Intelligent traditional Chinese medicine concentration scheme optimization system based on neural network

Abstract

The invention provides an intelligent traditional Chinese medicine concentration scheme optimization system based on a neural network, and the system comprises a multi-source data collection module which obtains liquid medicine concentration data, evaporation temperature data and vacuum degree data in a traditional Chinese medicine concentration process in real time; the data preprocessing module is used for carrying out standardization processing on the multi-source sensor data; the feature extraction module is used for extracting a multi-dimensional feature vector related to the concentration efficiency; the optimization model module comprises a traditional Chinese medicine component retention rate prediction neural network and a dynamic weight adjustment mechanism; the logic judgment module is used for executing an evaporation temperature-concentration-energy consumption multi-objective optimization decision; the parameter adjusting module is used for realizing iterative optimization of concentration process parameters; and the feedback control module is used for generating a steam valve opening control instruction and evaluating a liquid medicine quality index. The traditional Chinese medicine concentration efficiency and the effective component retention rate can be improved, and the stability of the concentration process and the reliability of the liquid medicine quality are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of traditional Chinese medicine pharmaceutical manufacturing, and more specifically, to an intelligent traditional Chinese medicine concentration scheme optimization system based on a neural network. Background Art

[0002] In the field of Traditional Chinese Medicine (TCM) pharmaceutical manufacturing, traditional TCM concentration processes rely primarily on manual experience and fixed parameter control. While this method can accomplish concentration tasks to a certain extent, it has numerous limitations. First, TCM components are complex, and the physical and chemical properties of different herbs vary significantly, making fixed parameters difficult to adapt to diverse concentration needs. Second, traditional processes lack the ability to monitor and dynamically adjust key parameters such as drug concentration, temperature, and vacuum level during the concentration process. This results in unstable active ingredient retention, low concentration efficiency, and high energy consumption. Furthermore, the uncertainty of manual operation easily introduces the risk of quality fluctuations, making it difficult to meet the requirements of modern TCM production for standardization and intelligentization. With the development of sensor technology, data analysis, and artificial intelligence, new ideas and methods have been provided for optimizing TCM concentration processes.

[0003] In the process of implementing the embodiments of the present invention, the inventors found that there are at least the following problems or defects in the existing technology: the existing technology lacks the ability to extract and analyze the multidimensional characteristics of the traditional Chinese medicine concentration process in real time, and cannot accurately predict the retention rate of effective ingredients; at the same time, it lacks a dynamic weight adjustment mechanism, making it difficult to balance the relationship between efficacy, energy consumption and process stability; in addition, the existing system has deficiencies in parameter adjustment and feedback control, and cannot achieve refined optimization of the concentration process, making it difficult to meet the needs of high-quality traditional Chinese medicine concentration production. Summary of the Invention

[0004] The present invention provides an intelligent Chinese medicine concentration scheme optimization system based on a neural network, comprising:

[0005] Multi-source data acquisition module, real-time acquisition of liquid concentration data, evaporation temperature data and vacuum degree data during the Chinese medicine concentration process;

[0006] Data preprocessing module, used to standardize multi-source sensor data;

[0007] Feature extraction module, extracting multidimensional feature vectors related to concentration efficiency;

[0008] Optimization model module, including a neural network for predicting the retention rate of Chinese herbal medicine ingredients and a dynamic weight adjustment mechanism;

[0009] Logical judgment module, which performs multi-objective optimization decision-making on evaporation temperature, concentration and energy consumption;

[0010] Parameter adjustment module to achieve iterative optimization of concentration process parameters;

[0011] Feedback control module generates steam valve opening control instructions and evaluates liquid medicine quality indicators.

[0012] Furthermore, the multi-source data acquisition module includes:

[0013] The drug liquid multispectral sensor group is used to monitor the changes in the light transmittance of the drug liquid;

[0014] Distributed temperature sensor array for measuring the vertical temperature gradient ΔT inside the evaporator Z ;

[0015] Near infrared online component analyzer for real-time detection of astragaloside IV concentration C herb and baicalin concentration C scu ;

[0016] A signal conditioning unit, used to convert the sensor analog signal into a digital quantity;

[0017] A time series database is used to store structured data packets at preset time intervals.

[0018] Furthermore, the dynamic weight adjustment mechanism of the optimization model module performs the following steps:

[0019] (1) Calculate weight sensitivity Where E represents the model loss function, which is represented by α·E herb and (1-α)·E energy Additive composition, α is the efficacy weight factor and 0<α<1, E herb is the effective ingredient retention rate error, E energy is the concentrated energy consumption error, W is the neural network weight matrix;

[0020] (2) Generate adjustment coefficient α(t) = η·tanh(β·||δ|| F ), where η is the baseline learning rate and η = 0.01, β is the concentration-related sensitivity coefficient, ||δ|| F represents the Frobenius norm of the weight sensitivity δ;

[0021] (3) Update weight W(t+1)=W(t)-α(t)·δ+γ·ΔW(t-1), where γ is the momentum factor and γ=0.9, and ΔW(t-1) is the previous weight update;

[0022] (4) Spectral normalization is performed when ||W(t+1)||2>θ, where θ is the weight norm threshold and θ=1.2, and ||·||2 represents the L2 norm.

[0023] Furthermore, the calculation formula of the concentration-related sensitivity coefficient β is:

[0024] β=1 / (1+exp(-k·(C herb / Ct arget -0.9)))

[0025] Where k is the shape factor and k = 2.5, C herb is the concentration of the active ingredient in the current liquid medicine, C target is the target concentration endpoint, and exp represents the natural exponential function.

[0026] Furthermore, the logic judgment module performs the following steps:

[0027] (1) Constructing a multi-objective function for Chinese medicine concentration

[0028] F=ω1·f temp +ω2·f herb -ω3·f energy

[0029] Among them, f temp =1 / (1+|T evap -T target |) represents the evaporation temperature approximation, T evap is the real-time temperature measurement value of the evaporation tank, T target is the target temperature and T target =65℃;

[0030] f herb =log(C now / C initial ) represents the effective ingredient retention rate, C now is the current concentration of the drug solution, C initial is the initial drug solution concentration;

[0031] f energy =∑(P steam t+P vacuum ·t) represents the total energy consumption of the concentration process, P steam is the steam generator power, P vacuum is the vacuum pump power, t is the running time;

[0032] (2) Verify process constraints:

[0033] T e vaping m ax <T safe , where T e vaping m ax is the maximum temperature of the evaporator, T safe is the safety temperature threshold and T safe =75℃;

[0034] ΔC / Δt <C crit, where ΔC / Δt is the concentration change rate, C crit is the phase transition critical value and C crit = 0.2 mg / ml / min;

[0035] (3) Generate a Pareto optimal solution set and select the optimization solution that prioritizes drug efficacy.

[0036] Furthermore, the weight coefficient ω i The adjustment formula is:

[0037] ω i =exp(s i ) / (exp(s1)+exp(s2)+exp(s3))

[0038] Where s1=0.8·E temp +0.2·ΔE temp , s1 is the temperature control item, E temp is the mean absolute error of temperature, ΔE temp is the temperature error change rate;

[0039] s2=1.2·E herb -0.3·ΔE herb , s2 is the efficacy priority, E herb =1-C now / C target is the concentration achievement rate error;

[0040] s3=0.5·E energy , s3 is the energy consumption suppression term, E energy is the normalized energy consumption error.

[0041] Furthermore, the parameter adjustment module includes:

[0042] Gradient correction unit, used to calculate steam pressure adjustment Where η is the learning rate, Q is the quality assessment index, and P is the steam pressure;

[0043] A phase change warning unit triggers a viscosity protection mechanism when it detects dC / dT > 0.15 mg per milliliter per degree Celsius, where dC / dT is the derivative of concentration with respect to temperature;

[0044] Iterative controller, automatically adjusts the step size η = 0.1 / (1 + 0.05·μ) according to the viscosity of the liquid, where μ is the real-time measured viscosity of the liquid;

[0045] Boundary protection unit, used to limit the steam pressure P in the range of 0.2 MPa to 0.8 MPa.

[0046] Furthermore, the gradient correction unit performs the following operations:

[0047] (1) Calculate the momentum gradient ▽′ = ρ·▽ + (1-ρ)·▽ prev , where ρ is the momentum factor and ρ = 0.8, ▽ is the current gradient, ▽ prev is the previous gradient;

[0048] (2) Add boundary penalty term ▽″=▽′+λ·sign(▽′)·max(0,|P|-P max ), where λ is the penalty coefficient and λ = 0.5, sign is the sign function, P is the current steam pressure, P max is the upper pressure limit and P max =0.8 MPa;

[0049] (3) Correcting parameters through projection operator:

[0050] If P old -η·▽″ <P min , then P new =P min , where P min =0.2 MPa;

[0051] If P old -η·▽″>P max , then P new =P max ;

[0052] Otherwise P new =P old -η·▽″.

[0053] Furthermore, the feedback control module includes:

[0054] Quality assessment unit, calculates enrichment quality indicators

[0055] Q=w1·(C herb / C target ) 2 +w2·(1-E energy )+w3·(1-T err )

[0056] Among them, w1=0.6 is the efficacy weight, w2=0.3 is the energy consumption weight, w3=0.1 is the temperature error weight, T err =|T evap -T target |;

[0057] Parameter updater, when Q th When adjusting the number of hidden layer nodes of the neural network, Q th =0.85 is the quality threshold;

[0058] ​A crystallization warning unit reduces the concentration rate when it detects dμ / dt>0.1 Pascal second per minute, where dμ / dt is the rate of change of viscosity;

[0059] Case database, storing the best evaporation curves of different medicinal materials.

[0060] Furthermore, the adaptive update formula of the drug efficacy weight w1 is:

[0061]

[0062] Where μ = 0.05 is the update rate, R = 0.95 is the efficacy benchmark value, Σ represents the summation operation for j = 1 to 2, E j It is divided into energy consumption and temperature error.

[0063] According to the above-mentioned embodiment of the present invention, there are at least the following beneficial effects: the system obtains the liquid concentration data, evaporation temperature data and vacuum degree data in the traditional Chinese medicine concentration process in real time through the multi-source data acquisition module, which can provide comprehensive and accurate data support for the concentration process. The data preprocessing module and the feature extraction module can effectively process and analyze these data, extract multi-dimensional feature vectors related to the concentration efficiency, and provide accurate input for the optimization model. The optimization model module includes a traditional Chinese medicine component retention rate prediction neural network and a dynamic weight adjustment mechanism, which can realize intelligent optimization of the concentration process, improve the retention rate of effective ingredients, and reduce energy consumption. The logic judgment module performs the evaporation temperature-concentration-energy consumption multi-objective optimization decision, which can balance the relationship between efficacy, energy consumption and process stability, and realize the optimization of the concentration process. The parameter adjustment module and the feedback control module can realize iterative optimization and real-time control of the concentration process parameters, ensuring the stability of the concentration process and the reliability of the liquid medicine quality.

[0064] Through intelligent optimization of TCM concentration solutions, this system significantly improves TCM concentration efficiency and active ingredient retention, reduces energy consumption, and ensures the stability of the concentration process and the reliability of the drug solution quality. Compared with traditional processes, this system enables refined control of the TCM concentration process, improving the standardization and intelligentization of TCM production, and meeting the needs of modern TCM production. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily apparent by reading the following detailed description with reference to the accompanying drawings, in which several embodiments of the present invention are shown by way of example and not limitation, in which:

[0066] Figure 1 A schematic diagram of the structure of an intelligent traditional Chinese medicine concentration solution optimization system based on a neural network provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0067] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided solely to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. Rather, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0068] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software.

[0069] It should be noted that any number of elements in the drawings is for illustration only and not for limitation, and any naming is only for distinction and does not have any limiting meaning.

[0070] Reference below Figure 1 , Figure 1 This is a schematic diagram of the structure of an intelligent Chinese medicine concentration program optimization system based on neural network provided by one embodiment of the present invention. Figure 1 As shown, a neural network-based intelligent Chinese medicine concentration solution optimization system 100 includes:

[0071] The multi-source data acquisition module 101 acquires the liquid concentration data, evaporation temperature data and vacuum degree data in real time during the Chinese medicine concentration process;

[0072] The data preprocessing module 102 is used to perform standardization processing on multi-source sensor data;

[0073] A feature extraction module 103 extracts a multi-dimensional feature vector related to the concentration efficiency;

[0074] Optimization model module 104, including a neural network for predicting the retention rate of Chinese medicinal ingredients and a dynamic weight adjustment mechanism;

[0075] Logic judgment module 105, performs evaporation temperature-concentration-energy consumption multi-objective optimization decision;

[0076] Parameter adjustment module 106, which implements iterative optimization of concentration process parameters;

[0077] The feedback control module 107 generates steam valve opening control instructions and evaluates liquid medicine quality indicators.

[0078] It should be noted that the core of this system lies in the multi-source data acquisition module, which is responsible for acquiring key data from the TCM concentration process in real time, including liquid concentration, evaporation temperature, and vacuum level. This data forms the basis for subsequent optimization processes. Liquid concentration data reflects the changes in the content of the active ingredient in the liquid, evaporation temperature data is the temperature within the evaporator used to concentrate the liquid, and vacuum level data represents the pressure environment within the evaporator. By collecting this data in real time, the system can dynamically monitor changes in the concentration process, providing an accurate basis for subsequent optimization.

[0079] Specifically, the multi-source data acquisition module consists of multiple sensors, including a multispectral sensor group for the drug solution, a distributed temperature sensor array, and a near-infrared online component analyzer. The multispectral sensor group monitors changes in the drug solution's transmittance in the 400-1100 nm wavelength range. By analyzing the transmittance characteristics of the drug solution at different wavelengths, it indirectly reflects changes in the concentration of its components. The distributed temperature sensor array measures the vertical temperature gradient within the evaporation tank, accurately determining the temperature distribution at different locations within the tank and ensuring the uniformity of the evaporation process. The near-infrared online component analyzer monitors the concentrations of astragaloside IV and baicalin in real time. These two components are common active ingredients in the concentration process of traditional Chinese medicine. Using near-infrared spectroscopy, their concentration changes can be rapidly and non-destructively detected. The analog signals collected by these sensors are converted to 12-bit digital quantities by a signal conditioning unit and stored in a time series database at 0.5-second intervals, providing structured data support for subsequent data processing and optimization.

[0080] Preferably, to further enhance the adaptability and accuracy of the system, the wavelength range of the liquid medicine multispectral sensor group can be adjusted according to the spectral characteristics of specific Chinese medicine ingredients to increase the detection sensitivity of specific ingredients. For example, for Chinese medicines containing specific pigments or alkaloids, the wavelength range can be extended to the ultraviolet-visible light region to better monitor changes in their composition. The layout of the distributed temperature sensor array can also be optimized according to the shape and size of the evaporation tank, for example, adopting a spiral or grid layout to more comprehensively cover the internal space of the evaporation tank and improve the accuracy of temperature measurement. In addition, the detection range of the near-infrared online component analyzer can be expanded to other active ingredients, such as ginsenosides, tanshinones, etc., according to actual needs, to meet the requirements of different Chinese medicine concentration processes. In the signal conditioning unit, a filtering algorithm can be introduced to further improve the stability and anti-interference ability of the signal, thereby improving the overall performance of the system.

[0081] In some embodiments, the multi-source data acquisition module includes:

[0082] The drug solution multispectral sensor group is used to monitor the changes in the transmittance of the drug solution in the 400 nm to 1100 nm band;

[0083] Distributed temperature sensor array for measuring the vertical temperature gradient ΔT inside the evaporator Z ;

[0084] Near infrared online component analyzer for real-time detection of astragaloside IV concentration C herb and baicalin concentration C scu ;

[0085] Signal conditioning unit, used to convert the sensor analog signal into a 12-bit digital quantity;

[0086] A time series database that stores structured data packets at 0.5-second intervals.

[0087] It should be noted that the multi-source data acquisition module is a key component of this system for achieving precise optimization. It uses multiple sensor groups to obtain key parameters in the traditional Chinese medicine concentration process in real time. The multispectral sensor group of the medicinal liquid is used to monitor the changes in the transmittance of the medicinal liquid in specific wavelength bands. This transmittance change is closely related to the concentration of the ingredients in the medicinal liquid. The distributed temperature sensor array can measure the temperature differences at different heights within the evaporation tank, thereby generating temperature gradient information. The near-infrared online component analyzer focuses on the real-time detection of the concentration of specific active ingredients in the medicinal liquid, such as astragaloside IV and baicalin. The analog signals collected by these sensors are converted into digital signals by the signal conditioning unit and stored in a time series database, providing basic data support for subsequent data processing and optimization.

[0088] Specifically, the multispectral sensor system for medicinal liquids operates based on spectroscopy, inferring the concentration of medicinal liquid components by measuring changes in transmittance within the 400-1100 nm wavelength range. This wavelength range covers the visible and near-infrared regions, effectively reflecting the absorption characteristics of various components in the medicinal liquid. A distributed temperature sensor array, by evenly distributing multiple temperature sensors vertically within the evaporation tank, measures temperatures at different heights, generating temperature gradient data. This temperature gradient information is crucial for optimizing the evaporation process, as it directly impacts concentration efficiency and energy consumption. The near-infrared online component analyzer utilizes near-infrared spectroscopy to measure the concentrations of astragaloside IV and baicalin in the medicinal liquid in real time. These two components are common active ingredients in the traditional Chinese medicine concentration process, and their concentration changes directly reflect the concentration effect. The signal conditioning unit converts the analog signals collected by the sensors into 12-bit digital quantities to ensure data accuracy and stability. A time series database stores these structured data packets at 0.5-second intervals, facilitating subsequent data analysis and model training.

[0089] Preferably, to further enhance the system's performance and adaptability, the wavelength range of the liquid medicine multispectral sensor array can be adjusted based on the spectral characteristics of specific TCM ingredients. For example, if the concentration process involves other ingredients with specific absorption peaks, the wavelength range can be expanded to a wider range, such as 200 to 2500 nanometers, to cover the ultraviolet, visible, and mid-infrared regions. For the distributed temperature sensor array, its layout can be optimized based on the shape of the evaporator and process requirements. For example, a spiral layout can more evenly cover the interior of the evaporator, thereby improving the accuracy of temperature measurement. Furthermore, the detection range of the near-infrared online component analyzer can also be expanded to include other active ingredients, such as ginsenosides and tanshinones, based on actual needs, to accommodate the requirements of different TCM concentration processes. More advanced filtering algorithms, such as Kalman filtering, can be introduced into the signal conditioning unit to further improve signal stability and anti-interference capabilities.

[0090] In some embodiments, the dynamic weight adjustment mechanism of the optimization model module performs the following steps:

[0091] (1) Calculate weight sensitivity Where E represents the model loss function, which is represented by α·E herb and (1-α)·E energy Additive composition, α is the efficacy weight factor and 0<α<1, E herb is the effective ingredient retention rate error, E energy is the concentrated energy consumption error, W is the neural network weight matrix;

[0092] (2) Generate adjustment coefficient α(t) = η·tanh(β·||δ|| F ), where η is the baseline learning rate and η = 0.01, β is the concentration-related sensitivity coefficient, ||δ|| F represents the Frobenius norm of the weight sensitivity δ;

[0093] (3) Update weight W(t+1)=W(t)-α(t)·δ+γ·ΔW(t-1), where γ is the momentum factor and γ=0.9, and ΔW(t-1) is the previous weight update;

[0094] (4) Spectral normalization is performed when ||W(t+1)||2>θ, where θ is the weight norm threshold and θ=1.2, and ||·||2 represents the L2 norm.

[0095] It should be noted that the dynamic weight adjustment mechanism of the optimization model module is an important part of the system to achieve precise optimization. Its core is to dynamically adjust the weight matrix of the neural network by calculating weight sensitivity, generating adjustment coefficients, and updating weights to adapt to the complex and changeable working conditions in the traditional Chinese medicine concentration process. Specifically, the weight sensitivity is obtained by derivation of the model loss function, which reflects the impact of weight changes on the model output. The model loss function consists of the efficacy weight factor, the effective ingredient retention rate error, and the concentration energy consumption error, which is used to measure the optimization effect of the model. By dynamically adjusting the weights, the system can achieve a balance between efficacy, energy consumption, and process stability, thereby improving the adaptability and accuracy of the optimization model.

[0096] Specifically, the implementation process of the dynamic weight adjustment mechanism includes the following key steps. First, the calculation formula of weight sensitivity is

[0097]

[0098] Among them, E is the model loss function, which consists of two parts: the efficacy weight factor α multiplied by the effective ingredient retention rate error E c , and (1-α) multiplied by the concentration energy error E e The efficacy weight factor α is a parameter between 0 and 1, which is used to balance the importance of efficacy and energy consumption. c It represents the difference between the actual retention rate and the target retention rate, while the concentration energy consumption error E e It reflects the deviation between the actual energy consumption and the target energy consumption. Next, the formula for generating the adjustment coefficient is γ(W)=η·σ(||S|| F ), where η is the baseline learning rate, usually set to 0.01, which is used to control the step size of weight update; σ is the concentration-related sensitivity coefficient, which is used to adjust the sensitivity of weight update; ∥S∥ F is the Frobenius norm of the weight sensitivity, indicating the magnitude of the weight sensitivity. Finally, the weight update formula is W(t+1)=W(t)-γ(W)·S+μ·ΔW(t-1), where μ is the momentum factor, usually set to 0.9, which is used to introduce the previous weight update amount ΔW(t-1) to speed up convergence and avoid falling into local optimality. When the L2 norm of the weight ∥W(t+1)∥2 exceeds the set threshold θ (such as 1.2), the system will perform spectral normalization to prevent excessive weights from causing model instability.

[0099] Preferably, in order to further improve the performance of the dynamic weight adjustment mechanism, optimization can be performed in the following aspects. First, the calculation formula of the concentration-related sensitivity coefficient σ can be refined as follows:

[0100]

[0101] Where k is the shape factor, usually set to 2.5, which is used to control the steepness of the curve; C eff is the concentration of the active ingredient in the current liquid medicine, and C tar is the target concentration endpoint. By adjusting the shape factor k, the sensitivity coefficient's response speed to concentration changes can be altered, thereby better adapting to the needs of different TCM concentration processes. Secondly, the baseline learning rate η and momentum factor μ can be dynamically adjusted based on actual operating conditions. For example, in the early stages of optimization, the value of η can be appropriately increased to accelerate convergence; in the later stages of optimization, the value of η can be reduced to improve optimization accuracy. Furthermore, an adaptive learning rate mechanism can be introduced to dynamically adjust the learning rate based on changes in weight sensitivity, further enhancing the optimization effect.

[0102] In some embodiments, the concentration-related sensitivity coefficient β is calculated as follows:

[0103] β=1 / (1+exp(-k·(C herb / C target -0.9)))

[0104] Where k is the shape factor and k = 2.5, C herb is the concentration of the active ingredient in the current liquid medicine, C target is the target concentration endpoint, and exp represents the natural exponential function.

[0105] It should be noted that the calculation formula of the concentration-related sensitivity coefficient σ is an important part of the dynamic weight adjustment mechanism in this system. It is used to dynamically adjust the sensitivity of the weight update according to the difference between the current concentration of the active ingredient in the drug solution and the target concentration endpoint concentration. The calculation formula of the concentration-related sensitivity coefficient σ is σ=1 / (1+exp(-k·(C eff / C tar -0.9))), where k is the shape factor that controls the steepness of the curve; C eff is the concentration of the active ingredient in the current solution, and C tar is the target concentration endpoint. By introducing the natural exponential function and shape factor k, this formula can dynamically adjust the sensitivity of weight updates based on the proximity of the current concentration to the target concentration, thereby better balancing the relationship between efficacy and energy consumption during the optimization process.

[0106] Specifically, in the calculation formula of the concentration-related sensitivity coefficient σ, the shape factor k is usually set to 2.5. This value determines the steepness of the curve, which in turn affects the response speed of the sensitivity coefficient to concentration changes. eff Close to the target concentration endpoint C tarWhen C eff Stay away from C tar When σ approaches 0, the sensitivity of the weight update decreases. This dynamic adjustment mechanism ensures that the system prioritizes energy consumption optimization in the early stages of the concentration process, while prioritizing efficacy improvement as the target concentration approaches. Furthermore, the 0.9 parameter in the formula is an empirical parameter used to adjust the balance point of the sensitivity coefficient. It can be fine-tuned based on actual process requirements to better suit the characteristics of different TCM concentration processes.

[0107] To further enhance the adaptability and flexibility of the concentration-dependent sensitivity coefficient σ, optimization can be performed in the following aspects. First, the value of the shape factor k can be adjusted based on the characteristics of different TCM ingredients. For example, for certain TCM herbs whose active ingredient concentration changes more slowly, the value of k can be appropriately reduced to slow the sensitivity coefficient's response to concentration changes, thereby avoiding prematurely shifting the optimization focus to efficacy. Conversely, for TCM herbs with more rapid concentration changes, the value of k can be appropriately increased to speed up the sensitivity coefficient's response, ensuring that the system can adjust the weights to optimize efficacy in a timely manner. Second, the empirical parameter 0.9 can be fine-tuned based on the actual concentration process requirements. For example, in concentration processes with high efficacy requirements, this parameter can be appropriately increased so that the system shifts optimization focus to efficacy improvement only when the target concentration is closer. In processes where energy consumption is more critical, this parameter can be appropriately decreased to prioritize energy optimization earlier. Furthermore, an adaptive mechanism can be introduced to dynamically adjust the shape factor k and the empirical parameter 0.9 based on the real-time monitored concentration change rate, further enhancing the system's intelligence and optimization effectiveness.

[0108] In some embodiments, the logic judgment module performs the following steps:

[0109] (1) Constructing a multi-objective function for Chinese medicine concentration

[0110] F=ω1·f temp +ω2·f herb -ω3·f energy

[0111] Among them, f temp =1 / (1+|T evap -T target |) represents the evaporation temperature approximation, T evap is the real-time temperature measurement value of the evaporation tank, T tar get is the target temperature and T tar get =65℃;

[0112] f herb =log(C now / C initial ) represents the effective ingredient retention rate, C now is the current concentration of the drug solution, C initial is the initial drug solution concentration;

[0113] f energy =∑(P steam t+P vacuum ·t) represents the total energy consumption of the concentration process, P steam is the steam generator power, P vacuum is the vacuum pump power, t is the running time;

[0114] (2) Verify process constraints:

[0115] T e vaping m ax <T safe , where T e vaping m ax is the maximum temperature of the evaporator, T safe is the safety temperature threshold and T safe =75℃;

[0116] ΔC / Δt <C crot , where ΔC / Δt is the concentration change rate, C crit is the phase transition critical value and C crit = 0.2 mg / ml / min;

[0117] (3) Generate a Pareto optimal solution set and select the optimization solution that prioritizes drug efficacy.

[0118] It should be noted that the logic judgment module is the core component of this system's multi-objective optimization decision-making for the traditional Chinese medicine concentration process. By constructing a multi-objective function and verifying process constraints, it generates a Pareto optimal solution set, thereby selecting an optimization solution that prioritizes efficacy. The multi-objective function comprehensively considers the evaporation temperature approach, the active ingredient retention rate, and the total energy consumption of the concentration process to balance the relationship between efficacy, energy consumption, and process stability. The evaporation temperature approach reflects the closeness of the current temperature to the target temperature, the active ingredient retention rate measures the retention of the active ingredient in the medicinal solution, and the total energy consumption of the concentration process reflects the energy consumption during the concentration process. By verifying the process constraints, the system can ensure that the optimization process is carried out within a safe and reliable range.

[0119] Specifically, the implementation process of the logic judgment module includes the following key steps. First, construct the multi-objective function F = w1·T for Chinese medicine concentration. app +w2·E eff -w3·E eng ,in Indicates the evaporation temperature approach, T realis the real-time temperature measurement value of the evaporation tank, T tar is the target temperature (usually set to 65°C); Indicates the effective ingredient retention rate, C real is the current concentration of the drug solution, C initial is the initial drug solution concentration; E eng =g(P steam t+P vacuum ·t) represents the total energy consumption of the concentration process, P steam is the steam generator power, P vacuum is the vacuum pump power, and t is the operating time. Secondly, verify the process constraints, including the maximum temperature of the evaporation tank T max Less than the safety temperature threshold T safety (usually set to 75°C), and the rate of concentration change Less than the phase transition critical value (usually set to 0.2 mg / ml / min). These constraints ensure the safety and stability of the concentration process. Finally, a Pareto optimal solution set is generated, and the optimization solution with drug efficacy priority is selected from it to achieve multi-objective optimization.

[0120] Preferably, in order to further improve the performance and adaptability of the logic judgment module, the following aspects can be refined or replaced. First, the settings of the weight coefficients w1, w2 and w3 can be adjusted according to the actual process requirements. For example, if the efficacy is the main optimization target, the weight of w2 can be appropriately increased; if energy consumption optimization is more important, the weight of w3 can be increased. Secondly, the target temperature T tar And the safety temperature threshold T safety Adjustments can be made to suit different TCM concentration processes. For example, for certain temperature-sensitive TCM ingredients, the target temperature can be appropriately lowered to reduce thermal damage. Furthermore, the phase transition threshold for the concentration change rate can be fine-tuned based on the physical properties of the actual medicinal solution to better suit the needs of different TCM concentration processes. When generating a Pareto optimal solution set, a multi-objective optimization algorithm (such as the NSGA-II algorithm) can be introduced to improve the quality and diversity of the solution set, thereby providing users with more optimization options.

[0121] In some embodiments, the weight coefficient ω i The adjustment formula is:

[0122] ω i =exp(s i ) / (exp(s1)+exp(s2)+exp(s3))

[0123] Where s1=0.8·E temp +0.2·ΔE temp , s1 is the temperature control item, E tempis the mean absolute error of temperature, ΔE temp is the temperature error change rate;

[0124] s2=1.2·E herb -0.3·ΔE herb , s2 is the efficacy priority, E herb =1-C now / C target is the concentration achievement rate error;

[0125] s3=0.5·E energy , s3 is the energy consumption suppression term, E energy is the normalized energy consumption error.

[0126] It should be noted that the weight coefficient adjustment formula is a key mechanism used in this system to dynamically balance the importance of each objective in the multi-objective optimization process. Through this formula, the system can dynamically adjust the weights of the temperature control item, the efficacy priority item, and the energy consumption suppression item according to the current optimization state, thereby better realizing the multi-objective optimization of the traditional Chinese medicine concentration process. Among them, the temperature control item reflects the stability and accuracy of the evaporation temperature, the efficacy priority item reflects the importance of the retention rate of the active ingredient, and the energy consumption suppression item focuses on the energy consumption during the concentration process. By dynamically adjusting these weight coefficients, the system can flexibly balance the relationship between the various optimization objectives in different stages and working conditions, ensuring the efficiency and adaptability of the optimization process.

[0127] Specifically, the adjustment formula of the weight coefficient is:

[0128]

[0129] Among them, w1, w2 and w3 correspond to the weights of temperature control item, efficacy priority item and energy consumption suppression item respectively. The calculation formula of temperature control item w1 is w1 = 0.8·T err +0.2·dT err , where T err is the mean absolute error in temperature, dT err is the temperature error change rate. The calculation formula for the efficacy priority w2 is w2=1.2·E eff -0.3·dE eff , where E eff is the concentration achievement rate error, dE eff is the rate of change of concentration error. The calculation formula of energy consumption suppression term w3 is w3=0.5·E eng , where E engis the normalized energy consumption error. These parameter settings enable the system to dynamically adjust the importance of each objective during the optimization process, better adapting to the complex and changing conditions of the TCM concentration process. For example, when the temperature error is large, the system increases the weight of the temperature control item to prioritize stabilizing the evaporation temperature; while when the active ingredient retention rate is low, the weight of the efficacy priority item is increased to improve efficacy.

[0130] Preferably, in order to further improve the flexibility and adaptability of weight coefficient adjustment, the following aspects can be refined or replaced. First, the weight coefficients in the temperature control item (such as 0.8 and 0.2) can be adjusted according to actual process requirements. For example, in the concentration process with extremely high temperature stability requirements, T err The weight of dE can be increased to more strictly control the temperature error. Secondly, the coefficients in the efficacy priority term (such as 1.2 and 0.3) can be optimized according to the characteristics of different Chinese herbal ingredients. For example, for some Chinese herbal ingredients that are sensitive to concentration changes, the concentration error change rate dE can be appropriately increased. eff The weights of the energy consumption suppression term (e.g., 0.5) can be adjusted based on actual energy consumption to better balance energy consumption and efficacy. In practical applications, adaptive mechanisms can be introduced to dynamically adjust these weights based on real-time optimization results, further enhancing the system's intelligence and optimization effectiveness.

[0131] In some embodiments, the parameter adjustment module includes:

[0132] Gradient correction unit, used to calculate steam pressure adjustment Where η is the learning rate, Q is the quality assessment index, and P is the steam pressure;

[0133] A phase change warning unit triggers a viscosity protection mechanism when it detects dC / dT > 0.15 mg per milliliter per degree Celsius, where dC / dT is the derivative of concentration with respect to temperature;

[0134] Iterative controller, automatically adjusts the step size η = 0.1 / (1 + 0.05·μ) according to the viscosity of the liquid, where μ is the real-time measured viscosity of the liquid;

[0135] Boundary protection unit, used to limit the steam pressure P in the range of 0.2 MPa to 0.8 MPa.

[0136] It should be noted that the parameter adjustment module is a key component of this system's optimized control of the traditional Chinese medicine concentration process. It dynamically adjusts and optimizes the concentration process parameters through the collaborative work of multiple units, including a gradient correction unit, a phase change warning unit, an iterative controller, and a boundary protection unit. The gradient correction unit calculates the steam pressure adjustment to optimize this key parameter. The phase change warning unit monitors the temperature derivative of concentration to promptly trigger the viscosity protection mechanism, preventing viscosity anomalies caused by phase changes during the concentration process. The iterative controller automatically adjusts the optimization step size based on the viscosity of the liquid to ensure the stability and convergence of the optimization process. The boundary protection unit limits the steam pressure to a safe range, ensuring the safe operation of the system.

[0137] Specifically, the gradient correction unit in the parameter adjustment module optimizes the steam pressure by calculating the steam pressure adjustment amount ΔP = η·(dQ / dP), where η is the learning rate, which is used to control the amplitude of the adjustment, Q is the quality assessment index, which reflects the quality level of the current concentration process, and P is the steam pressure. The phase change warning unit monitors the derivative of concentration with respect to temperature dC / dT. When it exceeds the set threshold (such as 0.15 mg / ml / degree Celsius), it triggers the viscosity protection mechanism to prevent the viscosity of the liquid from rising sharply due to rapid temperature changes. The iterative controller automatically adjusts the optimization step size α = 0.1 / (1+0.05·μ) according to the real-time measurement of the liquid viscosity μ to adapt to the optimization requirements under different viscosities. The boundary protection unit ensures that the steam pressure P always remains within the safe range of 0.2 MPa to 0.8 MPa to prevent process instability or equipment damage caused by excessively high or low pressure.

[0138] Preferably, in order to further improve the adaptability and reliability of the parameter adjustment module, it can be refined or replaced in the following aspects. First, the learning rate η in the gradient correction unit can be dynamically adjusted according to the actual process requirements. For example, in the early stage of optimization, a larger learning rate can be set to speed up the convergence speed, while in the later stage of optimization, the learning rate can be reduced to improve the accuracy. Secondly, the threshold value in the phase change warning unit can be adjusted according to the physical properties of different Chinese medicines. For example, for certain Chinese medicine ingredients that are easy to crystallize, the threshold value can be appropriately lowered to trigger the protection mechanism in advance. In addition, the step size adjustment formula in the iterative controller can be optimized according to the actual viscosity variation range, such as introducing a nonlinear adjustment mechanism to better adapt to high viscosity conditions. The pressure range in the boundary protection unit can also be adjusted according to the pressure resistance of the actual equipment and the process requirements to ensure that the system operates safely under a wider range of working conditions.

[0139] In some embodiments, the gradient correction unit performs the following operations:

[0140] (1) Calculate the momentum gradient ▽′ = ρ·▽ + (1-ρ)·▽ prev, where ρ is the momentum factor and ρ = 0.8, ▽ is the current gradient, ▽ prev is the previous gradient;

[0141] (2) Add boundary penalty term ▽″=▽′+λ·sign(▽′)·max(0,|P|-P max ), where λ is the penalty coefficient and λ = 0.5, sign is the sign function, P is the current steam pressure, P max is the upper pressure limit and P max =0.8 MPa;

[0142] (3) Correcting parameters through projection operator:

[0143] If P old -η·▽″ <P min , then P new =P min , where P min =0.2 MPa;

[0144] If P old -η·▽″>P max , then P new =P max ;

[0145] Otherwise P new =P old -η·▽″.

[0146] It should be noted that the gradient correction unit is an important part of the parameter adjustment module. Its core function is to achieve precise adjustment of steam pressure by calculating the momentum gradient, adding boundary penalty terms, and correcting parameters through the projection operator. This process not only optimizes the steam pressure, but also ensures that it is within a safe range, avoiding process problems caused by pressure anomalies. Among them, the calculation of the momentum gradient takes into account the influence of the current gradient and the previous gradient, the boundary penalty term is used to prevent the steam pressure from exceeding the set range, and the projection operator correction parameter ensures that the steam pressure adjustment meets the process requirements.

[0147] Specifically, the operation process of the gradient correction unit includes the following key steps. First, calculate the momentum gradient ▽′=β·▽+(1-β)·▽prev, where β is the momentum factor, usually set to 0.8, which is used to balance the influence of the current gradient ▽ and the previous gradient ▽prev. Second, add the boundary penalty term ▽″=▽′+λ·sign(▽′)·max(0,|P|-P max ), where λ is the penalty coefficient, usually set to 0.5, sign is the sign function, P is the current steam pressure, P maxis the upper pressure limit (usually 0.8 MPa). The role of the boundary penalty term is to suppress the further increase of pressure by increasing the penalty term when the steam pressure approaches or exceeds the upper limit. Finally, the parameters are corrected by the projection operator: If P-η·▽″ <P min , then set the steam pressure to the lower limit P min (usually 0.2 MPa); if P-η·▽″>P max , then the steam pressure is set to the upper limit P max Otherwise, the steam pressure is adjusted to P-η·▽″. This process ensures that the adjustment of steam pressure meets the optimization requirements and is within the safety range.

[0148] Preferably, in order to further improve the performance and adaptability of the gradient correction unit, the following aspects can be refined or replaced. First, the value of the momentum factor β can be adjusted according to the actual process requirements. For example, in a process that requires a fast response, the value of β can be appropriately reduced to reduce the influence of the previous gradient; and in a process that requires smooth adjustment, the value of β can be appropriately increased to increase stability. Secondly, the value of the penalty coefficient λ can be adjusted according to the sensitivity of the steam pressure. For example, for a concentration process that is more sensitive to pressure changes, the value of λ can be appropriately increased to enhance the effect of the boundary penalty. In addition, the correction logic of the projection operator can be optimized according to the actual process requirements, such as introducing a nonlinear adjustment mechanism to better adapt to the pressure adjustment requirements under different working conditions. It is also possible to consider introducing an adaptive mechanism to dynamically adjust the momentum factor β and the penalty coefficient λ according to the real-time monitored pressure change rate, thereby further improving the intelligence level and optimization effect of the system.

[0149] In some embodiments, the feedback control module includes:

[0150] Quality assessment unit, calculates enrichment quality indicators

[0151] Q=w1·(C herb / C target ) 2 +w2·(1-E energy )+w3·(1-T err )

[0152] Among them, w1=0.6 is the efficacy weight, w2=0.3 is the energy consumption weight, w3=0.1 is the temperature error weight, T err =|T evap -T target |;

[0153] Parameter updater, when Q th When adjusting the number of hidden layer nodes of the neural network, Q th =0.85 is the quality threshold; ​

[0154] A crystallization warning unit reduces the concentration rate when it detects dμ / dt>0.1 Pascal second per minute, where dμ / dt is the rate of change of viscosity;

[0155] Case database, storing the best evaporation curves of different medicinal materials.

[0156] It should be noted that the feedback control module is a key component of this system's quality monitoring and dynamic adjustment of the TCM concentration process. Working collaboratively, it assesses concentration quality in real time through a quality assessment unit, a parameter updater, and a crystallization warning unit. It adjusts system parameters based on the results to ensure the stability and consistency of the medicinal solution's quality. The quality assessment unit quantitatively evaluates the concentration process by calculating concentration quality indicators, taking into account factors such as efficacy, energy consumption, and temperature error. The parameter updater dynamically adjusts the neural network structure based on the evaluation results to accommodate different medicinal materials and process requirements. The crystallization warning unit monitors the rate of change in viscosity to prevent crystallization and ensure the smooth progress of the concentration process.

[0157] Specifically, the quality assessment unit in the feedback control module calculates the concentration quality index

[0158]

[0159] Among them, w1, w2 and w3 are the efficacy weight, energy consumption weight and temperature error weight respectively, which are usually set as w1 = 0.6, w2 = 0.3 and w3 = 0.1. eff is the current concentration of active ingredients in the liquid medicine, C tar is the target concentration endpoint, E eng is the normalized energy consumption error, T err Is the temperature error. The parameter updater determines whether it is necessary to adjust the number of hidden layer nodes of the neural network based on the value of the quality index Q. th (The quality threshold is usually set to 0.85), the system will adjust the number of hidden layer nodes to optimize network performance. The crystallization warning unit monitors the viscosity change rate. When it exceeds a set threshold (eg, 0.1 Pascal seconds / minute), the concentration rate is reduced to prevent the drug solution from crystallizing.

[0160] Preferably, in order to further improve the performance and adaptability of the feedback control module, the following aspects can be refined or replaced. First, the weight parameters w1, w2 and w3 in the quality assessment unit can be adjusted according to the actual process requirements. For example, in a process with extremely high requirements for drug efficacy, the weight of w1 can be appropriately increased; and in a process that is sensitive to energy consumption, the weight of w2 can be increased. Secondly, the quality threshold Q th ​Dynamic adjustments can be made based on the optimal concentration quality of different medicinal materials to better accommodate diverse process requirements. For the crystallization warning unit, the viscosity change rate threshold can be optimized based on the composition and physical properties of the medicinal solution. For example, for high-concentration solutions, the threshold can be appropriately lowered to provide early warning. Furthermore, adaptive mechanisms can be introduced to dynamically adjust weight parameters and thresholds based on real-time monitored quality indicators, further enhancing the system's intelligence and quality control capabilities.

[0161] In some embodiments, the adaptive update formula of the drug efficacy weight w1 is:

[0162]

[0163] Where μ = 0.05 is the update rate, R = 0.95 is the efficacy benchmark value, Σ represents the summation operation for j = 1 to 2, E j It is divided into energy consumption and temperature error.

[0164] It should be noted that the adaptive update formula for the efficacy weight w1 is a key mechanism used in the feedback control module to dynamically adjust the efficacy weight. This mechanism aims to optimize the weight distribution in real time based on the actual results of the concentration process, thereby better balancing the relationship between efficacy and other optimization objectives. The efficacy weight w1 is updated based on the difference between the current efficacy and the set baseline value. By incorporating feedback information on energy consumption and temperature errors, the weight value is dynamically adjusted to meet the optimization needs of different stages. This adaptive update mechanism ensures that the system balances energy consumption control and process stability while pursuing high efficacy, thereby achieving refined optimization of the traditional Chinese medicine concentration process.

[0165] Specifically, the adaptive update formula of the efficacy weight w1 is:

[0166]

[0167] Among them, α is the update rate, usually set to 0.05, which is used to control the amplitude of weight update; R is the efficacy benchmark value, usually set to 0.95, which represents the desired efficacy level; is the partial derivative of the quality index Q with respect to the efficacy weight w1, reflecting the impact of changes in the efficacy weight on the overall quality index. Through this update mechanism, when the current efficacy weight is below the baseline value, the system increases the weight to improve efficacy. Conversely, when the efficacy weight is too high, the system appropriately decreases the weight to avoid excessive pursuit of efficacy at the expense of energy consumption or other optimization objectives. This dynamic adjustment ensures that the system consistently maintains the optimal weight distribution under different operating conditions.

[0168] Preferably, in order to further improve the performance and adaptability of the adaptive update mechanism of efficacy weight, it can be refined or replaced in the following aspects. First, the update rate α can be dynamically adjusted according to the convergence speed in the actual optimization process. For example, in the early stage of optimization, α can be appropriately increased to speed up the weight adjustment speed, so that the system approaches the target weight faster; in the later stage of optimization, α can be reduced to improve the adjustment accuracy and avoid weight fluctuations caused by excessive adjustment. Secondly, the efficacy reference value R can be adjusted according to the requirements of different traditional Chinese medicine concentration processes. For example, for some traditional Chinese medicines with extremely high efficacy requirements, the value of R can be appropriately increased to ensure that the concentration process achieves a higher efficacy level. In addition, the partial derivative The calculation of the drug efficacy can be improved by introducing more complex numerical optimization methods, such as finite difference methods or automatic differentiation techniques, to improve the accuracy and efficiency of the calculation. It is also possible to consider introducing multi-objective optimization algorithms, such as Pareto optimization, to combine the update of drug efficacy weights with the update of weights of other optimization objectives (such as energy consumption and temperature error), thereby achieving more comprehensive dynamic optimization.

[0169] The above-described embodiments of the present invention have the following beneficial effects: The system uses a multi-source data acquisition module to acquire real-time data on liquid concentration, evaporation temperature, and vacuum during the Chinese medicine concentration process. Combined with a data preprocessing module and a feature extraction module, the system performs data standardization and multidimensional feature extraction, providing high-quality input data for the optimization model, thereby enabling precise monitoring and analysis of the concentration process. The Chinese medicine ingredient retention rate prediction neural network and dynamic weight adjustment mechanism in the optimization model module dynamically adjust weights based on actual operating conditions, balancing the relationship between efficacy, energy consumption, and process stability. Simultaneously, the logic judgment module performs multi-objective optimization decisions, generates a Pareto-optimal solution set, and selects an optimization solution that prioritizes efficacy. This effectively improves Chinese medicine concentration efficiency and active ingredient retention, reduces energy consumption, and ensures the stability of the concentration process and the reliability of the liquid medicine quality. The parameter adjustment module, through functions such as gradient correction, phase change warning, iterative control, and boundary protection, enables dynamic adjustment and optimization of concentration process parameters, further enhancing the intelligent level of the concentration process. The feedback control module, through functions such as quality assessment, parameter update, and crystallization warning, assesses concentration quality in real time and adjusts system parameters, further ensuring the stability and consistency of liquid medicine quality.

[0170] Furthermore, the system utilizes a multi-spectral sensor array, distributed temperature sensor array, and near-infrared online component analyzer in the multi-source data acquisition module to monitor changes in the liquid's transmittance, temperature gradients within the evaporator, and active ingredient concentrations in real time, providing comprehensive data support for the concentration process. The dynamic weight adjustment mechanism and concentration-dependent sensitivity coefficient calculation method in the optimization model module enable refined weight adjustments, further improving the model's prediction accuracy and optimization effectiveness. The construction of multi-objective functions and verification of process constraints in the logic judgment module ensure that the concentration process achieves optimization goals while meeting process requirements. The gradient correction unit and boundary protection unit in the parameter adjustment module effectively prevent anomalies during parameter adjustment, ensuring stable system operation. The quality assessment unit and parameter updater in the feedback control module dynamically adjust the neural network structure based on concentration quality indicators, further enhancing the system's adaptability. The introduction of a case database stores optimal evaporation curves for different medicinal materials, providing a reference for subsequent production and further improving the system's versatility and practicality.

[0171] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or a terminal device such as a computer, a server, a mobile phone, or a tablet.

[0172] The above descriptions are merely some preferred embodiments of the present invention and an illustration of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present invention is not limited to the technical solutions formed by a specific combination of the above-mentioned technical features, but should also encompass other technical solutions formed by any combination of the above-mentioned technical features or their equivalents without departing from the above-mentioned inventive concept. For example, a technical solution formed by mutually replacing the above-mentioned features with (but not limited to) technical features having similar functions disclosed in the embodiments of the present invention.

Claims

1. An intelligent Chinese medicine concentration scheme optimization system based on neural network, characterized in that: include: Multi-source data acquisition module, real-time acquisition of liquid concentration data, evaporation temperature data and vacuum degree data during the Chinese medicine concentration process; Data preprocessing module, used to standardize multi-source sensor data; Feature extraction module, extracting multidimensional feature vectors related to concentration efficiency; Optimization model module, including a neural network for predicting the retention rate of Chinese herbal medicine ingredients and a dynamic weight adjustment mechanism; Logical judgment module, which performs multi-objective optimization decision-making on evaporation temperature, concentration and energy consumption; Parameter adjustment module to achieve iterative optimization of concentration process parameters; Feedback control module generates steam valve opening control instructions and evaluates liquid medicine quality indicators.

2. The system according to claim 1, wherein: The multi-source data acquisition module includes: The drug liquid multispectral sensor group is used to monitor the changes in the light transmittance of the drug liquid; Distributed temperature sensor array for measuring the vertical temperature gradient ΔT inside the evaporator z ; Near infrared online component analyzer for real-time detection of astragaloside IV concentration C herb and baicalin concentration C scu ; A signal conditioning unit, used to convert the sensor analog signal into a digital quantity; A time series database is used to store structured data packets at preset time intervals.

3. The system according to claim 1, wherein: The dynamic weight adjustment mechanism of the optimization model module performs the following steps: (1) Calculate weight sensitivity Among them, E represents the model loss function, which is represented by α·E herb and (1-α)·E energy Additive composition, α is the efficacy weight factor and 0<α<1, E herb is the effective ingredient retention rate error, E energy is the concentrated energy consumption error, W is the neural network weight matrix; (2) Generate adjustment coefficient α(t) = η·tanh(β·||δ|| F ), where η is the baseline learning rate and η = 0.01, β is the concentration-related sensitivity coefficient, ||δ|| F represents the Frobenius norm of the weight sensitivity δ; (3) Update weight W(t+1)=W(t)-α(t)·δ+γ·ΔW(t-1), where γ is the momentum factor and γ=0.9, and ΔW(t-1) is the previous weight update; (4) Spectral normalization is performed when ||W(t+1)||2>θ, where θ is the weight norm threshold and θ=1.2, and ||·||2 represents the L2 norm.

4. The system according to claim 3, characterized in that The calculation formula of the concentration-related sensitivity coefficient β is: β=1 / (1+exp(-k·(C herb / C target -0.9))) Where k is the shape factor and k = 2.5, C herb is the concentration of the active ingredient in the current liquid medicine, C target is the target concentration endpoint, and exp represents the natural exponential function.

5. The system according to claim 1, wherein: The logic judgment module performs the following steps: (1) Constructing a multi-objective function for Chinese medicine concentration F=ω1·f temp +ω2·f herb -ω3·f energy Among them, f temp =1 / (1+|T evap -T target |) represents the evaporation temperature approximation, T evap is the real-time temperature measurement value of the evaporation tank, T target is the target temperature and T target =65℃; f herb =log(C now / C initial ) represents the effective ingredient retention rate, C now is the current concentration of the drug solution, C initial is the initial drug solution concentration; f energy =∑(P steam t+P vacuum ·t) represents the total energy consumption of the concentration process, P steam is the steam generator power, P vacuum is the vacuum pump power, t is the running time; (2) Verify process constraints: T e vaping m ax <T safe , where T e vaping m ax is the maximum temperature of the evaporator, T safe is the safety temperature threshold and T safe =75℃; ΔC / Δt <C crit , where ΔC / Δt is the concentration change rate, C crit is the phase transition critical value and C crit = 0.2 mg / ml / min; (3) Generate a Pareto optimal solution set and select the optimization solution that prioritizes drug efficacy.

6. The system according to claim 5, characterized in that The weight coefficient ω i The adjustment formula is: ω i =exp(s i ) / (exp(s1)+exp(s2)+exp(s3)) Where s1=0.8·E temp +0.2·ΔE temp , s1 is the temperature control item, E temp is the mean absolute error of temperature, ΔE temp is the temperature error change rate; s2=1.2·E herb -0.3·ΔE herb , s2 is the efficacy priority, E herb =1-C now / C target is the concentration achievement rate error; s3=0.5·E energy , s3 is the energy consumption suppression term, E energy is the normalized energy consumption error.

7. The system according to claim 1, wherein: The parameter adjustment module includes: Gradient correction unit, used to calculate steam pressure adjustment Where η is the learning rate, Q is the quality assessment index, and P is the steam pressure; A phase change warning unit triggers a viscosity protection mechanism when it detects dC / dT > 0.15 mg per milliliter per degree Celsius, where dC / dT is the derivative of concentration with respect to temperature; Iterative controller, automatically adjusts the step size η = 0.1 / (1 + 0.05·μ) according to the viscosity of the liquid, where μ is the real-time measured viscosity of the liquid; Boundary protection unit, used to limit the steam pressure P in the range of 0.2 MPa to 0.8 MPa.

8. The system according to claim 7, characterized in that The gradient correction unit performs the following operations: (1) Calculate the momentum gradient Where ρ is the momentum factor and ρ = 0.8, is the current gradient, is the previous gradient; (2) Adding boundary penalty items Where λ is the penalty coefficient and λ = 0.5, sign is the sign function, P is the current steam pressure, P max is the upper pressure limit and P max =0.8 MPa; (3) Correcting parameters through projection operator: like Then P new =P min , where P min =0.2 MPa; like Then P new =P max ; otherwise 9. The system according to claim 1, wherein: The feedback control module includes: Quality assessment unit, calculates enrichment quality indicators Q=w1·(C herb / C target ) 2 +w2·(1-E energy )+w3·(1-T err ) Among them, w1=0.6 is the efficacy weight, w2=0.3 is the energy consumption weight, w3=0.1 is the temperature error weight, T err =|T evap -T target |; Parameter updater, when Q th When adjusting the number of hidden layer nodes of the neural network, Q th =0.85 is the quality threshold;​ A crystallization warning unit reduces the concentration rate when it detects dμ / dt>0.1 Pascal second per minute, where dμ / dt is the rate of change of viscosity; Case database, storing the best evaporation curves of different medicinal materials.

10. The system according to claim 9, characterized in that The adaptive update formula of the efficacy weight w1 is: Where μ = 0.05 is the update rate, R = 0.95 is the efficacy benchmark value, Σ represents the summation operation for j = 1 to 2, E j It is divided into energy consumption and temperature error.

Images