Digital twin intelligent decision-making process prediction method and system for traditional Chinese medicine vacuum belt drying
Through digital twin technology and artificial intelligence, an intelligent decision-making model for vacuum belt drying was established, which solved the problem of lack of system data support for vacuum belt drying processes in the traditional Chinese medicine and pharmaceutical industry, achieved product quality prediction and process parameter optimization, and improved production efficiency and consistency of quality.
Patent Information
- Application Number
- CN202210671995.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-06-15
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2042-06-15
AI Technical Summary
The vacuum belt drying process in the traditional Chinese medicine pharmaceutical industry lacks system data support, resulting in high energy consumption, high cost and quality batch differences in production, making it difficult to achieve process optimization control.
Using digital twin technology and artificial intelligence, a vacuum belt drying simulation model is established based on the principle of discrete elements, combined with DQN algorithm and Markov decision-making process, an intelligent decision-making model is formed, and product quality prediction and process parameters are realized through online detection and optimization of process parameters. Dynamic adjustment.
The product quality prediction and process parameter optimization of the Chinese medicine vacuum belt drying process are realized, which improves production efficiency and product quality consistency, and reduces energy consumption and costs.
Smart Images

Figure CN114896821B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of intelligent technology for traditional Chinese medicine pharmaceutical manufacturing, and in particular relates to a digital twin intelligent decision-making process prediction method and system for traditional Chinese medicine vacuum belt drying. Background Art
[0002] The scale of my country's traditional Chinese medicine pharmaceutical industry has formed a momentum of vigorous development with the overall improvement of the manufacturing level. Restricted by the fluctuation of ingredients between batches of upstream raw materials, the "three difficulties" of modeling difficulty, control difficulty and optimization decision-making difficulty in the downstream traditional Chinese medicine pharmaceutical production process have become increasingly prominent. Due to the lack of in-depth and scientific production data support, the level of transformation of modern industrial technology in the traditional Chinese medicine pharmaceutical industry is still hovering at a low level, and the "three difficulties" seriously restrict the high-quality development of the traditional Chinese medicine industry. In order to solve the technical bottleneck of the traditional Chinese medicine pharmaceutical industry, it is necessary to establish a value transfer model of key quality attributes for the key pharmaceutical processes and their upstream and downstream combinations, and use artificial intelligence technologies such as reinforcement learning based on the model to dig out the transfer law of product quality, form a drug quality optimization control strategy with the characteristics of traditional Chinese medicine pharmaceuticals, and finally solve the problem of process optimization control of the pharmaceutical process. Through the gradual upgrading and transformation of the automation, digitization, informatization and intelligence of the pharmaceutical process, the optimized operation, control and management of the whole system can be realized, thereby building a high-quality development model for traditional Chinese medicine pharmaceuticals.
[0003] Traditional Chinese medicine vacuum belt drying is a key process in the production of traditional Chinese medicine, which directly restricts the quality of the final traditional Chinese medicine products. At present, its process operation is still mainly based on manual experience and lacks the support of necessary system data, which directly leads to high energy consumption, high cost and quality batch differences for production enterprises. Summary of the invention
[0004] The present invention proposes a digital twin intelligent decision-making process prediction method and system for traditional Chinese medicine vacuum belt drying, which forms intelligent decision-making based on the numerical simulation model of digital twins and artificial intelligence technology to improve the industry technical level of traditional Chinese medicine manufacturing.
[0005] To achieve the above object, the technical solution of the present invention is achieved as follows:
[0006] A digital twin intelligent decision-making process prediction method for traditional Chinese medicine vacuum belt drying, comprising:
[0007] S1. Based on the movement law of Chinese herbal medicine extract in the vacuum belt dryer, a simulation model is established for the entire vacuum drying process of Chinese herbal medicine extract to predict product quality parameters;
[0008] S2. Based on the DQN algorithm, the virtual production data generated by the simulation model is used for training to establish a vacuum belt drying intelligent decision-making model; the virtual production data includes product quality parameters corresponding to different production process conditions;
[0009] S3, online detection of physical parameters of Chinese medicine extract; online reading of process parameters of vacuum belt dryer;
[0010] S4, inputting the detected physical property parameters and process parameters into the simulation model to obtain predicted product quality parameters;
[0011] S5. Input the predicted product quality parameters and the detected process parameters into the vacuum belt drying intelligent decision-making model to obtain optimized process parameters.
[0012] Furthermore, the method for establishing the simulation model in step S1 includes:
[0013] Based on the discrete element principle, the Chinese medicine extract on the conveyor belt is decomposed into several discrete elements of the same size, and the solution space is discretized into a unit array composed of multiple discrete elements. The two adjacent discrete elements are connected through the change relationship of the moisture content of the discrete elements. The continuous vacuum drying process of the discrete elements corresponds to the process of the discrete elements moving in space. The state at the end of the previous second is the initial state of the next second. Based on this basic principle, the discrete elements are iteratively calculated and the entire vacuum drying process is modeled, including the three-stage model of boiling, transition and evaporation. The state of the discrete element in the last step is the product quality parameter of the discharge of the vacuum belt drying.
[0014] Furthermore, the iterative calculation of the discrete element includes:
[0015]
[0016]
[0017]
[0018] Q 吸热 =Cp*M*T Formula (4)
[0019] Q 放热 =M 水 *r Formula (5)
[0020] ΔQ=Q 吸热 +Q 1 -Q 放热 Formula (6)
[0021]
[0022]
[0023] M'=M 初始 -M 水 Formula (8)
[0024]
[0025] Among them, M' is the discrete element water content at the current moment; M 1 is the amount of water evaporated per unit time in the boiling stage, i.e., the boiling mass transfer flux; η is the material viscosity; r is the latent heat of vaporization of water; g is the gravitational acceleration; ρ l is the material density; v is the water vapor density; σ is the surface tension of the material; Pr is the Prandtl number; C pl is the specific heat capacity of the solute; C wl is the empirical constant for the heating surface-liquid combination; S is the contact area between the discrete element and the conveyor belt; t is the time when the discrete element completely leaves the current heating position;
[0026] M 2 is the amount of water evaporated per unit time in the evaporation stage, i.e., the evaporation mass transfer flux; D is the water vapor diffusion coefficient; T is the material temperature; P is the vacuum system pressure; C 溶质 is the solid content of the material; Q 1 is the heat transfer flux, which is used to calculate the heat change of the discrete element; λ is the thermal conductivity of the material; d is the thickness of the cake; Δt is the temperature difference; Q 吸热 is the heat absorbed by the discrete element; Cp is the specific heat capacity at constant pressure; M is the mass of the discrete element; Q 放热 Take away the heat for water evaporation; M 水 is the mass of evaporated water; r is the latent heat of vaporization of water; ΔQ is the heat change value; M 溶质 is the mass of the solute in the discrete element; M 初始 is the water content in the discrete element before heating begins;
[0027] The material in the above formula is Chinese herbal medicine extract. The heat change of the discrete element per unit time is shown in formula (6). After the above process, the water evaporation amount in the discrete element is continuously calculated in a cycle until the drying process is completed.
[0028] Furthermore, in step S2, the training process includes:
[0029] Referring to the Markov decision process, the Markov decision process M consists of four elements, M = (S, A, P, R). Where S represents the state set (s i ∈S),s i represents the state of the i-th step; A represents a set of actions (a i ∈A),a i represents the action of step i; P represents the state transition probability, which means that in the current (s i ∈S) state, after (a i∈A), the probability distribution of transferring to other states; R is the reward function, which gives feedback on the current state; the actions are the increase and decrease of various production process parameters, mainly including the adjustment of heating temperature, the adjustment of conveyor belt speed, and the adjustment of feeding speed; the adjustment range of the action is the action space, which is used to constrain the adjustment range of the action; finally, the reward function is determined, which is used to calculate the feedback of the environment on the action and comes from the moisture content of the product outlet.
[0030] Furthermore, the online detection of the physical properties of the Chinese medicine extract in step S3 includes detecting the viscosity, density and water content of the Chinese medicine extract through an online viscosity detector, an online density detector and an online moisture detector.
[0031] On the other hand, the present invention also provides a digital twin intelligent decision-making process prediction system for vacuum belt drying of traditional Chinese medicine, comprising an online physical property detection device, an FPGA, and a host computer; the online physical property detection device is connected to the FPGA, and the FPGA is connected to the host computer; the online physical property detection device inputs the detection data into the FPGA, and the FPGA simultaneously reads the online production process parameters, and predicts the product quality parameters through the simulation model; the FPGA inputs the predicted product quality parameters into the host computer; the host computer simultaneously reads the online production process parameters, and obtains the optimized process parameters through the vacuum belt drying intelligent decision-making model.
[0032] Furthermore, in the FPGA, the establishment of the simulation model includes:
[0033] Based on the discrete element principle, the Chinese medicine extract on the conveyor belt is decomposed into several discrete elements of the same size, and the solution space is discretized into a unit array composed of multiple discrete elements. The two adjacent discrete elements are connected through the change relationship of the moisture content of the discrete elements. The continuous vacuum drying process of the discrete elements corresponds to the process of the discrete elements moving in space. The state at the end of the previous second is the initial state of the next second. Based on this basic principle, the discrete elements are iteratively calculated and the entire vacuum drying process is modeled, including the three-stage model of boiling, transition and evaporation. The state of the discrete element in the last step is the product quality parameter of the discharge of the vacuum belt drying.
[0034] Furthermore, the iterative calculation of the discrete element includes:
[0035]
[0036]
[0037]
[0038] Q 吸热 =Cp*M*T Formula (4)
[0039] Q放热 =M 水 *r Formula (5)
[0040] ΔQ=Q 吸热 +Q 1 -Q 放热 Formula (6)
[0041]
[0042]
[0043] M'=M 初始 -M 水 Formula (8)
[0044]
[0045] Among them, M' is the discrete element water content at the current moment; M 1 is the amount of water evaporated per unit time in the boiling stage, i.e., the boiling mass transfer flux; η is the material viscosity; r is the latent heat of vaporization of water; g is the gravitational acceleration; ρ l is the material density; v is the water vapor density; σ is the surface tension of the material; Pr is the Prandtl number; C pl is the specific heat capacity of the solute; C wl is the empirical constant for the heating surface-liquid combination; S is the contact area between the discrete element and the conveyor belt; t is the time when the discrete element completely leaves the current heating position;
[0046] M 2 is the amount of water evaporated per unit time in the evaporation stage, i.e., the evaporation mass transfer flux; D is the water vapor diffusion coefficient; T is the material temperature; P is the vacuum system pressure; C 溶质 is the solid content of the material; Q 1 is the heat transfer flux, which is used to calculate the heat change of the discrete element; λ is the thermal conductivity of the material; d is the thickness of the cake; Δt is the temperature difference; Q 吸热 is the heat absorbed by the discrete element; Cp is the specific heat capacity at constant pressure; M is the mass of the discrete element; Q 放热 Take away the heat for water evaporation; M 水 is the mass of evaporated water; r is the latent heat of vaporization of water; ΔQ is the heat change value; M 溶质 is the mass of the solute in the discrete element; M 初始 is the water content in the discrete element before heating begins;
[0047] The material in the above formula is Chinese herbal medicine extract. The heat change of the discrete element per unit time is shown in formula (6). After the above process, the water evaporation amount in the discrete element is continuously calculated in a cycle until the drying process is completed.
[0048] Furthermore, the training process of the vacuum belt drying intelligent decision model in the host computer includes:
[0049] Referring to the Markov decision process, the Markov decision process M consists of four elements, M = (S, A, P, R). Where S represents the state set (s i ∈S),s i represents the state of the i-th step; A represents a set of actions (a i ∈A),a i represents the action of step i; P represents the state transition probability, which means that in the current (s i ∈S) state, after (a i ∈A), the probability distribution of transferring to other states; R is the reward function, which gives feedback on the current state; the actions are the increase and decrease of various production process parameters, mainly including the adjustment of heating temperature, the adjustment of conveyor belt speed, and the adjustment of feeding speed; the adjustment range of the action is the action space, which is used to constrain the adjustment range of the action; finally, the reward function is determined, which is used to calculate the feedback of the environment on the action and comes from the moisture content of the product outlet.
[0050] Furthermore, the online physical property detection equipment includes an online viscosity detector, an online density detector and an online moisture detector arranged on the vacuum belt dryer, and the viscosity, density and moisture content of the traditional Chinese medicine extract are detected by the online viscosity detector, the online density detector and the online moisture detector.
[0051] Compared with the prior art, the present invention has the following beneficial effects:
[0052] (1) The present invention uses digital twin technology and discrete element principle to establish a simulation model of the vacuum belt drying process of traditional Chinese medicine, realizes the function of predicting product quality, and provides data support for intelligent decision-making;
[0053] (2) The present invention establishes an intelligent decision-making model for vacuum belt drying and uses a reinforcement learning algorithm to dynamically adjust process parameters and improve production efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] Figure 1 is a schematic diagram of the hardware structure of the second embodiment of the present invention;
[0055] Figure 2 It is a schematic diagram of the software structure of the second embodiment of the present invention. DETAILED DESCRIPTION
[0056] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.
[0057] In order to make the purpose and features of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are further described below in conjunction with the accompanying drawings. It should be noted that the drawings are all in a very simplified form and use inaccurate ratios, which are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention.
[0058] The design concept of the present invention is to use a mathematical model of reinforcement learning to model and analyze the vacuum belt drying process based on digital twin technology and artificial intelligence technology, so as to form an intelligent decision-making method and system for vacuum belt drying.
[0059] Embodiment 1:
[0060] Embodiment 1 mainly proposes a digital twin intelligent decision-making prediction method for vacuum belt drying of traditional Chinese medicine, including:
[0061] S1. Based on the movement law of Chinese herbal medicine extract in the vacuum belt dryer, a simulation model is established for the entire vacuum drying process of Chinese herbal medicine extract to predict product quality parameters;
[0062] The method for establishing the simulation model comprises:
[0063] Based on the discrete element principle, the Chinese medicine extract on the conveyor belt is decomposed into several discrete elements of the same size, and the solution space is discretized into a unit array composed of multiple discrete elements. The two adjacent discrete elements are connected through the change relationship of the moisture content of the discrete elements. The continuous vacuum drying process of the discrete elements corresponds to the process of the discrete elements moving in space. The state at the end of the previous second is the initial state of the next second. Based on this basic principle, the discrete elements are iteratively calculated and the entire vacuum drying process is modeled, including the three-stage model of boiling, transition and evaporation. The state of the discrete element in the last step is the product quality parameter of the discharge of the vacuum belt drying.
[0064] S2. Based on the DQN algorithm, the virtual production data generated by the simulation model is used for training to establish a vacuum belt drying intelligent decision-making model; the virtual production data includes product quality parameters corresponding to different production process conditions;
[0065] The training process includes:
[0066] Referring to the Markov decision process, the Markov decision process M consists of four elements, M = (S, A, P, R). Where S represents the state set (s i ∈S),s i represents the state of the i-th step; A represents a set of actions (Q i ∈A),a i represents the action of step i; P represents the state transition probability, which means that in the current (s i ∈S) state, after (a i∈A), the probability distribution of transferring to other states; R is the reward function, which gives feedback on the current state; the actions are the increase and decrease of various production process parameters, mainly including the adjustment of heating temperature, the adjustment of conveyor belt speed, and the adjustment of feeding speed; the adjustment range of the action is the action space, which is used to constrain the adjustment range of the action; finally, the reward function is determined, which is used to calculate the feedback of the environment on the action and comes from the moisture content of the product outlet.
[0067] S3. Online detection of the physical parameters of the Chinese medicine extract; online reading of the process parameters of the vacuum belt dryer; the online detection of the physical parameters of the Chinese medicine extract includes detecting the viscosity, density and moisture content of the Chinese medicine extract through an online viscosity detector, an online density detector and an online moisture detector.
[0068] S4, inputting the detected physical property parameters and process parameters into the simulation model to obtain predicted product quality parameters;
[0069] S5. Input the predicted product quality parameters and the detected process parameters into the vacuum belt drying intelligent decision-making model to obtain optimized process parameters.
[0070] Embodiment 2:
[0071] Embodiment 2 proposes a digital twin intelligent decision-making process prediction system for vacuum belt drying of traditional Chinese medicine based on the method described in Embodiment 1, including a hardware part and a software part.
[0072] The hardware part is as follows Figure 1 As shown, it includes online physical property detection equipment, FPGA, and a host computer; wherein the online physical property detection equipment includes an online viscosity detector, an online density detector, and an online moisture detector arranged in the vacuum belt dryer; the viscosity, density, moisture content and other physical property parameters of the raw material (traditional Chinese medicine extract) are detected by the online viscosity detector, the online density detector, and the online moisture detector, and the relevant detection data are input into the connected FPGA for calculation, and the product quality parameters of the vacuum belt dryer under the current material parameters are predicted and output by the simulation model of FGPA, and the product quality parameters are input into the host computer connected to the FPGA, and the vacuum belt drying intelligent decision model is used to perform reinforcement learning operation, and finally the optimized process parameters are obtained.
[0073] The software part is as follows Figure 2 As shown, it mainly includes a calculation parameter extraction module, a main operation module, and a result output and feedback module.
[0074] 1. Calculation parameter extraction module:
[0075] The calculation parameter extraction module provides the main calculation module with the data required for calculation, including the physical property parameters of the raw materials (traditional Chinese medicine extract) and the online production process parameters. The physical property parameters of the raw materials are measured by the online physical property detection equipment, including the density, viscosity and water content of the raw materials; the online production process parameters are directly read from the process parameter setting records and process product quality records of the production equipment, including feed speed, conveyor belt speed, heating temperature of each zone, vacuum degree, etc.
[0076] 2. Main computing module:
[0077] The main operation module includes two parts, namely a simulation model set on FPGA and a vacuum belt drying intelligent decision model set on the host computer.
[0078] 1. Simulation model:
[0079] The simulation model predicts the product quality by reading the current physical parameters and process parameters obtained by the calculation parameter extraction module, and the predicted product quality parameters are obtained. The model construction of the simulation model is based on the movement law of the Chinese medicine extract relative to the conveyor belt in the vacuum belt dryer. The model includes three sections: boiling, transition and evaporation. This is because after the Chinese medicine extract material enters the vacuum belt dryer, the boiling point of the material is reduced due to the increase in the vacuum degree of the system, the material will boil violently, and the water will evaporate rapidly. At this stage, the boiling model is used to describe the material change state. After the boiling is over, most of the water in the material has been removed, and the wet material gradually becomes a dry cake, but there is still residual water in the cake that needs to be dried further. At this time, the cake can be regarded as a dry loose material. At this stage, the evaporation model is used to calculate the water loss rate of the material. The purpose of the transition section is to connect the boiling section and the evaporation section. The addition of the transition section makes the connection between the evaporation section and the boiling section smooth, and the data generated by the simulation is more in line with the actual production conditions.
[0080] The construction of the simulation model adopts the basic principle of discrete element analysis, regards the Chinese medicine extract as a discontinuous discrete medium, and then decomposes the material on the conveyor belt into several discrete elements of the same size, discretizes the solution space into a unit array composed of multiple discrete elements, and connects two adjacent discrete elements through the change relationship of the moisture content of the discrete elements. The interaction between different discrete elements in space is not considered, and only the change law of a single discrete element over time is considered. The process of discrete element movement in space corresponds to the continuous vacuum drying process of Chinese medicine extract. The state at the end of the previous second is the initial state of the next second. Based on this basic principle, the discrete element is iteratively calculated, and the entire vacuum drying process of Chinese medicine extract is modeled. The state of the discrete element in the last step is the discharge state of the vacuum belt drying.
[0081] The specific process of iterative calculation of discrete elements includes:
[0082] The iterative calculation of discrete elements comprises:
[0083]
[0084]
[0085]
[0086] Q 吸热 =Cp*M*t Formula (4)
[0087] Q 放热 =M 水 *r Formula (5)
[0088] ΔQ=Q 吸热 +Q 1 -Q 放热 Formula (6)
[0089]
[0090]
[0091] M'=M 初始 -M 水 Formula (8)
[0092]
[0093] Among them, M' is the discrete element water content at the current moment; M 1 is the amount of water evaporated per unit time in the boiling stage, i.e., the boiling mass transfer flux; η is the material viscosity; r is the latent heat of vaporization of water; g is the gravitational acceleration; ρ l is the material density; v is the water vapor density; σ is the surface tension of the material; Pr is the Prandtl number; C pl is the specific heat capacity of the solute; C wl is the empirical constant for the heating surface-liquid combination; S is the contact area between the discrete element and the conveyor belt; t is the time when the discrete element completely leaves the current heating position;
[0094] M 2 is the amount of water evaporated per unit time in the evaporation stage, i.e., the evaporation mass transfer flux; D is the water vapor diffusion coefficient; T is the material temperature; P is the vacuum system pressure; C 溶质 is the solid content of the material; Q 1 is the heat transfer flux, which is used to calculate the heat change of the discrete element; λ is the thermal conductivity of the material; d is the thickness of the cake; Δt is the temperature difference; Q 吸热 is the heat absorbed by the discrete element; Cp is the specific heat capacity at constant pressure; M is the mass of the discrete element; Q 放热 Take away the heat for water evaporation; M水 is the mass of evaporated water; r is the latent heat of vaporization of water; ΔQ is the heat change value; C max In this embodiment, the value is 0.2; C min In this embodiment, the value is 0.08; M 溶质 is the mass of the solute in the discrete element; M 初始 is the water content in the discrete element before heating begins; M' is the water content in the discrete element at the current moment.
[0095] As shown in formulas (1) to (6): M 1 (Boiling mass transfer flux) is the amount of water evaporated per unit time in the discrete element during the boiling stage. Relative to the current moment, the water content of the discrete element at the next moment is equal to M' minus M 1 , that is, the water content of the discrete element at each moment is equal to the water content at the previous moment minus the water content change. The water content of the discrete element will gradually decrease in this cycle. Similarly, M 2 (Evaporation mass transfer flux) represents the amount of water evaporated per unit time in the discrete element during the evaporation stage, and is used to calculate the trend of water change during the evaporation stage;
[0096] Q 1 is the heat transfer flux, which is used to calculate the heat change of the discrete element. The heat change of the discrete element per unit time is shown in formula (6). After the above process, the water evaporation amount in the discrete element is continuously calculated cyclically until the drying process is completed; the above is the iterative calculation.
[0097] 2. Intelligent decision-making model for empty belt drying:
[0098] The intelligent decision-making model for empty belt drying provides optimized process parameters through embedded reinforcement learning. The reinforcement learning model is developed based on the DQN (Deep Q Network) algorithm, and the virtual production data generated by the simulation model is used to preliminarily train the intelligent decision-making model for vacuum belt drying. With the help of the reinforcement learning algorithm, dynamic adjustment of the process is achieved and production efficiency is improved. The intelligent decision-making model for empty belt drying is trained using 100,000 pieces of data. The current production process state is read as the initial state in the application, where the process parameters are directly read from the production equipment, and the product quality parameters are calculated by the simulation model. The intelligent decision-making model for empty belt drying observes the initial state, calls the pre-trained weights based on the observed values to determine the action with the largest Q-value in the current state and executes the action. This step is repeated multiple times until the optimization result meets the conditions. The process parameter combination output in the last cycle is the recommended process parameter after optimization.
[0099] The intelligent decision-making model of empty belt drying is mainly divided into two implementation steps: offline training and online decision-making. For the model training stage, the Markov decision process can be used to explain the training process of the decision model. The Markov decision process consists of four elements M = (S, A, P, R). Among them, S represents the state set (s i ∈S),s i represents the state of the i-th step. A represents a set of actions (a i ∈A),Q i represents the action of step i. P represents the state transition probability, which means that in the current (s i ∈S) state, after (a i ∈A) acts on the probability distribution of other states that will be transferred to. For example, the probability of transferring to s' when executing action a in state s can be expressed as p(s'|s,a). R is a reward function, which gives feedback on the current state. In the offline training stage, the simulation model is first used to generate a large amount of background data for model training. This background data is mainly the product quality parameters corresponding to different production process conditions. A total of 100,000 data are used for training in the offline training stage. Secondly, the observation space of the reinforcement learning algorithm needs to be defined. The observation space mainly provides a basis for the action selection process of the algorithm. Here, the observation space mainly includes the process operation parameters at the current moment and the production results corresponding to this process operation parameter. Then, the action and action space need to be defined. The intelligent agent interacts with the environment and obtains feedback by making actions. In this model, the action is the increase and decrease of each production process parameter, mainly including the adjustment of heating temperature, the adjustment of conveyor belt speed, the adjustment of feeding speed, etc. The action space is the adjustment range of the action, which is used to constrain the adjustment range of the action. Finally, the reward function needs to be determined. The reward function is used to calculate the feedback of the environment to the action, which may include rewards or penalties. In this model, the reward mainly comes from benefit indicators such as the export moisture content of the product and the input-return ratio. The trained model will be packaged into software and installed in the industrial control computer for online prediction.
[0100] 3. Optimization result output and feedback module:
[0101] The optimization result output and feedback module is set in the host computer; the process parameters calculated by the main operation module are transmitted to the pharmaceutical equipment (vacuum belt dryer) by this module to guide the practice. At the same time, the production results under the recommended parameter state are compared with the predicted results. If the error range meets the requirements, the interactive comparison will continue. If the error range does not meet the requirements, the model will be considered for correction.
[0102] Through the decision-making system calculation, the heating temperature, conveyor belt speed, raw material processing efficiency and product outlet moisture content of the vacuum belt dryer zone 1, zone 2 and zone 3 can be optimized.
[0103] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A digital twin intelligent decision-making process prediction method for vacuum belt drying of traditional Chinese medicine, characterized in that: include: S1. Based on the movement law of Chinese herbal medicine extract in the vacuum belt dryer, a simulation model is established for the entire vacuum drying process of Chinese herbal medicine extract to predict product quality parameters; S2. Based on the DQN algorithm, the virtual production data generated by the simulation model is used for training to establish a vacuum belt drying intelligent decision-making model; the virtual production data includes product quality parameters corresponding to different production process conditions; S3, online detection of physical parameters of Chinese medicine extract; online reading of process parameters of vacuum belt dryer; S4, inputting the detected physical property parameters and process parameters into the simulation model to obtain predicted product quality parameters; S5, inputting the predicted product quality parameters and the detected process parameters into the vacuum belt drying intelligent decision model to obtain optimized process parameters; The method for establishing the simulation model in step S1 includes: Based on the discrete element principle, the Chinese medicine extract on the conveyor belt is decomposed into several discrete elements of the same size, and the solution space is discretized into a unit array composed of multiple discrete elements. The two adjacent discrete elements are connected through the change relationship of the moisture content of the discrete elements. The continuous vacuum drying process of the discrete element corresponds to the process of the discrete element moving in space. The state at the end of the previous second is the initial state of the next second. Based on this basic principle, the discrete element is iteratively calculated and the entire vacuum drying process is modeled, including the three-stage model of boiling, transition and evaporation. The state of the discrete element in the last step is the product quality parameter of the vacuum belt drying output. The iterative calculation of discrete elements comprises: Formula (1) Formula (2) Formula (3) Formula (4) Formula (5) Formula (6) Formula (7) Formula (8) Formula (9) Among them, M' is the water content of the discrete element at the current moment; M1 is the water evaporation of the discrete element per unit time in the boiling stage, that is, the boiling mass transfer flux; η is the material viscosity; r is the latent heat of vaporization of water; g is the gravitational acceleration; ρ l is the material density; v is the water vapor density; σ is the surface tension of the material; Pr is the Prandtl number; C pl is the specific heat capacity of the solute; C wl is the empirical constant for the heating surface-liquid combination; S is the contact area between the discrete element and the conveyor belt; t is the time when the discrete element completely leaves the current heating position; M2 is the amount of water evaporated per unit time in the discrete element during the evaporation stage, i.e., the evaporation mass transfer flux; D is the water vapor diffusion coefficient; T is the material temperature; P is the vacuum system pressure; C 溶质 is the solid content of the material; Q1 is the heat transfer flux, which is used to calculate the heat change of the discrete element; λ is the thermal conductivity of the material; d is the thickness of the cake; Δt is the temperature difference; Q 吸热 is the heat absorbed by the discrete element; Cp is the specific heat capacity at constant pressure; M is the mass of the discrete element; Q 放热 Take away the heat for water evaporation; M 水 is the mass of evaporated water; r is the latent heat of vaporization of water; ΔQ is the heat change value; is the mass of the solute in the discrete element; is the water content in the discrete element before heating begins; The material in the above formula is Chinese herbal medicine extract. The heat change of the discrete element per unit time is shown in formula (6). After the above process, the water evaporation amount in the discrete element is continuously calculated in a cycle until the drying process is completed.
2. The digital twin intelligent decision-making process prediction method for vacuum belt drying of traditional Chinese medicine according to claim 1 is characterized in that: In step S2, the training process includes: Referring to the Markov decision process, the Markov decision process M consists of four elements, M=(S, A, P, R); where S represents the state set ( ), represents the state of the i-th step; A represents a set of actions ( ), represents the action of step i; P represents the state transition probability, which means that in the current ( ) state, after ( ) will transfer to other states after the action; R is the reward function, which gives feedback on the current state; the actions are the increase and decrease of various production process parameters, mainly including the adjustment of heating temperature, the adjustment of conveyor belt speed, and the adjustment of feeding speed; the adjustment range of the action is the action space, which is used to constrain the adjustment range of the action; finally, the reward function is determined, which is used to calculate the feedback of the environment on the action and comes from the moisture content of the product outlet.
3. The digital twin intelligent decision-making process prediction method for vacuum belt drying of traditional Chinese medicine according to claim 1 is characterized in that: The online detection of the physical property parameters of the Chinese medicine extract in step S3 includes detecting the viscosity, density and water content of the Chinese medicine extract through an online viscosity detector, an online density detector and an online moisture detector.
4. A digital twin intelligent decision-making process prediction system for vacuum belt drying of traditional Chinese medicine, characterized in that: It includes an online physical property detection device, an FPGA, and a host computer; the online physical property detection device is connected to the FPGA, and the FPGA is connected to the host computer; the online physical property detection device inputs the detection data into the FPGA, and the FPGA simultaneously reads the online production process parameters, and predicts the product quality parameters through the simulation model; the FPGA inputs the predicted product quality parameters into the host computer; the host computer simultaneously reads the online production process parameters, and obtains the optimized process parameters through the vacuum belt drying intelligent decision model; In the FPGA, the establishment of the simulation model includes: Based on the discrete element principle, the Chinese medicine extract on the conveyor belt is decomposed into several discrete elements of the same size, and the solution space is discretized into a unit array composed of multiple discrete elements. The two adjacent discrete elements are connected through the change relationship of the moisture content of the discrete elements. The continuous vacuum drying process of the discrete element corresponds to the process of the discrete element moving in space. The state at the end of the previous second is the initial state of the next second. Based on this basic principle, the discrete element is iteratively calculated and the entire vacuum drying process is modeled, including the three-stage model of boiling, transition and evaporation. The state of the discrete element in the last step is the product quality parameter of the vacuum belt drying output. The iterative calculation of discrete elements comprises: Formula (1) Formula (2) Formula (3) Formula (4) Formula (5) Formula (6) Formula (7) Formula (8) Formula (9) Among them, M' is the water content of the discrete element at the current moment; M1 is the water evaporation of the discrete element per unit time in the boiling stage, that is, the boiling mass transfer flux; η is the material viscosity; r is the latent heat of vaporization of water; g is the gravitational acceleration; ρ l is the material density; v is the water vapor density; σ is the surface tension of the material; Pr is the Prandtl number; C pl is the specific heat capacity of the solute; C wl is the empirical constant for the heating surface-liquid combination; S is the contact area between the discrete element and the conveyor belt; t is the time when the discrete element completely leaves the current heating position; M2 is the amount of water evaporated per unit time in the discrete element during the evaporation stage, i.e., the evaporation mass transfer flux; D is the water vapor diffusion coefficient; T is the material temperature; P is the vacuum system pressure; C 溶质 is the solid content of the material; Q1 is the heat transfer flux, which is used to calculate the heat change of the discrete element; λ is the thermal conductivity of the material; d is the thickness of the cake; Δt is the temperature difference; Q 吸热 is the heat absorbed by the discrete element; Cp is the specific heat capacity at constant pressure; M is the mass of the discrete element; Q 放热 Take away the heat for water evaporation; M 水 is the mass of evaporated water; r is the latent heat of vaporization of water; ΔQ is the heat change value; is the mass of the solute in the discrete element; is the water content in the discrete element before heating begins; The material in the above formula is Chinese herbal medicine extract. The heat change of the discrete element per unit time is shown in formula (6). After the above process, the water evaporation amount in the discrete element is continuously calculated in a cycle until the drying process is completed.
5. The digital twin intelligent decision-making process prediction system for vacuum belt drying of traditional Chinese medicine according to claim 4 is characterized in that: The training process of the vacuum belt drying intelligent decision model in the host computer includes: Referring to the Markov decision process, the Markov decision process M consists of four elements, M=(S, A, P, R); where S represents the state set ( ), represents the state of the i-th step; A represents a set of actions ( ), represents the action of step i; P represents the state transition probability, which means that in the current ( ) state, after ( ) will transfer to other states after the action; R is the reward function, which gives feedback on the current state; the actions are the increase and decrease of various production process parameters, mainly including the adjustment of heating temperature, the adjustment of conveyor belt speed, and the adjustment of feeding speed; the adjustment range of the action is the action space, which is used to constrain the adjustment range of the action; finally, the reward function is determined, which is used to calculate the feedback of the environment on the action and comes from the moisture content of the product outlet.
6. The digital twin intelligent decision-making process prediction system for vacuum belt drying of traditional Chinese medicine according to claim 4 is characterized in that: The online physical property detection equipment comprises an online viscosity detector, an online density detector and an online moisture detector arranged in the vacuum belt dryer, and the viscosity, density and moisture content of the traditional Chinese medicine extract are detected by the online viscosity detector, the online density detector and the online moisture detector.
