such as mixing uniformity, silo feeding dynamics) for a period of time in the future is provided to capture the impact of uncertainty and external disturbances on system performance; MPC is based on the prediction information provided by DBN, rolling optimization of the control strategy within a short time scale, and generates specific control instructions by constraining the optimization objective function. DBN and MPC work together to perform dynamic prediction and real-time control strategy optimization. The execution layer adjusts the feeding rate, mixing parameters and other equipment actions by controlling the feeding valve, conveying equipment, stirring motor, silo control and other equipment according to the control instructions generated by MPC to complete the powder mixing processing task.

[0047] When the above-mentioned powder mixing processing control system is running, the powder mixing processing control method of this embodiment is executed. The executor of this method can be an electronic device such as a computer, a laptop, a tablet computer, a single-chip microcomputer, or a control chip embedded in an electronic device, and the present invention is not limited to this.

[0048] In order to facilitate the understanding of the present invention, specific embodiments of the present invention will be described in more detail below with reference to the accompanying drawings.

[0049] like Figure 2 As shown, the method comprises the following steps:

[0050] Step S100: real-time acquisition of target data during the powder mixing process;

[0051] Among them, the target data is the key data related to the powder mixing process, such as the state information of the powder (particle size, humidity, density, etc.) and the operation data of the equipment (speed, torque, etc. of the agitator). These key data can be collected in real time through sensors to obtain the target data. The target data mainly includes the system state X at the current time step t t and the control variable U at the current time step t The target data is used to input the dynamic Bayesian network to make probability distribution predictions on the status of the hybrid production line.

[0052] Step S200: inputting the target data into the prediction model to obtain the probability distribution prediction results of the state of the hybrid production line at several future time steps, wherein the prediction model is pre-constructed based on the hybrid production line using a dynamic Bayesian network;

[0053] Powder mixing involves dynamic interactions of multiple variables (such as particle size distribution, stirring speed, feeding order, etc.). There are complex nonlinear relationships between these variables, which are difficult to model and analyze with traditional methods. The present invention uses a dynamic Bayesian network to build a prediction model. In the dynamic Bayesian network, a state transition model is used to describe the relationship between the state of the current time step and the control variable and the state of the next time step, and a prediction model is built and trained. Then, based on the state X in the target data, the prediction model is trained. t and the control variable U t , use the trained prediction model to predict the state of the hybrid production line in the future T time steps, and obtain the probability distribution prediction results of the state. For example, predict the distribution and uniformity of particles in the next few seconds.

[0054] In this embodiment, the prediction model not only provides the predicted value of the state of the hybrid production line, but also provides the confidence interval of the predicted value. The confidence interval helps to quantify the uncertainty of the future state and provide more information to the subsequent MPC module for decision-making.

[0055] Specifically, the network structure of DBN consists of input nodes, hidden nodes and output nodes, and the dynamic prediction of the state of the mixing production line is realized through time series modeling. Among them, the data of the input nodes include material characteristics (such as particle size distribution, density, humidity), control variables (such as stirring motor speed, processing time, direction and other equipment operating parameters) and environmental conditions (such as the external environment that affects the mixing process). The hidden nodes are potential state representations that capture dynamic changes that cannot be directly observed during the mixing process (such as particle interaction characteristics and mixing stratification phenomena). The output node outputs the predicted target (such as uniformity wash UC, particle size standard deviation SD and other mixing uniformity indicators) and the recommended control range (such as future adjustment suggestions for stirring motor speed, processing time, direction, etc.).

[0056] DBN describes the relationship between the current state, control variables and the next state through the state transition model. The specific formula is as follows: P(X t |X t-1 ,U t )=Softmax(W·X t-1 +V·U t +b), where X t Represents the current state variable (such as particle distribution, uniformity index), for example: particle size distribution [0.1, 0.2, 0.3], X t-1 represents the state variable at the previous moment, U trepresents the control variable (such as stirring speed is 120RPM, stirring direction is clockwise), W represents the state transfer weight matrix, describing X t-1 X t Assuming that the system state X is an n-dimensional vector, representing n state variables (such as particle size distribution, uniformity index, etc.), then W is an n×n weight matrix, which represents the following dynamic relationships: 1. Time correlation of particle distribution, such as the proportion of large particles may gradually decrease while the proportion of small particles increases; 2. Uniformity evolution law, such as the current uniformity level is directly affected by the uniformity and particle distribution at the previous moment. The expression of W is:

[0057]

[0058] Among them, w ij Indicates the state of the previous time step The current time step state The impact strength, positive value indicates positive impact, negative value indicates negative impact, the larger the value, the stronger the impact.

[0059] V represents the control input weight matrix, describing U t X t Assume that the control variable U t is an m-dimensional vector (such as stirring speed and direction), the system state X t is an n-dimensional vector, then V is an n×m matrix. V is used to characterize the effect of stirring speed on particle distribution. For example, faster stirring speed may lead to more severe particle collision and more uniform particle distribution. It can also characterize the effect of feeding sequence on uniformity. For example, adding light powder first and then heavy powder can reduce stratification. The expression of V is:

[0060]

[0061] Among them, v ij Represents the control variable Status The intensity of impact.

[0062] b represents the bias vector, which corrects the prediction result.

[0063] In the prediction model, by optimizing the network structure and parameters of the dynamic Bayesian network, the probability distribution prediction results of the state of the hybrid production line in the future several time steps can be obtained, such as Figure 3 As shown, the specific steps of optimization include:

[0064] Step S210: Using the Bayesian Information Criterion to balance model complexity and fitting ability to optimize the network topology of the prediction model;

[0065] Starting from the fully connected network, the edges are gradually removed, and the BIC (Bayesian Information Criterion) value after each removal is calculated, and the structure with the smallest BIC is retained as the optimization result. The specific expression is: BIC = -2·log(L)+k·log(n), where L represents the likelihood function value of the model, k represents the number of free parameters (number of node connections), and n represents the number of training data samples.

[0066] Step S220: Calculate the weights of all edges in the network topology of the prediction model, and remove edges below a preset threshold;

[0067] Calculate the weights of all edges in the network topology of the prediction model, remove connections with lower edge weights to reduce redundancy, and perform edge weight screening. The specific expression is: Among them, θ ij represents the conditional probability parameter from node i to node j.

[0068] Step S230: Optimize parameters using maximum likelihood estimation and expectation maximization algorithms, iterate repeatedly until the parameters converge, and obtain probability distribution prediction results.

[0069] First, the maximum likelihood estimation is used to fit the state transition probability to maximize the prediction model's ability to explain the training data. The specific expression is: Optimize weights using gradient descent η represents the learning rate; it is iterated until the log-likelihood function L converges.

[0070] Then, the expectation maximization algorithm is used to optimize the parameters related to the hidden nodes and improve the model's ability to fit the potential state pairs. The specific expression is: Among them, H represents hidden nodes, and θ is the model parameter, including W, V, and b. Specifically, first calculate the posterior distribution of hidden nodes, and then update the parameter θ (k+1) Let Q(θ|θ (k) ) is maximized, and the prediction error and confidence interval are calculated in real time, and the back-propagation optimization parameters W and V are performed.

[0071] After updating, the parameters are used for the next round of prediction, the error feedback is corrected, and the iteration is repeated until the parameters converge to obtain the probability distribution prediction results.

[0072] Step S300: Based on the probability distribution prediction results and target data, a model predictive control algorithm is used to optimize the real-time control strategy to obtain a control instruction;

[0073] The model predictive control algorithm is a rolling optimization algorithm that dynamically adjusts the control variables through the following three steps: 1. Combine the system model and current state to predict the system behavior in the future time step; 2. Based on the prediction results, optimize the objective function and calculate the optimal control variables for the future time step; 3. Execute the optimization results of the current time step and perform real-time adjustment and rolling optimization.

[0074] Specifically, after receiving the prediction value and confidence interval from DBN, the model predictive control algorithm is used in combination with the target data to predict the system behavior in the next T time steps. Then, the optimal control variable for the current time step is calculated based on the objective function and constraints using methods such as QP (Quadratic Programming) or gradient optimization. Obtain control instructions to adjust the operation of equipment such as agitators. During the optimization process, physical constraints of the equipment (such as stirring speed range, energy consumption limit, etc.) and process constraints (such as mixing uniformity requirements, time limit, etc.) are considered to ensure that the optimization results meet the actual operating conditions.

[0075] In this embodiment, Figure 4 As shown in Figure 2, the specific steps of model predictive control algorithm optimization include:

[0076] Step S310: Based on the probability distribution prediction results and the target data, the system state model is used to predict the behavior of the hybrid production line in several future time steps, wherein the system state model is used to describe the state of the hybrid production line in the current time step and the influence of the control variables on the state in the next time step;

[0077] When modeling, the system state model is used to describe the state of the mixed production line at the current time step X t and the control variable U t For the state X of the next time step t+1 The state transfer formula is: X t+1 =f(X t , U t )+ε, where X t Indicates the state of the mixed production line at the current time step (such as particle distribution, uniformity index, etc.); U t represents the control variable (such as stirring speed, direction); f(X t , U t ) represents the state transfer function, which is supported by the DBN prediction results; ε represents the modeling error or noise term.

[0078] Step S320: Optimizing the objective function of the model predictive control algorithm according to the prediction results of the behavior of the hybrid production line in the future several time steps, and obtaining the optimal control variables in the future several time steps;

[0079] The optimization goal of MPC is to maximize mixing uniformity while minimizing energy consumption and time. The specific expression of the objective function is: Among them, UC(t) represents the uniformity coefficient, E(t) represents the energy consumption, T(t) represents the mixing time, and w1, w2, and w3 represent weight coefficients that control the importance of each optimization objective.

[0080] The core of MPC is to dynamically adjust the control variable U by optimizing the objective function. t To meet the process objectives. In MPC, the objective function is usually a quadratic function, and the constraints are linear. Therefore, under the constraints, this embodiment uses a quadratic programming algorithm to optimize the objective function of the model predictive control algorithm based on the predicted results of the behavior of the hybrid production line in several future time steps, and solves and outputs the optimal control variables. During MPC optimization, a rolling optimization strategy is adopted, using the optimization results of a limited prediction time domain to dynamically update the control strategy at each time step, which has strong real-time and adaptability.

[0081] Specifically, first set the prediction time domain. The prediction time domain is the time window (such as 5 seconds) used when MPC optimizes control variables. It means that the system will predict the state changes within the next T seconds and optimize the corresponding control variables. The longer the prediction time domain, the stronger the global optimization ability, but the real-time performance and computing cost may decrease; a time domain that is too short may cause "short-sightedness" problems. Short time domains (such as T = 3 seconds) are suitable for working conditions with fast dynamic changes, and real-time performance is prioritized; long time domains (such as T = 10 seconds) are suitable for more stable working conditions, which helps to improve the global optimization effect. After experimental testing in this embodiment, T = 5 seconds is selected. Then rolling update is performed. At each time step t, the future state X within T seconds is predicted based on the system state model. t+1 , X t+2 , ..., X t+T , optimize the control variable U t , U t+1 , ..., U t+T , at time t+1, get the latest state X t+1 , update the prediction model, and roll-adjust the optimization window (i.e., the prediction starting point and optimization variables).

[0082] When optimizing control variables, MPC needs to satisfy the physical and process constraints of the system, such as control variable constraints, state variable constraints, process time constraints, equipment safety constraints, and powder property constraints.

[0083] Among them, the control variable constraints include control variable range constraints and control variable change rate constraints. The control variable range constraint means that the control variable (such as stirring speed) needs to operate within the allowed range of the equipment: U min ≤U t ≤U max , where U tis the control variable vector of the current time step; U min ,U max The control variable change rate is the upper and lower limits of the control variable, and the specific value depends on the equipment capacity and material characteristics. The control variable change rate means limiting the change rate of the control variable (stirring speed) to avoid sudden changes in the system that may lead to a decrease in mixing effect or damage to the equipment. The specific expression is: ΔU min ≤U t -U t-1 ≤Δ max ,ΔU min ,ΔU max Indicates the upper and lower limits of the rate of change of the controlled variable.

[0084] The state variable constraints include system physical state constraints and dynamic Bayesian network prediction constraints. The system physical state constraints indicate that the system state variables need to comply with physical laws and equipment capacity limitations: min ≤X t ≤X max , where X t is the state vector of the current time step, X min , X max The upper and lower limits of the state variable, for example: the uniformity index of powder mixing must be higher than a certain threshold (such as UC min =95%) to meet product quality requirements. Dynamic Bayesian Network prediction constraints use the confidence interval provided by DBN as the dynamic constraint condition of the state variable: in, Represents the lower and upper confidence bounds of the DBN prediction.

[0085] The process time constraint includes the total time constraint and the mixing stage time limit. The total time constraint is the total time of the mixing process that needs to meet the production cycle requirement: T process ≤T max , where T process is the actual mixing time; T max It is the time limit of the production process. The mixing stage time limit means that the duration of certain stages (such as high-speed mixing stage) needs to be limited to a range: T min，stage ≤T stage ≤T max，stage , to prevent the mixing time from being too long, resulting in energy waste, or too short, affecting uniformity.

[0086] Equipment safety constraints include torque limit, temperature limit, and vibration amplitude limit. Torque limit means that the torque of the agitator must be limited within the equipment design safety range: τ t ≤τ max , where τ t Indicates the torque value of the current time step; the temperature limit indicates that the operating temperature of the device must not exceed the maximum temperature allowed by the device: T device≤T max , where T device The vibration amplitude of the mixing equipment must be controlled within the allowable range to prevent equipment damage or affect material properties: A t ≤A max , where A t Indicates the vibration amplitude at the current time step.

[0087] Powder property constraints include stratification and segregation limits and dust control. Stratification and segregation limits monitor the stratification phenomenon during the mixing process and limit the particle distribution difference: SDparticle≤SDmax, where SD particle is the standard deviation of particle distribution; dust control limits the amount of dust generated during the mixing process to ensure that the production environment meets safety regulations: D t ≤D max , where D t Indicates dust concentration (real-time monitoring).

[0088] Step S330: extracting the optimal control variable of the current time step from the optimal control variables of several future time steps to obtain the control instruction.

[0089] Run MPC, predict the state in the future time step T according to the system state model, optimize the objective function J, determine the optimal control variables for several future time steps, and obtain the control variable sequence Find the optimal control variable for the current time step from the control variable sequence Get control instructions.

[0090] Step S400: Control the mixed production line according to the control instruction.

[0091] Optimal control variables calculated based on MPC Execute system control (such as adjusting the agitator speed, changing the feeding order, etc.). The system status after execution is fed back to the system through sensors again, and new status data (such as particle distribution, mixing uniformity, etc.) is collected in real time as input for the next step of optimization.

[0092] After each time step, MPC will roll over the prediction window based on the latest system feedback and DBN prediction results to re-optimize the control variables. The rolling update process ensures that the mixing production line can respond to environmental changes and changes in process requirements in real time. As time goes by and new data is fed back, MPC dynamically adjusts the control variables to maintain the efficiency and stability of the powder mixing process.

[0093] In summary, this embodiment uses DBN to model the dynamic correlation between variables in the mixing process, predicts the state of the mixing production line in the future time step, and uses MPC to dynamically adjust the mixing equipment using the prediction results and real-time feedback. It combines real-time control with long-term optimization, and can perform real-time monitoring, dynamic prediction, and closed-loop optimization control, thus solving the problems of insufficient powder mixing uniformity and poor dynamic adaptability.

[0094] like Figure 5 As shown, based on the above powder mixing processing control method, the embodiment of the present invention discloses a powder mixing processing control device, including:

[0095] The acquisition module 600 is used to collect target data in the powder mixing process in real time;

[0096] A prediction module 610 is used to input the target data into a prediction model to obtain a probability distribution prediction result of the state of the hybrid production line at a certain time step in the future, wherein the prediction model is pre-constructed based on the hybrid production line using a dynamic Bayesian network;

[0097] A strategy optimization module 620 is used to optimize the real-time control strategy using a model predictive control algorithm based on the probability distribution prediction result and the target data to obtain a control instruction;

[0098] The control module 630 is used to control the hybrid production line according to the control instructions.

[0099] like Figure 6 As shown, an embodiment of the present invention discloses an electronic device, including a memory 401 storing executable program codes and a processor 402 coupled to the memory 401;

[0100] The processor 402 calls the executable program code stored in the memory 401 to execute the powder mixing processing control method described in the above embodiments.

[0101] The embodiments of the present invention further disclose a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the powder mixing processing control method described in the above embodiments.

[0102] The purpose of the above embodiments is to exemplarily reproduce and deduce the technical solution of the present invention, and to fully describe the technical solution, purpose and effect of the present invention. Its purpose is to make the public understand the disclosed content of the present invention more thoroughly and comprehensively, and it does not limit the scope of protection of the present invention.

[0103] The above embodiments are not exhaustive enumerations of the present invention, and there may be multiple other implementations not listed. Any replacement and improvement made without violating the concept of the present invention shall fall within the protection scope of the present invention.

Claims

1. A powder mixing processing control method, characterized in that: include: Real-time collection of target data during powder mixing process; Inputting the target data into a prediction model to obtain a probability distribution prediction result of the state of the hybrid production line at a certain time step in the future, wherein the prediction model is pre-constructed based on the hybrid production line using a dynamic Bayesian network; Based on the probability distribution prediction result and the target data, a model predictive control algorithm is used to optimize the real-time control strategy to obtain a control instruction; The mixed production line is controlled according to the control instructions.

2. The powder mixing processing control method according to claim 1, characterized in that: In a dynamic Bayesian network, a state transition model is used to describe the relationship between the state of the hybrid production line at the current time step and the control variables and the state at the next time step to obtain the prediction model; the probability distribution prediction result includes the predicted value of the state of the hybrid production line and the confidence interval corresponding to the predicted value.

3. The powder mixing processing control method according to claim 1, characterized in that: The method of obtaining the probability distribution prediction results of the states of the hybrid production line at several future time steps includes: Using the Bayesian Information Criterion to balance model complexity and fitting ability to optimize the network topology of the prediction model; Calculating the weights of all edges in the network topology of the prediction model and removing edges below a preset threshold; The maximum likelihood estimation and expectation maximization algorithms are used to optimize the parameters, and the parameters are iterated repeatedly until convergence, so as to obtain the probability distribution prediction result.

4. The powder mixing processing control method according to claim 1, characterized in that: Based on the probability distribution prediction result and the target data, a model predictive control algorithm is used to optimize the real-time control strategy to obtain a control instruction, including According to the probability distribution prediction result and the target data, a system state model is used to predict the behavior of the hybrid production line in several future time steps, and the system state model is used to describe the state of the hybrid production line in the current time step and the influence of the control variables on the state in the next time step; Optimize the objective function of the model predictive control algorithm based on the prediction results of the behavior of the hybrid production line in several future time steps to obtain the optimal control variables in several future time steps; The optimal control variable of the current time step is extracted from the optimal control variables of several future time steps to obtain the control instruction.

5. The powder mixing processing control method according to claim 4, characterized in that: The objective function of the model predictive control algorithm is optimized according to the prediction results of the behavior of the hybrid production line in the future several time steps to obtain the optimal control variables in the future several time steps, including: Under the constraints, the objective function of the model predictive control algorithm is optimized using a quadratic programming algorithm according to the predicted results of the behavior of the hybrid production line at several future time steps, and the optimal control variables are output.

6. The powder mixing processing control method according to claim 5, characterized in that: The constraints include: control variable constraints, state variable constraints, process time constraints, equipment safety constraints and powder property constraints.

7. A powder mixing processing control device, characterized in that: include: The acquisition module is used to collect target data in the powder mixing process in real time; A prediction module, used for inputting the target data into a prediction model to obtain a probability distribution prediction result of the state of the hybrid production line at a certain time step in the future, wherein the prediction model is pre-constructed based on the hybrid production line using a dynamic Bayesian network; A strategy optimization module, used to optimize the real-time control strategy using a model predictive control algorithm based on the probability distribution prediction result and the target data to obtain a control instruction; A control module is used to control the hybrid production line according to the control instructions.

8. The powder mixing processing control device according to claim 7, characterized in that: The prediction module includes a network topology unit, an edge unit and a parameter optimization unit. The network topology unit is used to use the Bayesian information criterion to balance the model complexity and fitting ability to optimize the network topology of the prediction model. The edge unit is used to calculate the weights of all edges in the network topology of the prediction model and remove edges below a preset threshold. The parameter optimization unit is used to use maximum likelihood estimation and expectation maximization algorithms to perform parameter optimization, repeatedly iterate until the parameters converge, and obtain the probability distribution prediction result.

9. An electronic device, characterized in that: It comprises a memory storing executable program code and a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the powder mixing processing control method according to any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the powder mixing processing control method according to any one of claims 1 to 6.

Images